Page 1

Richard F. Taitano

Micronesian Area Research Center Archaeological Excavations at

non nok tha northeastern Thailand 1965 - 1968

Part 1 by

donn bayard and

wilhelm g. solheim ii

An online publication of the Micronesian Area Research Center University of Guam

Aerial photograph of the Non Nok Tha region, showing the northeastern side of Phu Wiang, the Phong River, and the site itself (arrowed). The tree-covered village of Ban Na Di is located one km north of the site, across the lighter-colored rice paddies.

Archaeological Excavations at Non Nok Tha Northeastern Thailand 1965 – 1968 Region, Site, and Synthesis

By Donn Bayard and Wilhelm G. Solheim II with the assistance of Virgil Meeker and a contribution by R. H. Parker

Dedicated to the memories of Achan Chin You-di “the father of Thai prehistory” Hamilton Parker and Ludy M. Solheim

Contents PART 1 Preface By Donn Bayard


1. Introduction: The Thai-Hawai`i Archeological Salvage Program1 and Non Nok Tha Wilhelm G. Solheim II


2. The Phu Wiang Region and the Khorat Plateau


Donn Bayard 3. Excavations at Non Nok Tha: 1965 – 1966


R.H. Parker Introductory note by Donn Bayard 4. The 1968 Excavation


Donn Bayard 5. Chronology


Donn Bayard 6. Ceramics


Donn Bayard 7. Non-Ceramic Artifacts


Donn Bayard Wilhelm G. Solheim II 8. The Burials and Their Implications


Donn Bayard PART 2 9. Non Nok Tha in Economic, Social and Regional Context: Conclusions


Donn Bayard References




PREFACE By Donn Bayard This report has been an inordinately long time in preparation and publication; the first drafts of Chapters 1 and 2 were written in the early 1970s, while draft versions of Chapters 8 and 9 were not completed until the end of 1987. In his introductory chapter, Solheim outlines the factors leading to this thirty-year lag between excavation and publication, chief among them the 8,000 km gap between me at Otago and Solheim and the excavated material in Hawai`i. The delay is particularly regrettable because Non Nok Tha has generated considerable interest and debate due to the evidence for early metalworking that it produced and the controversial chronology supporting this evidence. In our defense, however I should add that a considerable body of publications-–almost 700 pages appeared about the site during the years 1967-1984 (see Chapter 5, footnote 4 for details). Given this delay, the volume of previous publications, and the fact that many excavations in Thailand since 1970 have yielded richer and more reliably provenanced and dated materials than those from the pioneer excavation at Non Nok Tha, one might ask why this report is being published at all. First, at 340 m2 it remains to the present day the largest excavation in terms of area in Northeastern Thailand, and as such it produced an enormous quantity of data, much of which has not yet been published in any detail. Secondly, as one of the readers of the manuscript stated, “its historical importance remains undiminished”: The real importance of Non Nok Tha lies in the fact that the unexpected data it produced and the controversies it spawned stand at the beginning of a new era in Thai and Southeast Asian archaeology. Non Nok Tha and the archaeologists working there have been instrumental in motivating the series of new and vigorous excavation and survey projects that have led to a substantial rethinking of the prehistory of Southeast Asia (Karl Hutterer).


Non Nok Tha could thus be viewed as having provided a “kick start” to modern Southeast Asian prehistory as far as interest by Western archaeologists is concerned; this is reflected by the number of entries in the bibliography of this report that postdate 1968 as opposed to those published earlier. However, I should note that no such impetus was needed for Vietnamese archaeologists, who were already engaged in a vigorous program of research into their own past Non Nok Tha was being excavated. The report is organized along traditional lines. Solheim’s introductory chapter discusses the background of the survey program that led to Non Nok Tha’s discovery by Chet Gorman and the other more minor excavations carried out during the 1963 – 1966 survey. My second chapter discusses the environmental and ethnological background of the Phu Wiang area. R. H. Parker and I outline the procedures, stratigraphy and problems we encountered in the 1965 – 1966 and 1968 excavations at the site in Chapters 3 and 4. The complex and still-controversial dating of the site is addressed in Chapter 5, and Chapter 6 provides full data on the large number of whole ceramic vessels and the less impressive volume of sherdage recovered from the site. Chapter 7 covers the non-ceramic artifacts; I discuss the metal and metal-related materials, while Solheim describes the stone, shell, and bone material from the site. In Chapter 8 I provide a fairly full summary of the site’s 224 burials, their contents, and implications for possible prehistoric social organization, and the concluding chapter attempts to provide a fairly traditional (non-nomothetic) culturehistorical syntheses of the Phu Wiang region during the period spanned by Non Nok Tha. The planned second volume of this report (all of which is already written in draft form) will include detailed grave-by-grave descriptions of the burials; specialist reports on the physical anthropology of the Non Nok Tha skeletal material; soils, ceramic, and bronze analyses; and details of the faunal remains from the site, with a list of the fauna exploited by contemporary villagers. A note on orthography: In translating Thai and Northeast Thai (Phasa Isan) words and site names, I have almost in all cases attempted to follow a simplified version of the standardized system used by the Thai Royal Institute and the American Library of Congress (personal and common place-names like Bangkok, Mekong, and Vientiane are the chief exceptions). Thus the nine vowels of Thai and Phasa Isan are translated as follows: i, e, ae (front); u’, oe, a (back rounded) and u, o, o (back rounded). For typesetting reasons, the last two could not be properly distinguished as o and o in the text, nor could aspirated and unaspirated ch; vowel length and tone are also not shown. In a few cases the use of this system results in a spelling different from that preferred by the excavators of a site; in such cases, the excavators; spelling is given in brackets when the site name first appears in the text, as with “Non Pa Kluai [Kluay]” of “Nong No [Nor]”. Acknowledgments In his introductory chapter, Solheim thanks the many people to whom he is indebted for planning and carrying out the 1963-1966 program and its aftermath. But there are also many people to whom I am grateful for help and advice in my part of the excavation and analysis. We are both thankful for the sound advice offered by the two readers of the draft manuscript; their comments did much to improve this report. Portions of my part of the manuscript were also read by Athol, Anderson, Charles Higham, Michael Pietrusewsky, Vincent Pigott, and Joyce White, and I am grateful for the useful suggestions they provided. I am particularly grateful to Brian Vincent for undertaking ceramic analyses of


some of the exotic sherds from Non Nok Tha, and to both Vincent and Joyce White for providing me with unpublished provisional data on sherd densities at Khok Phanom Di and Ban Chiang respectively. Over the past twenty-odd years I have received much sound counsel on metallurgical matters from the late Cyril Smith, from Robert Maddin, and particularly from Noel Barnard; I wrote my first letter to Noel from the porch of my village house just after the completion of the 1968 excavation and have valued his friendship and advice ever since. Correspondence and conversation over the years with archaeologists Bill Macdonald, Judy McNeill, David Welch, and Richard Wilen on Non Nok Tha and related matters have also proved of great benefit to me.

In Thailand there are many debts to acknowledge. My American assistant in 1968, Terry Marsh, was able to lift from me the burden of handling a four-wheel drive vehicle, as well as doing quite creditable field drawings. Among Thai officials, Dr. Kavi Chutikul, then Dean of the Faculty of Agriculture at Khon Kaen University, kindly offered me laboratory space and other assistance in 1968; the Nai Amphoe of Phi Wiang District was also most helpful. I greatly enjoyed working with the late Khun Viraj Khunnamas, the Thai Fine Arts Department representative on the 1968 excavation, and am grateful to Khun Charern Poltachar, his equivalent at the beginning of out fieldwork in 1965, for sympathetically introducing a somewhat culture-shocked Farang graduate student to village life in Phak Isan. But most of all, both Solheim and I wish to acknowledge the longstanding friendship and interest of the late Achan Chin You-di, who might with full justification be called, “the father of Thai prehistory.� We respectfully dedicate this report to his memory. On a more personal level, I am equally indebted to the villagers of Ban Na Di adjacent to Non Nok Tha during my eight months there; their tolerance of and interest in a young Farang have made a life-long impact in me. I dedicated my 1971 doctoral dissertation to them collectively, but particular thanks are given to Nai Bunma Ngaemdaeng, assistant


headman; to the late Nai Thiang Nuthao for sharing his ecological information with me; to my friend Khamporn Phithaksin, assistant in 1968 and during the later Pa Mong survey; and to the late Nai Bao Phimsi, my “foster father” who always provided me with humorously given but very sound advice. Finally, after thirty years of continuing friendship and correspondence, I, like Solheim, am particularly grateful for the support and fellowship of Thaweechai Uthaiwee. Chong hai thuk khon thuk than yu di mi haeng! In Hawai`i there are also people I must thank. Both Solheim and I are deeply grateful to our wives, Daisy Bayard and particularly the late Ludy Solheim, for many hours spent reconstructing pots, and for putting up with the family disruptions that the preparation of this report has occasionally entailed over the past 27 years. Personally, I am extremely grateful to my Hawaiian `ohana, Ruby and Leighton Donlin and Priscilla and the late Fred Cappis, for our Hawai`i stays in 1980 and 1987; their help with everything-–and their general aloha-–proved invaluable. Karen Essene, Judy Ito, and Walter Yee of the University of Hawai`i Computer Center provided valuable help to us on computing matters, as did Harry Partika. Geologists Larry Lepley and John Belshé supplied provisional classification of lithic materials from Non Nok Tha. Over the years, Peter Gilpin spent many hours in the lab and darkroom producing photographs of Non Nok Tha artifacts. I thank the many University of Hawai`i graduate students who helped us analyze the Non Nok Tha materials; Solheim mentions most of these in his introductory chapter, but I must add special thanks to Ragnar Schousboe for his careful and consistent analysis and computer coding of many of the Non Nok Tha restorable vessels. As far as the production of this report is concerned, both Solheim and I are deeply indebted to the wholly voluntary efforts of Virgil Meeker over the past twenty years: Out of interest alone, he undertook the heavy burden of organizing the figures for this report as they were gradually executed in Hawai`i and Otago, and he did a superb job, He also devoted his efforts to keeping the Non Nok Tha collections in orderly storage while they were being analyzed, as well as aiding Solheim in analysis of artifacts and vessel reconstruction. Without him this report would not have been completed. Finally, I also thank the drafting staff at the Otago Anthropology Department, who spent long hours over the years since the early 1970s producing many of the figures presented here. Peter Duncan, Sue Wilson, and Murray Webb were involved earlier on, but for the past fifteen years Martin Fisher has been struggling on and off with often muddy and sweat-stained field drawings, and I gratefully acknowledge his patience and persistence, Our secretarial staff and Computing Centre personnel have also offered much useful help and advice over the years, and I am of course grateful to the University of Otago for support during leaves in 1980 and 1987 to allow me to complete portions of this report. Funding for the field research was provided by a number of grants from the American National Science Foundation, as Solheim details in the introductory chapter.


Chapter 1 Introduction: The Thai-Hawai`i Archaeological Salvage Program and Non Nok Tha By Wilhelm G. Solheim II The salvage archaeology program undertaken by the Fine Arts Department of Thailand and the Department of Anthropology of the University of Hawai`i began in the field I northeastern Thailand in August 1963 with its primary support from a U. S. National Science Foundation Grant (GS-288) followed by a second grant (GS-523 and GS-956). Our work in the field came to an end in June 1966. Events leading up to this program go back to December 1958. The program evolved, and the first stage continued in the field until 1968. Our program was the first organized research in prehistoric archaeology in the plateau area of northeastern Thailand. We found much exciting and unexpected data that led directly and indirectly to an almost complete revision of Southeast Asian prehistory. This final report on one site discovered during the first year of our program in Thailand should support at least some of the claims we have made for what were at the time several elements of an astonishingly early and high development of culture in northeastern Thailand. My first visit to Mainland Southeast Asia was in October 1958, when I stopped in Thailand and Cambodia on my way from London to Kuching, Sarawak, to take up a Fulbright Fellowship at the Sarawak Museum. I met several of the Thai archaeologists at that time, including the late Achan Chin You-di, who was associated with out program in Thailand through all if its stages. In December 1958 I stopped in Saigon, Viet Nam for a week and there met Olov R. T. Janse, who had excavated at Dong Son in the late 1930s. He told me about the Mekong Valley Project, with its headquarters in Bangkok, we agreed on the need for a large-scale salvage program in the areas where various kinds of development would take place to rescue data on past and present peoples before these data were destroyed or irreparably disturbed. Upon completion of my Fulbright Fellowship in Sarawak in June 1959, I returned to the United States for a few months and then back to Kuala Lumpur in February 1960 for a five-month program with the National Museum of Malaya. In June 1960 I went to Bangkok, renewed acquaintances at the National Museum, and talked with Achan Chin and others about an archaeological salvage program for the areas to be disturbed by the Mekong Valley Project, without specific reference to Thailand. They agreed it was a program that should receive high priority. I was able to visit with C. Hart Schaff, the Executive Agent for the Mekong Development Committee. He was interested in the idea of an archaeological salvage program in connection with the project, and from that time on until he left that position, he gave me moral support and encouragement in my attempts to organize an international salvage program for the lower Mekong Valley. I spent considerable time and energy over the next five years or more trying to develop an international program along the lines of the Nile Valley Salvage program in connection with the building of the Aswan Dam, but I was never successful. I received moral support from many people, but without some impressive archaeological ruins to point to that would be destroyed by the Mekong Project, it was impossible to attract the


attention needed to have institutions and governments commit money and time. My major reason for wanting an archaeological salvage program for the area was that prehistorically, the whole area was almost completely unknown; even historically it was little known. We did know that the high, early historic civilizations of Funan, Chenla, and Cambodia were present in the Mekong drainage, so it was logical that there would have been people living in the area before this; but you don’t get commitments of money and time to work on an unknown quantity, at least in archaeology. Robert Hackenburg joined me for a time in trying to develop an interest in the people who would be directly affected by the Mekong development—-those who would be displaced or other wise directly affected by specific dams. At that stage, in the early 1960s, almost all work was concentrated on technical aspects of the project and there was little feeling among the technical experts that the people were of any relevance to its development. We tried to point out otherwise (Solheim and Hackenberg 1961), but we were not overly successful. The same attitude among members of the Mekong Committee was encountered by Bayard twelve years later. I began to realize in the fall of 1961 that while international programs were fine, I was getting nowhere and work on some dams would soon begin. The first area where Mekong dams were to be built was in northeastern Thailand. I decided that if I really believed in what I was saying, I would have to start developing a specific program for this area. I stopped in Bangkok twice in 1962 to talk with Thai archaeologists at the National Museum and officials of the Fine Arts Department. I talked with Dhanit Yupho, the Director-General of the Fine Arts Department, and he approved of the idea but indicated that their small budget was committed to other projects and that I would have to find the necessary financing. I was fortunate to gain two American allies who were living in Bangkok at the time to help with the development of the project. One of these was Elizabeth Lyons, an American specialist supported by the U.S. Department of State. She was an advisor to Director-General Dhanit Yupho. The other was Patricia Young, the wife of the American ambassador to Thailand at that time. Without their organizational assistance in Thailand, it is questionable whether the project would have come about. I submitted a proposal to the U.S. National Science Foundation in the fall if 1962 for the funds for a three-year salvage program in northeastern Thailand, to be a joint project of the Fine Arts Department of Thailand and the Department of Anthropology of the University of Hawai`i. The Program Director for Anthropology, to whose division the proposal was directed, let me know that the National Science Foundation rarely supported salvage programs as such. Because nothing was known of the area or nearby areas, there was no other way in which I could focus my proposal. It was rejected; primarily because they felt it too much of a gamble to support a three-year program when it was not known whether anything was to be found in the area. They suggested I apply for a one-year project and if anything was found I could reapply for a second and third year. This I did. In August 1963, in Thailand, after making the first exploration of our project, I received word that my proposal had been accepted.


First Year Chester Gorman, a graduate student in the Department of Anthropology, University of Hawai`i, arrived less than ten days after we received word of NSF approval of the project. He and Nai Vidya Intakosai, for the Fine Arts Department, began the systematic field survey in northeastern Thailand, while I had to return to Hawai`i for fall semester classes. I returned to the field with Allen Mosher, another graduate student in the University of Hawai`i Anthropology Department, to join Gorman in early March 1964. We continued working into May when the rains started. Twenty-one sites were found during the first season’s explorations in three different reservoir areas (Fig. 1-1). Two small rock shelter sites, both probably prehistoric, were found in the Lam Phra Phloeng Reservoir area, ca. 50 km southwest of Nakhon Ratchasima. From these sites simple tools of sandstone and utilized sandstone flakes and cores were recovered, in association with earthenware potsherds in the upper levels (Solheim and Gorman 1966:155–123). The Lam Pao Reservoir stretches in several directions from the subsequently constructed man dam, which is ca. 22 km north of Kalasin. We found ten sites in this reservoir Fig. 1-1 area, all of them historic. Three of these sites had standing or fallen carved stones that were probably from the Dvaravati Period, the earliest historic period of Thailand. One of these sites (Lam Pao 7) had a group of standing stones in an oval immediately adjacent to a wat compound (124-163) Ubonrat Dam (See Fig. 1-1) is about 47 km northwest of Khon Kaen and now backs up the Ubonrat Reservoir. In this area we found nine sites, three of them historic in their upper levels but prehistoric below, and two probably completely prehistoric. The three partly prehistoric sites—Nam Phong 6, 7, and 8-— had quantities of cordmarked pottery and some artifacts of bronze. Nam Phong 4 was a sandstone


shelter where very few artifacts were found but past of the rock surface had a number of geometric figures worked into it, including several that looked like bronze celts. Nam Phong 9 was a sandstone shelter with the red outline of nine human hands on the wall (Solheim and Gorman 1966:164–179). A final report of this first season’s discoveries has been published (Solheim and Gorman 1966), and several shorter articles include information on the program and the first season’s progress (Solheim 1962, 1963, 1964a, 1966a-c, 1967a). Additional financial support for the first season’s program came from the Mosher Fund for Southeast Asia Prehistory. Esso Standard Eastern furnished all petrol supplies, and these were considerable. This assistance we warmly welcomed. Second Year With further NSF funding, the plan for the second year of the project was to make small excavations in a few of the sites discovered during the first season’s explorations, carry out more intensive exploration in the neighborhood of those sites, and to test the Nam Phong sites 6, 7, and 8 to be able to select one of these, if they had any depth, for a major excavation during the third year of the project. As was the situation the first year, I had to be back in Hawai`i to teach the first term and could only return to Thailand in February the following year. Ms. Ernestine Green, a graduate student in archaeology from the University if Hawai`i, and Mr. Verapong Pengprecha for the Fine Arts Department began work in November 1964 at the Lam Phra Phloeng dam site. They excavated Lam Phra Phloeng 1 and 2, located two other small rock shelters in the same area and excavated these, and made further exploration up the valley, where they found nothing. (This move up the valley was the first time any of my students used elephants for transportation.) The artifacts, used stone, and unused stone from these sites (Solheim and Gorman 1966:116-123) were brought to Hawai`i where they were analyzed by myself, partly with the assistance of the archaeology graduate students in the University of Hawai`i, Department of Anthropology. The three Nam Phong sites were tested and the Nam Phong (NP) 7 was selected for the primary excavation to be made during the third season. A summary report on the test excavations at these three sites is presented below. I returned to the field in late February 1965 with my wife and daughter, and we joined the party at Lam Pao 6 (Nong Nok Tha, Solheim and Gorman 1966:160-161) shortly after they had arrived there. Here we excavated the total area within the oval of standing stones. The most common artifact from this site was roof tile. There were washed, weighed, and counted at the site during excavation. Because they were so bulky and heavy, many of them were placed together and left in the excavated area, which was refilled. The rest of the artifacts and tiles were brought back to Hawai`i for description and analysis. This work has been completed by Jane Allen and, it is hoped, will be published in the future. After Lam Pao 7 we moved to Sahatsakhan, where we excavated Lam Pao 6, a site that we thought was similar to Lam Pao 7. Very little material was recovered from this site. Summary of Test Excavations Made in Nam Phong 6 (NP 6)


Ernestine Green and Verapong Pengprecha conducted test excavations in Nam Phong (NP) 6 from January 8 to 17 1965. This site is somewhat irregular mound called Don Kok Pho (Bo Tree Rise; see Fig. 1-2). It is surrounded by small rice paddy fields and there is a small stream flowing just to the north of the mound. Its maximum measurements both north south and east west are about 300 m. In 1965 it was cut approximately in half by an east-west-running fence, which went across the highest portion of the mound. This northern half of the mound sloped downward toward the stream at the north side. The center of the mound was about 6 m higher than the northern side. The grayish black soil of the mound was distinct from the red soil of the surrounding paddy fields. A garden of watermelon and maize was present at the western side of the portion north of the fence at the time of testing 1965. The first reported cultivation of a portion of the mound was said to have taken place in 1963 (Solheim and Gorman 1966:175). Fig. 1-2


Fig. 1-3

This mound was recognized was an archaeological site by Gorman at the end of the first field season in May 1963 because of the considerable quantity of potsherds on its surface of both porcelain (from Sukhothai, Ayutthaya, and Ming China) and unglazed earthenware (Solheim and Gorman 1966:174-176). The earthenware sherds, including many that were cord-marked, indicated considerable variety in form for this pottery. A spindle whorl was found, as well as fragments of clay pipes, metal, a marine bivalve shell, and a broken tektite. Surface search before the testing was begun turned up more of the sherds, pipe fragments and pieces of broken grinding stones. The earthenware sherds were of several different pastes and tempers and included a few incised sherds. Ten test pits were excavated by 10 cm levels. The pits measured 1.0-1.5 m and were placed along the north side, not including the garden area, and across the center of the mound (Fig. 1-3). It had been intended to follow natural layers in excavation as soon as these could be distinguished, but in the small pits no stratigraphy was evident. Even with the one extended burial that was found, no indication of its pit was evident. The collections from the test pits have not been analyzed, so little can be said about the site of value for the comparisons with the obviously related sites of Nam Phong 7 (Non Nok Tha), Nam Phong 8, or the sites inside Phu Wiang. The earthenware pottery in the top 30 cm was primarily plain, with cord-marked pottery common from level 4 and below. From levels 6 and 7 in several squares along the north side, as well as level 7 in Square 7, were found a number of very thick sherds with laterite pebble temper, as much as 3 to 4 cm thick (refer to Chapter 6). Similar very thick pottery has been found at other sites in northeastern Thailand. Fragments of shell and animal bones were found in the top two levels of all squares except in Square 8, where shell and bone were also found in level 3; these remains were not found at lower levels. A few metal (probably iron) fragments were not found in level 3 of Square 10 and level 4 of Squares 7 and 8. A few small red glass beads were found in level 9 of Square 8 and level 10 of Square 9.


A burial with the skeleton extended on its back was found in levels 11 and 12 of Square 8. The body had been placed roughly NNW to SSE with the head towards the south. One earthenware vessel had been placed on top of the right led near the foot; another pot at the outside of the left thigh, and to the left near the head of the left femur was an iron axe of hoe blade. It could well be that the red beads found in level 9 were associated with this burial. Two unusual features were found. A few pieces of fired clay, apparently daub from wattle-and-daub structures, had been collected from the surface. Daub was found only one of the test pits, but in this pit it was found in seven different levels. In test pit 4, daub was recovered from levels 3, 4, and 9 through 13. In the lower set of levels, it was found scattered in considerable quantity. Wattle-and-daub structures are in use at the present time in Ban Na Di, a small hamlet nearby and neighboring Nam Pong 7, for granaries (P1. 3a) built totally above ground. If one of these buildings were burned, the fragmented fired-clay daub would form a small mound on the surface of the ground. If undisturbed for many years and it was not washed away, it would form a deposit that when excavated would be found in two or three 10 cm levels, but probably no more. The fact that daub was not found in any of the none other test pits suggests that the daub from levels 3 and 4 was not a random find but had some relationship with the daub found in the lower levels. Without further excavation, I can think of only two possible explanations. One would be that a wattle and daub structure, first built at the time level 13 was the surface, was rebuilt at some time following its destruction in the same location as the earlier structure, and that this cycle of construction and destruction through burning was repeated several times. The second possible explanation would be that at the time when level 3 (or level 9) was the surface, a pit was dug down into level 13 and lined with a wattle-and-daub frame. If considerable organic material was put into the pit and this was fired before the nest rainy season, the daub could have been fired as well, and what remained in the wall —-not falling into the pit-—would on excavation be found at many levels in the circumferential squares. The other feature was in Square 10. From level 10 into level 14, at least 17 earthenware vessels containing cremations were found, apparently originally placed all together at one time, or over a relatively short period of time, without noticeable individual placement. In level 10, fragments of bronze (?) were found associated with the pottery—the only bronze found in these ten test pits. The excavation of this pit had run behind the others. The other pits hit sterile soil in levels 9 and 10 for Squares 1, 2, and 3, in level 11 for Square 4, in 12 for Squares 6 and 8 and in level 13 for Squares 6 and 9. Square 5 contained very little cultural material in level 14, while Square 10 still had several vessels in level 14. They had to be left in the test pit, as the move to Nam Phong 7 had to be made without further delay. Nam Phong 7 (NP 7, Non Nok Tha) The purpose of this work is to present the final report for the excavation of Nam Phong (NP) 7, so it might seem superfluous to present a review of the testing that led to its being chosen as the site for the major excavation in this three-year program. It seems to me, however, that it may be of interest and value for the reader to understand why we


chose this site as the major focus of a three-year program that cost the American taxpayers, in the final analysis in the neighborhood of $300,000 and the Hawaiian and New Zealand taxpayers many tens of thousands of dollars to pay for various portions of the analysis and preparation for publication of the data recovered as a result of the excavation (Frontispiece). The test excavations, as for NP 6 and NP 8, were conducted by Green and Verapong from January 18 to 27, 1965. I start this review for the testing with a quote from the first entry in Green’s field notes: “The 10 workmen, I and the entire village of Ban NA Di began work at NP 7 this morning, There were at least 300 people here at one time; how many total, I’m not sure.” This is an indication of the sort of interest we usually had of our work in the field. We hope that in some way this final report will repay the support and assistance we had at all levels from the people of Thailand in the pursuance of this program. Fig. 1-4

The site and its physical setting will be described in detail in later chapters, so this information is not duplicated here (P1.1a). A very brief description of the site and surface finds there from was presented in the final report of the first year of our program (Solheim and Gorman 1966:176-177). The test pits were originally of two sizes; wither 1.5 by 1 m or 2 by 1 m. Three pits were extended when a burial was found in each; this was in order to recover the total burial so as to have a better idea of the contents and context of the burials. As with NP 6 and NP 8, the test pits were excavated in 10 cm levels, as natural levels could not be ascertained. The location of the seven pits, scattered on the mound, is presented in Figure 1-4. The most exciting result of the testing of NP 7 was the finding and depths of the metal fragments and artifacts. “Metal”


was reported from level 1 of Square 4, level 2 of Squares 2 and 4, and level4 of Square 6. These finds were probably iron, as starting with level 5, in Squares 1 and 4, bronze was reported and the term metal was not used. If the metal found in the top four levels had been bronze, it would most likely have been reported as such for bronze finds had been very rare up to this point in our work. More bronze was reported from level 6 in Squares 2, 3, and 4 and from level 7 in Squares 4 and 7. This indicated that bronze-— and no other metal-—was widely scattered in the site from levels 5 through 7. “Metal” was again reported from Square 4, levels 9 through 11, but in this case it probably meant bronze. In Green’s field notes for 21 January, clarifying some sketches concerned with level 9 of Square 4, just below where she had reported “metal” for that level, she wrote, “So far this site appears to be a bronze producing cremation mound.” Many cremations had been reported from all squares, often in jars (now usually badly crushed), but at times apparently not in jars. In the same squares and levels, human bone concentrations were often reported. Some or many of these concentrations may have been badly disturbed extended burials; five such burials were reported, and two of these when first discovered were called bone concentrations until further portions of the skeletons were found. A presumed secondary burial (called “Burial X” in the overall analysis; see Chapter 8) was found in levels 12, 13, and 14 of the extensions of Square 2, which were made to uncover the total length of Burial 5 (“Burial Y” in level 14. Considerable numbers of whole and broken but restorable earthenware vessels were found associated with burials and bone concentrations. Miscellaneous finds included small fired-clay balls and a few spindle whorls, as found at NP 6. A large painted potsherd with a yellow and red pattern on a black body was found in level 12 of Square 3, was well as thin, incised black potsherds. Several pieces of polished stone were found associated with burials of bone concentrations, including a portion of a black stone bracelet from level 11 of Square 7. No metal of any sort was found below level 11. Square 1 was sterile from level 9, Squares 6 and 7 from level 13, Squares 2, 3, and 5 from level 14, Square 5 from level 15 and Square 4 from level 18. Nam Phong 7 was obviously rich in burials of several different kinds. Green had remarked that the pottery from the lowest three or four levels of most of the squares was different from the pottery of the upper levels, though cord-marked pottery had been found at all levels. The single painted sherd was the only one that had been recovered from archaeological testing in northeastern Thailand up to that time. It also appeared that iron had been in use during the time the four levels were accumulating, that bronze may have been in use without iron while levels 11 through 5 were being deposited, and that the previous level 11 no metal was locally in evidence. If this appearance held true over the excavation, this would be the first known site in Southeast Asia to have clear and possibly lengthy use of bronze previous to and not in association with iron, This was an obvious site for careful excavation. Nam Phong 8 (NP 8) This site is on a large mound called Don Po Daeng (Red Kenaf Rise), which in turn is on a large knoll of decomposed laterite, about 2.5 km NNW of Ban Nong Waeng. The area of the mound is about 300 m east west by 500 m or more north south (see sketch map, Fig 1-5). Five test pits, each 2 by 1.5 m were put in and excavated by 10 cm


levels. Surface collections were similar to those from NP 6 (Solheim and Gorman 1966:177-178). Excavation was done from February 2 to 4 and 12 through 16: the site, sitting directly on Fig. 1-5

laterite, is quite shallow compared to the other sites and probably relatively recent. Square 1 came to sterile laterite in level 6. At the base of this square, in a pit dug into the laterite, was a burial that included glass beads. From a bone concentration immediately above this burial came an iron axe or hoe. Another iron axe was recovered from Square 3 in level 5; sterile laterite was encountered in level 8, while level 7 was a mixture of soil and laterite without artifacts. The sterile laterite was hit in Square 4 in level 6 and in Square 5 in level 7. Only Square 3 extended as deep as level 9, with level 10 sterile. In this square, heavy bone concentrations cut down into the laterite from level 5. Third Year Three excavations were planned for the third season’s work in 19651966. The first excavation at Nam Phong 7 (Ban NA Di) was to be from the middle of November until the end of January; the second, on Ban Sao Lao (Lam Pao 3), was to be in February and March; and the third, in Phimai in April until the rains started sometime in late May or June.

My family and I returned from a short vacation on September 1 and we started assembling equipment and supplies. The Fine Arts Department built a small screened-in laboratory on the ground floor of our house, which was in the Fine Arts Department compound at Phimai. The laboratory included ample storage space with shelves and worktables with good lighting. We could work late into the evening on days when it did not rain. The rain brought out swarms of termites in the evening and-—even with screens—-they would somehow get inside, in large numbers, to the light. Donn Bayard, at that time a graduate student at the University of Hawai`i arrived late in September. With the assistance of Jack Golson, I was able to add Hamilton (“Ham”) Parker from New Zealand to our crew. With the reputation he had in New Zealand for


being a first-class excavator, I brought him to take charge of the excavations. Not only did Parker turn out to be an excellent excavator, he was also a first-rate teacher. In spite of speaking no Thai at that time — and our workers and most of the Fine Arts University archaeology students understood and spoke little or no English — Parker was able to communicate his methods of excavation quickly, so that in a few days all of those who approached their tasks with a willing and open mind became competent excavators, and few-—among both the locally hired workers and the students-—became almost as good as Parker. Sadly, Parker died in 1994 after a long and productive life in Archaeology (Prikett 1994, Cooper and Sutton 1994); Bayard and I are sorry this report could not appear while he was still alive (P1.1b Solheim, Bayard and Parker in front of the field lab at Phimai). With the arrival of Parker in late October we started to go over the collections from the Nam Phong6 and 7 test pits to acquaint ourselves with the materials we could be excavating and to develop some idea of the possible strafgraphy. As the condition of the road for the last 6 km into Nam Phong 6, 7, and 8 sited was extremely bad, we delayed our departure to the field while waiting for missing parts for our winch. We finally gave up on the winch and left Phimai on December 9, arriving in Ban Na Di the morning of December 10 and starting to work on NP 7 the same day. Parker’s account of the site and its excavation comprises Chapter 3. I returned to Phimai after the beginning of the excavation and during that season’s excavations spent most of my time in the lab, working on the Lam Phra Phloeng collections. In November and December 1965 and first several months of 1966, I had to make at least one trip every month to Bangkok for visa extensions, some supplies, and cash for the payroll. Banditry of a traditional nature was still common. Every hamlet reportedly had two or three resident bandits. These men never robbed in their own hamlet but supposedly passed word on to their associates at some distance when any unusual amount of cash was expected to be on their hamlet. Unknown assailants on at least two different occasions relieved residents of hamlets where we were working (during our three-year program) of a considerable amount of cash that had been acquired the same day. We were never bothered, though we strongly believe that on at least one occasion we were looked over carefully by bandits-—visitors who were completely unknown to the local population. In anticipation of such a possible action, we never kept more than about $20 worth of Thai baht in our field camps. I had a secret compartment built into our vehicle for our payroll. I went to the site at least twice a month an as soon as possible after my arrival, we paid all the workers. There was no market at Ban NA Di, or at several other sites where we worked, so food was not easy to buy locally. I always brought in a considerable quantity of food and enough fresh met and vegetables to last for several days. When it cane time to move to Ban Sao Lao, we had not reached sterile soil in any of our squares because the site had been so rich-—and therefore slow in excavation. Our crew had been working seven days a week from about 9:00 in the morning straight through to about 2:00 in the afternoon. In the Philippines, Sarawak, and west Malaysia I had always started to a site where we were working around 6:00 in the morning so that we were at work at 7:00, taking advantage of the relative cool in the morning. We tried in Thailand to start early in the morning but we were never successful. The Thai in the


country get up at sunrise or before, like other farming people, but then they sit around and talk, have breakfast, and don’t start making preparations to move out until around 8:00. We’d usually get to the excavation site about 8:30 of so and it was usually 9:00 before everything was moving – and it was already hot. The Thai workmen would work straight through; with about a half-hour break sometime after 11:00 for a small snack, and by 2:00 they were exhausted-—as were we. The workmen were willing to continue working later than this, but they were so tired that both the quantity and quality of their work dropped off rapidly, so there was no sense in trying to work longer. With the sevenday-a-week schedule, everyone needed a break. On January 30 we closed down, to return to work again in February 11. The site continued to be rich, and by the end of February the bottom still was not in sight. I decided that we could continue the excavation until the end of March and reduce our other excavations to one month each. By the end of March we knew where the bottom of the site was, but we had reached it in only one or two squares. The bottom levels were so interesting-—and so much richer than the later levels-—that we could not stop at that point. I postponed our closing of the site again, this time until April 15. On April 13 when I arrived to help with the packing up and moving out, they had been able to complete only a few more squares, bring most of the rest of the squares down to the most interesting bottom levels, and remove some of the baulks. We could not stop excavation in that situation. I made an ultimate postponement. The rainy season would probably be starting sometime in May, so we had to be started at Bao Sao Lao by May 1 regardless of what we had to leave unexcavated in the site. We would close down excavation on April 27 and move out on April 28. This we managed, though we left ten of our earliest-—and richest-—burials unexcavated. After the first two months, several of our local laborers had become expert excavators. Only Parker and Bayard were able to photograph and record the features and determine the stratigraphic position of these features. The excavation kept getting farther and farther ahead of the recording, even with several Silapakon students pressed into service as draftspersons, and as a result more and more human burials were left on slabs awaiting recording and then removal. While in Bangkok on one trip I mentioned this to Dr. Sood Sangvichien, the anatomist who had worked with the human skeletal material of the Thai-Danish expedition. He offered to come up to the site for a few days to remove and pack the skeletons for us. We had arranged earlier for him to do the analysis of any human skeletal material we might find. He came to Ban Na Di with three students for three days, all in his own funds, and they took care of our backlog of recorded but unremoved skeletal material. As it developed later, Dr. Sood was so busy that he would be unable to get at the description of the skeletal material for a long time. I was able to make arrangements with Michael Pietrusewsky, out physical anthropologist at the University of Hawai`i Anthropology Department who had specialized in working with prehistoric skeletal material in Bangkok at Dr. Sood’s laboratory. We were greatly indebted to Dr. Sood for his assistance and interest in our program. We thank him very much. Two activities outside of normal excavations took place at Non Nok Tha during the excavation. (First referred to by its number as Nam Phong 7, we changed this to Ban Na Di when we started the excavation. Only later did we discover that there was a local


name for the mound itself and thereafter we called it by that name, Non Nok Tha, or “Partridge Mound”) We were visited by two geologists so that they could look at the soils of the site, and I did some experimentation in sampling the unexcavated portions of the site with a core soil sampler. The two geologists were H. Sawata, on ECAFE staff with the Mekong Valley Project, and Michael N. Alekseev, a visiting Russian geologist with an interest in tropical soils. They were able to be Way from Bangkok for only one weekend. Because of the bad roads (the Friendship Highway was not totally completed and the road from Khon Kaen was not well graded) and a Bangkok driver who did not know how to cope with washboard roads, they only had two hours at the site. We had excavated a special trench for them so that the soil in the wall would not be sunbaked before they examined it. I quote their brief report sent in a letter dated February 2, 1966: As the time of our stay at your excavation site a Bang Nodi [sic] was very short, we could not examine in detail the geologic conditions of the excavated profile. However, our observation at that time was that there are two layers of soil (upper is recent and lower is buried), which are separated by thin (10 cm or less thick) bed of slightly rewashed material. This bed, apparently, would be corresponding to some interruption in disposition and formation of soil [see P1. 1a]. The soils are considered to be related with the Swamp area in this origin as shown by their change in color form red to gray, and probably the rainy and dry seasons interchanged as the filled desiccation fissures suggest. The excavation, when the geologists were at the site, was only well down into the third soil so they did not see the total soil profile. While they did not mention it near in their report, Alekseev, while at the site, remarked that the main (lower) soil looked like a tropical podzol and that more research was needed on this type of soil in tropical areas. We are indebted to Sawata and Alekseev for their observations and thank them very much. Late in the first season’s field survey (mid-1964), Gorman and I began to feel that a tool for bringing an undisturbed core would be very useful in locating open sites and in working out the horizontal and vertical extent of a site. Late in 1964 I received a grant from the Wenner-Gren Foundation for Anthropological Research for purchase and shipment of such a kit to Bangkok. I thank the Wenner-Gran Foundation for this assistance. The kit, from the Acker Drill Company, arrived in Bangkok in April 1965, but I was unable to remove it from the customs warehouse underway Christmas Eve, 1965. Excavation of Non Nok Tha was well underway when I brought the kit to the field in January 1966. The first two cores extracted brought to light several problems in the use of the tools. But they also demonstrated that undisturbed cores could be used to extend the stratigraphy known from excavation to some distance beyond that are being excavated. The major problem with the corer was its propensity for making a slight compaction of the soil so that it became almost impossible to extract the core from the sleeve into which it had been forced without breaking the core into little pieces. The soil in Non Nok Tha is clayey, and during the dry season it becomes extremely hard. If you drive the corer in too far so that the soil starts to push against the end of the tube, it compacts the


already very hard soil and the core is most difficult to remove from the liner. A split liner probably would have solved this problem, but we did not have one of these. With these tools and the hard Non Nok Tha soil, I was able to extract in three days of work only five reasonably undisturbed cores that extended into sterile soil. Examination of these cores enabled Parker to extend a generalized stratigraphy well beyond the limit of the excavation in two directions. We took no further cores. We now had only the month of May, depending on when the rains started (we already had two or three rather heavy rains at Non Nok Tha that had done some damage), to make two excavations for which we had planned to use two months each. I decided that we should split up with Bayard in charge of the excavation at Ban Sao Lao (Pen Post Village) and Parker the excavation at Phimai. We therefore first moved directly to Ban Sao Lao (Lam Pao 3; Solheim and Gorman 1966:132-158) and helped Bayard get started on April 29 after which Parker and I returned to Phimai. I include here Bayard’s preliminary report on the Ban Sao Lao excavation (Solheim. Parker, and Bayard 1966); the materials recovered from this site have been published (Bayard 1977:93-96). Excavation at Ban Sao Lao, Satsakhan, Kalasin Fig. 1-6

During the 1965-1966 field season, the Program excavated a small site on the edge of the Lam Pao Reservoir, which is currently in the process of being filled. The site, designated Lam Pao 3, was tested in the spring of 1964 by Solheim and Chester F. Gorman. Results seemed productive enough to warrant further excavation, and it was decided to devote two weeks to a further testing of the site with an aim to: (a) obtain a wider picture of the stratigraphy encountered in the 1964 testing, and (b) procure a


sufficient sample of the pottery from the middle and lower levels to make analysis possible. Lam Pao 3 was excavated in the period from April 29 to May 14 by Donn T. Bayard, with the assistance of Thawee Uthaiwee, three students from Silapakon University, and four workmen brought from the excavation at Nam Phong 7 (Ban Na Di). The site itself is located about 300 m. northeast of the village schoolhouse (Rongrian Soem Sao Lao Witthaya); it is a low, oval-shaped mound about 300 m east-west by 100 m north-south. The edges of the mound are not clearly defined because the sides of the mound have been cut by rice terraces. An area between the two test squares excavated in 1964 was selected for excavation. Four 4 x 1 m trenches were laid out to form an “L”; the longer of the two sides was oriented to the north, the shorter to the west. Later a fifth 4 x 1 m trench was dug, extending the east-west section 5 m farther to the east in order to investigate further a low mound about 10 m in diameter located at the junction of the side of the original “L” (Fig. 1-6). All five trenches were excavated to the natural. The 11 x 14 m sections thus obtained indicated that the area excavated was formerly on the edge of the large mound; some ten of the twenty-one layers distinguished were water-deposited bands of brownish to pinkish sand containing water worn potsherds, separated by thin, dark brown lines of weathered sandy soil. Some six of the layers (including the modern ground surface) showed evidence of occupation in the form of eleven pits and postholes, as well as relatively extensive concentrations of sherds and charcoal. These layers, together with the sandy layers eroded from them, appear to define six occupational levels. An older seventh level is inferred from the presence of a few sherds and flecks of charcoal in several of the lower water-deposited sandy layers. Although it is difficult to draw firm conclusions from the limited area excavated (24 m2), this portion of the large mound seems to have been located on the edge of any settlement in the area and was apparently used primarily as a work area during most of its history. A brief summary of the evidence encountered for each of the occupation levels follows: I: No structural features were found, and only a limited number of water-deposited sherds were obtained from the lower sandy layers. II: An occupation surface was present only in the southeastern corner of the site; this contained a fairly high concentration of potsherds and one pit. The sandy layers eroded from this layer also yielded a considerable quantity of pottery. III: Occupation extended over a wider area of the southern and eastern portions of the site. The most striking structural feature encountered was a large bell-shaped pit some 130 cm wide at its bottom, narrowing to 110 cm near its top. The pit’s bottom was filled with charcoal, and pits of this type are still used in the area for making charcoal. Following its use for charcoal burning, this pit was used to deposit refuse on two separate occasions; large quantities of cow and/or buffalo bone were found. The upper part of this level yielded a sizeable amount of pottery, including a distinctive wide-mouthed jar some 40 cm in diameter. IV: This level is separated by a thin brown weathering band from the levels above and below it, but it has similar pottery. Little structural evidence was encountered.


V: The dominant feature of this level was a dense layer of irregular lumps of baked clay extending over the southeastern and south-central portions of the site; the small mound at the junction of the two lines of trenches is largely the result of this clay deposit. The clay may be refuse from kiln located in the vicinity. In addition, this layer yielded much charcoal, potsherds, and a cluster of badly rusted iron objects, some of which appear to be knife blades. VI: (“premier”): Little pottery of structural evidence was encountered. The area was apparently cultivated in rice following this period and then abandoned some decades ago; the dikes of the rice field that cross the mound are still present and in fairly good order. VII: (modern): Several pits and postholes were found, but very little pottery. The area is now in pasture and scrub brush. A brief preliminary analysis of the sherds from the occupation layers and the layers eroded from them seems to indicate that levels I-V essentially represent a continuous occupation by a single culture. In levels I and II a reddish gray sand-and-fiber-tempered ware is the most common domestic pottery; in levels III, IV, and V this gradually gives way to a thicker black pottery tempered with fiber only and often coated with a whitish or pinkish slip. Both of these types are rare and almost certainly intrusive in the upper two levels, the pottery that seems essentially modern. A similar progression of types in the lower levels and replacement by modern wares in the upper levels seems to occur in the frequency of several varieties of imported earthenware and stoneware. Two carbon samples, from levels II and III, have been submitted for dating this far. While it is hard to give an estimate for the age of the lower levels of the site, pieces of celadon and crudely made celadon imitations were found in level II, and it thus seems unlikely that the lowest levels in the site predate the Sukhothai period (see Bayard 1977a:96 for full discussion of the chronology). The two 14C dates from this site are: from E4, lower layer 10, 440±120 BP (Gak-990); and from E3, layer 7 pit bottom, 1600±80 BP (Gak-989). The few water deposited sherds found in the bottom level indicate that there is an earlier occupation elsewhere in the mound, probably toward the southeast as suggested by the distribution of artifacts in the second and third levels. The 14C date from level II, A.D. 1404 (1450) 1635, seems reasonable for that level considering the celadon sherds recovered. The date of A.D. 396 (440) 554 is unexpected for the bottom of the pit dug from level III. It may be that the charcoal was from heartwood of a large old tree, or it is possible that the sample was actually from the level of the bottom of the pit. In any case, while there may be prehistoric levels elsewhere on this site, none were located in the 1966 excavation. Phimai Excavations Phimai is noted for having the largest Khmer ruins outside of Cambodia. With our headquarters in the Phimai Fine Arts Department compound and our laboratory there, we were in close contact with the people working on the restoration program of a portion of the Khmer ruins, underway at that time. In 1862 we realized that there was a distinctive black pottery that extended to a considerable depth underneath the Khmer (Solheim 1965). In 1965 the Fine Arts Department told us about a site in the northeast


corner of Phimai where a quantity of this black pottery was being uncovered and expressed their interest in our making a small excavation there. We had visited the site several times before we started out excavation and made surface collections from an area where the town people were removing soil. Parker started the excavation on May 9, 1966 and continued until May 30, by which time heavy rains were becoming fairly common. All collections from the excavation were brought back to the University of Hawai`i. A preliminary report in the Phimai black pottery from this site appeared in 1979 (Solheim and Ayres 1979). The month of June 1966 was spent in organizing the collections, rebagging, and packing for shipment to Hawai`i. There had already been some major damage to collections made during the 1964-1965 excavation season. Level and artifact collections we had first put into heavy plastic bags with paper labels. The plastic bags were in turn placed in paper bags made from heavy cement bag paper. Until the lab was completed, a number of boxes of these bags had been stored on a cement floor on our kitchen. When we unpacked these bags to put them in boxes, we found that termites had started in one corner of the temporary storage pile and invaded about six boxes. They had gone right through the paper sacks, riddled the heavy plastic bags, and in many cases eaten up the labels. Our salvage from these boxes amounted to 10 to 25% of the bags in each box. We completed the packing, shipped the bags down to a bonded warehouse in Bangkok, and made all arrangements for shipping the collections to Honolulu by the end of July 1966. This completed our three-year salvage field program in northeastern Thailand. Progress of Research, September 1966 to June 1987 The two large crates and one small crate of our collections were to have left Bangkok early in August 1966 directly to Hong Kong. There they were to be off-loaded and transferred to another freighter bound for Honolulu. I expected the shipment to arrive in Honolulu around the middle of September. I had a new grant from NSF (GS-956) to cover expenses for a year and half of research on the collections. I had positions for two half-time graduate research assistants and a photographer for this length of time and enough money left from the field grant to continue Bayard half time for two terms. With the term beginning around September 26 and the shipment expected about the 15th, it seemed logical to fill these positions starting the fall term of 1966, which I did with Madge L. Schwede and Wei-Lan Wu. In spite of several urgent letters of inquiry to the shippers in Bangkok, I received no word about the shipment until the end of January 1967-—about a week before it arrived. Without artifacts to work on, I had the two students putting in their twenty hours a week assembling a bibliography of Southeast Asian prehistory. I don’t know why I didn’t develop an ulcer. The University if Hawai`i had turned over about half a temporary transit building for our lab. While there was a reasonable amount of space here for tables for washing, repairing, cataloging, measuring, and photographing, including a corner for a wellequipped, small darkroom, there was rather a little space for storage. The three crates were unloaded behind our lab, and one end of each was opened. Everything inside appeared to be in good condition. We quickly discovered that we didn’t have nearly enough storage room to unload even half of the smaller of the two large crates, so the


crates themselves would have to serve as storage space. Rain and variable winds up to 50 km per hour are common in Manoa Valley where the university is situated. While the crates were solid, they were not waterproof. We covered the open ends with sheet plastic but these were quickly torn apart by the wind, and we discovered that the crates were leaking in other places. We then completely covered both of the large crates with a heavy sheet of plastic, and this appeared to keep out the rain. By the time I left for Europe on sabbatical leave in August 1967, we had completed work on about two-thirds of the largest crate. The slightly smaller of the two large crates was removed to a university warehouse, and work continued after I left on the remaining material in the largest crate. Unhappily, we found that there was some damage from water to a few of the boxes nest to the end we had opened, which had happened during the first few days before we put on the heavy plastic sheeting. This amounted to less than 10% of about six boxes; we were able to salvage most of the collections from destroyed bags through partial information that we could make out on the labels and our bag lists. The great majority of our materials from Non Nok Tha were of two different sorts: either from burials or more rarely other kinds of features, or the non-feature contents of a layer. By far the most common item from the burials and other features was pottery, as was the case with the layer contents. The burial and feature pottery, however, was either whole or from considerably restorable vessels, while the layer pottery was sherds with relatively few that fit together. Many of the burial and feature vessels were packed in the field with their soil contents and brought to Hawai`i in this condition. While at Phimai we had tried to pack in such a way that when we unpacked in Hawai`i, everything would come out in order. Naturally enough, in packing the crates in Bangkok the boxes were all mixed up. Without a great amount of storage space in our lab to sort the boxes, we had to deal with each box as it came from the crates. We did do a bit of sorting so that we worked with the burial and feature pottery first, saving the layer bags for last. Our order of procedure was to unpack a box, sort out any layer bags, and then work with a pot at a time. The person working with the pot would remove the soil carefully to save any sherds, bones, or artifacts in the pot, wash the potsherds, after drying glue them together with a water soluble glue (so that we could later take them apart for returning to the National Museum in Thailand), catalog it, measure, and describe it. The vessel was then drawn and photographed. My first photographer was Bruce Erickson, followed later by Peter Gilpin. All the data on the vessel were then put on a 5� x 8� McBee card. The source of the vessel, measurements, and elements of form and surface treatment were punched in around the edge of the card, a description put on the face of the card, and a contact print of the vessel glued in one corner. While the McBee cards have been useful, this method of data storage was discontinued after about seven hundred cards had been completed because it was taking so much time. Smaller source and descriptive data were also put on IBM cards for each vessel, and later for each layer bag. (This was obviously before small computers came into use.) Bayard received an NSF grant (GS-1877) in support of his Ph.D. thesis research. In January 1968, he returned to Thailand to make a second excavation at Non Nok Tha to


concentrate on the earlier levels of the site that we had been unable to treat adequately in our 1966 rush to close down. After the 1965-1966 season with us, Parker had returned to New Zealand and joined the staff of the Anthropology Department of Otago University. Charles Higham became the Foundation Professor of that department and a made a decision that the department would include Southeast Asian prehistory as one of its special interests. Parker returned to Thailand in 1967 and 1968 with the British Commonwealth Expedition, led by William Watson, which was working primarily in central Thailand. They did include a brief further test of NP 6 but decided not to excavate there. I returned to Hawai`i from my sabbatical year in August 1968 to find an improvement in our lab situation. In the spring of 1968 Tom Maretzki, the Chairman of the Department of Anthropology, had been able to get a second half-Quonset building for archaeology and all of the Thailand materials from our first lab had been moved, making the first lab available for Hawaiian archaeology (and also mixing up what had been organized in our first lab). We could not move in the rest of the 1964-1966 collections still stored in the university warehouse, however, as the collections made by Bayard were on their way to Honolulu and we had to keep the space available so that they could all be unpacked and brought in as soon as they arrived. His collections, also greatly delayed, arrived in March 1969. During the university teaching year, we had a number of volunteers from the Anthropology 210 (Beginning Archaeology) and Anthropology 460 (Regional Archaeology: East Asia and the Pacific) courses, plus graduate seminar students, who worked with us on the Thai materials. With the return of Bayard’s collection, all of these resources were brought to bear on his materials, and the processing of the earlier collections slowed down to a trickle. The intention to collaborate on Thai prehistoric research between the Anthropology Departments of the University if Hawai`i and Otago continued to develop. In October 1968 I went to New Zealand as a guest of the department in Otago University, where we made specific plans to start a joint program in the fall of 1969. I submitted a proposal the National Science Foundation for a continuation of the Mekong salvage program in areas of Thailand, Laos, and Cambodia that we had not yet surveyed. For this project we also applied to the Wenner-Gren Foundation and the Breezewood Foundation for moderate funds. The NSF indicated firmly that they would no longer support salvage archaeology projects, and that proposal was turned down. I had included in that proposal, however, a request for further funds to support work in the Non Nok Tha collections at Honolulu that were not yet processed, and this portion of the proposal was approved (GS-2659). The Anthropology Department of Otago University, to further strengthen its program in Southeast Asian prehistory, invited Bayard to join their staff, and they wanted him there to start the first term in February 1970. This did not give him nearly enough time to process his complete collection. As we had concentrated on processing the burials from our 1965-1966 excavations, he decided to analyze the layer bag materials first and then do what he could with the burial pottery in the time left. While he did a good start on the burials, he had to leave most of the burial vessels unfinished when he left for Otago University. He completed his Ph.D. thesis on Non Nok Tha, using the materials he had been able to process and his excavation data, in 1971 (Bayard 1971a). A report on this


work was published in the same year (Bayard 1971b). The final report on his 1968 excavation and its stratigraphy, revised from his thesis and the published report, is presented in this volume as Chapter 4. The Otago team went to Thailand in November 1969, even though lacking the major portion of the funds that had been requested in the NSF proposal. The primary reservoir survey they had intended to make became impossible immediately after their arrival in Thailand: Terrorist activity was prevalent in this area for the first time. They changed their plans and made surveys and excavations in the Roi Et area and inside the outer ring wall of Phu Wiang (Hingham and Parker 1970). One of our undergraduate students who graduated in December 1969, Mei Mei Burke, joined this project in January 1970. Phu Wiang is a large saucer-shaped sandstone monadnock that lies to the west and south of Non Nok Tha. Fig. 1-2, Plate 1c The 1970 survey confirmed our suspicions that sites inside this formation are closely related to and probably completely overlap Non Nok Tha in time. One or two of the sites may fill in the long gap that we have at Non Nok Tha from the late first millennium B.C. to A.D. 1000-1200. Non Nok Tha, as well as NP 6 and NP 8, are physically dominated by the abrupt rise of the outer side of Phu Wiang and are clearly part of what can be called the Phu Wiang region. We decided to try to organize a long-term program inside Phu Wiang to develop a detailed local sequence of cultures in a relatively small and sharply bounded area that included all of the ecological zoned of which we were aware on the Khorat Plateau. It is an ideal situation in all ways. Unfortunately, Phu Wiang was included in the development plans for the Pa Mong Reservoir, a large reservoir resulting from a dam of the main course of the Mekong. Because of its saucer shape and the narrow gap through which it is drained, it would be easy to dam this gap and make the whole formation into an auxiliary reservoir for the Pa Mong Reservoir. In my proposal to the NSF for our Phu Wiang program, I made the mistake of mentioning the salvage nature of the project. As it turned out, neither the Pa Mong Dam nor most other planned subsidiary dams have yet been built. (The formation would not have been turned into a reservoir until at least ten years after the Pa Mong dam was completed.) This project was rejected, at least partly because it was considered a salvage project. It is obviously important, however, to present Non Nok Tha in regional context; therefore, Chapter 2 of this volume presents a description of the Phu Wiang region as a whole. In August 1969 we moved into new quarters where we had much more room than before. (Unfortunately I was again not present when this move was made, so further mix-up of the collections resulted.) Most if the materials on which we had completed work had been returned to the National Museum in Thailand. After the remaining unworked collections had been moved from our Quonset lab to our new quarters, we still had plenty of room left. Thus we were finally able to unpack the smaller of the two large crates that had been sitting in the warehouse for the previous two years. Many of the boxes in this circle were in poor condition. Apparently this crate had also taken in some rain before we had completely covered it with the heavy plastic. Fortunately there was very little in this crate from Non Nok Tha. In fact, quite a bit of it was material on which we had completed work in Thailand that was supposed to have been left in


Phimai but had mistakenly been packed with the unanalyzed materials and sent to Hawai`i. With the new grant from NSF in 1970, I was again able to hire two half-time graduate assistants. For shorter or longer periods for the nest three and a half years, Jean Kennedy, Ned Ewert, Bertel Davis, and Takeshi Ueki held these positions. In addition, Ragnar Schousboe carried out the assembly and computer coding of a large number of 1968 burial vessels, and Earl Neller voluntarily did considerable photographic work for me. Davis continued doing photography and the work he was doing as a graduate assistant long after the grant funds were used up. It was not until late 1973 that I realized how much burial material from the 1968 excavation was left to do. I could see that if we completed our analysis of the 1965-1966 collections, we would not even be able to start work on this other material for several years. In correspondence with Bayard, Parker and Higham at Otago University, we tried to locate funds to send this material to them, but we were unsuccessful. We reluctantly decided that, rather than delay the final report for another several years while I completer the work on the 1068 material with only what voluntary labor I could find, we could send it back to Thailand unprocessed and not include it in our final report. We had complete coverage of the burial materials from the 1965-1966 excavations and would have had complete coverage of the layer bags for both excavations (if our plans had been carried out), plus a scattering of the burial data from 1968, so we felt that we would not be doing major damage to the final site report. During the spring term of 1974 I had students in one class completely process one vessel each of the 1968 burial pottery to acquaint them with earthenware pottery. They didn’t all complete their pots, so in the summer my wife and I completed them. Previously the soil inside the jars had been removed and examined quickly for microremains. After completing the pots started by the class, my wife Ludy decided that she wanted to do a few more. She has had no archaeological training, but after hearing some of us talking about screening and flotation of soil to find seeds and other botanical remains, she decided to examine the soil of a pot by putting it through a very fine screen in water and then carefully going over the material that was too coarse to pass through the screen. In the first pot she examined, she found what appeared to be a rice grain and several other possible seeds, as well as some very small bones, probably from fish. Nothing like this had been recovered from any of the other several hundred vessels from which the soil had been examined, except for one pepper seed and one rice grain from a much later pot. This first pot she water-screened was from a burial dug from the bottom layer of the 1968 excavation, and thus, it was the oldest rice grain we had. She then went ahead and did some ten more jars finding more possible seeds and bones. In one jar, also from a bottom (Early Period or EP 1) 1968 layer burial, she found two small nodules that appeared to be bronze. If these were bronze, this would mean that bronze was present-—and probably being cast-—from the beginning of the site instead of the third level from the bottom (EP 3). The nodules and a sherd from the jar were, for safety, hand carried to a mainland laboratory for examination and thermoluminescence (TL) dating. Unfortunately, the two nodules and the sherd disappeared (see Chapter 7 for a full discussion). The twenty or so vessels that we had gone over


that spring and summer included variant forms that I had not seen before. The new data that we had recovered from these few vessels forced me to reconsider my decision made earlier. We would have to delay our final report for several more years to that we could include the full data from the 1968 burials. During the spring of 1975, I was close to completing a first draft of this volume of the final report. Instead of using my sabbatical year of 1975-1976 to work on the final report, I gave in to my twenty years intention of making an archaeological survey of coastal Irian Jaya. As a result, not only was this first volume not completed, but also work on the undescribed portions of the Thailand collections came to a halt. During my absence a major portion of the completed collections were packed and shipped back to the National Museum in Thailand by Bert Davis, and the rest of the collection was reorganized in the process of redistributing our departmental laboratory space (considerably more mix-up of the collection predictably resulted). With several moderate-sized grants from the University of Hawai`i Research Council, work was resumed and continued, primarily on the 1968 burial pottery. During the summer of 1980, Bayard, on sabbatical leave from the university of Otago, spent five months in Hawai`i completing the measurements and description of the burial pottery and some of the miscellaneous finds that had been put aside until the last. During the spring 1981 term, two graduate students, Somsuda Rutnin and Diane Trembley, made detailed descriptions of some of the remaining miscellaneous finds and wrote reports on this material for a research seminar, and with two other much appreciated and needed University of Hawai`i research grants, Gilpin completed much of the photographic work while Judy McNeil, with the assistance of Jack Morin, two more or my graduate students, made drawings and attended to the other loose ends. Further grants from the University of Hawai`i have resulted in more drawings and photographs for the final report, while Bayard and drafting staff at the University of Otago have put in many hours of analysis, illustration, and writing on pottery and burials since the late 1960s. In Hawai`i, Virgil Meeker has been doing volunteer work for close to twenty years, first on reconstruction of pottery vessels and then from 1985 on the organization of the non-pottery materials for final description. Lately he has worked with us on its analysis. Bayard again came to Hawai`i for eleven weeks during the northern summer of 1987 to work with me. From May to August 1982 he had written the ceramics and bronze chapters. While in Hawai`i, from September 1987 to January 1988, he wrote the burials, chronology, and conclusions sections. Although not all of the 1966 excavation level bags have been analyzed, we reluctantly decided to go ahead with the write-up and publication of the final report without delaying further. With the 1968 level bags totally covered and many of the 1966 level bags completed (together comprising over 90% of the site total), we felt that we could move ahead without great damage to our conclusions. A former undergraduate anthropology major at the University of Hawai`i and then graduate student at the University of Michigan, William K. Macdonald, used our pottery data for his Ph.D. thesis at Michigan. Macdonald, Bayard, and (to a lesser degree) I were in correspondence for several years on this matter with enrichment resulting for all concerned. Macdonald’s results are discussed in Chapter 6; see also Macdonald 1980.


The early bronze at Non Nok Tha has been a subject of great interest to many, and numerous papers specifically on the bronze have been published. This is discussed in detail in Chapter 7 so I will not go into it here. The early dates for bronze use and manufacture suggested for Non Nok Tha have naturally led to controversy, particularly because the dates do not present a clear chronology for the site or for the bronze recovered from the site (Loofs-Wissowa 1983; Bayard and Charoenwongsa 1983; Solheim 1983; Higham 1984a-b; Bayard 1984a-b, 1987). A full discussion of the site chronology and the development of the controversy are presented in Chapter 5. Final Report on Non Nok Tha — 1987-1993 There have been several changes in our planned organization of the final report on Non Nok Tha. While more and more archaeological research is being done in the world today, I have the impression that, with the outstanding exception of Higham’s commendable final reports on Thailand sites that have been excavated under his cooperative direction with Thai archaeologists, fewer and fewer final reports are being published and many of these reports include much summarization. Because of the cost of publication it has become impossible to publish the data fully without either a very large subsidy or an extremely expensive publication, or both. With expensive publications, only a few libraries and a very few individuals can afford the reports; they should be available to a considerably wider, though still rather small audience. With this in mind, and in consultation with the University of Hawai`i Press Editorial staff, we decided that we wanted our final report to include all the data available, with a large number of illustrations. When accepted for publication, we would seek subsidation so that those people and libraries that should have these data available would be able to afford to buy the report. When Bayard returned to Otago in 1987, we hoped to have the report completed and ready to submit to the University of Hawai`i Press in early 1988 for possible publication. Bayard had his part competed in January 1988, but I was far from completing my small part. I had intended to retire at the end of the spring term in 1990 so that I could put full time into completing the report, but in negotiating my retirement plans with the Chairman of the Anthropology Department and the Dean of Social Sciences, I found they were willing to let me have one more sabbatical leave and retire at the end of the year following my return from sabbatical. As a result, I didn’t get back to completing my portion of Chapter 7 until my return from sabbatical in January 1991. The completed text was turned over to the University of Hawai`i Press in June 1991 for review. Most of what remained to be done at this point was to complete any missing illustrations, of which there were several. I would like to thank Jo Singer and Conrad Iger for their help with the photography, Hemantha Jayatilake from Sri Lanka and my stepson Ed Solheim for their drawings and Ed for his final computer preparation of the illustrations, and Virgil Meeker for all the help he has given in getting not only the illustrations but the entire final report ready for publication.


Because very few people are acquainted with northeastern Thailand, it took quite some time to find reviewers, but it was finally approved for publication late in 1991 and a Memorandum of Agreement for publication was signed on February 27, 1992. Everything was ready to go. It is extremely embarrassing to again have to explain why this did not come out in late 1992. After having spent close to $1,500,000.00 in various grants from many different sources to undertake the fieldwork, description and analysis of the collections and complete the text and illustrations we felt it would not be a problem to find the needed smalls subsidy to start publication of the final report. It turned out that this was not so. All sources we could think of were approached, but no funds were forthcoming. At the end of the millennium, we explained to the Editor of the University of Hawaii Press that we had not been able to find a subsidy to publish the final report. On January 9th, 2001 William H. Hamilton, the Director of the Press, sent the following document stating: The University of Hawai`i Press hereby returns to you all the fights granted to the Press through the Publishing Agreement dated February 27, 1992, for the unpublished work entitled Archaeological Excavations at Non Nok Tha, Northeastern Thailand, 1966-1968… Bill has indicated that you would now like to pursue publication opportunities with a Thai publisher who has expressed interest. We agree that it would be in the best of interests of the project for us to withdraw our interest at this time.” We here express out profound appreciation to the University of Hawaii Press for their understanding support of this project, its Director and Pam Kelley the editor in charge of seeing this volume through to publication, contributing hundreds of hours of time and major funding to produce the copyedited manuscript which was turned over to Bill Solheim in June 2001. I picked up the two heavy boxes containing the manuscript and illustrations in early September 2001 to hand carry back to the Philippines, where I am now living. I am making a final proofreading before sending the manuscript on to Pisit Charoenwongsa, Director of SPAFA (SEAMEO Project in Archaeology and Fine Arts) and a close and valued friend. With this help we hope to get a grant in Thailand to finally have this published at a reasonable price. Our intention is to publish a second volume that will present more detailed data. This will include the burials and their contents; the levels and their contents; the human skeletal remains as analyzed by Michael Pietrusewsky and Sheilagh T. Brooks; the funeral remains, both associated with burials and from the levels, by Charles Higham; and chemical soil analysis by Mary Nelson. I would like to thank the Fine Arts Department of Thailand and its several Director Generals since we started our program for their great patience in waiting for out completion of the work on the collections, Bayard’s subsequent excavation at Non Nok Tha and the ultimate return of the artifacts to the National Museum of Thailand. I hope when the final report has appeared that they will feel that their patience has been rewarded. Our particular appreciation is due to Achan Chin You-di without whose support and continuing assistance over many years it is likely this work would never have been done, at least by us. We are very sorry that this could not have come out in time for him to have had the chance to read it. We thank those people named above for the illustrations, research


and reports on this research which we have used and many officials of the Fine Arts Department of Thailand, Thai Government officials, Thai students and students of the Universities of Hawaii and Otago who have helped us in many ways. We are indebted to the hospitality of the Thai in whose homes and villages we have stayed, to our Thai workmen at the many different sites and areas where we have worked, and we are very much in debt to Thawee Uthaiwee who became our foreman during the first year’s excavation at Non Nok Tha. His good humor, ready smile, and constant assistance have been memorable (P1. 1c). Editor’s Note: After all these years, the Richard F. Taitano Micronesian Area Research Center of the University of Guam agreed to e-publish this historic and important work in Guampedia, Guam’s Online Encyclopedia in 2010 to allow for free public access. Due to the passage of time and changes in technology, some of the original information was not found or inaccessible. If you have photo caption information please contact Guampedia by using the contact links at the bottom of the website. Thank you!


Chapter 2 The Phu Wiang Region and the Khorat Plateau By Donn Bayard Introduction To the traveler entering Northeastern Thailand after the scenic drive through the Phetchabun Mountains from Bangkok, the initial impression of the Khorat Plateau is rather disappointing. As the Friendship Highway descends the escarpment east of Pak Chong, one is confronted with a wide panorama of gently rolling land, covered with open scrubby forest interspersed with rice fields. As one drives on, particularly during the dry season, the impression of monotony and uniformity is increased; broad expanses of dry paddy fields covered in yellow-gray stubble or blackened after burning off alternate with undulating areas of sandy red soil inadequately clad with scruffy, dry forest. Farther on, one can sometimes glimpse the bluish outlines of long, low mountains through the dry season haze; more rarely the highway itself ascends hills of 50 or 60 m, providing some welcome relief. But the overwhelming impression gained is of dry, dusty landscape with minimal environmental variation all along Highway No. 2, until its end-—and the end of Northeastern Thailand-—is reached at Nong Khai, 450 km from Pak Chong. This landscape comes as a surprise to anyone expecting tropical verdure, dense rain forests, or glistening green fields; at least it did to me in 1965, as a graduate student with two years’ residence in Hawai`i behind me. But in the course of two further years spent in the Northeast, the impression of monotony and uniformity predictably waned as I gained sensitivity to local variations in landforms, vegetation, and rainfall. In this I was aided immeasurably by the valued and valuable friendship of those born and raised in the area; for it is precisely these local variations that to a large extent still mean the difference between relative wealth and poverty to the subsistence farmers of the Khorat Plateau. As our knowledge of the archaeology of Northeastern Thailand has grown over the years since 1965, it has become even more apparent that these environmental considerations must be taken into account to aid in explaining such crucial developments as agriculture, metallurgy, intensive wet-rice cultivation, and state formation. Environmental constraints seem to have been at least partly responsible for the marked contrast between a longlasting Hoabinhian tradition in the mountainous regions west and north-west of the Khorat Plateau and the “progressive” (i.e., agriculturalist) cultures of the Plateau itself (Higham 1979; Bayard 1980c:128-132). Similarly, more local environmental variations on the Plateau now appear to have had a major influence on the settlement pattern of early rice agriculturalists and their expansion following the development of intensive wet-rice cultivation (Higham and Kijngam 1979; Higham 1979); and on the subsequent rise of states in the area (Ng 1979; Bayard 1980b:105-107; Kijngam et. al. 1980:79-80; Welch 1984, 1985). The relationships of both phenomena to environmental factors have been discussed at length by Higham and his colleagues (Higham and Kijngam 1984:546-575, 721-724; Higham 1989).


If this report had been written in the early 1970s, when Non Nok Tha was still the only major excavation in Northeast Thailand, it would have perhaps been permissible to limit our discussion of the environment of Non Nok Tha largely to the adjacent Phu Wiang region, as in fact I did in the earlier unpublished version of this section (1971a). However, research has subsequently flourished in the Loei, Chi, Songkhram, and Mon Basins. Excavations at Ban Chiang, Ban Na Di, and other important sites, the rest of Thailand, Viet Nam, and Southern China make it imperative that the Phu Wiang region be placed in a larger context, both as a western appendage to the Khorat Plateau and as a border area lying close to the major ecotone between the Plateau and the Phetchabun Mountains to the west. Accordingly, while the following discussion still concentrates on the Phu Wiang area, it will also contrast the region with the rest if the Khorat Plateau and with the western mountains, outlining those features that will assume considerable importance in the conclusions of this volume. Geomorphology and Environment Non Nok Tha and the district of Phi Wiang surrounding it lie on the western margins of the Khorat Plateau (Fig. 2-1), which forms the major part of northeastern Thailand. The plateau, described by Cruijis as one of the major physiographic provinces of Thailand, “is a blocklike structural and geomorphplogical unit that comprises an area of about 170,000 km2, and lies between the parallels of 14 and 19 North latitude, and between the meridians of 101 and 106 East longitude.” (Cruijis 1978:7). As Ward and Bunnag (1964:3) have pointed out, the term plateau is an inaccurate description of the region, which possesses a plateaulike descending escarpment only on its southern extremity (the Dang Raek Range). The region is made up of a series of Mesozoic sedimentary deposits (the Khorat Series) that form two almost level synclinal basins-—the Khorat to the south and the smaller Sakon Nakhon to the north-—separated by the anticline of the Dong Min 1 Hills, which run roughly northwest-southeast. Both basins lie between 125 and 200 m above m.s.1 and drain east into the Mekong. The Khorat Basin is drained by the Mun River and its tributaries, the Chi and Phong, while the Sakon Nakhon Basin is drained by the Songkhram, The central portions of the Khorat Basin, to the east of the Phu Wiang region, lie at an elevation of about 150 m, and consist of deposits of Quaternary and somewhat older sediments (“low” and “middle terraces”; Cruijs 1978:18-19) along the courses of the Mun, Chi, and Phong Rivers; outcrops of a still older “high terrace” occur fringing all three rivers. Higham [1989:94] dates all three terraces as Pleistocene. These deposits (not true alluvial terraces; Löffler et al. 1983) immediately overlie the youngest of the Mesozoic sedimentary deposits, the so-called Jurassic Salt Formation (actually late Cretaceous; see United Nations 1968:16), which Cruijs more aptly terms the Maha Sarakham Formation (1978:15). Although this formation does not extend into the Phu Wiang region itself, it has almost certainly played a part in the prehistory of the whole Khorat area as a source of salt. Brackish ponds and salt wells are common in the central area of the Khorat Basin, and have been exploited by the area’s inhabitants for a considerable time as a source of brine for the manufacture of salt by evaporation (Higham and Parker 1970; Brown et al. 1951:153-154). Fringing this central area, most of the rest of the Khorat Basin proper is



Fig. 2-1


Fig. 2-2

underlain by the Khoh Kruat Formation, a slightly earlier Cretaceous deposit that also contains thin beds and patches of salt as well as gypsum.

Non Nok Tha and the Phu Wiang region are located on the basin’s rim, which is formed by exposures of three later Triassic and Jurassic sedimentary beds; in order of increasing age these are the Phu Phan, Phra Wihan and Phu Krad’ung Formations (Fig. 2-2). Non Nok Tha itself overlies Phu Krad’ung bedrock but is only 4-5 km distant from the mountain of Phu Wiang, which gives the district its name. This is a large sandstone monadnock made up of Phu Krad’ung sandstone at its base, overlaid by a thick medial layer of Phra Wihan sandstone and capped by a layer of resistant Phu Phan sandstone. As might be expected, these three formations account for most of the lithic material encountered on the site, both in terms of weight and number; hence a brief description of the three formations (following Cruijs 1978:11-13) and the lithic materials derived from them (based on personal observation) is warranted. Phu Phan Formation: “Consisting mainly of resistant and massive layers of gray to palered, cross-bedded, arkosic and conglomeratic sandstone and conglomerate. The very thick-bedded conglomeratic sandstone and conglomerate at the top of the formation are the layers which cap numerous hills around the rim of the Khorat plateau, as well as eroded anticlines in the Phu Phan hills.” Material recovered at Non Nok Tha was in the main quite hard, well-cemented red to pink to white sandstone, nonfriable and often metamorphosed; quartzite was also recovered. This material found little use in the tool making at the site, although occasional grindstones



and hammer stones of Phu Phan sandstone were found; it was also used in the form of relatively large cobbles as probable grave markers. Phra Wihan Formation: “Consisting mainly of resistant and massive layers of white and gray, cross-bedded and cross-laminated quartz sandstone with minor and red siltstone and shale, and locally, lenses of lignite and jet.” Material classes as Phra Wihan at Non Nok Tha was in the main medium to medium-hard gray to gray-green sandstone, somewhat friable and occasionally micaceous, and seemingly little metamorphosed. This was the material most commonly chosen for tool making; most of the bronze casting molds are made from it, as are half of all grindstones. Phu Krad’ung Formation: “The main types of the predominantly soft rocks of the formation are red and gray sandy (micaceous) siltstone or shale, with some micaceous and calcareous sandstone, and intraformational conglomerate and conglomeratic sandstone.” As observed at Non Nok Tha, this material was generally medium to medium-hard and red to reddish brown, somewhat friable and almost always micaceous. The fine-grained sandstones and siltsones were often used for the manufacture of molds and grindstones, but less so than the Phra Wihan sandstones. The numerous pieces of ferruginous siltstone used for making red pigment that were recovered from Middle and Late Period levels may also derive from the Phu Krad’ung formation. (More general descriptions of the formations and of the region’s geology in general may be found in Brown et al. 1951:22-23, 36-38, and in Haworth et al. 1964; see Cruijs 1978 and United Nations 1968:15-18 for more recent classification of the formations). As mentioned above, the Phu Wiang region is basically a western appendage of the Khorat Basin, separated from it by the long, thin, and almost continuous ridge of Phu Phan. Landforms and vegetation to the west of Phu Phan are almost indistinguishable from the middle terrace area of the basin proper, save for Phu Wiang, Phu Kao and the other sandstone monadnocks that rise dramatically from the rolling countryside. However, if one proceeds due west from Non Nok Tha some 25 km, or (more comfortably) drives the same distance from the Chumphae District center on the paved road to Loei, the Khorat Basin comes to an abrupt end as the road suddenly crosses into the more densely forested karst country resulting from exposures of the Paleozoic Ratburi Formation; the contrast is even more startling from above (P1. 1b). This major ecotone separates the Khorat Basin from the 100-150 km wide Phetchabun Mountains, themselves a southern extension of the complex series of ridges that form the Golden Triangle of northern Thailand, Laos, and Burma. The ecotone is in the main easily traceable on geological maps from the Khorat northward to the Mekong; at that point it has almost merged with the sinuous, westward-curving ridge of Phu Phan, which crosses into Laos at the Pha Mong water gap and continues north, forming the western margin of the Vientiane Plain. For most of its length in Thailand, the ecotone is marked geologically by the junction of the Khorat Series with the Paleozoic Ratburi and Kanchanaburi Series, although to the north, in Udon Thani Province, outcrops of Tertiary igneous intrusions replace the much earlier Paleozoic deposits in places. This ecotone appears to mark an environmental boundary of considerable importance in the region’s prehistory. Surveys carried out after the Non Nok Tha excavations in the Phu Wiang area and northward (Higham and Parker 1970, Marsh 1969) indicate a settlement


pattern and economy quite closely tied to other parts of the northern Khorat Plateau over the past four to five millennia; other surveys carried out only 10 to 50 km west of the ecotone (Bayard 1980c, Penny 1982, Wilen 1986-1987) suggest a radically different prehistoric orientation, with sparse populations continuing an essentially Hoabinhian way of life until much more recently-—perhaps even as late as A.D. 1000 (Bayard 1980c:123). Hence the areas immediately to the west of the ecotone appear at present to have more in common with the Hoabinhian gathering and hunting orientation of northern Thailand and the rest of the Golden Triangle than with the early agriculturalists, cattle tenders, and bronze smiths of the northern Khorat Plateau. The first tentative intrusion of rice cultivation into the mountain zone took place only in the last few centuries B.C. or later (Bayard 1980c:47-48, 80), and the ancestors of the present rice farmers of the Loei area probably arrived only four to five centuries ago (Bayard 1980c:134). Nonetheless, the montane zone and its outliers-—the densely wooded caps of Phu Wiang and the other monadnocks-—were doubtless exploited for hunting and forest products by the prehistoric inhabitants of Non Nok Tha and related sites just as they are today. The montane zone proper may also have been the source of fine-grained stone used in adze manufacture, and importantly the copper, tin, and lead ores necessary for the early development of bronze working in the area (see Chapter 9). Landforms While the montane/lowland division is of course the most distinctive one in the area (and in Southeast Asia as a whole, where it has become enshrined as a basic tenet of the region’s cultural geography; e.g. Burling (1965) three-way classification of landforms has proved to be more profitable for the purposes of Northeast Thai prehistory. In addition to the montane zone on the one hand and late Pleistocene and flat recent alluvium on the other, much of Northeast Thailand is made up of the gently rolling terrain described at the beginning of this chapter. Higham (1977a:387) earlier termed this zone the Piedmont, which is apt enough when discussing the area immediately around Non Nok Tha for the paralleling of the major ecotone some 25 km to the west (note, incidentally, that the site is not located on the ecotone proper as stated by Higham 1975:251). However, much of the rolling land is not adjacent to the montane zone (Fig. 2-3), and I prefer the term rolling lowlands to distinguish this zone from the true alluvial plains. This latter zone extends approximately from the city of Nakhon Ratchasima eastward to Roi Et and Ubon, with a lobe along the Chi reaching Khon Kaen and a further pocket on the upper Chi. Smaller pockets are found in the Sakon Nakhon Basin and along the Mekong, as well as to the north of Udon city (this last is a southern extension of the Vientiane Plain). Following this classification, the middle and high alluvial terraces, along with the Khorat soil group, are classed with the rolling lowlands rather than the plains; this sharply reduces the area of the latter when compared to other classifications (e.g., United Nations 1968:3), but reflects a basic division in soil types between the two zones (see below), as well as fitting what we have learned to date about the prehistoric demography of the area. It became clear to several of us in the early 1970s that there was a quite close association between landform type and prehistoric site distribution, with nonagricultural occupation limited to the montane zone; early (pre-3000 BP) agricultural sites on the rolling lowlands; and later agricultural communities, almost certainly dependent on advanced wet-rice



Fig. 2-3

agriculture, present on the plains only after 3000 BP or later. Gorman emphasized the importance of a presumed shift of population density from the mountains to the lowlands ca. 6500-3500 B.C. quite likely brought about through the development of cereal agriculture (1971:316); I outlined two important transitions-—from a Hoabinhian-type economy in the montane zone to one based (possibly) on swidden rice cultivation on the rolling lowlands and from this to a developed wet-rice economy on the plains (Bayard 1971a: 230-236); and Higham elaborated on these foundations to produce a three-stage sequence of prehistoric eras (1972).

Archaeological surveys carried out since then (e.g., Bayard 1980c, Kijngam et al. 1980) have modified this simplistic initial-stage model. Sites of the type of early Ban Chiang and Non Nok Tha are still largely limited to the rolling lowlands of the middle terrace; however, initial utilization of the lowterrace and flood plain alluvium prior to the introduction of iron and buffalo has now been pushed back to about 1500 B.C. (Higham and Kijngam 1984: 38-39). Higham’s 1970s model, which saw the introduction of iron and water buffalo allowing agricultural intensification and post-1000 B.C. movement to the alluvium, has since been superseded by his and his students’ subsequent work. In addition, his most recent scenario (1988a) downplays the possibility of an earlier transition from a Hoabinhian gathering economy to an agricultural one on the hilly margins of the Plateau (as he postulated in 1979). He would now favor the other alternative postulated in his 1979 article, migration of agriculturalists “from a more distant place” (1979:678). This scenario envisions the origins of agriculture in the coastal regions of Thailand and Viet Nam, followed by movement up the major river valleys into the interior (1989:86-87, 358)1. Hence the source of the earliest agriculturalists on the Plateau remains unclear; equally unclear is the nature of the agricultural techniques they originally employed and the type of cultivation involved in the presumed intensification that appears to have taken place in the first millennium B.C. These questions will be discussed briefly below, and in more detail in the concluding chapter of this report.


The Phu Wiang region is largely made up of middle terrace deposits (excluding, of course, Phu Wiang itself and other sandstone mountains), affording a clear example of the rolling lowland terrain: a gently undulating landscape ranging from 180 m to 210 m in elevation, crossed by a complex network of meandering seasonal watercourses that are flanked by (or in some cases converted into) rice paddies (see P1.2a). The intermediate upland areas are clad in scrubby, dry monsoon forest and brush, although sometimes used for cultivation of dry crops (see below). There are only three perennial streams in the area: the relatively large Nam Phong, and two smaller tributaries, the Lam Phaniang and the Huai Bong/Huai Khaen (which drains the interior plain of Phu Wiang). The landscape is thus one of considerable variability (Plate 2b) when compared to the uninterrupted expanse of paddy fields characteristic of the plains zone. Adding to this diversity is the bulk of Phu Wiang, comprising an outlier of the montane zone easily accessible from the adjacent lowlands; Non Nok Tha is only some 2-3 km from the foot of the mountain, and an additional 1 or 2 km fairly steep walk will take one into the triple-canopy forest that caps the flat top of the mountain, affording access to a wide range of resources used by past and present inhabitants. Soils Fig. 2-4

The dominant soils of the region are mainly the Khorat and Phon Phisai Series characteristic of the middle terrace (Table 2-1); however, these are largely unsuitable for rice cultivation due to their sandy nature and porosity (Thai Ministry of Agriculture 1973), and the bulk of rice agriculture in the area is carried out on Roi Et sandy clay loams. This soil series is more characteristic of the low alluvial terrace, and some of the larger pockets of this soil (such as the one in which Non Nok Tha itself is situated) may in fact be outlying deposits of this terrace paralleling the Phong River above the Phu Phan water gap (an area now largely flooded by the Nam Phong Reservoir), But many of the smaller pockets are more likely the result of human modification. The possibility of a close correlation between soil type and site location was first suggested by Parker in 1970, but not published until a decade later (1980:56-57). Meanwhile, the idea received elaboration by Higham (1975a), who pointed out the extremely close correspondence



between the occurrence of Roi Et soils and the land currently under wet-rice cultivation, not only for Non Nok Tha, but for Ban Chiang and other Khorat sites. While this would appear to be the case for sites located on the low-middle terrace ecotone, it presents a somewhat oversimplified picture for the Phu Wiang area. Here and in similar areas, many of the smaller pockets of Roi Et soils are probably the result of diking and paddy formation on concave areas of Khorat soils (Pendleton and Montrakun 1960:15, 27), and hence are an artifact of human activity rather than an environmental determinant in settlement. It would thus be unwise to assume a universal one-to-one relationship between soil type and land use, with the latter as the dependent variable, although the principle certainly has general validity. Climate The climate of the Phu Wiang region, along with most of the rest of the Khorat Plateau, can be described as “tropical wet and dry” (United Nations 1968:13) or “tropical savanna (Köppen Aw)” (Pendleton 1962:122). However, the situation with climatic variations parallels that of soil distribution; while on a gross regional level the plateau area shares the three-season dry monsoon climate of much of the rest of the country, several factors produce considerable local variation in this overall pattern. Most important among these is the rain shadow effect; with the exception of the northeastern quadrant, the Khorat Plateau lies within the shadow of the Phetchabun and Dangrek Mountains. The winds of the MayOctober southwest monsoon, having lost much of their moisture in crossing the ThaiBurma ranges and the Cardamom Mountains, deposit a large amount or the remainder on the Dangrek and in particular the Phetchabun Mountains. Hence most of the western and central Khorat Plateau receives less rain than Bangkok (1,418 mm mean), itself in the shadow of the Thai-Burma ranges. Most of the eastern plateau receives about the same amount as Bangkok (e.g., Ubon Ratchathani with a mean of 1,466 mm), while its northern and northeastern margins receive considerably more (Vientiane 1,717 mm; Sakon Nakhon, 1,600 mm; The Knèk, 2,451 mm; all figures from United Nations 1968:6, 10). The rain shadow effect is most marked in a zone north and east of Nakhon Ratchasima from which it tapers off gradually. The sandstone monadnocks of Phu Wiang and the other mountains also create miniature rain shadows in the areas immediately north and east of them, but this is mitigated to some extent by runoff from the slopes.


Variation in temperature follows the three-season pattern of the rest of the country. Temperatures are lowest in the November-February cool season, ranging from average lows of 17˚C to mean highs of about 31˚C at the weather station near Khon Kaen (Some 60 km southeast of Phu Wiang). Temperatures reach a peak in the March-April hot season (22-36˚C) and decline slightly during the May-October summer rainy season due to greatly increased cloud cover (24-33˚C). Daily extremes exceeding these figures are common, however, and highs of over 40˚C occur from February to May, while the lows below 10˚C are occasionally experienced during the cool season. Altitude plays a part in local variation, as temperatures on the summit plateaus of Phu Wiang and the other monadnocks seem to average 3˚C or more below those on the rolling lowlands. Humidity also varies seasonally, ranging from 62 to 72º during the dry season to a range of 73-82º during the rainy season (again, these figures are based on the Khon Kaen data presented in United Nations (1968:11). Of the above variables, it is rainfall that is the most crucial to the inundation wet-rice farmers of the area today, and precipitation appears to have played a significant part in the region’s prehistory as well. It is necessary here to consider not only rainfall over the plateau as a whole, but in the immediate vicinity of Non Nok Tha, since as implied above the presence of Phu Wiang and the other mountains creates considerable microclimatic variation. Looking first at the whole Khorat Plateau, Higham has emphasized in several articles (1975a, 1977a, 1977b, 1979) the contrast between the well-watered Sakon Nakhon Basin and the “more arid and inhospitable southern Khorat Plateau” (1979:679); i.e., the Khorat Basin proper. He put forth convincing arguments, based on faunal data from Ban Chiang and related sites, for a transition from some type of swidden cultivation of wet-rice growing based on inundation and using iron tools and tractive power of the water buffalo; this appeared to have taken place in the Sakon Nakhon Basin by about 1600 B.C. (Higham and Kijngam 1979). He hypothesized that these adaptations, stimulated by an increase in population pressure, allowed “movement across the important ecological divide which follows both the line of the Dong Mun Range and the 1,250 mm isohyet” (1977b: 106) onto the more arid Khorat Basin by about 500 B.C., lying the groundwork for state formation in the first millennium A.D. Further research has seen considerable modification of this view. But the thesis as it stood in 1980 appeared plausible, although rainfall distribution maps published in Brown et al. 1951 (P1. 2) indicate a considerably smaller arid zone (limited to the southwestern Khorat Basin) than that assumed by Higham, based on the United Nations Mekong Atlas map (United Nations 1968:14, as published in Higham 1977b:105). It also should be noted that there is nothing particularly significant about the 1,250 mm isohyet; much more significant would be the bundle of isohyets covering the range 1,500-1,700 mm that appear to run closely together north of the Dong Mun Hills along a line running from Vientiane to Savannakhet (based on average annual rainfall figures for all available stations in the region; U.N. 1968:6, 10; see Table 2-2). However, as the soil types, the situation is more complex than described by Higham in 1977. First it is necessary to consider the water requirements of both the type of swidden agriculture that appears to have been practiced earlier on at Ban Chiang and Non Nok Tha and of the inundation wet-rice system that replaced it. Secondly, we must consider these in relation to seasonal variation in rainfall, both average and minimum, rather than to overall annual means.



Higham and I are in agreement that the type of rice agriculture characterizing the economics of early Non Nok Tha and Ban Chiang was probably not the inundation system of wet-rice cultivation used in the region today (Bayard 1971b:32-33; Higham 1979:676). Gorman (1977), however, felt that the inundation system has considerable antiquity, preceding the development of slash-and-burn cultivation. The evidence available at present from both sites is inconclusive, but taken as a whole it is possible to arrive at some tentative conclusions. First, if swidden was employed, it was almost certainly not the classic type used by Southeast Asian hill tribes today; the gentle relief and sparse forest cover of the rolling lowlands do not lend themselves to this type of land-extensive cultivation of rice (although it is practiced today in Loei Province, just to the west of the montane/lowland ecotone). The dry monsoon forest and lack of relief would lead to problems in obtaining a complete burn off and sufficient amount of ash (and hence nutrient) yield. Much more likely, as Higham has pointed out (Higham 1979; Higham and Kijngam 1979), is some form of swamp cultivation using swidden methods. Seavoy (1973) has outlined a sequence from classic forest swidden to continuous wet-rice cultivation in Kalimantan through an intermediate stage of paya (mud) swamp swiddening. This involves the selective draining of ground that is swamp for most of the year and fed by perennial streams; the clearing and burning of perennial grasses; and the planting of rice in the cleared mud that remains. Fields are fallowed for three to five years to allow the natural grasses to regenerate. The system requires minimal water control (simple cross-valley dikes to permit drainage and prevent flooding of the growing rice seedlings) and produces yields two to three times as great as dry swidden, but is of course more labor-intensive. Moreover, a clearly defined dry season of enough length to allow the cut grasses to dry is required, and excessive rain and flooding can destroy the rice seedlings (Seavoy 1973:222).


Higham cites this article as a possible model for early agriculture at Ban Chiang (1979:676), but later in the same article he states that cultivation was perhaps carried out by planting in cleared “seasonally flooded lake and river margins, where wet swiddening of rice could be undertaken” (1979:678; cf. also Higham and Kijngam 1984:172). This would imply a slightly different and more basic type of paya cultivation-—one that consists of simply enlarging the native habitat of wild rices by partial clearance of other intrusive plants, and broadcasting of seed at the most opportune time in the natural regenerative cycle (White 1995). The two different types may be distinguished as burn paya and clear paya, and each would seen to require somewhat different annual rainfall patterns. Clearly, a uniform dry season is necessary for burn paya, and Seavoy implies that the crop is grown during the dry season; heavy rains and uncontrolled flooding during this period are harmful to the crop in that they encourage faster growth of the perennial grasses that seeded at the same time the rice was planted. Wet-season variations are unimportant as long as the streams remain perennial. Rainfall requirements for clear paya, on the other hand, would seem to be an even and dependable wet season, with no undue peaks or droughts to produce fluctuations in cleared lake and river margins where the crop is growing; as with burn paya, permanent watercourses or year-round swampy lakes are a necessity. The rainfall requirements for inundation wet-rice cultivation more closely resemble those for the clear paya, but they become very critical where few perennial streams are present, as in much of the Khorat Basin, as White (1995) notes, the prolonged dry season of Northeastern Thailand is quite different from Kalimantan. The transplanted inundation techniques used currently require some rain in May for preparation of nursery beds and the plowing of the paddies themselves, and they further demand a sufficient rainfall during the crucial June-July period. If these fail (“as they often do), the crop is lost; the heavy rains that come in September in the western plateau are of no avail. Sowing of the seed broadcast-—which presumably preceded the development of transplanting-—would omit the need for May rains but would require an even more dependable rainfall during the June-August period. These differing rainfall requirements, taken in conjunction with the figures in table 2-2, suggest a number of points. First, the type of agriculture practiced by the inhabitants of both early Non Nok Tha and Ban Chiang was probably some form of paya; the crucial factor here was probably not the rainfall (i.e., Higham’s distinction between the wellwatered Sakon Nakhon Basin and its more arid Khorat counterpart). but the presence of perennial streams in the Ban Chiang area. There is evidence in Higham and Kijngam;s faunal analysis (1979, 1982a) of the presence of perennial lakes and clear-flowing streams in the Ban Chiang area (but cf. White 1995), dependent perhaps on runoff from the Dong Min hills as much as from the heavier rain indicated for the Sakon Nakhon area; the area may thus have been well suited for clear paya cultivation. The Phu Wiang area, however, does not seem to have had perennial watercourses in such abundance, but runoff from Phu Wiang would permit dry-season burn paya to be carried out; it is also likely that some areas of paddy fields today were at one time at least seasonal swamps, and large areas of seasonally fluctuating swampland are still to be found along the course of the Nam Phong and in the interior of Phu Wiang.



But heavier rain and the backing up of streams in the Ban Chiang area (Hingham and Kijngam 1979:224) on the one hand and very low minimum monthly rainfall pattern on the other may have made the Ban Chiang area less favorable for paya cultivation than Phu Wiang. In contradistinction to Higham’s view of the Sakon Nakhon Basin as optimal vis-àvis the regions to the south, the undependability of the Ban Chiang rainfall pattern, when compared to that at Ubon, Savannakhet, or even Roi Et (all of which have higher minimum rainfall figures than Sakon Nakhon), may well have been a factor leading to experimentation with various water control systems, including inundation wet-rice cultivation. Higham postulates that such experimentation was due to expanding population, but the relatively heavy yet undependable rainfall could have been a factor as well. As Schauffler put it in an unpublished paper, “arguments for insufficient rainfall in the northeast as a limiting factor in early rice cultivation are less relevant than those questioning the ability of these early settlers to control the water” (1978:32). The entwined questions of absolute rainfall, dependability of rainfall, and population pressure have been discussed in considerable detail in Welch’s work (1984, 1985); it is worthwhile to quote his conclusions at some length: The alternative model that I propose here sees the adoption of the intensive techniques of bunding, plowing, and transplanting as in large part an adaptive response to conditions of unpredictable rainfall. I would argue that population pressure was not a significant causal factor in this transformation. . . Future evidence may support a population pressure based explanation of intensification, but present data do not. (1985:388) In any event, whatever the causal factors and techniques involved, it seems unlikely that one system would have immediately superseded the other. Paya and wet-rice cultivation were doubtless practiced side-by-side for some time (rather paralleling the simultaneous bet-hedging use of dry-field and inundation wet-rice techniques by contemporary Loei lowland farmers and many hill tribes). For populations living in the Khorat Basin proper, conditions may have given permanent wet rice an edge over paya; chief among these would be the lesser number of perennial streams and swamps that could be cleared for wet-season paya or burned for dry-season cultivation. Permanent paddy fields not requiring fallowing (and hence three to five times the total area) were apparently a positive adaptive response to these conditions, particularly in the eastern Khorat Basin, which receives almost as much rain during the growing season as Sakon Nakhon and in fact has slightly higher minimum growing-season figures than the latter area. In fact, White (1995) has questioned the use of “swidden” for these presumed systems and asked why such paya tracts could not have been used perennially, with no fallowing. “Swamp rice” may be a better term than “wet swidden”. Conditions were more severe in the western basin, in the zone between Khon Kaen and Nakhon Ratchasima and west to the slightly wetter Phu Wiang area; even today, villagers expect one crop in three to fail due to insufficient rain in the June-July period. One could speculate that paya swiddening was more dependable form of cultivation than wet rice in this zone but would then have to explain its abandonment for permanent inundation cultivation. Although I agree with Welch that it is an unsatisfactory general causal factor


(Bayard 1985), local population pressure in the Phu Wiang region may have been a factor (cf. Welch 1985:389). In any event, I will return to these questions in the concluding chapter of this work. The major point of this discussion is to illustrate the complexity of such environmental factors as rainfall and to emphasize the need for consideration for more than just the annual mean figures and global rather than local causal factors when discussing explanations for prehistoric economic alternatives. Vegetation Finally, let us consider the vegetative cover of the region. As might be expected, this is determined by the interaction of landforms and rainfall. Although some local variation occurs in conjunction with distribution of different soil types, by and large these are closely correlated with landform type and assume importance only when considering differentiation at a local level. In the regional sense, it is fair to limit our description to three types of vegetative communities, each associated with one of the three gross zones outlined above. The montane zone (represented in the Phu Wiang area by the tops of the Phu Wiang itself, Phu Kao, and the other monadnocks) is uniformly covered in what Pendleton has called a slightly deciduous variant of Evergreen Tropical Rainforest (1962:86). Holdrige et al. prefer to call it Subtropical Wet Forest, but are obviously referring to the same forest type in their survey of the Khon Kaen region; as they note, this type is limited to areas above 400 m in elevation (1971:703). The association is dominated by large dipterocarps (D. alatus-— called yang in Thai, nyang in Lao-—and other species) and other large trees like takhian (Hopea odorata). This upper canopy towers over an understory of young trees and smaller species, which in turn overlie a dense ground cover of herbaceous plants, shrubs, and vines (Nuttonson 1963:76). This triple-canopy forest is not as dense as that found in Peninsular Thailand or on the southeast coast, and it is in fact transitional between archetypal examples of this type and the Moist Monsoon Forest that dominates northern Thailand, extending to Loei Province as well. This latter type has more deciduous species, and species such as Xylia (daeng), Lagerstroemia (puai daeng), and-—of commercial importance-—teak (Tectona grandis; sak) are more common than the dipterocarps (Pendleton 1962:90-93, Nuttonson 1963:76-77). Those portions of the rolling lowland zone not converted to rice paddies are in the main clad in what Pendleton and others have called Dry Monsoon Forest. Holdridge et al. place this in their Subtropical Moist Life Zone (1971:703). Pendleton further recognizes three subdivisions of this category: Dwarf Dry Forest, Bamboo Dry Forest, and Thorn Bush Roadside. The first is by far the most common in the Phu Wiang area, where it is made up of an assemblage of deciduous and semideciduous dipterocarps, teng (Shorea obtusa), and rang (Pentacme siamensis). These average about 8-10 m in height and are widely spaced; as Pendleton says, “The usual stand is so thin that the crowns barely touch, and the foliage is so light that during the full development of the rainy season the sun’s rays can still strike the ground and evaporate much of the rainfall” (1962:94). Ground cover consists of grasses, low herbaceous plants, and shrubs interspersed with patches of bare ground covered with dead leaves in the dry season. As Higham and Kijngam have pointed out, such a forest is far more attractive to large ground-dwelling herbivores than the moist evergreen forest, and it is also amenable to manipulation by burning off; this stimulates the



formation of grassy glades that attract herbivores such as gaur, banteng, sambar, and Eld’s deer. However, “There is also more likelihood of a lean season in the dry deciduous habitat, and therefore more reason to amass wet season surpluses and develop techniques of food storage” (Higham and Kijngam 1979:224). Taken as a whole, this habitat would seem less favorable for broad-spectrum hunting and collecting economy than the moist evergreen forest; while richer in large ungulate fauna, it lacks the wide variety of species available in the moist forest. This would seem to be substantiated by the absence to date of any preagricultural sites or even artifacts (e.g., Hoabinhian tools) on the rolling lowland zone, despite systematic surveys of significant parts of the Khorat Plateau. The original vegetative cover of the alluvial plains zone, now almost all utilized for wet-rice cultivation, is more problematical. Gorman believed they were originally covered with “tropical deciduous woodlands and mixed savannas, a vegetational zone considered to be the optimal exploitable environment for hunting and gathering populations” (1971:306). As with the rolling lowlands, however traces of such populations are so far lacking. Moreover, relict areas not yet brought under rice cultivation due to flooding problems suggest that the original cover was grassy seasonal swamp-—e.g., the extensive area covering large parts of Roi Et, Surin, Sisaket, and Ubon Provinces (Pendleton 1962:98; cf. also Higham 1989:95). Similar areas were widespread in the area north of Bangkok-—on the Chao Phraya alluvium-—until the last century, when khlong construction allowed water control and efficient farming there. Hanks believes such areas were vast, flat spaces “covered with tall grasses, sedges, and in this region, on the higher points, with thick bushes” (1972:73). Travel by either boat or foot would have been extremely difficult, and the area would not have been an optimal one for hunters, as large ungulates were absent (although swamp deer would have been available, at least seasonally). Despite this, Hanks mentions that a “few Khmer-speaking hunters” inhabited the area during the last century, suggesting that earlier hunters might have frequented the area as well; seasonal flooding of these areas is not conductive to preservation of the campsites of such groups, which are of low visibility even in good conditions of preservation. In summary, the three vegetative types-—moist evergreen forest, dry deciduous forest, and swamp grassland-—would appear to offer diminishing amounts of resources to a preagricultural population, and this is reflected by the apparent absence of sites exhibiting such an economy (although the fit at present is somewhat too neat, and I would not rule out the discovery of at least a few such sites in the future). On the other hand, the three biozones would afford increasing possibilities to groups familiar with wet-rice technology. The montane zone, with its narrow valleys and forest cover, is not suited for large-scale paddy production as demonstrated by its present low population density and the late movement of wet-rice farmers into the area (Bayard 1980c). The rolling lowlands and dry deciduous forest offer an opportunity-—albeit a rather risky one-—for inundation paddy cultivation. Finally, the swamp grasslands of the alluvium presented a quite hostile environment for most economies, but once water control techniques were developed they allowed vastly improved wet-rice production, enabling these areas to support the majority of the present population of the region: “The consequences were an increased potential for expanding the land under cultivation, population growth, and a potential increase of yields” (Welch 1985:388). However, the three vegetation zones are gross ones; of greater


impact on the present inhabitants-—and doubtless on the prehistoric population as well-— are small-scale local variations in these divisions that allow for differential utilization and resource collection. These are outlined for the Phu Wiang region in the next section. Local Resource Zones Local variability in rainfall, altitude, and particularly soil type have produced a series of zones that are explicitly recognized and used differentially by at least some of the area’s contemporary inhabitants. The seven basic zones described here were arrived at through my observations during more than two years residence in Northeast Thailand (eight months of this in the Phu Wiang area), coupled with the views of one of the several spirit doctors (mo du) of the village of Ban Na Do near Non Nok Tha. Nai Thiang Nuthao was a trained pham or Brahman as well as a traditional doctor (he died in 1971); in the course of practicing both professions he often had recourse to a pharmacopia of sixty-odd medicinal plants. While he was engaged in making a collection of these plants for me to ship to the Bishop Museum in 1968, I noticed that the habitats he described for them fell into a fairly small range of environments that coincided quite closely with my own observations of topographic and vegetative variations in the area. As neither of us had any formal training in ecology, the zones as I presented them in my 1971 thesis and as presented in a subsequent article by Higham (1975a) were of necessity subjective. Some amount of objective support has become available since then in the form of the detailed soil maps of Khon Kaen Province published by the Thai Ministry of Agriculture (1973); these show a very high correlation between the original zones as I mapped them in my 1971 thesis and the distribution of the different soils of the area. More support is now available in the form of a much more extensive 20-month ethnological survey undertaken by White in the Ban Chiang area during 1979-81 (White 1995). Like me, White worked with a knowledgeable Thai-Lao informant and found a complex system of soil and associated vegetative land classification reflecting the inhabitants’ long standing familiarity with and dependence on their environment. As White points out, the local classifications sometimes differ from official soil or forest classifications, as the local residents reasonably take both factors into account in arriving at their classifications. Nor do the classificatory systems suggested to me by Nai Thiang and to White by her informant wholly agree; this is not surprising, as mentioned above the two areas are environmentally different. It is not difficult to find such differences even in the Thai Ministry of Agriculture’s ratings of various soil types as suitable for agriculture from one provincial soils map to that of another province (Bayard 1992:19-20); White (1995) also points out such discrepancies between soil maps and her informant’s classifications. My inclination is to trust the people whose livelihood depends on these classifications, rather than maps made by specialists from Bangkok and the West, and I feel sure that this knowledge of local ecology was as well-developed among the prehistoric population as it is today (if not even more refined). Hence, I feel that the seven divisions I (and Nai Thiang) describe below for the Phu Wiang area have a high degree of reliability, and are applicable to both current and earlier patterns of exploitation in the Phu Wiang region (Fig. 2-4)



Fig. 2-4

ZONE 1: River Bottomlands (lam nam in the Lao dialect of Phu Wiang); elevation ranges from ca. 175 to 205 m. This zone is found immediately adjacent to the few permanent watercourses in the area; the dominant soil types are the loamy sands of the Nam Phong and Alluvial Complex Series. The most notable feature of the vegetation is the presence of many tall yang trees that are more typically found in the true rain forest but are found here because of the more or less constant supply of water. This supply also ensures a dense cover of undergrowth, including many water-loving plants not found elsewhere in the region. ZONE 2: Ricelands (thong na); elevation varies from 180 to 215 m. This zone consists almost exclusively of wet-rice paddies and the dikes separating them; the soils are almost exclusively sandy loams or loamy sands of the Roi Et Series. Vegetation aside from rice itself and stray grasses growing on the dikes is limited to small-to-medium trees atop the 2-4 m termite mounds that occur commonly in the fields; like White (1995), I was also told that these mounds are broken down to provide soils rich in


nutrients. Interrupting the broad expanse of the paddies are occasional small mounds called non-—actually outliers of the following zones--raised slightly above the level of the fields; Non Nok Tha is one of these mounds. They are mainly utilized for the growing of garden crops. ZONE 3: Uplands (don); these higher areas, separating the zones of paddy cultivation, are between 190 and 210 m above sea level. Soils are mostly the permeable Khorat and Khorat Stony Series unsuitable for paddy cultivation. As noted above, however, it is likely that at least some of the Roi Et soils in Zone 2 have actually resulted from the diking and repeated cultivation of low-lying Khorat soils; hence the division between these two zones is more an artifact of human activity than a natural feature of the environment. Dominant vegetation on these upland areas of reddish Khorat soils consists mainly of small, large-leafed trees rarely more than 10 m in height; these are mostly teng and rang (or hang) as described for the Dwarf Dry Monsoon Forest type by Pendleton (1962:94-95). Trees are fairly widely spaced and do not form a canopy: undergrowth is fairly heavy and consists of shrubs rather than grasses. ZONE 3A: Laterized Hilltops: (the term khok is used in the area for large, prominent mounds in general; Pendleton and Montrakun state that the term talat phi or “spirit marketplace” is used for these bare areas, but I have never encountered this term locally); this subzone consists of enclaves within Zone 3 on hilltops between ca. 210 and 225 m and is composed of shallow deposits of Phon Phisai soils overlying a layer of laterized concretions. In many cases the laterite is exposed, producing rather dismal bare areas of up to half a hectare in extent, covered with small laterite pebbles and almost completely devoid of vegetation. ZONE 4: Foothill Forest(tin khao): this zone rings the base of Phu Wiang (and presumably the other sandstone mountains in the region as well) between about 200 and 230 m. Soils are a mixture of Khorat Stony Phase and (on and adjoining the sandstone outcrops that appear in this zone) the Slope Complex of shallow soils derived from the sandstone base material. Vegetative cover is noticeably more dense that in Zone 3 (probably due to increased runoff). Many species of plants occur here that are not found in Zone 3; these seem to represent a mixture of species from this zone with those from the following Mountainside Forest Zone, including a rather short (ca. 8 m) type of thorny bamboo quite common in the lower portion of Zone 5. ZONE 5: Mountainside Forest (pa or dan khao); covers the steeply sloping outer edge of Phu Wiang from ca. 230 to 500 m in elevation and a considerably greater area on the more gently sloping interior of the mountain. The forest cover approaches that found in the following zone but is notably less dense, with slightly smaller trees and less underbrush. As mentioned above, bamboo is common in the lower reaches of this zone, but drops out rather suddenly with increased altitude. Soils are universally the shallow, stony ones referred to by the Thai Ministry of Agriculture maps as the Slope Complex. ZONE 6: Mountaintop Forest (pa or thang thoeng khao); covers the relatively level top of Phu Wiang above 500–600 m, as well as sheltered areas at somewhat lower elevations on the gently sloping interior of the mountain. This true triple-canopy forest is



markedly denser than that of the mountainside and seems to approach what Pendleton has termed a Moist Monsoon Forest (1962:90). Trees over 30 m tall are relatively common and undergrowth is quite heavy, with the exception of a few more open grassy areas that approach a savannah-like condition. The zone as a whole seems to receive markedly more rainfall than any of the previously described zones. As described in the section above, the soils are quite different from those at the foot of the mountain in Zone 4, at least on the northern plateau where I observed them. There they were largely sandy yellow-orange clay with laterite below, similar to but shallower than the soils for this area on the Thai soil maps. ZONE 6A: Mountain Clearings (lat, not shown on Fig. 2-4); meadows covered with grass and occasional small trees, often surrounding bare rock outcrops (not shown on Fig. 2 – 4). Some of these may be due to relatively recent human activity (e.g., cutting timber). Soil profiles seem to be very shallow, but similar to Zone 6 in general. ZONE 7: Villages (ban); these are of course artificial microenvironments, usually built in Zone 3 near the margins of workable ricelands (i.e., Zone 2). Vegetative cover is extensive and consists largely of trees having some economic value (kapok, tamarind, coconut, jackfruit, mango, etc.) and garden crops. Soils are as described for Zone 3; i.e., Khorat and Phon Phisai Series. During my eight months in Ban Na Di, I was able to gain information informally (and unfortunately not always by personal observation) on the patterns of exploitation currently practiced in the above zones by the local inhabitants. Although my observations were not as systematic as those of White at Ban Chiang (1984, 1995; see above), they do provide a general picture of contemporary subsistence activities. With the exception of Zone 3A, all zones are exploited differentially, providing the local population with a variety of different resources. While there is little direct evidence indicating the extent to which these zones were exploited by the prehistoric inhabitants of the area, a brief review of the current pattern will provide an ethnographic baseline against which this and the indirect evidence present in the site may be viewed. Zone 1 at present finds its greatest utility as a producer of fish, the dominant source of protein for the villagers. Evening net-fishing expeditions to the Nam Phong are frequent occurrences during the dry season. The river bottoms are also valuable source of aquatic and semiaquatic vegetables, form convenient locations for those garden crops that require frequent watering. The yang timber is also a valuable resource. Zone 2 by definition provides the major food resource of the local population, glutinous rice (khao niao; nonglutinous rice, or khao chao-–the variety preferred by the central Thai-–is also grown for sale). As mentioned above, the small mounds rising out of the rice fields are used for the cultivation of field crops: jute, cotton, mulberry, banana, pepper, mango, papaya, cabbage, onion, beans, and other valuable crops are grown. Supplementary protein in the form of rice crabs, small fish, lizards, frogs, and certain insects (mainly red ants, water bugs, and May beetles) is obtained through leisure-time gathering by women and children.


These animals (excepting crabs, fish, and frogs) are also gathered in Zone 3, as are some wild vegetables. More importantly, however, the area is used for cultivation of dry crops on a fairly large scale. Jute and kenaf are probably the two most important crops; they are planted after the area selected is cleared, burned over, and plowed. It should be noted that extensive planting of cash crops such as these is a relatively recent phenomenon; prior to about 1960, cultivation in this zone was probably limited to occasional gardens. Some small timber is also obtained from the scrub forest predominant in this area, as is wood for charcoal manufacture. As mentioned above, Zone 3A at present is not exploited at all, as it contains very little to exploit other than laterite pebbles. However, these areas were almost certainly an important source of iron ore in prehistoric and historic times; Dobby’s statement that these lateritic concretions “have at times been smelted, but extensive use for this purpose awaits development of a suitable process” (1964:80) would seem belied by the many villages in the area bearing names containing the word “iron slag” (ban or non khi lek is perhaps the most common, although I should note that khi lek is also the name for a kind of tree). The Foothill Forest Zone (Zone 4) finds extensive use not only as a source of small game and some medium timber, but as an area in which true slash-and-burn agriculture (hai) can be practiced, despite the apparent unsuitability of Khorat Stony Phase soils. Both cash crops (cotton, tobacco) and food crops (squash, beans, pepper, onions, watermelon) are grown. Zone 5, being too steep and with soils too shallow for agriculture, finds its primary use as an area for the hunting of medium-to-large game, (barking deer, wild pigs, etc.), as well as the primary source of large timber; the bamboo growing on the lower parts of the mountainside is also used frequently. The triple-canopy forest of the mountaintop (Zone 6) is the main hunting ground of the male villagers living around the foot of the mountain, as well as those living on the interior plain. A wide variety of medium and large game is hunted: langurs, monkeys, giant squirrels, pangolins, monitor lizards, gibbons, wild pig, chevrotains, barking deer, samber deer, gorals, several varieties of civet, slow lorises, and formerly sun bears and gaur (now extinct in the area). Many species of birds are also hunted and eaten, including the wild jungle fowl. The projected second volume will contain list of animals and birds hunted and collected in the area compiled by Nai Khamporn Phithaksin, a local resident. Although hunting is viewed as largely a leisure-time recreational activity, hunting expeditions are quite frequent occurrences, particularly during the dry season, and a considerable quantity of the game shot is dried or smoked and taken back to the village. It seems to form a small but significant part of the protein intake of those villages close enough to the mountain to allow frequent hunting trips. As implied above, domesticated plants and animals raised in the village proper form an important source of food; in addition to the food plants listed, chickens and ducks are raised in all villages, and a number of villages raise pigs when sufficient surplus food is available to support them. Tamarinds and kapok are sold as cash crops. Given these seven more or less distinct resource zones, to what extent can we infer a similar pattern of utilization by the prehistoric inhabitants of Non Nok Tha? Due to the



scarcity of even general paleoclimatological and palynological data for the Indochinese peninsula and a complete lack from northeast Thailand, any inferences must be extremely tentative. It does seem possible to assume, however, that conditions at the time of the first occupation of the site were far from identical to what they are today. As yet we have no data on when in the Non Nok Tha sequence inundation wet-rice cultivation first became dominant; it is probable, based on the large amounts of game animal remains and the scarcity of water buffalo in the Early and Middle Period levels, that this form of cultivation was not introduced until sometime toward the end of the Middle Period, or perhaps even later. However, on the basis of paleopathological evidence from the Non Nok Tha skeletons, Pietrusewsky has suggested that perhaps wet-rice cultivation may have been introduced as early as the beginning of the Middle Period, the second large division of the sequence (1974a:135). It is difficult to imagine the exact nature of the impact of wet-rice cultivation on the landforms of northeast Thailand; looking down from the top of Phu Wiang at present, almost half of the visible landscape consists of rice paddies (P1. 2b). The effects of vegetation clearance and diversion of natural drainage patterns necessary to create these have had a marked effect on the entire ecology of the area presently included in Zones 1-3 and possibly 4 as well. The effect of paddy building in changing soil types has already been mentioned; its effect on the flora and fauna of the area and perhaps on overall climate as well must have been equally marked. It is possible that prior to the introduction of wet rice into the area, the rolling plains were covered by a more or less uniform forest of a type similar to the dwarf dry scrub now present in Zone 3. The presently intermittent watercourses may possibly have been swampy areas, representing an extension of the present Zone 1 or something similar to it (cf. Higham and Kijngam 1979, White 1995). Given the heavier forest cover and the wider spread of swampy ground, the climate in general may possibly have been slightly more moist that at present (cf. discussion of views of soil specialists who visited Non Nok Tha in Chapter 1). Evidence supporting the above speculations is scanty; as far as the immediate vicinity of Non Nok Tha is concerned, some evidence that the vegetative cover has changed significantly in the near past is provided by the accounts of the villagers who first settled the area beginning in 1892. They state that the region surrounding the site was covered in dense forest, containing some trees with trunk diameters of up to 3 m. No sign of prior human occupation was visible in the way of rice dikes, obviously artificial clearings, or the like. The forest as described sounded rather like a somewhat more dense version of the scrub forest (mixed to dry Deciduous Dipterocarp forest) present in still-unpaddied areas of Zone 3; i.e., don or khok in my and Thiang’s terminology, but called din khok or din dong (“hill land” or “forest land”) by White’s (1995) informant. It was apparently not a version of the mainly evergreen forest found at present on either the top or the sides of Phu Wiang itself. Evidence that may indicate a more pervasive change earlier on is present in the contrast between the soil laid down during the earlier periods: the latter soil (Soil II; see Chapter 3) is notably more moist and a darker brown than Soil I above it and seems to contain more humus. Supporting this is the fact that Soil II contains more organic carbon than the soils below or the lower part of Soil I, as would be expected due to the yearly plowing of the site (report by Nelson to appear in the projected volume of specialist reports on the site). The


more moist, humic Soil II may possibly have been laid down under conditions somewhat wetter than at present, and in the presence of a heavier vegetative cover. All that can be concluded in summary is that the microenvironmental resource zones available for exploitation by the prehistoric inhabitants of Non Nok Tha particularly those living in the Early and Middle Periods were probably somewhat different, at least in extent, than those used today. They perhaps only gradually assumed their present state from the late (or post-) Middle Period onward through the creation of paddies and the clearing of the forests that this entailed. Increased drainage into the paddies, plus deforestation, may have been at least partly responsible for the present state of Zones 3 and 3A through the processes of podzolization and laterization respectively (Dobby 1964:77–80). It would seem equally likely, however, that little or no change has taken place in the higher zones (aside from some slight deforestation and local extinction of some game species). The ultimate answers to these questions await the collection and interpretation of adequate pollen samples from the area; those from the site, the foothill forest zone, and the mountaintop collected in the course of this program were unfortunately destroyed by an overzealous graduate student. The Present Population The contemporary inhabitants of the Phu Wiang region, along with most of the approximately twelve million other northeasterners, form an ethnic group distinct from the Thai of the Central Plains. In terms of language, diet, dress, and customs they have far more in common with the lowland farmers of Laos than with the Siamese of the Chao Phraya Plain. Linguistically, the Tai2 dialects of all but the southwestern quadrant of the Khorat Plateau are clearly Lao, not Thai, although it should be noted that the two standard languages differ to about the same degree as Spanish and Portuguese (there are of course other Tai and non-Tai languages spoken by smaller groups in the northeast: Lue, Khmer, etc.). The use of glutinous rice (khao niao) as a staple food and the distinctive pha sin worn by women also emphasize the ties to the Lao proper across the Mekong; as the villagers put it, they are “phu lao tae sat thai–-Lao people but Thai citizens” (Keyes 1966a: 354). However, this is not to say, as one ethnographer has, that the northeasterners are simply “’Laotian’ Thai” (Tambiah 1970:31). The villagers sometimes identify themselves as Lao, although they outnumber the ethnic Lao of Laos by at least six to one and are Thai, not Lao, citizens; but they never view themselves as ethnic Thai, for to be ‘Thai’ meant ceasing to be a Northeastern villager” (Keyes 1966b:368). Most often they prefer to see themselves as Khon Isan (“Northeasterners”) and their language as Phasa Isan (“Northeastern language” of Phasa Hao (“our language”), as distinguished from the Phasa Kang (“Central language”) of the Siamese Thai. This “sinn fein” outlook is a relatively recent phenomenon resulting from events in the last century. There are, however, many continuities (and a few surprising discontinuities) present between the modern population and the later levels of Non Nok Tha. Thus it is desirable to present a brief overview of the society, economy, and technology of the current occupants of the Phu Wiang area to determine to what extent the present patterns of subsistence and technology are related to those of the upper third of the Non Nok Tha sequence. In doing so, I rely on the few available ethnographic accounts of the area, chiefly those of Keyes (1966a,b; 1977) and Klausner (11974). The work of Tambiah (1969, 1970) contains some valuable data, but it is problem-oriented and has many inaccuracies,



mainly owing to the imposition of assumptions based on central Thai society. I also draw on my own observation during my two years’ residence in the region during the period September 1965 to April 1975. In contrast to the strip villages characteristic of the Chao Phraya delta, Isan farmers live in nucleated villages averaging between five hundred and one thousand in population (Sternstein 1965); these are subdivided into fairly well defined neighborhoods (khum), which in turn are made up of compounds surrounding one or more houses. The normal residence unit consists of parents, one married daughter and son-in-law, and unmarried children. As the society is normatively uxorilocal, other houses often in the same compound in time are occupied by one or more older married daughters and sons-in-law who establish their own households as younger daughters acquire husbands (Keyes 1975:282-287). A daughter and son-in-law may also set up a household after a customary period of residence with and service for the wife’s parents as partial payment of the bride price (Klausner 1974:70). Each compound contains one or more granaries (sang or lao) of bamboo and mud construction (P1. 3a). The only public buildings to be found are the village spirit houses (ban phi); sometimes a platform for holding village meetings (sala kang ban)-–both usually located in the center of the village; and the school and wat (Buddhist temple compound), both usually built on the edge of the village. Larger villages may be split into neighborhoods by the presence of one or more wat compounds. These last two structures are the only ones likely to be detectable archaeologically, and schools in the Phu Wiang area are a relatively recent phenomenon (post-World War II). The wat compound contains a pavilion (sala) and one or more dormitories for monks (kuti); in more prosperous villages a bot or ground-level hall of stuccoed brick is also present. A necessary adjunct for all but the smallest village is the pond (nong or bung). These are usually found on the margins of villages and supply water for drinking, washing, and bathing. They are either modified from natural perennial ponds by deepening and regularizing their margins, or if necessary excavated by the villagers. However, deepwater wells (bo) are becoming increasingly common in the area. Rubbish deposition zones are rather ill defined and are usually found in unused corners of the compounds and along disused pathways; Calder’s ethnoarchaeological study (1972) of pottery use and discard patterns in a northeastern village provides full details, and will be discussed further when analyzing the refuse sherds from Non Nok Tha. The regional political system is in many ways a modernized carryover from a system with considerable antiquity in Mainland Southeast Asia. Each village is governed by a headman (phu nyai ban), aided by an assistant (phu suai); these are (at least in theory) democratically elected by the villagers. The lowest level of the hierarchy is thus the village (mu ban). Several villages form a tambon (“commune”), administered by a kamnan elected by the headmen from among their number. The central government first makes its authority apparent at the third level; the amphoe (“district”) is governed by an administrator (nai amphoe) responsible directly to the governor of the province (changwat) who is in turn appointed and directed from Bangkok. As the nucleated village pattern of the northeast makes for a natural division corresponding to the governmental mu ban (not the case with the dispersed strip villages of Central Thailand), the system is an efficient one, based on generally democratic principles at the grassroots level and on the unidirectional flow of centralized bureaucracy from Bangkok down to the amphoe level.


The five tiers of this system have developed during the past century from what was originally a three– or at most four-tiered system. Until the late nineteenth century, the Khorat Plateau was made up of a number of what Keyes refers to as “principalities” (1966a:82). These mu’ang were intermediate in area between large amphoe and small changwat and were governed by a “lord” or chao, assisted by a small group of “ministers” forming a krommakan mu’ang or privy council. The term mu’ang in a general sense covers a number of meanings in English, ranging from “town” to “nation” (as in mu’ang Thai, Thailand); here it is used in the sense of principality and the seat of the principality as well. The mu’ang were spaced at intervals of one or two days’ journey by oxcart from one another (cf. map in Kennedy 1970 discussed below), and boundaries between them were doubtless rather vague. Such vagueness also characterized the ties of the mu’ang to the top of the hierarchy. The mu’ang were loosely affiliated in vassalage to one or another of the four kingdoms in the region (Louang Phrabang, Vientiane, Champasak, and Ayutthaya/Bangkok), but patterns of allegiance were constantly shifting as the power of the four kingdoms waxed and waned (Keyes 1966a, loc. Cit.) in a pattern typical of Southeast Asian states that Wolters aptly terms “concertina-like” (1982:17). Only after the 1827 revolt of Chao Anu, king of Vientiane, and the consolidation of all the Khorat Plateau by Rama III and his successors was a fourth level of regional administration (monthon; later phak) introduced. With the subdivision of the region into changwat and the addition of the tambon/mu ban distinction during the first part of this century, the present five-level system was arrived at (Keyes 1977:152). The village-mu’ang-kingdom pattern is obviously one with a long history in the area (Wolters 1982, Bayard 1992); it is known to have been the dominant political system of wet-rice farming peoples from Laos and Thailand to northern Burma (the Shan mong governed by saobwa; Hall 1964:158 – 159). It almost certainly goes back to the earliest Indianized kingdoms of the region, although it is uncertain whether the many political entities mentioned in Chinese chronicles prior to A.D.700 (Smith 1979b:444)) should be viewed as large mu’ang or small kingdoms (i.e., “complex chiefdoms” or what Wolters (1982) terms mandala; cf. Wilen 1984). Archaeological surveys carried out in the southern part of the Khorat Plateau (Higham and Parker 1970, Kijngam et al. 1980, Welch 1984, 1985) have added to our knowledge of large numbers of moated sites that were apparently established throughout the Mun and lower Chi Basins by at least the first millennium A.D. These were first detected in aerial photographs by Williams-Hunt (1950), subsequently investigated by Thai archaeologists (e.g., Saengwan 1979; Vallibhotama 1984), and have now been collated and classified by Moore (1988). Although inadequately dated, there is evidence for a pre-Indianization date for some of these sites in the Phimai and Roi Et areas, as well as Khon Kaen itself (e.g., Bayard et al. 1982-1983; Welch 1984, 1985). Hence some form of state-level organization based on the mu’ang model may well have laid the groundwork for the introduction of Indian politico-religious concepts (Bayard 1980b:106-107; 1986-1987). This and the whole complex question of “chiefdom” vs. “state” will be discussed further in the conclusion of this report. Turning to the Phu Wiang region specifically, we have little concrete evidence of its early history; village records do not seem to have been kept until after World War II. The region was until quite recently an isolated one; an all-weather road into the district center was completed only in 1963, and few westerners had visited the area prior to the Non Nok Tha



excavations. Henri Mouhot passed close to Phu Wiang on his way from Phu Khiao to Loei in 1860 (Mouhot 1868), but the first Eurpoean to visit the area in recent times were Solheim in 1962 and Chester Gorman, the discoverer of Non Nok Tha, in 1963. Mention of the Phu Wiang region in historical records is also rare. It is known, however, that Phu Wiang was a municipality belonging to the kingdom of Wiang Chan (Vientiane) during the eighteenth and early nineteenth centuries: it was one of the southern border municipalities annexed to Siam following the suppression of Chao Anu’s rebellion in 1827 (Vella 1967:86 and Map 2). A Thai map dating to the early part of the nineteenth century shows Phu Wiang as a fortified settlement surrounded by the enclosing walls of the mountain (Kennedy 1970:P1. 8). Given the limited resources available to me, I have been unable to locate any further written reference to the area either before or after this period. However, some additional evidence as to its importance, at least some time in the past, is afforded by two place-names in the area. The first of these is the name of a village on the plain inside the mountain: Ban Mu’ang Kao, “the village of the old city”. This name may refer to the fact that the district administrative center (and supposedly the one shown on the Thai map(Kennedy 1970:330) was formerly located there prior to being moved outside the mountain to its present location, a move that I was told took place after World War II. This name tends to support the possibility of the presence of an important center there prior to initial establishment of the Thai government district center; the establishment of the latter in this rather inaccessible location makes the previous presence of a community of some size there very likely. The second and more important name is that of the mountain itself: Phu Wiang, “City Mountain”. Wiang has a more specific meaning than “city”, however, the Dictionary of the Royal Academy of Thailand defines the word as follows: “A city or town (mu’ang) which has encircling walls” (Ratchabanditsathan 1973:847). The walls referred to are of course the natural walls of the mountain itself; no physical remains of the old city or town, however, have been located. We can only add that the area of the interior plain encompasses some 100 km2, most of which is suitable for wet-rice cultivation (ca. twothirds of this area is currently under cultivation), and which at present supports some twenty villages and a population of over ten thousand. The ring wall surrounding the plain presents a steep, easily defendable face to the outside; this face is at least 100 m high at all points save for 1 km on each side of the single gap in the wall, and the gap itself is only 100 m wide. From outside the mountain, the interior plain is of course invisible and the mountain appears to be a single long ridge similar to the mountain of Phu Phan some 25 km to the east. In short, given the isolation and lack of firm governmental control by any kingdom in this area of northeast Thailand until 1893 (Keyes 1966a:85), it would be surprising if this interior plain was not the seat of some local authority during part or most of the past two millennia and perhaps even earlier than that. It has certainly been an administrative center for at least the last two centuries. As Non Nok Tha is located some 10 km north of the interior gap, it seems likely that if such a local authority did exist, the prehistoric inhabitants of Non Nok Tha would have been subject to it to some degree. Traditional accounts of the inhabitants of the nearby village of Ban Na Di are also unfortunately of little help in attempting to establish a link between them and the prehistoric


population. The village was founded only some seventy-four years prior to the time of the 1968 excavation, in Rattanakosin Sakarat (Bangkok Era) 112, or 1894; I was told by one of its oldest inhabitants, Nai Bun Phithaksin (born shortly after its initial settlement), that no trace of prior occupancy was evident in the area. The sole exception noticeable to the early settlers was the presence of traces of an ancient road running along the southern edge of the present village; some of these traces are still visible at present, although badly obscured by paddies and garden crops. The road may have considerable antiquity, as it was apparently in disuse at the time of preparation of the Thai map mentioned above, which shows no road to the northeast of the mountain (Kennedy 1970:P1. 8). The road is of interest chiefly because-–if Nai Bun’s impressions are correct-–it seems to have connected a series of archaeological sites that appear through limited testing to have been important ones. To the north the road curves to the west of Ban Na Di through the site of Don Kha, then continues north to the site of Don Po Daeng (NP 8, see chapter 1), and on to cross the Phong River at the northern end of Phu Wiang. South of Ban Na Di the road apparently swung to the southwest of Don Kok Pho (NP 6) to intersect a small site near the foot of the mountain called Nong Sa Phang; from there it proceeded south paralleling the foot of the mountain to a point past the present district center called Phu Khi Bao, or “mountain of crucible slag”, a large heap of slag apparently from iron smelting. Although Nai Bun believed the road continued south from this point, he had no knowledge of its further course. Unfortunately a crowded work schedule left no time for further investigation of this road, nor were Higham and Parker able to gain more information in 1970. However, its presence certainly supports the likelihood of contact between the late prehistoric inhabitants of Non Nok Tha and the presumed administrative center inside the mountain. Although the immediate area around Non Nok Tha may have been deserted at the time the present population settled there, we have no evidence supporting a widespread depopulation of the entire region. The founders of Ban Na Di and the larger nearby village of Ban Nong Na Kham “migrated” a distance of only about 5 km from the older, wellestablished village of Ban Khok Na Fai on the banks of the Phong River, which at that time was threatened by an epidemic of “coughing fever”-–possibly tuberculosis. Many other villages in the area have been in existence for well over one hundred years, and the expansion into the Non Nok Tha area probably represents merely a local movement resulting from the slowly increasing population. This process continues to the present, as indicated by the large number of Ban Noi and Ban Mai place-names (“little village” and “new village” respectively). The interesting point to be drawn from this expansion, when taken in conjunction with the large number of iron-period sites in the area, is that population density in the area may have been somewhat greater five or six hundred years ago than it was a century ago and than it is in fact today. The depopulation of the region in general may well have been a result of military activity in the area (Kennedy 1970:340, Fig. 2) and the forced resettlement of large numbers of people by both sides during and after the Vientiane revolt (Vella 1957:81– 88); however, we have no direct evidence of this, and no traditions of forced resettlement and repopulation exist. In fact, as nearly as I could ascertain, the present inhabitants view themselves as having always occupied the area. Evidence that this is in fact generally true over at least the past two or three centuries is available from linguistic data. Judging from what sources are available on Lao dialects (primarily Brown 1965), plus my own observations in the area, Lao as spoken in Laos and Northeast Thailand (this does not include the “Lao” or Kam Mu’ang of Northern Thailand)



would seem to include three principal dialects. These are based chiefly on tonal distinctions rather than other phonological or lexical variations. The first of these extends south along the Mekong from Louangphrabang and Saygnaboury into Loei Province; it is distinguished by an HO tone rather resembling the H2-M2-L1-L3 tone in Central Thai3; i.e., more or less falling (see Brown 1965:87–94). The second area extends from north of Vientiane south across the Mekong to include western Udon and Khon Kaen as well as most of Phetchabun and Chaiyaphum; this dialect is usually characterized by a merger of the MO tone with a more typical rising HO tone, as well as by a high rising LO tone (Brown 1965:95–102). The third zone includes the rest of central and southern Laos and northeastern Thailand east of a line running somewhat west of the cities of Udon and Khon Kaen. This line, at least from southern Udon Province south, may well coincide with the natural barrier of Phu Phan. This last and numerically most important dialect is marked by distinct HO and MO categories and a typically “high falling” LO tone category (Brown 1965:103–110). Thus the speech of the present inhabitants of the Phu Wiang area has many features more like the speech of Vientiane than that of the city of Khon Kaen a mere 60 km away. These three dialect areas, forming as they do three north-south bands cutting across the Mekong and the Thai-Lao border, would seem to represent a pattern of distribution of considerable antiquity; I can estimate that two or three centuries would be a reasonable minimum. Thus it would seem that the present population--or one very closely related to it–-has inhabited the Phu Wiang region for some time. This is in contrast to the population west of the ecotone already described; the Loei population appears to have entered the area from the north only in the sixteenth century (Bayard 1980c:128–134). Similarly, the economy of the inhabitants of the area reflects a largely successful and longstanding adaptation of wet-rice subsistence farming to the often undependable rainfall. While the Khorat Plateau is the poorest region of Thailand with its low rainfall and poor soils, its villagers are better off than subsistence farmers in many other parts of the world, and government development programs (including irrigation systems) instituted over the past two decades have considerably improved their position relative to the rest of the kingdom. Traditionally, however, the Isan farmer has been dependent on cultivation of the glutinous rice staple by inundation methods, and no irrigation is practiced in the Phu Wiang area. Unfortunately, no account of the yearly cycle of subsistence activities is available for the central Khorat Plateau; however, a detailed account is provided by Klausner for Ubon Province (1974:27–43). Ubon differs from the Phu Wiang area in that rice is grown during the dry season by sowing in natural ponds (na saeng) and during the rainy season in water-retentive depressions (the khao bao crop) and dry swiddens fields (na hai), as well as in inundation paddies (na); this is apparently made possible by Ubon’s somewhat greater and more even rainfall (see Table 2-2). In the Phu Wiang region only the main wet-season crop is feasible, although cash crops and vegetables are grown both in swidden patches at the foot of Phu Wiang and in areas adjacent to permanent water supplies. Preparation and planting of the nursery bed takes place in late April and May, after the arrival of the first rains; plowing of the paddies (using buffalo), harrowing, and transplanting of the seedlings are accomplished mainly during June. Two anxious months follow as the farmers await the often undependable midseason rains; if these are not forthcoming, the heavy September rains are of little help in salvaging the crop. The fields are gradually drained as the crop matures during the slackening rains, and harvesting takes place in November and December. Threshing is done by beating the sheaves on a


prepared cow-dung floor, after which the crop is stored in granaries to be husked and winnowed as required. I would estimate that glutinous rice provides some 80% of the caloric intake of the average villager; much of the rest consists of fruits and vegetables (such as the famous som tam salad of unripe papayas). These are mainly cultivated in the village and some of the other local resource zones described above; however, a significant proportion (I would guess over 20%) is not grown but gathered, either purposely or in the course of another activity. This is also the case with protein sources; the major protein staple of the Isan villagers, of far more importance than chicken, pork, or any other domestic animal, is pa daek (pla ra in Thai), a fermented fish paste made by layering small fish with salt and allowing a selective rotting process to occur. Although available in the market, much of the pa daek consumed is prepared in the village for local consumption or barter using fish caught as the waters recede following the rainy season, with the catch being consumed fresh or preserved by drying. In addition to fish, well over fifty species of mammals and birds have been hunted in living memory in the Phu Wiang area (a list provided by my friend and local informant Khamporn Phithaksin will be included in the projected volume of specialist reports), although several of these are now locally extinct. Many kinds of lizards, snakes, frogs, snails, and insects are gathered, primarily by young boys but also by women while gardening and so on. Such hunting and collecting activities are viewed as recreational as well as useful by the villagers; however, I would say that a significant portion of their protein is gained by such foraging activities. In short, the Isan villagers of Phu Wiang must be viewed as part-time hunter-gatherers as well as wet-rice agriculturalists; as such, they continue a pattern of subsistence with its roots in the Hoabinhian (Gorman 1971), and one that has seemingly declined in importance only slightly since the first occupation of Non Nok Tha and Ban Chiang (Higham 1979) by swiddens farmers-cum-foragers (it is interesting to note that this ancient Southeast Asian foraging pattern has been continued in the parks and riversides of America and New Zealand by Indochinese refugees). Indeed, one can fairly say that the northeast Thai economy has undergone more change in the past two decades than it has in the previous two millennia. With the construction of all-weather roads and more reliance on a cash economy and cash crops, the Isan farmer is becoming more specialized; increasing numbers of young men are leaving the fields–-not for the traditional few years’ employment in Bangkok before returning home but to settle permanently in the towns and cities of the northeast as well as the capital. The traditional economic pattern, still very much a reality in the Phu Wiang region when I first arrived there in 1965, has moved well along the way toward specialized farming and interdependence on urban centers. Nonetheless, many aspects if the old way of life remain and will continue to do so for some time to come. Continuities are also apparent between the technology of the modern farmers and that evidenced in the upper levels of Non Nok Tha; however, these are of a general sort, and there are some surprising specific differences. For example, pottery continues to be made throughout northeast Thailand (although not in the Phu Wiang region proper) in much the same way as that of basal Non Nok Tha and Ban Chiang; use of the paddle and anvil is universal in those villages specializing in pottery manufacture (Solheim 1964b, Bayard 1977b), although differences are certainly apparent in specific traits like temper type,



carved vs. plain paddle, and the like. Similarly, the contemporary pottery of Phu Wiang (purchased from potting villages in neighboring areas and now rapidly being replaced by aluminum vessels) is quite distinct from that of Non Nok Tha. It is smooth-finished and tempered with fired ground clay and chaff, in contrast to either the sand-tempered, cordmarked ware that occurs in all layers of Non Nok Tha or the chaff-tempered ware that predominates in the upper layers of the site. Nonetheless, the general impression is one of continuity in form and manufacture (but with two important exceptions noted below). The iron tools from the upper levels of Non Nok Tha also find close parallels in contemporary villages (although the traditional forms are being replaced by Western models at present). Iron digging-stick blades (siam) and scissors recovered from the Non Nok Tha Late Period are virtually identical with ones in use in neighboring villages today, and even the socketed bronze axes of the Middle Period have a general similarity in hafting and use to the traditional contemporary ones. These latter are made of iron, with the socket formed by forging of lugs to form a sleeve rather than by casting; otherwise the use of a counterbalance shaft to which the handle proper is attached is quite likely the same method of hafting as employed in the Non Nok Tha Middle Period. It is difficult to envision the prehistoric axes being used at all efficiently if hafted directly to the handle in the Western fashion. Some general continuity in house style is also evident. The type of dwelling viewed as “typically” Isan by the villagers (or, as Tambiah puts it, “orthodox”) is a large double-roofed structure forming two large rooms of ca. 6 by 3 or 4 m; these are separated by a 2 m wide passage, making an overall size of about 6 by 9 m2. A cooking and washing platform (where water jars are kept) is usually attached to one side of the dwelling, and an unroofed porch (san) is often built onto the front of the house (cf. Tambiah 1970:19). The entire structure is elevated between 1.5 and 2 m on posts spaced about 2 m apart (P1. 3b). One or both of the large rooms are subdivided. Ideally, the parents and young children sleep in one of these subdivisions, with an older unmarried daughter or daughters in the other; the other half of the house is to be occupied by a married daughter and son-in-law. This is the normative pattern; but while excavating Non Nok Tha, I undertook a survey of dwellings in Ban Na Di and found that of the 157 houses, only 23 (15%) fit this pattern. The vast majority were smaller single-roofed dwellings measuring about 6 by 6 m and subdivided into a living area (6 by 3 m) and two sleeping rooms (3 by 3 m each); most had porches and cooking platforms attached (P1. 3c). This provides some support for my impression that neolocal residence is in fact more common than the traditionally approved matrilocal pattern. Most houses were constructed of hardwood boards, but some of the smaller and less substantial ones were sided with woven bamboo screens. Roofs were traditionally made of wooden shingles, but even in 1966 a large minority were roofed with the “modern” but heat-producing iron; the number is doubtless far greater today. Archaeologically, such structures would of course leave no trace of foundations, but they should be detectable by posthole patterns and differential distribution of rubbish (Calder 1972). Despite the opening of a large area at Non Nok Tha (340 m2), no complete outline of an entire dwelling was recovered. Substantial portions of several structures were recorded, and these will be discussed in full in Chapters 3 and 4. Houses of the Late Period do appear to have been similar to the smaller variety found in the Phu Wiang area at present, with a 4 by 4 post construction suggesting an overall size of 6 by 6 m; in at


least one case, traces of a cooking platform were found. None of the partial structures seemed to correspond to the larger “typically Isan” houses, but it could well be that only portions of such structures extended into the excavations area, making definition impossible. As mentioned in Chapter 1, concentrations of fired clay were frequently encountered at Non Nok Tha. These may be remains of burned wattle-and-daub granaries, as postulated for the test pits discussed in Chapter 1; on the other hand, I believe that many in the area excavation proper result simply from scorching by groundlevel fires (the charcoal having been washed away during rains), or from special-purpose pits such as the apparent lime-burning kilns encountered in 1968 Late Period level 4. In addition to these continuities, there are a number of rather surprising discontinuities present between the assemblage recovered from the Non Nok Tha Late Period and the artifacts in use today. Some of these can be explained convincingly; for example, the presence of large numbers of small baked-clay marbles, 1–2 cm in diameter, throughout the Middle and Late Period levels of the site and their complete absence in contemporary villages. Although Parker and I were initially baffled in 1966, our older workmen were able to confirm these as ammunition for pellet bows (see Chapter 3 and White 1982a); their absence today is due to the introduction of Western-style slingshots in recent years. Similarly, the absence of lumps of hematite used for red pigment, common in Middle and Late Period levels, may be due to the introduction of superior coloring agents. Turning to the pottery, it is more difficult to explain the apparent absence of fragments of the distinctive vessels (mo nung) used for steaming glutinous rice; this staple is always steamed (never boiled) in a basket placed in the wide neck of the steaming vessel. None of the perforated sherds found at Ban Na Di and tentatively described by Higham and Kijngam as rice steamers (or alkali filters? 1984:202, 684) were recovered from Non Nok Tha. It seems unlikely that a dietary shift from regular to glutinous rice (a highly significant one for rice eaters) took place in the past two centuries; it is far more probable that the small size of sherds from the upper Late Period levels has precluded our making a positive identification of such vessels. Less likely, but also feasible, is the possibility that another sort of vessel was used formerly. Also puzzling is the absence from the contemporary village scene of the pierced biconical clay objects we have called spindle whorls. The spinning wheel is universally used in the Phu Wiang and Khorat regions today, and our workmen were unable to ascribe a function to the clay objects, which were recovered from all Middle and Late period levels at the site. We can only conclude that the spinning wheel was a relatively recent introduction to the area; spindle whorls drop out of the Non Chai sequence in the late first millennium B.C. (Bayard et al. 1982– 1983). In summary, there are obvious economic and technological continuities apparent throughout the Non Nok Tha sequence and into the contemporary society. Chief among these are an apparent reliance on rice as a staple, supplemented by broad-spectrum gathering, from the earliest occupation of the site until the present, and an overall continuity in ceramic and metal technologies. Such continuity has almost certainly existed in the important bamboo and hardwood technologies as well, although it is not reflected in the archaeological record. However, it must be emphasized that society in the Phu Wiang area was far from static; important and far-reaching changes appear to have taken place at the end of the Early Period and during the latter portion of the Middle Period; significant changes also seem to have occurred during the gap separating the Middle and Late



Periods. Finally, enormous changes have taken place in the economy and society since the establishment of the constitutional monarchy in 1932; these changes have accelerated rapidly since improvements in communications and road transport were carried out in the 1960s. Thus, while the present population can provide a relatively reliable baseline for comparison with the Late Period at Non Nok Tha, extension of analogies to earlier periods at the site must be made with caution. Any such analogies to be drawn will be discussed fully in the conclusions of this report. Footnotes 1. Attracted to the recent renaissance of the Austric hypothesis, Higham’s most recent scenario would have the bearers of early rice cultivation migrating southward “down the lines of lest resistance, the rivers which radiate out from the region of the upper and middle Yangzi Valley” (1966:357). The linguistic evidence on which this is based is certainly suggestive (Bayard 1996)[Higham 1996, Bayard 1998 added to bibliography] 2. “Tai” is a generic term for a language family which includes not only Central Thai, but Northern Thai, Lao, Shan, and many “tribal” languages spoken by groups in Laos, Vietnam, and South China. 3. “H”, “M”, and “L”, in the tonal descriptions refer to the “high”, “middle”, and low groups of consonants in the Thai alternative tones which can be assigned to a syllable; to more possibilities are available with “dead” syllables ending in a stop, depending on vowel length.

Chapter 3 Excavations at Non Nok Tha: 1965-1966 By R. H. Parker Prefatory Note (DTB) Parker wrote this contribution in the early 1970’s; subsequent analyses of both seasons’ work have of course changed our interpretation of certain aspects of the site. However, it is important to include Parker’s report to provide historical background on the problems encountered in the pioneer 1965-1966 excavation and to document the initial stages and evolution of our present interpretation of the site’s stratigraphy and cultural sequence. While some readers would doubtless have preferred a single synthesized presentation of the excavation procedures and stratigraphy, such a presentation would necessarily have to simplify and gloss over some of the difficulties Parker and I encountered in both seasons. Much of Non Nok Tha’s importance lies in its pioneering attempt to apply large-scale excavations based on natural stratigraphy in a Southeast Asian context, and I believe it is important to report in full the difficulties and apparent failures as well as the successes we encountered. In addition, Parker’s retirement from Otago in 1977, before much of the ceramic and none-ceramic analyses had been carried out, made a single joint synthesis impracticable. Hopefully the overall schema of Early, Middle, and Late Period levels (abbreviated respectively as EP, MP, and LP; (see fig. 4-2) employed in the rest of this report alongside the specific 1966 and 1968 designations will aid the reader in coping with what is admittedly a stratigraphically complex and confusing site. I have added these overall period designations in parenthesis to Parker’s following account. Other added comments are in brackets, some with my initials (DTB) and some with Solheim’s (WS). Also, as Parker did not provide a general description of the site proper, I have supplied one at the beginning of Chapter 4. While it is obvious that Parker’s interpretations of the late 1960s and early 1970s are not identical to that presented in Chapters 5, 8, and 9 of this report, there are few if any major discrepancies. There are both advantages and disadvantages in having different portions of the same site excavated by different directors. On the negative side is the inevitable likelihood of differing interpretations of stratigraphic features, burial and ceramic typologies, and the like; however, these are in my view outweighed by the greater confidence and security in interpretation arising from two independent (semi-independent, in this case) appraisals. Introduction The following is an account of the excavations on the site of Non Nok Tha in the season of 1965-1966. Essentially it is final revision of the text prepared by me for inclusion in Solheim, Parker and Bayard’s preliminary reports on excavations at Ban Na Di, Ban Sao


Lao, and Phimai. It is concerned solely with an account of the excavations and description and analysis of the stratigraphy and structural remains. It suggests deductions and conclusions only insofar as they can be drawn directly from the excavation evidence. In general, I have not thought it proper to make extensive alterations to the text on the basis of Bayard’s 1968 excavations on the same site nor to attempt a synthesis of evidence from the two seasons. It seems to me better to use this report to set down the evidence and conclusions as they seemed to us in 1966. In fact, changes in interpretation suggested by the 1968 excavations are in all cases relatively minor and Bayard has already published a plausible reconciliation of the two seasons, necessarily tentative, as he was unable to obtain permission from the landowner to cut trenches linking the two areas stratigraphically (1971b). It is regrettable that unavoidable delays have occurred in the preparation of this final report and equally regrettable that analysis of the stratigraphical and artifactual evidence has been carried out in institutions situated five thousand miles apart. The delays in this report resulted in the 1968 excavations being published first, while much of the evidence that was in Bayard’s mind and used by him in making his own analysis was not publicly available. The layer-level number problem arises from the fact that the stratigraphy in the area excavated in 1968 was significantly different from that in the 1966 area. In each area there were layers present that were entirely absent from the other. Hence we have a set of layer (and level) numbers for 1966, a somewhat different set for 1968, and finally Bayard’s tentative reconciliation of the two systems (see Fig. 4-2). Outline of the Excavations Fairly extensive text excavations were conducted on the mound on behalf of the expedition in January 1965. This testing was conducted by Ernestine Green (under the general direction of Prof. W. G. Solheim II) and was part of a general program of tests conducted on sites located by Dr. C. Gorman in the previous season. In all, Green dug seven small tests pits on various parts of the mound and, as a result of the evidence derived from these (see Chapter 1), the mound was selected by Solheim for more complete investigation during the current season. The expedition returned to the site on December 10 1965, and a new square was opened close to the present center of the mound. The excavations on this occasion were conducted by myself and D. T. Bayard, assisted by Nai Thaweechai Uthaiwee, again under the general direction of Solheim. On December 26, 1965, the party was joined by ten students from Silapakon University (University of Fine Arts), Bangkok, and the new test square became a 3 x 2 m unit in a 4 x 4 m grid of squares laid out to the west and south of an arbitrary datum close to the center of the mound. The pegs of this grid were numbered from east to west and lettered from north to south, each square being designated by the same letter/number as the peg marking its northeast corner (Fig. 3-1). The new test square thus became square C3 in the grid. It and other squares in the C line measured 3 m east=west by 2 m north=south; the remaining squares were 3 by 3 m. With the students available for supervision, on the force of local labor was increased to ten squares C4, C5, C6, D4, E4, and 14 m east west). At the same time the survey for the preparation of a detailed site plan was begun by three students under the guidance of Solheim.


Fig. 3-1

The first party of students left on 8 January 1966, and were replaced by a second party of twenty students. The local labor force was then expanded to twenty men and squares G4, C7, D6, D7, E5, and E6 opened, expanding our initial sections into an areal excavation of about 204 m2 with long sections of 17 m (Pl. 3a). This second party of students remained with us until 22 January. With their departure, the labor force was reorganized, the detailed supervision of squares (previously undertaken by senior students) now being undertaken by senior workmen Nai Inta, Nai Chaloem, and Nai Thawee. At the same time, to ensure the completion of our two major long sections, work in squares G3, D5, D6, D7, E5, E6, and C7 was halted at the layer 9 surface and our activities confined to C3, C4, C5, D4, E4, and F4. Work continued on this basis until 30 January, when the site was backfilled and a short break taken to enable Bayard and myself to visit Bangkok and to carry out some excavation to the Lam Phra Phloeng sites (Pl. 4A). Non Nok Tha was reopened on 12 February and work there continued until 28 April when the season’s activities in the Phu Wiang District ceased. During the last phase the excavation of all squares except G4, E3, D7, and E6 was completed to the surface of the natural, although the time available did not permit our completely clearing more than a limited sample of the complex disturbances from the earliest levels (Pls. 4b and 5a-d). We would like to take this opportunity of expressing our gratitude to the Fine Arts Department, Thailand, to the Nai Amphoe of Phu Wiang District for permission to excavate, to the Police Major at Phu Wiang for the offer of police protection, and to the


officials of the Fine Arts Department at Khon Kaen for their kind cooperation and Assistance. We thank the students who worked with us. And most gratefully acknowledge people of Ban Na Di for their warm hospitality and for their help and assistance so freely given. Stratigraphy In all, twenty-two major layers were recognized (Fig. 3-2 to 5). These, from the top downward, are: (1) The modern topsoil. Black to dark gray [note: no Munsell chart was available to us in the field in 1965-1966; see Chapter 4 for Munsell values of the equivalent 1968 layers— DTB]; heavy root mat; contains humus; much disturbed by recent plowing. Little cultural material, most of it modern. About 7 cm in average thickness. (2) Black, sandy; very compact and hard. No cultural material. Limited to western side of site; lenses out in squares C6, D6, E5; 0—10 cm in thickness. (3) Dark gray but distinctly lighter in color than layer 1. Contains some but not much raw humus; progressively lighter with drying. Upper part of layer disturbed by modern agricultural activity. Local concentrations of potsherds scattered throughout layer rather patchily. Compact and hard. About 10 cm in average thickness. (4) Concentration of cultural material in matrix of material similar to but darker than layer 5 below. About 4 cm in average thickness. (5) Dark red brown, rather lumpy. Layer as a whole is moister and softer than in layer 3 and roots are generally larger. Thickness about 6 cm but rather variable. (6) Concentration of cultural material in matrix similar to layer 5 above but slightly darker. About 2 cm in thickness. (7) Red brown, lighter than layer 5 above and softer and moister when freshly exposed, but dries very hard. Compact and even textured; rather fine grain size. No cultural material. Lenses out to west in E6, D5, on top of layer 8 below. Almost 7 cm in maximum thickness. (8) Brown, sandy, even-textured material restricted to western part of site. A concentration of cultural material lies on its upper surface. The rest contains a fairly even scatter of rolled and badly weathered small sherds more heavily concentrated toward the bottom of the layer. Variable in thickness up to 15 cm in thickest parts, as in square D7. (9a) Brown to dark brown, similar to layer 9 (below) in soil type and behavior on drying. Limited to the central area of the site where it lies in, and partially fills, the concave upper surface of layer 9. Contains comparatively little cultural material except for fair numbers of relatively small sherds. (9) Red brown on the east to almost black on the western side of is site; rather similar to layer 5 above but very lumpy. As it dries, cracks develop and widen to an even more marked extent than in layer 5. Very heavy concentrations of cultural material on the surface especially in the east. Pottery characterized by relatively large sherd size. Fairly large numbers of largish roots tending to run horizontally near or on the surface of the layer and to break sharply down vertically through patches of weakness. Quite variable in thickness but generally between 3 and 15 cm. (10) Similar to above but lighter and more reddish in color. Little cultural material. Root penetration for the most part vertical reflecting the generally softer condition of the layer. About 5 cm in average thickness.


Fig. 3-2

Fig. 3-3


(11) Similar to layers 9 and 10 above. A concentration of cultural material, fairly dense in places but on the whole less so than in layer 9. Root channeling again tends to be horizontal. About 3 cm in average thickness. (12) Dark brown loam, fairly soft, even textured dries; dries out fairly hard but always remains markedly softer and easier to cut than any of the layers above. Slightly yellowish tinge when dry, especially on surface. Little occupational material; 6 to 10 cm in thickness. (13) Similar but slightly darker material with heavy concentration of cultural material on and in the layer. Rather more compacted than layer 12 and shows some cracking when dry, but is never lumpy. Thickness is 4 to 6 cm. (14) A darker, harder zone with much diminished cultural material; darkens further and bakes very hard on drying with marked cracking of exposed surface. Unlike layers 5 and 9 above, cracks never develop horizontally to isolate discrete lumps. About 10 cm in average thickness. (15) A very heavy concentration of occupational debris in a matrix intermediate between layers 14 above and 16 below. About 6 cm average thickness. (16) A soft-textured medium brown, sandy loam with little cultural material. About 6 to 10 cm thickness. (17) A rather thin layer of cultural material in a matrix similar to layer 16 above but containing somewhat more cultural debris and probably for that reason rather darker in color. About 4 to 8 cm thick. (18) A zone of generally yellowish soils resembling layer 19 in soil type but layer 17 in the nature of the cultural inclusions. Contains a moderate amount of cultural material. Thickness variable but averages about 5 cm. (19) A yellowish gritty layer that seems to be the natural, minimally modified by cultural inclusions. Contains some local lenses of blackened soil, especially in C4, C5, D4, E4. Averages about 10 cm in thickness. (20) Light yellow to light yellow brown. Similar to layer 19 but less gritty in texture. Contains fairly substantial quantities of cultural materials, especially sherds (often of large size), bones, and other materials probably derived from the layers burial mounds. Frequent patches of shells of freshwater snails. About 10 cm in average thickness. (21) Very similar to layer 20 but on the whole darker in color. Cultural inclusions similar. Content of snail shells markedly less. About 20 cm in average thickness. The considerable thickness of this layer, the variability of the pottery, the considerable typological differences between burials cut from the layer, and the fact that the burial mounds located at this level seemed to be based at somewhat different absolute depths within the layer all combined to suggest that layer 21 was not a single layer but a complex of perhaps three layers. We failed to resolve these postulated layers stratigraphically, however and the indications as we saw them in 1966 were that layer 21 was a single formation possibly built up over a protracted period. In 1968 Bayard successfully resolved these three suspected layers and established their stratigraphical reality (Bayard 1971b). (22) The natural lateritic soil unmodified by cultural activity. Only in a limited area-—in squares C4, C5, and D-—was it possible to define all twenty-two of these layers. Over the rest of the site erosive processes to be described later had, from layer 9 down, removed the whole or large parts of particular layers. In this situation the layer numbers are necessarily discontinuous in most squares and were allocated by relative position within the particular soil (see next section), checked by matching the layers across the baulks with those in adjoining squares, and cross-checked by confirming the match as the baulks were removed.


Fig. 3-4

Fig. 3-5


Note on layers 7, 8, 9a, and 9: Over the whole central area of the site (especially in squares C5, C6, C6, D5, E5, E6), the soils below the layer 9 occupation have been heavily truncated, probably by stream erosion in the period immediately after the close of the layer 11 occupation. In this region, therefore, all layers below 9 down to 19 or 20 are either entirely absent or limited to residual lenses on the east and west sides of the eroded area. The broad but shallow depression left by the erosion was first partially filled by the materials of layers 9 and 10, leaving the surface of the former dipping in a wide, even curve. Then at the end of the layer occupation (see below: “level IX”) rapid erosive processes seem to have stripped layer 9 materials from the higher ground (especially to the east) and redeposited them over the concave surface of 9, partially—-in places almost completely—-leveling the surface and rapidly sealing the debris of the Level IX destruction of the settlement. This leveling up of the surface was completed, probably somewhat more slowly, by the formation of layers 7 and 8. The Soils The above layer can conveniently and significantly be grouped into five distinct formations. From the top downward, these are: Soil I. Hard, lumpy, brownish, becoming sandier and more even textured and conspicuously harder and more compact close to the surface, especially on the western side of the site. The gray-black topsoil is in situ throughout and the whole formation provides essentially the complete profile of the modern soil. Soil II. A very similar but fossil soil separated from the soil above by layer 7 on the east and layer 8 on the west. On the west the upper part of layer 8 has been in part incorporated into the modern profile of Soil I. To the east the upper part of Soil II seems to be missing, giving a truncated profile. On the west it is complete and sealed by materials derived from the erosion of Soil II on the east. Soil III. A rather deep, medium brown loam, somewhat sandy in places and up to 50 cm in thickness. Again a fossil soil. It gives a rather abrupt interface with Soil II and the archaeological indications are that the whole of Soil III was in situ for a long period before a part of Soil II began to form. The clearest evidence for this is the approximately 6 to 10 cm of soil that is largely sterile culturally overlying the latest occupational level within Soil III and the very sharp cultural break between this occupation and that in the lower part of Soil II (layer 11). Soil IV. A deep (20 to 50 cm thick) complex of materials representing the natural soil minimally modified by cultural activities. Throughout the period when Soil IV was forming, virtually the whole of the excavated area was in use exclusively as a burial ground. Almost the whole of the considerable amount of disturbance that modified the natural materials to form Soil IV results from the digging of graves, the building of grave mounds, and the subsequent leveling up of the site by the erosion of the mounds. The cultural inclusions of Soil IV seem, almost without exception, derived from the erosion of grave mounds. The net result of these complicated and long-continued processes has left the detailed


stratigraphy of Soil IV very intricate and difficult to decipher. The general overall pattern of the major layer formations was, however, relatively clear and our diagnosis of these is substantially confirmed by the typology of the burials. Soil V. The natural lateritic soil, reddish brown in color and somewhat gritty in texture. It is moderately hard and compact and contains many small, irregularly rounded concretions of iron. The significance of these soil formations in archaeological terms seems to rest of the following points: a. The radically different character of Soil III, which suggests (i) That it may have been formed under fundamentally different conditions of plant cover and perhaps climate from those prevailing when Soil II and I were formed; (ii) That there may be a considerable time gap between the formation of Soil III and the formation of Soil II. b. The similarity between Soils I and II, which suggests that they were formed under similar conditions of climate and plant cover. The presence and character of layer 3 is consistent with at least some degree of podzolization. c. The probability that Soil II is truncated on the east of our site but not on the west and the existence of layers 7 and 8, which-–both in their character and in the nature of their cultural inclusions (or lack of them in the case of layer 7)—suggest that they were laid down by the fairly rapid water erosion of the higher parts of the site at the end of the period during which Soil II was being laid down. This is likely to have occurred under the conditions set up by the abrupt and complete abandonment of the site shortly before the onset of the rainy season. The archaeological evidence for this period in the site’s history seems to be highly consistent with this supposition. d. The uniformity in the character of the build up of Soil I (broken only by the laying down of layer 2 quite recently in the site’s history) suggests continuity and uniformity in the conditions prevailing on the site throughout the period of its formation. This supposition again seems to be supported by the available archaeological evidence. Soils of the Western Part of the Excavation Area Over the whole of the central and western half of the area, Soil I and II occur in forms that are to a marked degree more sandy, more compact, and much harder than the “normal” form of these soils to the east and the question arises as to whether these materials should be regarded as distinct formations. However, a careful scrutiny of the sections failed to find any definite break; the layers seem to be exactly stratigraphically equivalent to and continuous with the layers to the east and the whole appearance is one of a gradual transition from one form to the other, becoming progressively more marked as the layers dip into the hollow of the central part of the site. The increased sandiness is marked in Soil I, much less so in Soil II, and in Soil II becomes progressively less marked with depth, so that it is barely noticeable in layer 9 and absent below layer 9. The most probable explanation on the evidence available at the moment lies in the manner of deposition of these materials. If it is true, as has been suggested above, that the central


hollow was largely filled by the erosion of layer 9 and 10 materials from the eastern parts of the site, it is likely that the lighter and more friable elements in the soil would become the main components of the materials being shifted, while the heavier and stickier elements would tend to be shifted to a much smaller extent. This explanation accords well with the character and supposed manner of deposition of layers 9, 9a, and 7. The strongly sandy character of layer 8 and of Soil I in this region is consistent with their having been derived at least in part from somewhat sandier materials that seem to make up the upper levels of the area to the west of our excavation. It is again consistent with their being thus derived that layers 2 and 8 (which are restricted to the western parts of the site) both lens out to the east, while most other layers lens out, or tend to lens out, to the west. Description of the Excavation Area by Archaeological Levels: The Early and Middle Periods Within the stratigraphical framework described above it is possible to distinguish thirteen archaeological levels, most of which provide clear and well-defined stratigraphical horizons over the whole area exposed in excavation. These are, beginning with the earliest: Level I to V: (Layers 21–17. Early Period, Middle Periods 1–5 in overall sequence; see Fig. 4-2) Throughout the period Levels I to V, the area excavated by us was used almost exclusively as a burial ground for inhumation burials. Except for a few restricted patches, mainly along the northern margin of the excavation, there was no evidence for any other kind of activity. Our view of the culture of this protracted period is therefore a severely limited one. On the other hand, the large number of burials (Fig 3-6) and the ordering of these into a reasonably reliable sequence of types provide a basis for interpreting the development of burial custom. The graves themselves provided a fairly wide selection of complete pots, a considerable quantity of animal bone, a rather more limited but not uninteresting selection of other objects, and more than sixty complete or substantially complete human skeletons. Sufficient charcoal was obtained from the three earliest levels to run carbon dates, although unfortunately the samples are all small in quantity and-–on this heavily disturbed and badly root penetrated site-–the risk of contamination or stratigraphical error is high (see Chapter 5 for a full discussion of provenance and interpretation – DTB).


Fig. 3-6

Level I: (EP 1 – 3 and MP 1 in tentative overall sequence; Fig. 3-7). In all, fourteen burials of this period were located. Three of these were situated marginally to our area and could not be completely excavated without cutting extensions of considerable size for which we had neither the time nor the labor available. The clearing of the graves of five others was still incomplete on April 28, when time and deteriorating weather conditions forced the closure of the excavations. Of the remaining six, three were infant burials included within Level I burial mounds and seemingly contemporary with the construction of the mound. The following description of the burial custom at this level therefore depends of the three fully excavated burials, on the excavation of the mounds and upper parts of the grave fills of five other burials, and on the partial clearing of the mounds of three marginally situated burials (see Chapter 8 for description and analyses of burials from both seasons’ excavation–DTB).


Fig. 3-7

Of these eleven burials, no two were precisely alike. Indeed, the range of typological variation in the burials assigned to this period is far greater than for the burials of any other period in the site’s history. The following characteristics seem common to all: a. The graves (where they were defined) were from 70 cm to 1 m deep. b. The body lay on its back, both hands by the sides and the feet usually together. There was no evidence to suggest the orientation was of any significance. c. Pots (sometimes of quite large size) were placed in the bottom of the grave, especially close to the head and feet. d. The grave fill frequently but not invariably contained further offerings of animal skulls, animal long bones, and whole or smashed pots. The character and arrangement of these “fill offerings� differed in each burial investigated and at least one they were entirely lacking. In the three burials completely cleared, the pots of the fill offerings in one were placed about half way up the fill above the skull and extended onto a ledge beyond the skull; in a second there were no fill offerings, and in the third the fill contained a buffalo skull and some blocks of stone, but no pots.


e. f.




In one of the graves, incompletely cleared, the fill seems to have contained only smashed pots. The infilling was halted at ground level and its upper surface covered with the sherds of two or three large smashed pots. Above this was mounded up an often-considerable furniture of animal and human bones, whole pots, and a variety of other objects, and the whole covered with a substantial mound of earth. As found, the greater number had eroded away almost completely, exposing part of the “mound offerings” to weathering and displacement. Two mounds, however, survived to a height of nearly half a meter and there were indications that they had formerly been at least half as high again. In their original form the majority of the grave mounds seem to have been about 3 m x 4 m in area or a little smaller. The shape and size of the graves seem to have been variable and adjusted primarily to the amount and character of the grave furniture to be placed with the body. In general the graves seem to have been from about 2 to 3 m in length and from 50 cm to 1 m in width at the bottom. The decorated pottery found with Level I burials include vessels cord-marked on the lower part with incised designs forming a band around the shoulder and buff-colored vessels without cord-marking decorated with painted designs in red around the upper part. Decorated vessels form a very small minority of the total vessels, but these two types are characteristic of Level I and limited to it. The human bones, found included in most of the mounds of this level and in some of the mounds of Level II, present a problem. The following points are clear: (i) They are not bones accidentally disturbed from earlier burials, as in many cases they were found still in articulation and often laid in some kind of deliberate arrangement. (ii) They are not partial secondary burials made at a later date, as in some cases they are found mingled in the closest association with animal bones that clearly were funerary offerings. Further, these assemblages of assorted human bones were never found outside a burial mound and, in spite of careful search, in no case did we encounter any disturbance capable of accounting for their later deposition. (iii) Typical groups included: skull alone, usually without lower jaw (six examples); skull and long bones arranged as described by Green, the skull in both our instances being crushed and in one the long bones being split longitudinally (two examples); and upper arm, shoulder, and some ribs more or less in articulation (two examples in a single mound). In addition, scattered ribs and whole or broken long bones were frequently found. In one mound a fragment of pelvis was found with the upper three inches of the femur still in articulation. (iv) In addition to these groups, however, other groups (though not from this level) were found that could not be assigned to the burial mound. In at least one grave, the bones–-though arranged in roughly correct relative position–had clearly not been in articulation when buried. It seems to me that the most probable explanation of these groups is the reburial of bones disturbed by the digging of later graves. (See Chapter 8 for a discussion of EP 3 and MP 1 secondary burials–DTB.)


For the above reasons, it is concluded tentatively that the inclusion of hacked-up portions of human bodies in the burial mounds formed a frequent though not invariable part of the funerary ritual at this period. j. In three mounds of this period and in one of the succeeding period, a complete infant burial was included in the burial mound. The bodies lay in each case at the base of the mound, and in one case lay directly on top of undisturbed sherds of the sherd sheet covering the top of the grave fill. In one case the head of the infant lay on a large shell. In none of the four cases was a separate grave (cut into the mound to receive the infant) detected, but on the other hand these were clearly normal burials of a complete individual and in no way resembled the interment of dismembered human remains discussed above. It is quite uncertain whether these infant burials were contemporary with the construction of the mound or made at a later date, though the former possibility seems slightly more likely. No bronze was recovered from any Level I context and only one small stone adze (of the type with flat cross-section) came from a firm Level I situation. The body in one of the graves (Burial 6) was found elaborately adorned with strings of small, flat, disk-shaped shell beads around the waist. Three large and finely polished stone beads (tubular in form, two of them with the ends cut at an angle) were found. Two of these came from secure Level I contexts. Level II: (MP 1-2; Fig. 3-8) In general, this level continues closely the tradition of Level I and is probably historically continuous with it. The incised and painted decorated wares disappear and indeed may have already done so before the end of Level I. They are in some sense replaced by vessels cord-marked over the whole surface with very simple curvilinear incised designs superimposed on the cord-marking (type 2C; see Chapter 6). Vessels not cord-marked but with red slip are fairly common, and small, footed, cord-marked jars (types 2C, 2E, 4D) are markedly common. Two or three jars decorated with rounded knobs on the shoulder of the vessel were found (type 1E) and these (though somewhat different in ware or firing) may be related to the “horned� pots (types 1R, 6R) of Level III. Burial practice was markedly similar to Level I, except that: a. The mounds were usually (invariably in late Level II) smaller and flatter. b. The graves were narrower and very shallow, the body lying immediately under the sherd sheet. c. The pots of both grave furniture and mound offerings were usually laid on a single level at the bottom of the mound. In one instance--a Level II or very late Level I burial-–the mound offering was in the upper part of the mound. d. The content of both human and animal bone was greatly reduced. Complete ruminant legs, fairly common in Level I burials, were absent from Level II burials and neither human nor animal bones were found in the grave fills.


Fig. 3-8

In all, eleven burials of this period were recognized. Eight of these were completely cleared, one was marginal to our area, and two others were still incompletely excavated when the season’s work ceased. In addition, four other burials, stratigraphically insecure but belonging to either Level I or Level II, were found. None of these were completely cleared and in none of them was the actual interment found. A number of small fragments of bronze were recovered from late Level II context. From the outskirts of two late Level II burial mounds, three pairs of intact bivalve molds for casting bronze axes were lifted [note: one pair was recovered from the mound of B. 32, the other two from a disturbed area where MP5? B. 10 cut into the edge of the mound of MP 1? B. 8–DTB]. No intact or recognizable bronze objects were found in this level. Apart from the burials, the only other large concentration of finds from Level II came from a moderate-sized patch of pottery midden in square C7. Our excavations caught only the southern edge of this midden, which extended over the northeastern part of C7 and covered about a quarter of the area of the square. It yielded a number of nearly complete


and restorable vessels and a considerable quantity of potsherds. (I would consider this concentration of pottery a feature, as there were two unique vessels included; see Chapter 5–WS.) Other finds include numbers of spindle whorls in baked clay1 and-–from two different contexts–-a number of small bone rods pointed at both ends and with circular crosssection. These last seem too big to be fish gorges and are possible spear or arrow points. One body was accompanied by half a dozen shell bracelets, all broken when found, and one had half of a broken shell anklet lying immediately under the left ankle. Level III: (MP3; Fig. 3-9) The material culture continues well within the broad tradition defined by the previous two levels. Indeed, this broad tradition seems to persist virtually throughout the site’s history so that the overall picture is rather one of development within a single framework than of revolutionary and disruptive changes of direction. In spite of this, however, there seem to be, with Level III, minor but significant changes in such things as burial practice, in the character of the pottery, and in the form of the socketed bronze axe that may well mark a change in the ownership of the site. Fig. 3-9


The graves of Level III were narrow trenches (about 40 cm in average width and about 3 m in length). The body lay with the head at one end of the trench; in the other end of the trench one or more complete cow legs were laid, sometimes overlapping the lower part of the body as far as the pelvis. One to three pots were normally placed in the bottom of the grave beyond either the head or the feet, and in three instances two fair-sized bowls were placed on top of the fill at one end of the grave-–the head end in one case and the foot end in the other two. The mound seems to have been much reduced and to have contained no funerary offerings. The body itself lay on its back and in a majority of instances the lower jaw lay on the chest with the skull fallen back from it. This seems to indicate that in burial the head was propped up on some material that subsequently decayed. No indication of the nature of this material was found. In two instances the skull was propped up against a pot and had retained the raised position. (See also note below on the disturbed articulation of bones in burials of Levels I to V.) One body of this period wore two bronze bracelets on each arm. A second body was accompanied by a socketed bronze axe (with the short blade flared out into horns each side in marked contrast to the long-bladed axes with simple curved or angled cutting edge represented by the Level II molds). This axe lay on the left thigh and had perhaps been hafted when deposited. However, no trace of the wood of the suspected haft was found, the hafting being suggested merely by the position and angle of the axe, which could be explained by supposing that the haft was held in the right hand at the time of burial. In sum, thirteen burials were found belonging to the Level III period. None of these were completely cleared, and there were situated marginally to our excavation and incompletely cleared. Levels IV and V: (MP 4 - 5; Fig. 3-10) Level IV and V burials form a continuum, as indeed the layers (18 and 17) seem to. In general, however, it was possible to distinguish upper and lower levels within this and burials of Level V were cut with graves markedly deeper than those of the earlier period. As a result of this, the later graves in several instances cut through earlier. Wherever this occurred, it provided a measure of confirmation for the hypothesis on the significance of orientation in the Level IV and V burials advanced here. In graves themselves, the trench form of Level III seems to have been retained but the Level IV graves were dug somewhat shallower than in the previous period. Grave furniture was now in general limited to from one to three pots placed immediately beyond the head or feet, although in one grave there was also a smashed pot under the pelvis and two others had a pot between the knees-–a feature found also in one or two Level II and III burials. There is some evidence to show that the small mounds raised above the graves sometimes contained nests of up to five small pots. The orientation of burials of Level IV and V was quite variable–-in almost every possible direction from northeast through west to south. The earliest burials of Level IV, however,


Fig. 3-10

were uniformly oriented with heads to the south, and wherever the relative sequence of two burials could be clearly established (as in the case of intersections), the later burial was oriented farther toward the north. This suggests that as the Level IV and V period went on, the direction of orientation of the burials changed continuously from south through west to north. This, if confirmed, suggests that the period was of relatively long duration (see Chapter 8 for discussion of MP 4-–6 burial orientation–DTB). The two level together yielded twenty nine burials, twelve of which were marginal to our area and incompletely cleared, the remainder being completely cleared. In both levels the body was placed on its back; in earlier Level IV burials, the right hand is uniformly above the pelvis while the left hand is below. In later Level IV and Level V burials, both hands are under the pelvis. Bronze was in use throughout the period, many small fragments being recovered from both levels. One of the skeletons (from about the middle of the period) was adorned with sixteen bronze bracelets (superficially, at least, similar to those from the Level III burial) on the left arm and one on the right wrist. A small, heavy vessel, presumed to be a crucible, was recovered from Level V and found to have small nodules of bronze adhering to the


interior and traces of green staining around the lip; a sherd similarly stained came from the same level. A fair number of small stone adzes with flattened rectangular cross-section (rather similar to the adze from Level I) were recovered from these levels. In one burial, a young child from Level V wore a single shell bracelet on the right arm and in two others there was a string of shell disk beads around the neck. Patches of occupation-blackened soil occurred along the east and northeastern margins of the site at these levels, extending in some instances up to 1.5 m into the area excavated, and a fair quantity of general occupational debris was recovered from the surface of layer 17 (Level V), making it likely that at these points our excavation was close to the western border of the main occupation area. Note on the Disarticulation and Disarrangement of the Bones of Burials in Levels I to V: To a marked extent in burials of Levels I and II, most conspicuously in burials of Level III, and to a lesser degree in burials of Level IV and V, the bones of the skeleton were found disarticulated and seriously displaced from their normal alignment. This appears to have been due to settling of the body after the decay of the flesh and is assumed to have been caused by the decay of some thick layer of organic material on which the body had been placed. In spite of very careful searches under several bodies, no hint could be found of the nature of this material – whether matting, leaves, straw, or something else. Most probably, matting would have left definite and recognizable traces, so that this is the least likely of the possibilities. In the Level III burials where the settling was particularly marked, it was not unusual for the body to have twisted somewhat to one side, and in one instance the pelvis was tilted at an angle of almost 450. In one burial (probably late Level II), the feet and lower legs had remained at the original level at which the body had been placed while the upper part, from the pelvis upward, had sunk by about 20 cm, leaving the femora sloping down sharply from the knees to the pelvis. Summary of Levels I to V (Overall EPV (Overall EP, MP 1–5)) The whole period represented by our Levels I to V (from the available evidence) exhibits a high level of continuity in the context of a larger degree of cultural change than is found at any other period in the site’s history. New elements, however, like bronze or certain pot types, are introduced, and other elements, like the decorated pots of Level I or certain burial practices, disappear without any indication of major or sudden general cultural revisions. The only exception to this occurs at the beginning of Level III where there is a reversion to the practice of placing complete cow legs in burials, some small but significant changes in pottery, and perhaps some change in the external relations of the society reflected in the introduction of a new type of bronze axe. The period is certainly prolonged, and the overall picture suggested by our evidence (as far as it goes) is one in which both continuity and change are expressed in terms of the


continuous development of a single tradition, the direction of this development being only occasionally modified (and then only in a minor degree) by the introduction of new cultural elements. The evidence is still too scanty for confidence, but the above suggestions seem worth advancing with the utmost caution. As will be seen, they seem to remain true of this site almost throughout its history. The somewhat complicated problems arising from the thirty-six carbon and thermoluminescence dates that have been run on samples from both excavations at this site have been fully discussed elsewhere (Bayard 1971b:26ff; Bayard 1970; Parker 1980). Here it is enough to suggest that the evidence as it stands quite strongly favors the “long dates”, although there are enough anomalies to make this somewhat less than completely certain [we now have fifty dates available; see Chapter 5 for discussion of the current situation– DTB]. There is little evidence bearing on the economics of the society. In Level I and throughout the greater part of Level II, bronze seemed to us in 1966 to have been absent from the cultural equipment. In 1968, however, Bayard found metal down to his Level III, equivalent to the 1966 upper Level I. By Level III, bronze appears to have become relatively common. Small stone adzes probably continued to be used throughout the period, the type showing little or no modification from Level I to Level V. There is at present no evidence bearing on the presence or absence of agriculture, but from the selection of a lowland site and the long continuity of occupation we may perhaps presume it without being able to define the crops grown. If the presence of spindle whorls is confirmed among our finds, this would greatly increase the probability that cotton was being produced. On the other hand, it is clear that hunting and probably animal husbandry were important elements in the economy of food production. Animal bones recovered include those of pig (possibly domesticated; see Bayard 1971b:32), cattle (probably domesticated; see Higham and Leach 1971), and several almost certainly wild species. Small game was hunted with a “pellet bow”2. The period seems to have closed at the end of Level V with a fairly prolonged abandonment of the site indicated by our largely sterile layer 16. Level VI and VII (Overall MP 7 and 8) Level VI: (MP 7; Fig. 3-11) The occupation of this period was one of the most massive located within the area of our excavation. It is characterized by a very dense scatter of cultural debris over the whole area excavated. Within this scatter, three major zones are defined by the character of the material found in them. 1. Spread over an area covering the eastern three-quarters of square C4, the whole of D4, the eastern half of E4, and the northeastern one-third of F4 was a dense scatter of large animal bones, irregular lumps of sandstone, and other debris thought to be characteristic of refuse from a cooking area. No actual fireplaces were located and it seems probable that these lay outside our area of excavation to the east.


Fig. 3-11

2. To the west of this area the level was characterized by a scatter with much diminished bone content, rather small sherds, some sandstone lumps, three or four small and roughly made heavy bowls with flat bottoms, and a scattering of small irregularly shaped lumps of bronze generally found within a radius of 2 or 3 m of the bowls. The bronze lumps look like waste from bronze smelting or casting, and the bowls have been definitely identified as crucibles (Smith 1973). As partial confirmation of this hypothesis, a similar bowl and a small rim fragment of another similar bowl from Level II showed traces of green staining on the inner surface near the lip owing to the presence of bronze, and small lumps of bronze were found adhering to the interior of the bowl. 3. Still further to the west was an area containing inhumation burials. A few postholes from this level were located, but these form no clearly defined pattern and most are too small to have been house posts. It is possible that our excavations were too far to the west to touch the dwelling area of this period.


The burial area of Level VI encroached only on to the western half of our site. In all, six rather scattered burials of this period were found. Two of these were marginal to our area and incompletely lifted, the remainder being completely cleared. The typical grave seems to have been a trench 30 to 40 cm wide at the bottom and 30 to 40 cm deep. The body invariably lay with the head to the west. The orientation in our six examples varied by no more than 10°. The body in each case lay on its back with hands by the sides. The bones were in articulation and without the evidence for post burial settling that characterized burials of earlier levels. A single fair-sized pot was usually placed by the right shoulder and one body had the sherds of a scattered pot under the pelvis. Two graves contained lumps of red pigment. There was no evidence to suggest the presence of a burial mound. There was some evidence for stone working in the form of fragments of polished stone, and a few small, flat sandstone slabs of a type thought to have been used in polishing stone. No adze fragments of definable types were found, although some of the chips polished on one face may have come from stone adzes. Level VII: (MP 8) This occupation is similar in general character to Level VI, differing in the following respects: a. The “frontier” between the “kitchen waste” area and the “bronze working” is farther to the west by about 4 m and somewhat differently aligned. The kitchen waste covers the whole squares C4, D4, E4, three-quarters of F4, and two-thirds of C5. This strongly suggests a front curving in to the east, with its most westerly extension in the northern part of the excavation, as against a front curved in the opposite direction in Level VI. b. Several substantial fragments of stone adzes were found. These are of two fundamentally different types, one with a flattish cross-section and somewhat skewed cutting edge (very similar to the stone adzes of Levels I to IV and quite probably intrusive here) and the other with a nearly square cross-section. Several substantial lumps of stone suitable for adze making were found, some of it the same material used for one of the square cross-sectioned adzes. In addition there were a number of flat sandstone slabs, at least one of which was worn smooth over almost the whole of one surface. c. Two of the presumed crucibles were found, but the total quantity of bronze seems to have been markedly less than in Level VI (this requires checking in precise quantitative terms). d. Three large postholes in alignment across the southern edges of squares E3 and F4 were found. It is assumed that a fourth lies under the baulk and that the whole represents one side of a house structure. Some smaller postholes to the north and northwest of this alignment are thought to be associated with this structure. e. Large numbers of clay disks, interpreted as either net weights or spindle whorls, were found in Levels VI and VII. Also found were a number of somewhat larger


disks with unequal curvature of opposing faces that are thought to be probable weights for spindle whorls. If this attribution of function can be confirmed, it would almost certainly indicate the cultivation of cotton in the area. The burial area of Level VII penetrated into our site only along the eastern squares, with two isolated graves in the next most eastern line of squares. Seven burials of this level were found and all were completely cleared. The grave was similar in size and shape to the graves of Level VI and the body lay in a similar position, except that the right hand generally lay on the right thigh. There was again no evidence for post burial settling of the body. The skull invariably lay on the sherds of a shattered pot, the feet frequently on the sherds of another. In one grave three pots were found beyond the skull. One skeleton had an irregular lump of bronze lying among the bones of the right hand –-possibly a heavily corroded bronze ornament. As with Level VI burials, there was no evidence for the existence of a mound over the grave. All Level VII burials were oriented within narrow limits with the head to the south, but the range of variability was a little greater than in Level VI. Summary of Characteristics of Levels VI and VII: Each of these levels has provided evidence for bronze working, substantial in Level VI and diminished in VII. Level VII, on the other hand, has given more substantial evidence for the making of stone tools than Level VI. It seems that there may have been some reversion to the use of stone tools at this time. Other finds belonging to the Level VI and VII period but not yet clearly located within it include fragments of stone bracelets. These were found in Levels VI and VII (as well as in earlier levels) but in stratigraphically insecure situations and may belong to almost any of the earlier periods in the site’s history and be present as high as Levels VI and VII merely as the result of disturbance. The Late Period Level VIII: (Overall LP 1; Fig. 3-12) After the end of Level VII occupation the site seems to have been abandoned for a prolonged period, sufficient to allow the accumulation of up to 10 cm of culturally sterile soil under conditions of a reversion to natural plant cover (the limited amount of cultural material found in layer 12 is thought to be intrusive). When occupation is resumed at Level VIII, the evidence suggests that a profound and pervasive cultural change has taken place. The Level VIII occupation is known almost entirely from a scatter of sherds and from a number of unfortunately fragmentary cremation burials. The most complete burials were two found in square C3 and one found in E4. In addition, two further but more fragmentary burials of this type were found in C3 and two in E3. In no case was a complete burial pit recovered and-–except for the two relatively complete burials in C3-–in no case could a complete wall of the burial pit be traced downward. All seven burials, however, are


Fig. 3-12

typologically similar, yet they differ from the burials of any higher occupational level, and their absolute depth excludes the possibility that they could belong to any level earlier than VIII. The two relatively complete pits in C3 were clearly and indisputably cut from Level VIII. The characteristics of the groups are as follows: a. A square pit (generally about 40 cm square) was dug to a depth of about 30 cm. b. Lumps of baked clay, sometimes with one (the outer) side flattened, were placed at the four corners (sometimes with additional lumps along the sides). c. Sometimes, but not invariably, the bottom of the pit was covered with small lumps of unbaked clay. d. The pit was partially filled, with the upper surface of the fill somewhat sloping (there does not seem to be any regular direction for this slope, but it was present in at least four instances-–one down to the north, one to the southeast, one to the southwest, and one to the west). e. Four more lumps of baked clay were set on top of the fill at this level. f. A pottery urn containing charred bones and presumably ashes was set on top of the fill and deliberately smashed so that the sherds and contents spilled down the slope.


g. The pit fill was completed. The urns used for this type of burial seem, from a superficial examination, to be somewhat variable in type, but none of the types used for cremation burials of later periods were found in any of the seven burials described here. (In terms of the vessel typology presented in Chapter 6, these vessels were types 5A (3), 5B (2), 2L (1) and 2A (1)–DTB) Level IX (Overall LP 2; Fig. 3-13) From this period came the first clear (though incomplete) evidence for a house type. Running along the northern side of square C4 and C5 was a firm alignment of four square postholes spaced about 2.5 m apart. The easternmost posthole was larger and deeper than the other three. (We have been told that in modern practice in the area, one of the eastern corner posts of a house is regarded as the main post and is often larger that the other house posts. The spacing of posts in modern houses in the area is usually from a little over 2 m to 2.7 m.) to the south of this the ground sloped to the south and provided a number of interesting features: Fig. 3-13


a. Near the southwest corner of square C5, a shallow depression held a small quantity of charcoal and ash around a substantial scatter of fairly large potsherds. b. In the southeast corner of the square, the ground surface was slightly sunken, darker in color, compacted, and troweled off to give a smooth, “clayey� surface. Pote Gueagoon, one of the students, suggested that this patch represented the spot above which the water jars had stood, and the charcoal and potsherds in the southwest corner represented the ground under the cooking area. c. One or two smaller postholes suggested that the greater part of both squares had been covered by a platform extending the floor level of the house to the south that would have been of much the same type as the platforms in front of modern houses that carry the cooking stone and water jars. The front posts of this platform lay along the line of the baulks C4/D4 and C5/D5. A similar house was located in squares D7 and C7 and another (for which unfortunately we have only fragmentary evidence) in squares G4 and F4. The sum of the above evidence suggests very strongly that the houses of this period were very similar to the houses in use today and of much the same size. To the south of the house in square D4, the surface of the level falls by about 10 to 15 cm and offers a highly compacted surface with many small sherds imbedded in it. This surface resembles closely the hard-trodden area found in front of modern houses in its general character, in the degree of depression, and in the presence of numerous small sherds trodden into its surface. In the modern villages, such surfaces are masked by a thin layer of dust, but this can be readily brushed or blown off to expose the compacted surface. Lying on this compacted surface in squares C4, D4 and E4 was a scatter of large sherds, and a number of large bones, the sherds being from a pot with a large, sharply everted rim. Those lay on the surface with the rim protruding upward. One of the long bones lay on a slight slope, its lower end embedded in the surface of layer 9 but its upper end protruding for several centimeters into layer 7. If the bones come from a single skeleton, whether human or animal, they can only belong to one that died on the site and remained unburied. To the north and west of the group of large shattered pots in square D6 and extending back across D7 was a large sheet of potsherds mingled with blocks of burnt clay, some of them showing flattened outer surfaces. In the small spaces between and under these were many small lumps of charcoal. This was interpreted as the remains of a clay-lined hearth together with the domestic pottery standing on and around the hearth, and again is thought to have been deposited in the condition in which it was found by the collapse of the Level IX house in this area. Many of the postholes of the three houses found at this level contained some charcoal in the upper few centimeters and several had charcoal on the Level IX ground surface immediately around them. Against and extending under the baulk D6/D5 about 4 m to the north of the house in D7 and C7 was the skeleton of what appeared to be a small woman or adolescent [bones were too fragmentary for examination by Pietrusewsky-DTB] in a shallow hollow in the Level IX surface in an oddly contorted position, twisted half over onto the right shoulder with the head bent sharply back. It was clearly not a burial for it lay absolutely and stratigraphically above any surface belonging to a period in which inhumation burial was practiced, was not contained within any detectable grave, and was unaccompanied by grave furnishings of any kind.


The total situation suggests strongly that at the end of the Level IX period the settlement was sacked and burned and the inhabitants massacred. The destruction must have been virtually total; otherwise one would expect the survivors to have returned to cremate the dead and rebuild the houses. In fact, everything in the lower-lying parts of the site seems to have been buried quickly and effectively under a thick sheet of rapidly silted material (layer 9a), and this in turn to have been sealed by the perhaps more slowly deposited layers 7 and 8. This in turn suggests that the destruction took place at the beginning of the rainy season. Toward the western side of our site, layer 7 was replaced by layers 8 and 9a, thought to represent an earlier and more violent stage of the same erosive process that produced layer 7. Layer 8 contained many small, heavily weathered (and probably rolled) sherds. In squares D5 and D6, however, layer 9a had submerged the shattered fragments of two or three large pots. These were lying piled on top of one another in confined areas. This situation is interpreted provisionally as having been produced by the collapse of house platforms spilling the pots that stood on them to the ground surface, and again this is thought likely to have occurred at the end of the Level IX period. One of these pots in square D5 contained a quantity of bones from a small animal or bird [apparently unanalyzed by Higham-DTB]. Two presumed cremation burials from this period were found in square E4. The later of these had cut more than halfway into the earlier and resulted in the urn of the earlier burial being displaced, broken, and partially removed. The urn of the later burial was intact. The presumed urn was standing upright, slightly inclined to the northeast, at the bottom of a square pit a little over 40 cm square at the top and 40 cm deep. The base of the urn was placed in the center of the pit bottom. There was no other grave furniture and no other special features were detected. The urn was a vessel distinct in type from any of those used in Level VIII burials and equally distinct from those used in Level XI窶的I burials. In square D6, E6, and C7 a number of small bronze fragments were found, including fragments of several bronze rings. A number of pieces of heavily corroded iron were recovered from this level, including a square iron nail from C3 and a small iron fishhook from the same square, together with a somewhat damaged socketed iron axe found in this level in the C5/D5 baulk. Levels X, XI, and XII: (Overall LP 4; Fig. 3-14) These levels may be conveniently discussed together, as the evidence suggests there is little difference between them. In most squares two or three major levels of pottery concentration were found, but there is a likelihood that these levels cannot be matched into three distinct stratigraphical horizons throughout the site. If this proves to be the case, the picture will rather be one of a large number of interleaving local layers produced by a slow, progressive buildup extending unbroken throughout the whole period. Four houses belonging to the two levels XI and XII penetrated in whole or part into our area. One lay in the southern part of squares D6 and D7; one side of a second lay in the eastern part of squares G4 and F4; the third angled across squares C4 and the northeast


Fig. 3-14

corner of D4; and the fourth (probably a later structure than the third) covered almost the whole of square D4 and extended for an undetermined (but probably not large) distance to the east. The postholes of this period were nearly all circular, with a few doubtfully square. They are smaller and set closer together (at less than 2 m apart) than those of Level IX. Although we have no complete plan, it seems likely that (as in the Level IX period and in many modern houses) these houses were based on a plan requiring sixteen posts arranged in evenly spaced lines of four. The cooking areas of two houses were located and seem, as in Level IX and today, to have been situated at one corner of the exterior platform. The general impression is therefore of houses essentially similar to the modern houses and to those of Level IX but smaller and flimsier that either. As in Level IX, the ground surface between the houses was depressed by 10 to 20 cm and heavily compacted, with many very small sherds packed into its surface. Four burial areas belonging to this period were located. In the first, in C3, at least six and perhaps as many as ten pits had been dug, the later in each case cutting away the greater part of the earlier so that only the three last pits survived complete enough to contain more than fragments of pots. The second, in E3, again provided two intact cremation burials.


Both pits were half-sectioned in the baulks of this square. The third, in D4, provided one intact burial. The fourth, in C5, again was a tightly interlocking series of pits, three of which (all badly damaged) lay within the square. From these various burials it was possible to reconstruct the typical burial practice of the period: a. The pot is a highly standardized vessel thought not to have been used for any other purpose at this period. Only minor variations in shape, ware, or size can be detected from among the five intact and probably twelve damaged burials located. [The five reconstructable vessels were type 2C, following the typology used in Chapter 6–DTB.] b. The pot is set at the center of the bottom of an almost cubical pit about 45 x 45 cm. It is invariably placed upright, although occasionally tilted slightly in filling the pit. The grave pit displays no other special features. c. The pot, like those used at Level IX, is uncovered. d. The pot is never deliberately broken. e. The last two pots buried in square C3 had their pits surrounded by a light fence supported on four stakes held in small square postholes. One other similar postholes from the same square indicates it is probable that the earlier pits had been isolated in the same way. f. It is tempting to relate each of the major burial areas to one of the available houses and suggest a household burial plot standing close to the northeast or southeast corner of the house. It is a curious and at the moment inexplicable fact that these pots are wholly indistinguishable from the similar footed, cord-marked jars found in burials of the Level II to Level V period [i.e., overall Middle Period 2–5; the vessels recovered were type 2C–DTB]. Yet of the pots of the Level X–XII period, all three in C3 and at least two in C5 were stratigraphically secure and contained within pits cut from Levels XI and XII; the two in E3 were almost as secure, although the pits were less clearly defined, and only the single pot from D4 was doubtfully within the limits of the pit to which it was assigned. All lay at absolute depths, which left them separated by at least 20 to 30 cm from even the highest points of Level V occupation and by somewhat more in their immediate vicinity. They do not appear to occur at all in Levels VIII and IX. We can perhaps conclude that the manufacture of these jars continued in the area over a very long period of time with minimal change of either form or fabric and that their use was reintroduced to the site by the people of Level X when they occupied the site after the period of abandonment, which followed the destruction of the site at the end of the Level IX occupation. Alternatively and more probably, these were ancient jars dug up by the later people and reused as cremation pots, perhaps precisely because of their antiquity. It seems not unlikely that many of the pits dug at this period, to which no other purpose can be assigned, were in fact robber pits dug to recover these small, globular, footed jars. Some iron but little bronze was found in all of these levels.


Fig. 3-15

Level XIII: (Overall LP 5; Fig. 3-15) Two houses of this period penetrated into our area – one in C3 and one in F4 and G4. In neither case was a complete posthole pattern recovered even for the area exposed. The available evidence, however, suggests that the postholes were for the most part square-– with a few being circular–-deep and large enough to support fairly solid posts, and spaced about 2 m or a shade more apart. The cooking areas were again, in each case, close to one corner, and the general impression is of houses much like those in use today and perhaps only slightly smaller. Again, in the area between the houses, the ground surface was depressed, compacted, and characterized by the presence of numerous small sherds. Three burial pits of this period were found. Each was a square pit about 45 to 50 cm square at the top and about the same depth. There was no burial urn, the charred bone and ash being placed instead in a bowl-shaped depression more or less in the center of the bottom of the pit and occupying about half the total area of the pit bottom. In one case the sides of this depression and in another the whole bottom of the pit had been covered


with small lumps of unbaked clay. One of the pits appeared to be enclosed by a light fence supported by four posts, exactly as with the Level XII burial areas. The total quantity of potsherds from this level was markedly less than for Level X and XII, either because we were close to the outskirts of the settlement or because the occupation as a whole was thinner and possibly more short lived than the occupations of the two preceding periods. There is no clear evidence to suggest any sharp break in continuity of occupation between them except in the preference for square postholes in Level XIII, the slightly larger size of the houses located, and the change in burial practice. The presence of a few lumps of iron slag near the western extremity of the area excavated may suggest that iron was being smelted close to the edge of the mound a little to the west or southwest of the area excavated. Summary of Levels VIII to XIII Inclusive: In spite of considerable cultural differences, there is a fairly high degree of continuity between the character of all these occupations. There are probably clear temporal breaks of fairly long duration between Level VIII (LP 1) and Level IX (LP 2) and between Level IX and Level X (LP 3), but from there on to the top of XIII (LP 5) the occupation may well have been virtually continuous with only slow, gradual, and progressive cultural change to distinguish one period from the next. As a whole, the period from the beginning of Level VIII till the end of XIII (the Late Period) has certain marked characteristics that distinguish it from the earlier period of Levels I to VII (the Early and Middle Periods): a. The kitchen contains comparatively little Bone and it is nearly all in small fragments. There has clearly been a sharp decline in the importance of meat in the diet and probably a decline in the importance of hunting as a means of providing it. It is tempting to correlate this change with the introduction into the area of wet-rice cultivation. b. The house type throughout the period, in spite of minor changes, is essentially similar and essentially the same as the type in use today. The cooking seems invariably to have been done on the corner of the house platform, whereas it is likely that in the Level VI – VII period it was done on the ground surface with the cooking pots supported on lumps of sandstone. c. Burial practice throughout the Level VIII to XIII period involves one form or another of cremation with subsequent burial of a few small chips of bone and some ash. The sequence of the various forms used for the interment of bone and ash indicates development and change within the framework of a basically unitary concept. This tradition stands in sharp contrast to the burial practices of Levels I to VII. There is little evidence for the dating of this period, but celadon has been found in Levels IX and X and doubtfully in VIII, so that at least Levels IX and X must lie within the period of its manufacture. If this leaves us with an unduly wide range, it nevertheless provides a definite limit within which to look for dates. It is hoped that a fuller examination of the imported pottery will establish key dates with reasonable precision. [This proved not to be the case-DTB]


Level XIV: (Overall LP 6) Level XIV is sharply separated from Level XIII by the black layer 2 thought to have been laid down by erosive forces during a fairly long period of abandonment of the site. Level XIV contains little but modern material (and not much of that) and is clearly to be correlated with the reoccupation of the site and its use for garden cultivation, which began about eighty or ninety years ago. In squares C4, C5, and the northern part of D4, rows of the small holes dug to plant mulberry shoots were found. Two patches of clay were found, one in C5 on the surface of the layer, thrown out from one of the test pits dug in 1965, and one in F4 on the bottom of the layer from some previous digging. Note on Changes in the Surface of the Mound: When a progress report on the excavation was compiled early in February 1966, we had not penetrated below the layer 9 surface at any point in the central and western parts of the excavation. Since this layer 9 surface was dipping to the west from squares C5 and D5 and since the soils above it were for the most part of the compact sandy types we associated with comparatively recent formations, we assumed as likely that our excavation lay across the edge of the original mound and that the whole of the area to the west represented a cultural buildup, most of it post-Level IX. Our subsequent excavations in the western parts of the site did not support this hypothesis. The evidence as it now stands is as follows: a. The Level I occupation was found concentrated in two main areas: to the east, a concentrated burial area in C3, C4, C5, D4; to the west, three burials that are probably either late Level I or early Level II but whose stratigraphical position is rather insecure. b. On the available evidence, the bedding of the natural surface seems to rise to the east, west, and southeast corners of the site leaving a shallow depression in the central part of the site that has a central axis more or less north to south. In this central depression, no evidence of Level I occupation was found. c. This evidence needs more careful checking, but as far as it goes it suggests quite strongly that the initial settlement was on two patches of high ground to the east and west of our area, with a shallow depression – possibly a small runoff gully – between them and burial and midden areas on the sloping sides of the depression. Unfortunately, the natural was nowhere sectioned deeply enough or on a long enough face to confirm the existence of this gully. d. The Level II occupation spreads completely across the whole area excavated, its upper surface nearly level over most of the site but rising somewhat sharply in square F4 in the southeast corner of our excavation and much less sharply in the southwest corner (square E6). From there up to Level VIII the buildup seems, layer by layer, to have maintained essentially the same direction of slope as the Level II surface, gradually changing the contours to an almost level surface. e. At some time between the Level VIII and IX occupations, the central part of the site was again heavily scoured down, probably by water erosion. This erosion in general cuts down into layer 19 along its deepest part, biting into and in places


partially removing burials of the Level VI and VII period but leaving the Level II (layer 20) occupation substantially undisturbed. f. The Level IX occupation seals over this eroded hollow but maintains the contour set by it. g. In the period of abandonment that followed the destruction of the Level IX occupation, the central depression was again almost completely filled up, apparently with rapidly deposited silted materials. h. From this point to the modern ground surface, successive layers maintain this almost level surface with only very minor modifications. The changes in the drainage pattern of the surface of the mound, which account for the alternative cutting down and partial infilling of the center and parts of the western side of the site, are illustrated in Figs. 3-15, 3-16, and 3-17. It seems to be this process that accounts for the events postulated above and for the rather different character of the soils in the areas affected [see discussion of 1968 soil zones in Chapter 4 – DTB]. The changes are modifications of the microdrainage patterns. The basic pattern seems to have been determined by the original surface and the mounds of the burials of the early periods. Subsequent events merely modified this basic pattern.


Fig. 3-16

Fig. 3-17

Fig. 3-18


Conclusions There are two major conclusions that can be drawn from the 1965–1966 excavations. First, the occupation of the site has been prolonged but not continuous. Second, it may be subdivided into the following five major periods: 1. The Early Period is thought to be in the general sense neolithic throughout. Goodquality pottery was being made, most of it cord-marked but some of it attractively decorated with incised, impressed, or painted designs. Small stone adzes were in use; although heavier woodworking tools must have been available, none were found in the present excavation. The practice of agriculture has not been proven, although tentatively it may be assumed from the prolonged occupation of a lowland site. Hunting and probably animal husbandry are likely to have been important sources of food. Inhumation burial was practiced and the graves were normally fitted out with a somewhat elaborate equipment of grave furniture. 2. During a second Middle Period, the culture contained the manufacture of bronze – actually introduced in the closing stages of the previous period. Stone tools continued to be manufactured throughout this period to at least some extent, and the available evidence may suggest that there was some decline in the use of bronze at the end of the period and a revival in the use of stone tools. No iron of any kind was recovered either stratigraphically or absolutely below Level VIII. Hunting and probably animal husbandry maintain their importance. Inhumation remains the normal form of burial, with a progressive simplification of the grave furniture as the period advances. When this period closed there was a probably prolonged interval during which the site remained unoccupied. 3. At the end of this break, the site was reoccupied by an iron-using people who practiced cremation [LP 1- 2]. The practice of cremation in the area is not likely until after the introduction of Buddhism, and the evidence available suggests that this reoccupation of the site is not likely to have been much before A.D 1200 (celadon appears in the next level above it) and may well be less than nine hundred years ago. It is possible that these people introduced wet-rice cultivation into the immediate area. 4. After another perhaps relatively short break, the site was reoccupied [LP 3–5] and remained so until about 150 years ago, with only short breaks in continuity. During this period there was some slight degree of cultural change, which could imply a change in ownership of the site, a change in the cultural affiliations of the occupants, or perhaps merely a degree of internal cultural development. This period ended again in the abandonment of the site. 5. After several decades, the site was reoccupied by its present owners [LP 6] and developed for garden cultivation. This is known to have happened about seventy or eighty years ago. Note on Methods and Difficulties Throughout the season, the excavation represented an attempt to apply the methods of stratigraphical excavation used in New Zealand to the tropical soils of northeast Thailand. The excavations were thus to some extent experimental and our success limited by our lack of experience in using these methods on soils of this type. Enough was achieved, however, to convince us all that this approach can be employed usefully, even in extreme


conditions in which soil colors fade rapidly to a uniform gray, and in which the dried soils bake to a hardness that makes use of most trowels impossible. Our difficulties came from three main sources: (a) the linguistic problem of explaining to the labor force exactly what was required; (b) the difficulty of obtaining suitable tools; and (c) the character of the soils. With the passage of time the best of our workmen became very good indeed at locating soul changes, at following a layer accurately, and at isolating the fill of a disturbance from the material into which the disturbance cut. Most became adequate; a few are never likely to become proficient. It did, however, take time to reach even this stage and undoubtedly much information was lost, especially in the early stages, simply through bad digging. The second class of problems was more difficult to overcome. To cut through the often very hard soils, tools must either be of high-grade steel or relatively massive. Use of the heavier tools resulted in a good deal of inaccuracy in digging and involved the risk of damage to finds. It also made it impossible to dress down the face clear. On the other hand, really good trowels and spades were not obtainable in Thailand at that time and the cost (and delay) of importing them would have been prohibitive. As a result, it was unavoidable that within a matter of days almost all our trowels were either broken or so buckled as to be useless. Three high-quality trowels brought from the United States and one brought from New Zealand stood up to the work well and proved adequate for handling all but the very hardest materials. We experimented with the use of heavy knives with a blade about 12 1/4 cm long, and these proved reasonably satisfactory, although the lack of offset in the handle limited their usefulness. Spades of sufficient quality proved quite unobtainable. The best substitutes we could find were square-mouthed shovels with the open edge sharpened. The soft steel of these required constant resharpening and the wide blade made them cumbersome to use. Thus much of the digging had to be done with the local hoes (chop) and long-handled digging sticks (siam) tipped with a narrow and strongly curved blade. With these it is very difficult to cut or maintain a really square face. Future expeditions would be wise to make arrangements to bring in with them sufficient quantities of good trowels and spades and to supplement these locally with knives, hoes, digging sticks, and so on. The final category of difficulties is inherent in the nature of the soils and can be overcome only with time and the building of a body of archaeologists and workmen experienced in handling the special problems involved. After about six weeks we found that we were working with much greater certainty and were able to detect variations in the soil that were quite invisible to us when we started. At the same time, these particular soils are never likely to become altogether easy to handle. Soil color and hardness undergo considerable modification in drying; root penetration over much of the site, together with the annual drying and cracking that alternates with soaking and refusion of the blocks, has resulted in much modification of the structure and has rendered it almost impossible to follow a disturbance through certain materials (this is especially true of layers 4 to 11 inclusive). All too frequently, we located pits or postholes at a particular level only to find it impossible to follow them down with any degree of certainty or precise definition of the edges. All too frequently, in consequence, we had to resort to the cutting of horizontal and vertical sections to relocate our features. That some disturbances were missed altogether is


indicated by the incompleteness of the posthole patterns in some of the presumed house structures (especially those found in layers 4 and 11). Notes: 1. Use of the Pellet Bow (krasun or kathun): Particularly common finds from every level of the site’s history were large numbers of small pellets of baked clay about the size of and much the shape of clay marbles. These were immediately identified by our workmen as pellets for use with a pellet bow still widely used in many parts of Thailand by boys for hunting birds and small game. Nai made one of these bows for us and he and others demonstrated its use. The bow is made of bamboo tapered toward each end and notched near the horns to receive the strings. The bow “string” is doubled and near the center the two strings support between them a small square frame made of thin strips of bamboo and grass that receives the pellet. In use, the pellet is supported on the frame with finger and thumb and the bow drawn somewhat to one side. When the string is released the pellet travels past the bow to one side or the other just above the point of grip. We were told that both fired and unfired pellets are used, the latter being preferred for very small birds. Two fine (modern) bows of this type in the possession of Mrs. Foord of Dunedin, New Zealand, both of which come from the peninsular area of southern Thailand, have hand grips carved in the form of stylized birds [cf. also White 1982a:24 for photographs.-DTB] 2. Pottery Disks. One flat, circular 4 cm pottery disk apparently trimmed to shape from a large sherd was found in a probably secure Level II context. Similar disks have been reported from other sites in Thailand. This our workmen identified as used in a game called len lum (“playing holes”) and now played only during the festivities of the Buddhist New Year [Songkran]. In this game a small hole slightly larger than the diameter of the disk and about half depth is made by twisting the disk edge down in the ground. The contestants then attempt to pitch the disk from a few paces away into the hole. Old coins are often used instead of the pottery disks. [Exactly the same sort of disks made from earthenware, stoneware, or porcelain sherds are recovered from archaeological sites in the Philippines and the same games are played by boys there today.-WS] Disks of this type have elsewhere been described as lids. Certainly they could be used as lids for very small-mouthed vessels, but I know of no vessels from this early period with such small mouths.

Chapter 4 The 1968 Excavation By Donn Bayard As mentioned at the beginning of the previous chapter, Parker and I thought it neither advisable nor feasible to present a single synthesized account of the two excavations and their stratigraphy. However, the frequent cross-referencing of 1966 and 1968 layers and levels to the system of overall Early-Middle-Late Period (EP, MP, LP) levels throughout the text of both chapters, together with the chart presented here as Fig. 4-2, will hopefully allow the reader to understand the sequence as a whole. This EP-MP-LP system will be used almost exclusively in the following chapter. The Site of Non Nok Tha For a site exhibiting a sequence spanning an estimated four to five thousand years, Non Nok Tha presents a most unprepossessing appearance (P1. 2a). The top of the mound lies only some 80 to 150 cm above the level of the fields; it is barely noticeable as a mound at all from a distance, distinguished only by its cover of garden crops and a patch of tall bamboo at its eastern end. A shallow seasonal stream, the Huai Nyai or “Big Creek,” lies to the north of the mound. During the dry season this is a series of mud holes, many of them completely dry; even during the rainy season little movement of the water appears to occur. The first modern inhabitant of the site was an old man who moved there shortly after the founding of Ban Na Di in 1894 to trap partridges; hence the name of the side, “Partridge Mound.” For the past several decades the site has been cultivated in garden and cash crops. At present the site is divided by lines of banana trees into four plots owned by four families living in Ban Na Di some 500 m to the north of the mound. The only structural features present on the mound at the time of both the 1966 and 1968 excavations were three field shelters (thiang na) and a rice threshing floor made of cow dung. The area excavated in 1966, located in the approximate center of the mound, had been planted in jute prior to excavation. The 1968 area was planted primarily in cotton, but its western end intruded into a field of small mulberry trees grown for feeding silkworms. Neither of these crops entail much disturbance of the underlying soil; in fact, plowing for all garden crops is rarely deeper than about 15 cm. The absolute location of the site is lat. 16º47’57” N and long. 102º18’17” E, as estimated on the 1:50,000 U.S. Army Map Service series (UTM coordinates are 48 QTD 02138-18592). At the time of the excavations, access is not easy. Ban Na Di was connected to the district center only by a cart track running north from the Phu Wiang district center to pass through Ban Na Kham some 2 km east of Ban Na Di; a branch track leads off this to the village itself. Neither track was passable to any vehicle larger than a bicycle during most of the rainy season, but during the dry season they could be managed by motorcycle, jute and timber trucks and smaller four-wheel drive vehicles such as the one used during the 1968 excavation. Roads have since been considerably improved. From the Phu Wiang district center, a more-or-less all-weather road leads to the paved


Loei-Khon highway. The distance from Khon Kaen proper is some 86 km, or about two hours’ driving time; prior to 1960 the trip would have taken at least a day by oxcart. Excavation Prior to 1968 As already mentioned, the site was located by Gorman at the end of the dry season in April 1964; the onset of heavy rains precluded any investigation beyond the collection of a few sherds from the surface and exposed sides of the mound (Solheim and Gorman 1966:166, 176). The following season, during the second year of the Thai Hawai`i Program, Ernestine Green returned to the site in January 1965 for nine days of extensive testing. (See Chapter 1 for a summary of Green’s testing.) The results of this limited testing, and the fact that it seemed to indicate an extensive burial area (only the marginal squares 1, 3, and 5 failed to produce burials), led the director of the Thai-Hawai`i Program, Wilhelm G. Solheim II, to devote four months of the third and final season of the program to a large-scale area excavation at the site; Parker has described this excavation in the preceding chapter. The site proved far more dense in burials than we had anticipated: The preparation, photographing, mapping, and lifting of the eighty five complete or partial burials encountered proved so time-consuming that the deadline for completion of the excavation had to be extended several times. Finally, due to commitments to excavate elsewhere, the site had to be closed in late April, with several baulks unexcavated. Despite this, a total area of 150 m2 was excavated almost all of it down to a sterile base (some of the bottommost layer remained in the southern and western extremities of the site). Some sixty of the eighty-five burial pits were completely cleared, mapped, and lifted. All major sections were drawn and features mapped for each layer prior to completing the excavation. The 1968 Excavation: Procedure The variety of materials recovered in 1966 raised a number of interesting questions; Chief among these was the nature and origin of the quite sophisticated bronze technology so evident in the lower levels (particularly Layer 20 or Level II in the 1966 sequence [MP 2; see Fig. 4-2]). This excavation produced the first double molds for bronze casting ever recovered in Mainland Southeast Asia in a dated context, it also afforded the first clear stratigraphic evidence of a “Bronze Age” period in Southeast Asia during which iron was not is use (Solheim 1967b;899-900), 1967c, 1968). Even more startling, of course, were the dates obtained from a level above that containing most of the molds, suggesting that the technology was present in the third millennium B.C. (Solheim 1968; Bayard 1972). Finds of several crucibles and an actual axe from the dated level and ones slightly above it tended to support the conclusion that the molds were not merely a chance introduction, but part of an indigenous traition. More importantly, the forms of the crucibles and double molds bore no relationship to the metal technology of the Indus Valley civilization, which was only beginning to develop by 2300 B.C. (Solheim 1968:62); nor was there any relation to the apparently slightly later Chinese technology (N. Barnard, personal communication). The obvious but still unproven conclusion we drew was of course that this development of


bronze metallurgy was indigenous to Southeast Asia, although the specific area of origin was (and is) unknown (Bayard 1980b). The indications of this early and apparently indigenous development, however, raised questions of a more general nature. The evidence of faunal remains in the burials of the pre-bronze and bronze periods seemed to indicate a different subsistence base than that of the present population, whole the remains from the upper layers (late Period 1 to 6) provided tentative evidence of a lesser reliance on game and bovines, implying a wet-rice subsistence base similar to the modern villagers. The presence of water buffalo remains mainly in these upper levels (Higham 1975b) and their near absence in the earlier levels would tend to imply an earlier subsistence pattern based on something other than wet rice, for which most cultivation techniques require buffalo for plowing. This was most probably slash-and-burn cultivation of rice, although at least one other worker has argued for inundation cultivation of wet rice (Gorman 1977:339). A second general question related to subsistence was what effect the presumed transition from swidden to paddy cultivation of rice may have had on the settlement patterns and social structure of the prehistoric population during the period 500 B.C.–A.D. 1000. I returned to Non Nok Tha in 1968 hoping to obtain additional data pertinent to both of these general questions, as well as to get further carbon samples to corroborate our early dates for metallurgy and attempt to clear up our somewhat contradictory series of radiocarbon dates from the 1966 season. As both of the general questions could best be solved by the investigation of a bronze period occupation area and comparison of this with the occupational evidence we had recovered from the upper iron period levels, I decided to excavate an area to the north and east of the 1966 excavation, where evidence from that season’s work had indicated that an occupational or refuse area might adjoin the cemetery (Solheim, Parker, and Bayard 1966:39). During the course of the excavation I was assisted by Nai Thaweechai Uthaiwee from Ban Koeng, Maha Sarakham province, an accomplished excavator who had been associated with the Hawai`i program since 1964; Mr. Terry Marsh of Honolulu, an experienced potter and surveyor; and Nai Khamporn Phithaksin and Nai Buapan Bunsom, both of Ban Na Di. Liaison with the Fine Arts Department and assistance in the field was provided by Nai Viraj Khunnamas of FAD Unit No. 7, Khon Kaen. The 1968 excavation began on February 11, following the setting up of field headquarters and hiring of twenty-four workmen, most of them veterans of the 1966 season. The area selected for excavation lay 18 to 40 m north of the 1966 area and extended from its 4 grid line eastward for 20 m. This grid line was chosen as the baseline for the 1968 excavation, and 4 x 1 m trenches and a central 4 x 4 m square were laid out to the west of this line. To the east, seven 4 x 4 squares were laid out and excavated; 1 meter baulks were retained between all squares (Fig. 4-1); this enabled completion of sections measuring 24 m E/W by 22.5 m N/S. Squares were named after their northeastern grid peg, with number and letter reversed relative to 1966 squares to avoid confusion. Excavation continued on a daily basis into the beginning of the rainy season in late April, which made the work more difficult but still possible with frequent bailing. Fortunately, the weather cleared in early May, allowing us to complete removal of all baulks and draw all sections with the exception of the west side of the 4-line trenches (this section would have closely duplicated the


Fig. 4-1

eastern sides of the trenches). The excavation was finished and backfilling begun on May 28 (see Pls. 6a-e). During this period, seventy-eight days of excavation enabled us to remove an average of 140 cm of cultural deposit plus a large number of deeper burials over an area of 189.5 m2 (the northern 1.5 m of trench 4H was not excavated due to the presence of a large ground termite nest). The 1968 excavation thus covered a somewhat greater area than the 150 m2 excavated in 1966. To ensure that sterile natural had been attained over the width and breadth of the site, the narrow 4-line trenches and a 1 meter strip along the north sides of the F-squares were dug down to a depth of at least 210 cm below the surface. The natural stratigraphy of the site was followed throughout the excavation, both in square and baulk removal. When layers thicker than 10-15 cm were encountered, spits of 10 cm were used; in a few cases, such as the thick layers 5 and 6 and in baulk removal, spits of 15 cm were used. But in all cases, whenever the surface of a new layer or spit was cleared, all detectable disturbances were mapped and either completely cleared (if


postholes or small pits) or carefully defined (if large pits or burials lying deeper than 10 cm below the layer). These methods proved considerably easier to apply in 1968 than in 1966, largely because both the director and the labor force were by now familiar with the soil conditions on the site. Several other factors were important however. The chief difficulties in 1966 had been threefold: the basic problem of communication with the workmen; hard soils; and the extreme difficulty of stratigraphic interpretation of the soils themselves (Solheim, Parker, and Bayard 1966:6). Improvement in the first and last areas of difficulty in 1968 was marked; the excavation was conducted almost wholly in Thai and Phasa Isan rather than English, and the workmen and I had arrived at a sufficiently precise and detailed archaeological vocabulary to enable work to proceed with a minimum of misunderstanding. Moreover, in addition to the stratigraphic experience gained during the five months of the 1966 excavation, we were very fortunate in that most of the 1968 area consisted of sandier, softer soils that were considerably easier to interpret than those to the south. The difficulty with tools persisted, however, and as in the earlier excavation we were forced to rely chiefly on the local short-handled, broad-bladed hoe or chop for gross removal of layers. We used good-quality New Zealand trowels and small, locally purchased knives for the fine work such as burial clearing, posthole excavation, and final definition of layer surfaces (e.g., P1. 6f). As in 1966, the use of the chop resulted in a small amount of unavoidable damage to pots and burials, but the workmen rarely removed more than 5 cm per stroke of the hoe and on encountering something would switch to more suitable tools at once. A more severe difficulty remained unsolved in 1968; as of yet, no effective method has been developed for screening of deposits. The soils of Non Nok Tha and other sites in the Phu Wiang area have high clay content and can be broken into particles of less than 1 cm only with difficulty. Hydraulic sifting proved even less feasible when we tested it in 1966, as the soils soften slowly and become very sticky when wet. In addition, water is scarce; the nearest source lies some 150 m from the site and is needed for human and animal consumption during the dry season. It was apparent from viewing our dumps that a number of small sherds (and occasional large ones overlooked by the workmen) were being lost due to this lack of screening, and this must necessarily bias the sample collected and analyzed to some extent. Still, the vast majority of sherds over 2 cm across were recovered, as well as more easily visible metal fragments less than 5 mm in diameter. The same bias is of course present in the representation of microfaunal remains; the sandier soils of sites like Ban Chiang and Ban Na Di allowed sieving and the recovery of large amounts of fish and frog bones (Higham and Kijngam 1979, 1982a); such remains were doubtless present at Non Nok Tha (as indicated by the analysis of contents of funerary vessels; see Chapter 6). Although the above methods certainly imposed limitations, I feel that in general we were able to follow stratigraphic distinctions and record the provenance of artifacts with a high measure of success. I believe we were also able to detect, define, and clear well over half of the extensive disturbances present. However, I also feel certain that as many as 30% of these disturbances went undetected, particularly those originating in the upper two soils, where any disturbance larger than a large posthole (e.g., graves) proved extremely hard to detect unless we encountered it cut into more distinctive Soil III at a lower level. Nonetheless, I feel that in spite of these difficulties the 1968 excavation produced a body


of data as carefully excavated and accurately provenanced by natural stratigraphy as any open site excavated in Southeast Asia at that time, although obviously techniques have improved since. Hence, I cannot agree with Clark’s intimation that the stratigraphy is unreliable because it was “so heavily disturbed by burials, rootholes, termites and washouts” (1977:346). The site is not atypical in this regard, and similar problems have been encountered at subsequent excavations in the area-—for example, the bioturbation obscuring all but the gross stratigraphy at Ban Chiang (Higham and Kijngam 1984:22). The site of Ban Na Di, Udon (not to be confused with the village of the same name near Non Nok Tha) Proved to have much more easily interpretable stratigraphy, at least in the lower sand-lensed layers (op. cit., 29); however, even at the latter site, plans of postholes and other structural features from the upper levels (1-6) have yet to be published. Stratigraphy As mentioned above, the soils at Non Nok Tha are far from easy to interpret stratigraphically; an exposed freshly cut section would seem to present only three or perhaps four distinct layers over a depth of some 150 cm if viewed by one lacking experience with soils of the area. The effects of cultural modification within these gross divisions are difficult indeed to determine, as the site lacks distinct bands or extensive lenses of sherds and other occupational debris to serve as level markers. The modifications to the soils as a result of occupational activity that are so easily observed in many temperate or tropical volcanic soils are largely submerged in the general soils profile: They become detectable only after considerable experience in looking at and handling the soil. Given these factors, the temptation to abandon any attempts to follow natural stratification and fall back on unit-level excavation is a strong one; a glance at Figures 4-3 through 4-8 will demonstrate the gross misinterpretation that would have resulted had this been done. Fortunately Parker, while directing the 1966 excavation, exercised the caution, restraint, and patience necessary to enable the excavation team to become familiar with the soil distinctions present; for example, over two weeks were spent in a careful investigation of the stratigraphy in a single test square prior to laying out the excavation area proper. The 1968 excavation, carried out by essentially the same team, was thus able to proceed with considerably more speed and confidence due to this four months of previous experience on the same site. Subsequent excavations at Northeast Thai sites with much deeper stratigraphy, beginning with Ban Chiang (Gorman and Charoenwongsa 1976, White 1982a) and continuing through Non Chai (Bayard et al. 1982-1983) and Ban Na Di (Higham and Kijngam 1984), have been able to follow a gross layer, 10 cm spit excavation system with considerable success, although in all cases final assignation of spits to overall layers took place only after excavation was completed and final sections were drawn. At Non Nok Tha, somewhat different techniques were necessary in following stratigraphic distinctions horizontally as opposed to the recording of sections on completing excavation. While texture (to the eyes and trowel as well as to the fingers) was important in both instances, it was of more importance in distinguishing layer boundaries during removal, as were color contrasts, many of which vanished quickly upon drying and hence were of less


use during section drawing. The latter proved to be a quite complex process at Non Nok Tha: In addition to the rapid fading of some color distinctions, other equally valuable distinctions appeared only after the soil had dried from two to five days, then faded again after eight to ten days; exposure. As a result, constant attention had to be paid to the section profile as it was exposed layer by layer, and absolute depth readings of each layer were taken to be checked against the final section drawings. In recording the sections, light conditions proved to be the most important factor: Distinctions were clearest in early morning or late afternoon sun, and they were almost as clear under light overcast; however, direct noonday sunlight, open shade, or heavy overcast all made interpretation and recording extremely difficult. Given these problems, it should be clear that section drawing at Non Nok Tha was even more a subjective interpretation of the actual situation than I believe it usually is, particularly with regard to the recording of sectioned disturbances. Despite this, the matching up of baulk sections as the baulks between them were removed proved to be quite simple. In a few cases some of the layers recorded were incorrectly labeled (Bag labels were of course changed in such cases), but in almost all instances every distinction made on a particular section matched very closely those made in the adjoining squares-— even though these may have drawn up to six weeks apart. Given the overall fit of all layer distinctions and the observations made during the excavation, I feel the sections as recorded give good representation of the actual situation. In all, some 174 m of sections were drawn (the western sides of the 4-line trenches were not drawn, as they closely duplicated the eastern sides); a selection of these is shown in Figs. 4-3 to 4-8. As was the case in 1966, five fairly distinct soils were distinguished within the 150 cm profile, but since the soils of the two excavations areas differ more than would be expected, these two sets are not identical; the necessarily tentative correlation of soils, levels, and layers is shown in Fig. 4-2, along with the overall system of Early, Middle, and Late Period levels used here and in previous publications on the site. Soil I (layers 1-5; see Figs 4-4 and 4-5) is a gray podzolic clayey loam containing considerable amounts of fine sand; it is hard, compact, and very lumpy due to intensive root and insect penetration. Layer 5, which makes up the bottom 30 cm of this soil, is more brownish than the upper 20 cm (Layers 1-4). This soil seems quite firmly equivalent to both Soil I and Soil II in the 1966 sequence (See Chapter 3), but no fossil soil zone equivalent to 1966 layer 7 was encountered in the 1968 profile. The water-deposited layers present on the western portion of the 1966 area are also absent, with the probable exception of layer 3, as discussed below. Soil II (layer 6) is a brown clayey loam containing somewhat less sand than Soil I; like the latter, it is hard, compact and root-broken. This soil is equivalent to Soil III in the 1966 scheme and similarly gives an abrupt contrast with the layer above when exposed horizontally, but it is extremely difficult to detect in well-dried vertical exposure. In the 1968 area this soil had been truncated by erosion prior to deposition of Soil I; this is particularly marked on the eastern part of the site. Soil II has an average thickness of about 25 cm, but it is quite variable.


Fig. 4-2

Soil III (layers 7 and 8) is markedly softer sandy loam with a fossil humus zone present at the II/III interface. At this point, Soil III is brownish gray with motlings of Soil III interpenetrating it, probably due to earthworms. Soil III proper, however, is brownish-yellow in the southern part and brownish red in the northern third of the area. The average depth of this soil is 20 cm, but it varies in thickness even more than Soil II. Fig. 4-3

Soil IV (layers 9 and 10) is a modified version of the natural soil that forms the base of the mound. It is considerably harder than Soil III and contains a higher proportion of clay, as well as many small lumps of brownish-black laterite in its lower portion.


This soil varies from a moderately hard, rather loose, reddish sandy clay on the northern part of the excavated area to a very hard, compact whitish or yellowish clay with little sand on the south of the area. Once again, thickness is variable, ranging from 10 to 30 cm. Both soil III and Soil IV seem to be derived primarily from the weathering of the natural (probably aided by repeated exposure due to extensive grave digging); they thus contrast with both Soil I, modified significantly by occupational use, and Soil II, the loamier texture and increased humus content of which may be due to the presence of more extensive vegetation cover during its formation. Soils III and IV together seem to be equivalent to 1966 Soil IB, but differ from it in their sandier texture and the greater relative ease with which the layers within them may be distinguished. Soil V (layers 11 and 12) is the unmodified natural base material of the mound, and seems essentially identical to Soil V in 1966. It is moderately hard, lumpy clay, interspersed with many laterite nodules. The hardness increases some 20 cm below its surface and continues down to the lower limit of the excavation in the 4-line and F-line trenches to at lest 240 cm below the present surface. Within the above five soil divisions, ten occupational or weathering layers were distinguished that extend over all or a significant part of the excavation; in addition, three sublayers were present over limited portions of the excavation area. Unfortunately, as in 1966 it was impossible to procure a Munsell soil color chart prior to departure for Thailand, and hence the color descriptions given below for the layers as they appeared when freshly exposed are necessarily subjective. Munsell color determinations were later made on dry soil samples in the University of Hawai`i laboratory (a full report on soil analyses will appear in the volume of specialist reports), and these should suffice to give a general idea of the range of colors involved. In point of fact, readings taken in the field would create a false impression of precision because a large number of different readings could be taken at a single spot, depending on the length of time since exposure, groundwater seepage, light conditions, and the proximity of large roots, to mention only a few of the factors involved in color variability at the site. The layers defined are described as follows: Fig. 4-4


1. A quite thin, loose and crumbly layer made up of humus and plant debris derived from cotton and mulberry cultivation. It is medium gray in color and varies in thickness from 0 to not more than 2 cm except where it fills recent holes from planting. Fig. 4-5

2. This layer consists of the compact, lighter gray humus underlying layer 1; like the latter this layer is thoroughly mixed by yearly plowing of the garden to a depth of some 10-15 cm. The depth of the layer is from 3 to 6 cm; it contains some small amount of plowed-up sherds and little else. Recent planting holes were the only disturbances detected. The Munsell notation on samples of this layer (mixed with a small amount of layer 1) form the eastern side of the site (OF/1F baulk) and the western side (3F/4F baulk) were E: 10 YR 4/2; and W: 7.5 YR 4/2. Fig. 3-4

3. This layer is more compact and sandy than layer 2 and is dark brownish gray in color (E: 10 YR 4/2; W: 7.5 YR 4/2). The layer is variable in thickness, ranging from 3-15 cm. In what at present are the lower portions of the site (squares OF, 1F and 4H) this layer contains many small, evenly worn sherds that may have been waterdeposited, perhaps deriving from layers similar to or identical with 1966 layers 2 trough 6 to the south. However, there is no lensing out of sublayers of layer 3 on the higher portions of the site under other sublayers of 3 on the lower portions, and there is no clear distinction within the layer in the two areas; hence it is treated as a single layer, although it is unlikely that the thick portions of the layer, in squares 4F


and 3F (which contain very few shards) were water-deposited. No structural features-—aside from two or three possible pits-—appear to have derived from this layer, although other features (as well as any signs of water redeposition) may have been masked by the fact that it lies in the bottom of the plow zone. 4. This layer is very similar to 3 but somewhat more brownish and noticeably lighter after a period of drying (E: 10 YR 5/1; W: 7.5 YR 4/2). It is also distinguished by many small red and yellow flecks of burned earth, larger pieces of burned earth (usually yellowish), and numerous small pieces of charcoal. The three-quarter interface is marked by many sherds, pieces of animal bone, and some fairly heavy concentrations of charcoal. This layer varies in thickness from about 5 cm in southern 4D to about 15 cm in the lower areas of the site; the average thickness is about 10 cm. A larger amount of the sherds and several postholes and pits were encountered in this layer, as well as many large root holes. 5. A lighter brown layer, somewhat sandier and less lumpy than 4 but also containing many red and yellow flecks (E: 10 YR 5/2; W: 7.5 YR 4/2). The layer varies from 15 to 40 cm in thickness with many sherds and structural features present and probably represents a long period for fairly regular-—if not continual-—occupation. Two particularly dense zones of occupational debris seem to be present at or near the surface of the layer and about two-thirds of the way down through it. However, neither of these zones extend uniformly over the site, nor do they form distinct living surfaces in contrast to the rest of the layer. Fig. 4-6

6. A dark, almost chocolate brown layer (when first exposed) containing less sand and more clay than the layers above (E: 10 YR 6/2; W: upper, 7.5 YR 4/2, lower 10 YR 5/1). On drying, this layer is even harder and lumpier than those above it. Three distinct cultural horizons seem to be present in this layer. The first of these is represented by posthole butts and burials in shallow graves presumably truncated by the erosion of the upper part of the layer, particularly on the N and E edges of the area; this horizon is designated Level VII/MP 6. The second horizon is comprised of structural features and some burials provenanced to an occupational surface on or very near the present top of the layer (Level VI/MP 5). A few postholes and numerous burials derived from an earlier cemetery level some 15 to 20 cm below the present surface of layer 6 make up the third horizon (Level V/MP 4). This


deepest horizon is marked in places by discontinuous lenses of sherds. The layer as a whole ranges from 15 to 40 cm in thickness but is thinnest on the eastern part of the site, where erosion of its upper portion has been most marked. 7. This layer contains much more sand and much less clay than the layers above it; it is also markedly looser in texture. The 6/7 interface is marked by the fossil humus zone mentioned above, which averages about 5 cm thick. Below this grayish brown mottled zone, layer 7 ranges in color from a brown tinged with gray and yellow in the southwestern part of the site (soil zone C) to a more clayey grayish yellow version containing large amounts of snail shells on the east (10 YR 6/2; zone B) and a reddish brown sandy version on the northwest quadrant of the site (7.5 YR 7/2; zone A). Thickness varies from 5 to 30 cm, averaging about 10 cm on the west side but thicker to the east. Many burials are provenienced to this layer but relatively little in the way of structural features and occupational refuse (aside from that derived from tops of graves) was recovered from it. 8. Somewhat lighter in color and more compact than layer 7, this layer varies in color and sand/clay content in a fashion similar to it (E: 10 YR 6/2; W: 5 YR 6/2). This layer varies sharply in thickness, averaging 10 cm on the west and about 20 on the east. Despite its greater depth on the east, this layer appears to have been truncated by erosion there, as well as on the extreme western margin of the site; its depth in the east is apparently due to its having been laid down over several low points in the surface of layer 9 below it. In this area layer 8, like 7, contains a fair amount of snail shells, although less than the upper layer. In both cases the presence of the snails seems to be related to the areas of the site that were lowlying during the formation of these layers. The burials associated with layer 8 are attributed to two cultural levels, II and III, on the basis of typology alone, since no distinction is evident within the later itself. Layer 8A, a more clayey version of 8, over part of square 1E below 8; this sublayer is about 15 cm thick and almost sterile. 9. The color and sand-to-clay ratio of this layer vary as with layers 7 and 8 (E: 7.5 YR 6/2; W: 7.5 YR 5/4); however, layer 9 is always harder, somewhat lighter in color, and more like the natural than 8, which allows for relatively easy detection of 8 burial pits at the 8/9 interface. Thickness varies from about 10 to 25 cm. Layer 9A, a very hard whitish yellow layer, lies between layers 9 and 11 in squares 4D, 4E, and 3E; it is an average 20 cm in thickness but attains a depth of 40 cm in places. Layer 9B, a stickier, grayer version of 9A (e: 7.5 yr 6/2), lies under 9A in square OF; two additional lenses (9C and D; see fig. 4-7, F/N section) under the western end of 9B may represent a water runoff gully cut into layer 11. Aside from the Level I burials derived from this layer and a few associated postholes, layer 9 is almost sterile; the sublayers are completely sterile. Fig. 4-7


10. This is a fairly hard, red, lumpy layer extending north from northern 3F and 4F (W: 7.5 YR 5/4). It does not appear to be of any more cultural significance than the sublayers 9A and 9B described above, but as it covers a fairly large portion of the site and was assigned a distinct number early on in the excavation, it is convenient to retain the number. Like layer 9 it contains a small amount of laterite nodules and like layers 9A and 9B it is sterile. Its thickness averages about 25 cm. 11. This is the top 20 to 60 cm of the natural soil that forms the base of Non Nok Tha; it is clayey, lumpy, and moderately hard, with many laterite nodules. Its color varies from reddish to yellowish brown, depending on location, as with the layers above it (E: 7.5. YR 5/4; W/YR 6/3). Fig. 4-8

12. This is a harder, more reddish version of the natural; when moist, its color is affected little if at all by the soil zones discussed below (W: 5 YR 6/3; E: 7.5 YR 6.2). It extends for at least 30 cm below the bottom of layer 11 over the entire site and like layer 11 it contains a large amount of small laterite concretions. Fig. 4-9

Late in the excavation, as we reached Soil III in all squares, it became apparent that considerable areal variation was present at the site; this was indicated by the color, sand/ clay ratio, and relative hardness of Soils III and IV, as well as the relative ease of stratigraphic interpretation of these soils and Soil II. Based on these criteria, the site can be divided into the following three zones: Zone A extends over 4H, 4G, and the extreme northern margins of 4F and 3F (See Fig. 4-1); and isolated pocket of similar soil conditions also occurs in 1E, covering all but the southwest corner of the square. Within this zone Soil III and Soil IC are soft, sandy, and generally reddish in color; the stratigraphy is relatively easy to interpret. Zone B occupies the remainder of 4F and 3F with the exception of their adjoining southern corners, as well as all of 2F, 1F and OF plus 1E and the eastern third of 2E. In this area the


lower soils are notably more yellowish and clayey, although still softer than they are in Zone C; this is probably due to their apparently higher moisture content. In addition, Soils I and II here become harder and more insect penetrated (mainly by ants and termites); since the number of insects encountered in this zone did not seem any greater than that met with in Zone A, I assume the effects of tunneling tend to last longer in the harder soil. In any case, interpretation of sections becomes a more difficult task in this zone. Zone C, which extends over the remainder of the excavation area (i.e., the southwest quadrant) is even worse from an archaeological point of view. Here Soils III and IV are even more clayey, hard and compact than in Zone B; insect and root disturbance is even more marked and apparently more persistent. Difficulties in section interpretation and general soil conditions approach the extremely refractory soils lying to the south in the area of the 1966 excavation, which, with the exception of the westernmost four squares (C7, D7, C6, D6; probably in Zone B), would seem to lie wholly within Zone C. While the ultimate causes for the formation of these different zones are unknown, and while their formation may predate any human activity on the mound (see Bayard 1971b: 12), it nonetheless seems likely that the soils zones had at least a slight effect on the human use of the mound. One example is afforded by burial distributions in the various cultural levels encountered; thus almost all Level III/EP 3 burials are confined to Zone C, while those of Level IV/MP 1, although occurring almost everywhere else in the 1968 area, are notably absent (with two marginal exceptions) from Zone C (see Figs. 3-6 and 4-1). This latter distribution pattern would seem to hold true for the Level IV/MP 1 equivalents in the 1966 excavation as well, most of which occur on the western (B equivalent) edge of the area. Zone C would seem to represent a particular area of Non Nok Tha that has remained high ground throughout its history of use. Examination of contour maps of the surface of Soil V in both excavations would indicate that this area was originally a low ridge running roughly north and south, with drainage to both east and west. Throughout its history the excavated area has never varied in surface elevation more than about 70 cm; however, this plus the soil variations resulting from it were apparently enough to have at least some slight effect on the activities of the inhabitants. Cultural Levels and Correlation with the 1988 Sequence Given the difficulties in detecting large disturbances such as when encountered horizontally, particularly in Soil II, it is fortunate that there were a large number of intersecting grave cuts to allow relative sequences of burial to be established and to tie in stratigraphically many of those burials isolated from the vertical section of a baulk. On the basis of this stratigraphical evidence, plus numerous and marked differences in burial typology during the period in which it was being used as a cemetery, it is possible to divide the sequence at Non Nok Tha into a series of more or less discrete cultural levels. Burial intersections were present between all adjacent levels save III (EP 3) and IV (MP 1). However, due to erosion of parts of the site and spatial variability in its utilization by the prehistoric population, the sequence of levels in the 1966 and 1968 areas differs more than would be expected from the 18 m that separate the two excavations. The some fourteen more or less distinct levels were distinguished in 1966, as compared to the eleven arrived at in 1968. Overall level designations (EP, MP, LP) are given the following local 1968 level numbers.


Of these eleven levels, I-II (EP 1-2) apparently predate the appearance of bronze working in the immediate area; that is finds of stone tools are common and no metal has been encountered (except two small fragments-—possibly bronze-—that were found in the basal level in one of the vessels associated with a Level I/EP 1 burial; see Chapter 1). No metal was found in Level II, but in Level III (EP 3) a single bronze tool and two small nodules of the same metal were recovered. Levels IV through VII (MP1-6) may be described as bronze period, or what I have called elsewhere “General Period B” Bayard 1984b:164); bronze axes, molds for casting them, and crucibles for melting the metal were recovered, as were numbers of small lumps of bronze, apparently from casting spillage. Following Level VII (MP 6) and two later levels encountered in the 1966 area (MP 7 and 8), the site was apparently unused for a considerable period and then occupied by an iron-using people who practiced cremation burial rather than simple inhumation. In the 1968 terminology these iron-period levels are designated VIII through XI (LP 2-6), with the last level representing the period of use by the present population of Ban Na Di from 1984 to the present. Detailed descriptions of the burial typology and mortuary analyses will be given in Chapter 8. The lack of exact correspondence between the two areas makes it impossible to relate dates from any one level in the 1966 excavation and the corresponding 1968 level with anything more than a moderate amount of confidence. Figure 4-2 gives some idea of the degree of correlation possible between the two excavations. It will be noted that a number of levels present in the 1966 sequence are not present in the 1968 area (1966 Levels II, III, IV, VI, VII, VIII, and perhaps a portion of IX); similarly, an exact equivalent for level VII in 1968 is not present in the 1966 scheme, and there is some question about the placing of 1968 Level II in the 1966 sequence. 1968 lower Level VIII may well be more equivalent to the 1966 Level VIII than IX. Part of this lack of correspondence is doubtless due to erosion of one or the other areas of the mound at various times during its history. A more apparent factor during the premetal and bronze periods, however, was probably the pattern of use of the site. With the exception of Level VI (MP 5) and perhaps IV (MP 4), it was apparently utilized almost solely as a burial ground during these first two periods. A burial area used by one generation would gradually become filled and succeeding generations would be forced to extend the burials over a wider area, returning to the original site only after sufficient time had elapsed to allow the memories of those buried there to fade. Hence the picture of a sequence of discrete periods (EP1, EP2, etc.) of burials of different types that both excavations have produced is quite likely a spurious one in some cases. At least some of these discrete periods probably represent the return of the same population to the same burial area after a lapse of several generations-—in some cases perhaps well over a hundred years-—when ties to those buried there had been forgotten. However, a radically different interpretation has been postulated by Higham (1996-1997); this will be discussed in Chapter 8. At least a partial solution to the problem of correlation would have been afforded by the obvious method of driving a trench south from square 4D to the old excavation area; unfortunately, a line of banana trees and uncooperative landowners made this impossible (similar problems were encountered in connecting areas in later excavations like Ban Chiang and Ban Na Di). Hence, while comparison of the 1966 and 1968 sections and


burial pottery typologies from both areas allows a moderate degree of correlation, this does not seem sufficient to allow all data from the fourteen 1966 level to be combined with those of the eleven 1968 levels and treated as a single unit, although some combined ceramic analyses will be presented in Chapter 6. In point of fact, it is very likely that neither area presents anywhere near a complete picture of the history of use of the mound as a whole. Even if the two areas alone are considered, it seems clear that, rather than either eleven or fourteen levels being present, the actual number is closer to seventeen. Analyses of ceramic and non-ceramic artifacts and the burials from both excavations will be presented in terms of these overall seventeen levels in Chapters 6 through 8. Evidence of Structures from the 1968 Excavation As mentioned above, the soil conditions at Non Nok Tha made excavation according to the natural stratigraphy of the site difficult; this difficulty was also present in defining the evidences of cultural activity in the form of pits and postholes. Further complicating attempts to define postholes was the presence of numerous small ground termite nests, which in horizontal section bear a close resemblance to medium-diameter postholes. These are quite shallow in vertical section, however, and were easily distinguished from postholes by half-sectioning. In some cases, remains of old termite nests were found inside postholes, making exact measurement of diameter and depth difficult. While posthole fill rarely showed any distinctive color differences from the surrounding soil, differential drying usually proved a contrast, as the slightly loosed soil of the postholes became dry and rather crumbly. Differential drying was of little use in the case of pits, however, where larger dimensions and more compacted fill tended to minimize the contrast (Fig. 10a-f). Thus, while I believe that we were able to record a majority of the postholes present, the percentage of pits that we were able to detect accurately is considerably lower. This is particularly true for those pits (as well as postholes) lying wholly within layers 5 and 6, in which disturbances were extremely difficult to detect. Those disturbances originating in these layers but extending into layers 7 and 8 were, on the other hand, easier to define, and those extending into layers 9, 10, and 11 were quite visible due to the contrasting soils. Figures 4-10a to 4-10f present the postholes and pits classified as definite and probable for the major levels; possible postholes are dotted in, and those features considered to be doubtfully present at the time they were recorded are not shown on the diagrams. Outlines of structures suggested by the distribution of postholes are shown by heavy and light dashed lines, the former being used to indicate the more probable alignments (usually consisting of three or more postholes). Pits and inhumation burials are shown, as are areas of the level surface disturbed or removed by subsequent activity. Despite the fact that structural feature maps for each level were updated at frequent intervals during the course of the excavation and the areas inside the structures suggested by the alignments examined and compared with areas outside them, no differences in soil level, compaction, or coloration could be detected; nor were we able to discover any other evidence suggesting intensive occupation of enclosed, ground-based dwellings. Thus I am forced to conclude tentatively that the structures we are dealing with were all built on piles


Fig. 4-10a

Fig. 4-10b


well above the surface of the ground, as are almost all structure in the contemporary villages of the area (these include almost all houses, rice barns, the sala or wat pavilion, kuti or monks’ quarters, schools, and field shelters. The only ground level structures found currently are the bot, rice mills, small village stores, and toilets. The last three are recent introductions to the area. An examination of the diagrams reveals that all of the possible structures suggested by the posthole alignments are either rectangular or (in a few cases) more or less square. This again is paralleled by modern structures in the area (with the exception of rice barns, which are sometimes circular). Similarly, posthole spacing seems to range between 1.5 and 3 m, with a tendency to cluster at about 2 m; this is very close to modern posthole spacing. One apparent difference that may be of some significance is that of overall size. The approximate dimensions of the structures shown on the diagrams are given below by overall level; the 1968 levels are shown in parentheses. EP 1 (Level I): one structure, 5 by more than 4 m. (Fig. 4-10a) EP 2/3 (Level II/Level III): two structures (both probably from Level II), 6.5 by 11 m and 7.5 by more than 7 m. A possible 5 x 5 m structure in 4F, 3F, and 3E may derive from Level III. (Fig. 4-10b) MP 1 (Level IV): one tentative structure, 7.5 by 15 m. (Fig. 4-10c) Fig. 4-10c

MP 4 (Level V): no measurable structures apparent. MP 5 (Level VI): three superimposed structures, two of them 5.5 by 11.5 m, the third ca. 5 x 8 or more. (Fig. 4-10d)


Fig. 4-10d

Fig. 4-10e


MP 6 (Level VII): no structure apparent. LP 2 (Level VIII/L): two structures, 5 x 8.5 m and 5.5 by more than 5 m. (Fig. 4-10e) LP 3 (Level VIII/U): one structure 7 x 7 m. (Fig. 4-10f) LP 4 (Level IX): one structure 5 x 8.5 with what may be a cooking porch measuring 2.5 by 3 m on the northwest corner. (Fig. 4-9f) Fig. 4-10f

LP 5 (Level X): no structures apparent. To examine the probability of continuity between at least the LP structures and contemporary ones, it is worthwhile to consider modern village architecture briefly. House structures in the nearby village on Ban Na Di are of several types. The house style viewed by the inhabitants as typically Lao or Isan is a composite structure consisting of two houses of about 10 x 6 m joined lengthwise by a narrow porch (P1. 3b). However, houses of this type (including more complex variants) comprise only about 10% of the village total. Far more common are single house units ranging from 7 to 10 m in length and from 6 to 8 m in width; the wider houses usually result from the addition of an unroofed porch along one side of the house. A third type of small, square dwelling measuring about 4 x 4 m with a 3 x 3 post pattern is said by the villagers to be a recent introduction from Central Thailand or the Khorat area. The single house units thus approximate the dimensions suggested by the LP structural evidence. It seems reasonable to conclude, however, that the houses in use by the occupants of the site prior to the long gap in the sequence were substantially longer than those in use


during the later phase of occupation through to the present, although width seems to have remained in the range of 5 to 8 m. This increased size could be explained by any number of factors: strict uxorilocal residence by all the daughters of a family rather than by a single (usually younger) daughter or several daughters in sequence, as is the common practice in the area at present (Keyes 1966a: 64-65); residence by limited extended families rather than the nuclear family unit predominant today; or merely by a cultural preference for larger dwellings. In any event, further speculation in the absence of more concrete data is of no value. Of greater interest, however, is the considerably larger incidence of what would appear to be superimposed structures in the levels preceding the gap in occupations; MP 5 (Level VI) is a particularly striking example. The similarity in width (and perhaps in length as well) of the three structures and the closely parallel alignment of the two houses oriented to the northwest would tend to suggest that the structures were constructed one after the other, with a gap in time between them small enough to allow the builders to recognize and reuse the same location, including in some cases the same postholes (or perhaps the posts themselves, if they were left in position rather than dug up and used elsewhere). This is admittedly a speculative conclusion, but when taken in conjunction with the factors mentioned at the end of the previous section, it tends to imply a settlement pattern and economy significantly different from that of the present day. If, as we suspect, the rarity of water buffalo remains implies the absence of wet-rice agriculture, then it would seem reasonable to assume that the pre-iron inhabitants of the area were practicing some form of swidden cultivation of rice (see Chapters 2 and 9). While swidden cultivation does not necessarily entail the periodic movement of villages (cf. the Hanunóo [Conklin 1957] and the Land Dayaks [Geddes 1954]), it is nonetheless tempting to think of parallels at Non Nok Tha during this period with the dampa system of the Iban (Freeman 1955). Although we have no evidence for longhouses with seasonal occupation of dampa by each work unit near its own fields, some sort of semi migratory settlement with a return to the original dwelling site after an interval of one or two decades might well explain the superposition of several of the structures in the pre-iron levels as well as the apparent rotation of the areas of the mound used for burial. It might also explain the lack of relationship between burials of a particular level to the structures associated with that level, either in terms of superposition and presumed burial under dwellings or of complementary distribution implying a discrete cemetery area. One may imagine several semisedentary villages of swiddeners occupying the Phu Wiang area that utilized Non Nok Tha both for burial and for limited occupation, but at different times; thus village A might use the mound for burial for a space of ten years or fifty while it is unoccupied, then move its dwellings away shortly before a closely related village B moves into the area and either established itself on the mound or continues to use it as a cemetery. Given the small area of the mound excavated, it is impossible to state that both activities were not in fact carried on in different areas of the mound at the same time. Wilen has proposed a similar scenario—-considering such sites as regional ceremonial centers—based on his excavations at Non Pa Kluai on the western side of Phu Wiang and his site survey of that region.


A fairly plausible case for this periodic occupation and abandonment would seem to be afforded by level MP 5. Although its graves are clearly related to MP 4 and MP 6 both on the grounds of burial typology and funerary pottery, MP 4 and 6 nonetheless have much more in common with each other than they have with MP 5. The intrusion of the slightly more distantly related MP 5 population into the immediate area and their use of the mound for burial, followed somewhat later by occupation (based only on the disturbance of Burial 23; the converse is also possible), was in turn followed by their subsequent movement out of the area and its reuse during MP 6 by people who were more closely related to the population using the mound during the MP 4 period. This might be an example of the sort of population movements prevalent during the period prior to the introduction of the water buffalo, iron, wet-rice cultivation, and permanent villages; that is, the complex of traits that I believe characterized “General Period C� in the Mainland Southeast Asian archaeological sequence (Bayard 1984 b). While this interpretation involves much speculation, it does seem to fit the data more closely than other interpretations that would necessarily involve an equal amount of imagination given the limited data at hand. We will return to this question in the concluding chapter.


Chapter 5 Chronology By Donn Bayard Introduction1 “Thus while the evidence for the presence of metallurgy in Southeast Asia at a date of 2500-3000 B.C. or perhaps earlier is at present not totally conclusive, it seems very probable. Hopefully further dates from other bronze-period sites throughout Mainland Southeast Asia will settle the question conclusively in the near future.” (Bayard 1971b: 31). “In short, one cannot altogether escape the impression that a veritable conspiracy is afoot among journalists, certain scholars, and officials in Thailand, the U.S.A., and maybe other countries, to keep dates for early bronze up and critics of them down.” (Loofs-Wossiwa 1983a:10). “But it is necessary to reaffirm Bayard’s own hope, expressed in 1971, that further dates from other sites would settle the question. . . . He rightly points out that one-third of all dates at Non Nok Tha fall before 2000 B.C. and feels that the site was occupied during the third millennium B.C. Without provenanced samples, I am not prepared to accept such a conclusion.” (Higham 1984a:233). As these quotes illustrate, the dating of bronzes and other artifacts from Non Nok Tha is without any doubt the most contentious aspect of the site. The presumed early dates on cattle, pigs, dogs, and rice have also drawn comment (Reed 1977), but the possibility of bronze working in the third millennium B.C. or earlier has been at the center of the controversy since Solheim released the first dates from the site in 1967 (Solheim 1967a, b). Over thirty years the debate has ranged from careful, reasoned analysis (Smith 1979a, Reed 1977, Macdonald 1980, Higham 1984a) to often vituperative and unproductive exchanges. The question has been described by one of the more extreme commentators as “probably the most divisive issue in archaeology since the Piltdown and Glozel cases” (Loofs-Wissowa 1983a:4)-—a flattering if greatly exaggerated judgment. While Solheim and I certainly expected some amount of comment when the first dates were released in 1967, we were frankly surprised at the furor they caused. At the time, very few area excavations up to modern standards had been carried out in Mainland Southeast Asia (none at all in Northeast Thailand), and it could truly be described as terra incognita (Higham 1972). Further, the region has the greatest abundance of tin in the world and ample supplies of copper ores, so the two metals used to make bronze were readily available. Hence, while we thought others might find our advocacy of an early and largely indigenous bronze technology there somewhat surprising, we also felt the presence of such a technology was not altogether unexpected and were ourselves in turn startled by the volume of debate that resulted.


Sadly, in spite of my optimistic statement in 1971, debate continues without signs of a clear resolution, although the chronological range of such debate has significantly narrowed. [Higham is convinced that the dates for bronze before ca. 1700 B.C. are wrong—WS] Given the large amount published and the number of contributors to the debate, I think a review of the arguments advanced for and against the chronological evidence is necessary, not only to see what these “extravagant and unfounded claims for early bronze and iron” (Higham et al. 1986-1987:155) actually were and who made them, but also as an example of the ebb and flow of a contentious “myth” in the human sciences in general. Debates over such myths-—whether it be Freeman versus Mead on Samoan adolescence, Robarchek and Dentan on the stereotype of the Semai as either “non-violent” or “bloodthirsty” (1987), or Martin’s debunking of the oft-quoted example of Eskimo snow terms (1986)-—have a usefulness to the profession beyond their specific content. They tell us something about the subjective convictions and preconceptions of anthropologists themselves. Workers in Southeast Asian archaeology obviously also have differing attitudes and preconceptions not only toward their specific field of study, but also on the more general question of diffusion versus independent invention. Solheim and I are no exception, and the following account obviously reflects our point of view. I do hope to show, however, that the presence of an early, largely indigenous bronze metallurgy in Southeast Asia is no myth. This is easy enough to do if one accepts “early” as meaning prior to ca. 500-800 B.C., the date commonly accepted in the 1960s; it is considerably more difficult to carry out if “early” means pre-2000 B.C.! I present only summaries of the arguments, with references to sources that will provide full documentation. Historical Review Non Nok Tha Stands Alone (1967-1973) When the first dates from Non Nok Tha were published in 1967, Solheim postulated a date somewhat earlier than 2300 B.C--“say about 2500 B.C.” (1967b:899)--for bronze working at the site, and an earlier date for its spread into Southeast Asia from (presumably) the Middle East. One of the goals of the 1968 excavation was to obtain more adequate charcoal samples and further dates to either confirm or refute the earlier ones. At first the combined dates from both excavations seemed to indicate two sequences: one of the eleven dates supporting a very early appearance of bronze in EP 3, and a second “late sequence” of five or six dates suggesting a traditionally late arrival of bronze at about 1000 B.C. I believed in 1969 that “the early sequence is the correct one based on the internal evidence from the site that is available to date, although I am certain that many will arrive at alternative conclusions from the date presented” (1970:134). Indeed, even at this early stage Pittioni, in a brief report on analyses of some of the bronze items, felt constrained to add parenthetically that, “they can hardly be put before the eight century B.C.” (1970:158). It was at this time that the first thermoluminescence (TL) dates on Ban Chiang red-onwhite pottery were made available to us by Achan Chin You-di (subsequently published by Bronson and Han 1972), introducing what ultimately turned out to be two red herrings into the debate: An astoundingly early date of 4630 B.C. for the few red-on-white ware, which were rumored to be associated with bronze; and a postulated similarity between Ban


Chiang red-on-white ware and the few red-on-white vessels recovered from EP 3 and MP 1 at Non Nok Tha. My view in 1969 was that a pre-3000 B.C. date was probable, but not conclusively proven (Bayard 1970, loc. cit.). However, this statement was qualified by a footnote reporting four new TL dates from the site that implied a somewhat later date for EP 1 and the first use of this site. In my short monograph on the excavation and chronology of Non Nok Tha, I gave a revised probable date of 2500-3000 B.C. for the presence of metallurgy (1971b:31). Solheim similarly dated the appearance of metallurgy at 2300-3000 B.C. in several publications (1969a:135; 1969:37; 1970a:224-225). Higham was more positive about “the recognition of a unique and indigenous bronze-working tradition by about 3500 B.C.” (Higham and Leach 1971:54). In 1972 the combined impact of the Ban Chiang TL dates and the application of secular corrections to the carbon dates led Solheim, Higham, and me to postulate an even earlier fourth millennium B.C. date for metallurgy in the region in general, with the agriculture earlier still (Solheim 1972:35, 41; Higham 1972:46; Bayard 1980b:197 [written 1971]). Subsequently I returned to a more “conservative radical” stance, believing it was “possible to say with a high degree of confidence that a well-developed bronze technology was present in mainland Southeast Asia prior to 2000 B.C.” (1972:1412). The chronology of bronze in Mainland Southeast Asia was discussed at considerable length at the London Colloquy on Early Southeast Asia in 1973, where I presented a paper examining in detail the Non Nok Tha dates, their proveniences, and possible contamination. I concluded that the site provided “not entirely conclusive” evidence for distinct bronze and iron periods rather than a single “Metal Period” and the first appearance of bronze there prior to 2000 B.C. (1979:25). Smith also provided the first of several careful and objective critiques of the Non Nok Tha dates, emphasizing the coherent group of five dates for MP 4-5 which suggested that “a bronze-using culture flourished in this part of mainland South East Asia by at the latest the second half of the second millennium B.C.” (1979a:40) In fact, general (but in some cases reluctant) agreement was reached at the 1973 Colloquy that bronze was present in the region by 1500 B.C.; I stated at the time and subsequently that “the tendency to support a pre-2000 B.C. date for bronze is quite clear, although admittedly not conclusive” (1975:168) Apparent Support: Additional Dates from Ban Chiang (1974-1980) In 1974 and 1975, extensive excavations at Ban Chiang were undertaken by the Thai Fine Arts Department and the University of Pennsylvania under the direction of Pisit Charoenwongsa and Chet Gorman, respectively. These seemingly revealed deep deposits containing bronze to the very bottom of the site at 3600 B.C. while also making it apparent that the now-famous red-on-white pottery was much later that the fourth millennium date suggested by the earlier TL dates. Loofs-Wissowa had also published a critique of the Ban Chiang and other Thai TL dates, correctly pointing out that it would be “unwise to take every such date as a cornerstone solid enough to support major hypothetical edifices” (Loofs 1974:61). Based on lengthy conversations and correspondence with Gorman and Pisit on the Ban Chiang sequence during the 1974-1975 Pa Mong survey, I grew more confident in the mid-1970s about the presence of bronze in the region prior to 3000 B.C. (1980a:105;


original version presented 1977).2 Higham was also influenced by the apparent evidence for bronze in the basal levels of Ban Chiang; in fact, he was actively involved in the Ban Chiang Program at this time and collected some of the charcoal samples from the site (Higham 1988a:75). While correctly pointing out that “the dating of Non Nok Tha has been both intrinsically difficult and controversial,” he nonetheless felt that “the recently obtained dates from Ban Chiang strongly support Bayard’s proposed early chronology [for Non Nok Tha]” (1975a:245-246). Similarly, his contribution to a symposium on the origins of agriculture held at this time supports the “acceptable hypothesis” that Non Nok Tha was first occupied in the fifth millennium B.C., with bronze technology appearing in the fourth millennium (Higham 1977a:387, Fig. 2; see also Higham and Kijngam 1979:230-231, 227; Higham et al. 1980:159, 149). Indeed, he suggested that the Ban Chiang “dates, with one or two exceptions, provided a much clearer and more consistent pattern than at Non Nok Tha, and suggest that Ban Chiang was occupied by the middle of the fourth millennium B.C.,” with bronze “in burials of the site’s earliest phase” (Higham 1979:674-675). Others were also expressing their skepticism during the middle end late 1970s. Clark stated in the third edition of his World Prehistory (1977) that “The published evidence is unimpressive. The radiocarbon dates for layers containing metal objects at the settlement mound of Non Nok Tha is as anomalous as one might expect of a sequence which to judge by the published section of the 1968 excavation was so heavily disturbed by burials, rootholes, termites and washouts” (as mentioned in Chapter 4, Non Nok Tha is by no means unique in its complexity). Seemingly more important for Clark, however, were the implications of the early dates: “If metallurgy was flourishing in Thailand by the middle of the fifth [sic] millennium, one may ask how it was that the Thai smiths should have marked time for four thousand years and why having gained so great a start in technologically on China and India the peoples of south-east Asia should have fallen so far behind these territories in the race to literate civilization” (Clark 1977:345-346). In addition to his “midfifth millennium” date for bronze, I find it very puzzling that Clark should ascribe such a casual role to metallurgy in Southeast Asia while in the same work he minimizes such a role for Europe (p.151). In the same year, Reed published a much more detailed and scholarly review of the chronological evidence for early agriculture in Southeast Asia, devoting a good proportion of it to the Non Nok Tha chronology. After an objective and fair assessment of the dates and problems in their interpretation, the only conclusion he could reach “with regard to the time of first occupation of Non Nok Tha (and the time of first evidence for rice in southeastern Asia) is that we do not know” (1977:915). He also correctly noted that in several cases others have interpreted my analyses in a far more positive light than I ever intended: “Stories do get changed in the telling, and my conclusion is that in this way many people have managed to convince themselves and others that valid evidence exists to support the concept of very early agriculture in Southeast Asia” (1977:914). However, he, like others—-including myself and Higham—-found himself positively influenced by the emerging evidence from Ban Chiang in the middle 1970s (Reed 1977:917, 953). But the Ban Chiang evidence proved to be far less conclusive than was then believed, and debate broke out afresh in the early 1980s. Before discussing this, however, we should turn briefly to the only rigorous statistical analysis carried out on the Non Nok Tha dates—-that of Macdonald in his 1980 thesis on


Non Nok Tha social organization and ceramics. He carried out ANOVA analyses on sets of dates deriving from each layer having two or more dates, eliminating those with F-values too high to have derived from the same population; he then combined these with single dates and eliminated those that seemed not to fit. The resulting sequence of twelve dates suggested the first occupation of the site at about 2000 B.C.; it is not possible to give a detailed comparison of his chronology for the site, as it is based on a phase system of his own derivation from the ceramic data, in some cases relying on an incorrect reading of the criteria used by others in analysis (see Chapter 6). In any case, once secular corrections are applied, his chronology suggests an initial occupation of the site in the mid-third millennium B.C., within the range that I currently believe to be most reasonable (2500-3000+ B.C.) Ban Na Di and Revision of the Ban Chiang Chronology (1981-1987) Thus, at the beginning of the 1980s the Non Nok Tha chronology and the possibility of very early bronze metallurgy were seemingly reinforced by the Ban Chiang radiocarbon dates (listed in detail in a 1978 paper given by Gorman and Charoenwongsa and published by Hurst and Lawn in 1984. The first telling attack on the credibility of the Ban Chiang chronology came from Stargardt (published in 1981, but pointed out to Higham at a 1978 seminar); she emphasized that in many cases the proveniences of the charcoal samples from Ban Chiang left much to be desired and in some cases were obviously contrary to the position assigned the date in the sequence. “Phases I and II are lumped together as spanning 3600-2900 B.C. . . . where the most notable artifact was a socketed bronze spearhead found in a flexed burial. Yet underneath this burial, a charcoal sample was found which gave a C14 date of 700 B.C. My point is that the association of all these samples with the burials cannot be established and we must look elsewhere for datable materials. (Stargardt 1981:333; italics author’s) Non Nok Tha also came in for criticism, based on my 1979 article: “For Bayard, the association of the charcoal samples is-—at Non Nok Tha—-established by depth and proximity, but his method allowed an astonishing latitude: ‘Given the rarity of charcoal at the site, several samples were submitted which consisted of scatters of small fragments present at the same depth in the layer rather than a single large piece. These were not collected over an area greater than 3 m2, however’ (p. 21).”3 This argument was taken up with considerably more precision and detail by Higham following his and Kijngam’s excavations at Ban Na Di in 1980-1981. This excavation is a significant one; not only did it overlap the latter part of the sequence at nearby Ban Chiang and provide the first in situ documentation of bronze-casting furnaces (as described in Chapter 7), but even more importantly, Higham and Kijngam were able to collect excellent charcoal samples in secure stratigraphic contexts. The results of the first five samples to be dated from the site suggested that “there is no evidence to sustain the exceptionally early chronology which has been advanced for Ban Chiang, but rather the data parallels recent chronologies obtained for the Vietnamese sites, where bronze was first found in the Phung Nguyen culture during the early 2nd millennium B.C.” (Higham 1981:2). Eight more samples were dated; four of these, from a single layer 7 pit, are clearly too late (dating the layer to the early centuries A.D.). The remaining four agreed closely with the original five and led Higham to conclude that the Gorman-Charoenwongsa chronology for


most of the Ban Chiang sequence was: too early by about a millennium” (Higham et al. 1982:21). Their data instead led them to “prefer figures of 1000-1750 B.C. for early bronze, and c. 400-200 B.C. for the appearance of iron” (Higham and Kijngam 1982:103, fn.). From the fourth or even fifth millennium dates suggested for the Ban Chiang red-on-buff ware by the earlier TL dates (of course much more moderately dated by the site’s excavators to 700-300 B.C.; Gorman and Charoenwongsa 1978), Higham and his colleagues now proposed a date beginning ca. A.D. 100. The proposed revision of the Ban Chiang sequence naturally had important implications for the dating of Non Nok Tha, which soon came under a considerably more exuberant attack on quite different grounds by Loofs-Wissowa (1983a,b). To my admittedly unobjective eye, he appeared to use four principal arguments: 1) We have yet to uncover clear evidence of a polymetallic experimental state—-specifically one using copper rather than bronze alloys-—in Southeast Asia (since disproven by Pigott and his colleagues’ research); 2) early dates for bronze in the region should be rejected because this does not cohere with the traditionally accepted early first millennium B.C. date; 3) radiocarbon dates are not as reliable as “archaeologic” dates; and finally 4) as outlined in the quote heading this chapter, Solheim, Gorman, and I are involved in a cabal with the journalistic media to misinform the world on this question. Owing to Gorman’s untimely death in 1981, Pisit Charoenwongsa and I replied to Loofs-Wissowa on his behalf as well as our own (Bayard and Charoenwongsa 1983); Solheim also made a separate reply (1983). This often rather nasty exchange unfortunately generated more heat than light; I will deal with the question of polymetallism in Chapter 7, and will deal with the myth of journalistic promulgation of early bronze below. It is difficult to see any purpose in considering his remaining two arguments; archaeology must of course deal with shifting paradigms, and radiometric dating is one of our basic tools. However, interested readers are invited to consult the full debate and arrive at their own conclusions. A much more cogent and coherent critique of the Non Nok Tha and Ban Chiang chronologies was put forth in the same year by Higham in his Mortimer Wheeler lecture (1984a). Following Stargardt’s earlier criticisms of the provenience of the Ban Chiang charcoal samples and by now no longer associated with that program, Higham devoted the first part of his overview of later Thai and Vietnamese prehistory to a similar, more detailed critique of the apparent inconsistencies and insecure proveniences of the Non Nok Tha dates and samples, as well as discrepancies between the Gorman-Charoenwongsa chronology for Ban Chiang (but omitting reference to White’s 1982 revision of the dating) and the well-provenanced and consistent dates from Higham’s own site of Ban Na Di. After taking into account the newly available dates from northern Vietnamese sites, he concluded that “the final review of the Ban Chiang radiocarbon dates will reveal the same problems of provenience and association as were encountered at Non Nok Tha” (1984:238). White’s catalog of the Smithsonian Ban Chiang Exhibition (1982) had in fact already pointed out that the site’s chronology and ceramic sequence was indeed in need of revision; although the overall time span of the site remained unchanged, there was no evidence for the presence of bronze in the basal levels. Rather, the earliest bronze artifact appeared at a date of 2000 B.C. or somewhat later, although some apparent inconsistence


in White’s dating of bronze at the site were pointed out in a review by Higham for Antiquity (1984e). The revision of the Ban Chiang sequence removed the chief prop in the earlier arguments put forth by Gorman, Solheim, Higham, and myself for a possible fourth millennium date for bronze. However, rather than adopting the rather radical revision of Higham down to an approximate mid-second millennium date for the arrival of bronze (Higham 1984a:fig.8), I returned to my earlier estimate of a later third millennium date. The reasons for this were outlined in two often productive exchanges between Higham and myself based on his Wheeler Lecture (Bayard 1984a, 1986-1987; Higham 1984b, 1986-1987); These concerned the specific questions of provenience and chronology of interest to us here, as well as the wisdom of Higham’s drawing of close parallels between Southeast Asia and the Aegean or Shang (to be discussed in the concluding chapter). Briefly, my arguments on the acceptability of the Non Nok Tha and Ban Chiang dates focused on Higham’s stringent criteria for the sources of carbon samples: “Clearly, then, all radiocarbon samples must come from in situ contexts such as hearths, charcoal from within bronze-casting furnaces, or wood cut from prehistoric coffins. Charcoal removed from the matrix of grave fill, or scattered charcoal collected from a given layer is useless” (1984:231). I pointed out that the contexts of most of the Vietnamese dates considered by Higham were as “valueless” as those from the two large Thai excavations: “I believe we are led toward two alternatives: either to consider (with qualifications, of course) all reasonable dates from Viet Nam, Non Nok Tha, Ban Chiang and other Thai sites; or to follow Higham’s stringent criteria and discard them all, and hence accept Higham’s 17 consistent dates as our only reliable witness from mainland Southeast Asia. At present I prefer the former alternative” (Bayard 1984a:5). Clearly Higham’s criteria are ideal ones that all archaeologists would prefer; as he said, “These guidelines are, in my view, nothing other than common sense” (1986-1987:145). However, carried to an extreme we would have to defer comment on Southeast Asian chronology for at least another decade or two to enable a corpus of totally secure samples to be dated, and I would agree with Bronson that “many of Higham’s fellow archaeologists . . . feel that his criteria for a valid date are unnecessarily strict” (1985:207) White’s doctoral thesis (1986; see also White 1988a) also addressed Higham’s critique of sample proveniences, discussing in some detail possible sources for the charcoal dated at Ban Chiang. She drew extensively on ethnographic analogy, including a review of funerary procedures that suggested the charcoal recovered from Ban Chiang graves may have been deliberately added and hence contemporary with the burial dated: “In short the ethnographic examples support the position that the grave-associated charcoal need not be arbitrarily dismissed as redeposited” (1986:164). I think both White and I would agree with Bronson when he wrote that “unless evidence to the contrary exists, we should be safe in assuming that he majority of radiocarbon dates derive from samples which reached their final resting place in the soil no more than a few decades after the death of the plant or animal from which the sample came.” (1985:207). Thus by about 1986-1987, a consensus had been reached among most workers on a late third to early second millennium B.C. date for bronze in Southeast Asia; early bronze was


no longer a myth, save in the minds of a few specialists who have carried out only limited fieldwork in the region. It is necessary, however, to turn briefly to two additional myths that have grown up around Non Nok Tha and Ban Chiang: the myth of suppression of data about the sites and the myth that promulgation of “early bronze” has been carried out largely through the popular printed media. In 1986, Watson remarked: “We had heard much of third--or second--millennium dates for bronze in inner Southeast Asia, but only bare C-14 dates circulated beyond the pale of a small band of scholars to whom all mention of external relations in East Asia, or worse, comparison with ‘old world” standards, was anathema” (1986:410). This is simply not the case as far as Non Nok Tha is concerned. Almost seven hundred printed pages devoted to the site and its relationships had appeared by that date (including three monographs), making it one of the most copiously published sites in the region.4 In addition, the controversial dates and their equally controversial proveniences had been publicly available from me in the form of a 5,000-word manuscript since 1984 (Bayard 1984a:5; 1986-1987:135; Higham 1986-1987:141). This wealth of published material (relative to sites like Ban Chiang and Watson and LoofsWissowa’s own Khok Charoen site) also belies the statement that Solheim, Gorman, and I have been waging a campaign using the popular press rather than more academic media, as gently suggested by Watson (1986) and stridently proclaimed by Loofs-Wissowa (as in the quote heading this chapter). Certainly Non Nok Tha and Ban Chiang have received their share of coverage in popular and semi-popular publications, but none of the articles were written by us, and they often contained exaggerations and errors that we were powerless to correct, although in some cases we certainly tried.5 More Recent Evidence and Further Fluctuations (1988-present) During the past five years the pendulum had continued to swing over the dating of Non Nok Tha, Ban Chiang, and the appearance of bronze in the region, but now within the narrower range of 1500-2500 B.C.. Higham and Bannanurag’s important major excavation at Khok Phanom Di, southeast of Bangkok, has provided a beautifully stratified and dated sequence of burials and occupational remains from 2000 to 1500 B.C., from which bronze is completely absent (Higham et al. 1986-1987; Higham and Bannanurag 1990). Of at lest equal importance has been the work of the Thailand Arhaeometallurgy Project (TAP) under the direction of Vincent Pigott and Surapol Natapintu, which has given us our first detailed insights into non-ferrous mining and metalworking techniques, first at Phu Lon in Northeast Thailand, and subsequently in the Khao Wong Prachan Valley near Lop Buri (Pigott and Natapintu 1988, 1996-1997; Pigott et al. 1992; see Chapter VII). The total absence of bronze at Khok Phanom Di led Higham to lower his earlier consensus date for bronze of 2000-1500 B.C. (Higham 1989:98-9) to a date of 1500 B.C. or later in a paper presented at the 1988 Kiola conference on ancient Chinese and Southeast Asian Bronze Age cultures (Higham 1996-1997). However, the arguments I presented in rebuttal there (Bayard 1996-1997), plus Pigott’s announcement at the conference of two seemingly secure dates of about 2300 B.C. at Non Pa Wai for a basal context containing a burial with a copper axe (plus a second burial with a pair of bivalve ceramic molds for casting a socketed axe), led me to continue to opt for a late third millennium date for bronze.


This appeared to gain conclusive support from Higham and Bannanurag’s 1991 excavation of burials containing bronze at Nong No (Nong Nor), near Khok Phanom Di. A series of five securely provenanced dates with narrow standard errors and midpoints at about 2450 B.C., in conjunction with Pigott’s two dates, led Higham to the conclusion that bronze must date back to at least 2500 B.C. in Southeast Asia (Higham and Bannanurag 1992:188). While solving one question, this of course raised the equally knotty problem of explaining the absence of bronze at Khok Phanom Di, only 14 km away and dating several hundred years later. However, further close examination of the Nong No stratigraphy in 1992, plus a series of accelerator dates on burial pottery, has suggested to Higham (1996:276) that the site’s burials date from ca. 1200-800 B.C., over a millennium later than the shell midden which provided the five third millennium dates. Judging from the beads illustrated in Higham and Bannanurag 1992, I would concur. Of equal importance is Pigott’s recent reinterpretation of the basal context of Non Pa Wai as presented at Kioloa in 1988; he now believes that copper production began in the Khao Wong Prachan Valey at about 1500 B.C., with the working of non-ferrous metals beginning in Thailand “in the first half of the 2nd millennium B.C.” (Pigott and Natapintu 1996-1997). Spriggs, a co-developer of the “chronometric hygiene” enthusiastically supported by Higham (1996:12), was kind enough to carry out this exercise on the Non Nok Tha dates after the Kioloa conference. He concluded that Non Nok Tha was probably initially occupied by 3,000 B.C. with bronze appearing some 500 years later (Spriggs 1996-1997; See discussion point 6 below). At the 15th Congress of the Indo-Pacific Prehistory Association held in Chiang Mai in 1994, Higham presented a vigorous verbal critique (still unpublished) of the early chronology arguments for Ban Chiang and Non Nok Tha dates. In his recently published Bronze Age of Southeast Asia (1996) he is more cautious, still strongly supporting a post-1500 B.C. date for the appearance of bronze in the region, but accepting the possibility that earlier dates may yet appear: “AMS dating of mortuary ceramics from Ban Chiang, no less than the excavation and dating of further sites, will contribute to this issue and it remains possible that dates earlier than 1500 BC will be obtained” (1996:240-241). In fact White published the first batch of such dates in the same year. These dates, although still small in number, do support a pre-1500 B.C. date for bronze at Ban Chiang (White 1996). I had hoped to be able to offer the readers of this report a fairly conclusive statement on the appearance of metallurgy in the region, but as the above makes clear, the question is obviously still very much open to debate. In the conclusions of my latest paper in the subject (Bayard 1996-1997, original version presented in 1988), I suggest that the prudent reader might wish to opt—-like Pigott and White—-for a range of 1500-2000 B.C. for the appearance of metallurgy in Southeast Asia, which appears to be about the closest we have to a consensus at present. Until still more data are published it would be imprudent of me to postulate anything more definitive. In conclusion, I hope that this fairly lengthy review of the debate over “early bronze” has at least shown that Higham’s “extravagant and unfounded” claims have been made by a range of scholars over the years, including Higham himself’ the debate also provides a very good example of just how inadequate and insecure our data from the region still are. In any case, it is now time to turn to the controversial Non Nok Tha dates themselves.


The Non Nok Tha Dates and their Proveniences In spite of the uncertainties and continuing debate, it is safe to conclude that metallurgy was present in Mainland Southeast Asia by at least the mid-second millennium B.C., and I continue to believe that it may well have arrived several hundred years prior to that date. While absolute dates and their varying interpretations have in my mind little to do with what I believe is a fairly secure relative stratigraphic sequence at Non Nok Tha, they are of course of considerable importance to regional (and indeed world) prehistory. Hence I hope that the detailed presentation and commentary that follows will bring out clearly the positive as well as negative aspects of the Non Nok Tha dates and illustrate the fact that the inconsistencies of the sequence are as much an artifact of dating technology as of the proveniences of the material dated. As most of the insecurely provenanced dates come from the 1966 excavation, it would be well to provide some background on our expectations and the resources available when we began excavating there in December 1965. Non Nok Tha was only one of several projects that remained to be cleared up in the final season of Solheim’s three-year program, and as a result we were operating under considerable time pressure. Trained personnel available consisted of Parker and myself, as Solheim was committed to other projects. Non Nok Tha was of course the first area excavation ever conducted in Northeast Thailand, and thus we had little idea of what to expect in the way of stratigraphy or finds. Parker had had extensive experience excavating in New Zealand with Jack Golson before being engaged to direct at Non Nok Tha, and he quite properly refused to excavate using the unit-level method applied at Ban Kao and in the previous year’s test pits at Non Nok Tha. As a result, the first month of the 1965-1966 season was devoted to slow, meticulous exploration of the site’s difficult stratigraphy, must of it in a single 2 x 3 m square (Square C3, at the approximate center of the mound. In hindsight, it would have been more fortunate if we had begun the excavation in a more marginal area that had seen less intensive use). The previous year’s test pits also led us to believe that we could expect several burials; I recall Parker and I hoping that we might recover as many as ten or twelve. As it turned out, we recovered eighty-nine. The pressures of recording and lifting these burials and associated artifacts were heavy, even though we were able to use the considerable drafting talents of some of the Silapakon University undergraduates who joined the excavation for part of its duration; we were also fortunate to receive some much-needed help from Professor Sood Sangvichien and his students during the busy final days before the onset of the rainy season. The richness and abundance of burials hence proved to be something of an embarrassment, and as described in Chapter 1, Solheim was forced to extend the deadline for completion of the excavation three times; even with the extra eight weeks this gave us, time and pressure was considerable. The necessity to obtain material for dating became very strong as well once it was known that Non Nok Tha-—unlike any other site previously excavated in Southeast Asia-—gave clear evidence of a distinct bronze period during which iron had not yet appeared.6 Unfortunately, charcoal was rare at the site, and


we were forced to rely on small fragments in many cases, both in the form of diffuse concentrations in levels (see note 3) and small pieces from burial pits. As I hope the discussion of the TF-651 date below indicates, we were aware of the problems involved with such samples; however, these collection methods were those in general use in Thailand at the time and for the next fifteen years (including some of the samples collected by Higham in his 1970 excavations-—Higham 1977b). When our first dates came back in 1966 and 1967, it was apparent to me that a lack of secure provenience was indeed a major problem. One of my objectives in the 1968 excavation was to obtain samples of more secure provenience, and I think that by and large I was successful in doing this. As I have discussed elsewhere (1979) and summarize below, however, the Florida State University series proved highly contradictory throughout, and the undoubtedly secure collagen and apatite dates compounded this. In short, I agree with Higham (1984a) that Non Nok Tha cannot in itself stand as an independent witness to third millennium bronze metallurgy. This, however, is not merely the result of hasty and inconsiderable submission of samples, but rather of the exploratory nature of the first excavation and dating problems beyond the control of the excavators. The problem is not unique to Non Nok Tha (cf. Higham’s aberrant dates from layer 7 pit at Ban Na Di mentioned above), nor to Thailand, and if we were to discard the Non Nok Tha evidence as failing to meet Higham’s strict criteria, many other Thai dates and a large number of those from Viet Nam and China would doubtless have to go as well. As stated above, I continue to believe we should consider-—with due evaluation of relative reliability, or course-—all available evidence, including that from Non Nok Tha. In any case, here are the proveniences of the Non Nok Tha dates, “warts and all,” as I presented them in the 1984 memorandum announced in South-East Asian Studies Newsletter 16:5, with commentary from my 1968 and Parker’s 1966 field notebooks and my own conclusions on their adequacy. The two subsequent bone dates from Arizona and the eleven accelerator dates on potsherds form Oxford referred to in this paper have also been added, to bring the total of dates from the site to fifty. The dates are presented in this order: charcoal, sherd organic content, bone, and thermoluminescence for each year’s excavation, alphabetized by the following laboratory abbreviations: AA FSU GaK GX I N OxA PT TF Y

University of Arizona (1968 only; bone) Florida State University (1968 only) Gakushin University (1966 only) Geocron Laboratories (1968 only) Teledyne Isotopes (1968 only; bone) Rikagaku Kenyusho (1968 only; bone) Oxford University Research Laboratory (1968 only; accelerator) University of Pennsylvania (thermoluminescence) Tata Foundation (1966 only) Yale University

Several corrections are supplied as well as the basic Libby half-life and 1-sigma error; these are provided merely to facilitate comparisons with other series of dates published earlier, not to imply a false sense of precision. They are the corrected 5730 half-life with 1sigma error in calendar years; MASCA secular correction; secular correction and 2-sigma


error range published by Klein et al. (1982); and finally, the corrections arrived at through application of the University of Washington computer calibration program distributed by Pearson and Stuiver (1986). At this last correction in the form similar to Klein at al., a 1sigma range is given in the form recommended by Stuiver and Pearson: AD/BC lower-limit (midpoint[s]) upper-limit. The error ranges in the uncorrected dates are those supplied by the various laboratories; no multiplier was used in the Pearson-Stuiver corrections. Lowercase bp, bc, and ad are used for the uncorrected dates. Non Nok Tha Radiocarbon and Thermoluminescence Dates 1966 Dates GaK-908 1966 LAYER 9 SQUARE D6 NW? LEVEL: LP 2 Wt. 3 g7 BAG NO. none8 DATE COLLECTED: 26/2/66? DATE SUBMITTED: 8/3/66 Provenience: Probably collected from “potters’ workshop,” a burned and collapsed cooking platform of layer 9 provenience lying on layer 11 surface. NB VIII, 26/2: “5/ clay in NW quadrant is associated with broken pottery and charcoal.” Sample picked up by WGS on 28/2. (5568)710±90 bp (5730) 1220±90 ad MASCA CORRECTION: 1220±90 AD KLEIN CORRECTION (2 sigma): 1195-1400 AD PEARSON/STUIVER CORRECTION (1 sigma): AD 1244 (1277) 1379 GaK-956 1966 LAYER 19SQUARE D5 LEVEL: MP 3Wt. 3 g BAG NO. none8 DATE COLLECTED: 22/3/66 DATE SUBMITTED: 14/4/6 PROVENIENCE: NB IX 3/22: “12/ d5: Charcoal-fair quantity.” See note 2/ of same date for discussion of level; also cf. TF-651 date on charcoal in bottom of grave of 19 B. 55 in adjacent Square E5. 5568) 4120±90 bp (5730) 2290±90 bc MASCA CORRECTION: 2800±90 BC KLEIN CORRECTION (2 sigma): 290-2410 BC PEARSON/STUIVER CORRECTION (1 sigma): BC 2883 (2867, 2817, 2693, 2684, 2662, 2633, 2626) 2510 GaK-957 1966 LAYER 13 SQUARE E5 LEVEL: MP 8Wt. 3 g BAG NO. none8 DATE COLLECTED: 4/3/66 DATE SUBMITTED: 14/4/66 PROVENIENCE: NB VII, 4/3/66: “2/ E5-removing top of layer 19? Layer 13?” NB VII 5/3: “18/ E5-layer 13,18, or 19 as case may be. This is obviously not a reliable date. 5568) 1780 bp (5730)180±80 ad MASCA CORRECTION: 240±80 AD KLEIN CORRECTION (2 sigma): 45-565 AD PEARSON/STUIVER CORRECTION (1 sigma): AD 227 (265, 281, 333) 412 Gak-958 1966 LAYER 9 SQUARE D7 LEVEL: LP 2 Wt. 30 g BAG NO. none8 DATE COLLECTED: 23/3/66 DATE SUBMITTED: 14/4/66


PROVENIENCE: NB IX 23/366: “1/ D7-removing 11.” “7. D7-two layer 9 postholes with charcoal. Both postholes filled with charred wood. Postholes 170 cm apart at centers.” The early date is unexplained, despite secure provenience. 5568) 2220±110 bp (5730) 340±110 bc MASCA CORRECTION: 320±110 BC KLEIN CORRECTION (2 sigma): 545-20 BC PEARSON/STUIVER CORRECTION (1 sigma): BC 400 (362, 282, 258) 124 GaK-595 1966 LAYER 21? SQUARE C4 LEVEL: EP 2 Wt. 2 g BAG NO. none8 DATE COLLECTED: 4/3/66 DATE SUBMITTED: 14/4/66 PROVENIENCE: NB VII 4/3/66: “3/ C4-some charcoal in ‘natural’ below level of ‘chief’s burial.’ 4/ C4-fair quantity of charcoal in pit 130-170 A.D.” NB VII 4/3: “2/ C4-going down in possible deep disturbance. . . 7/ C4-charcoal still coming up in deep cut being made. The actual ‘chief’s burial’ lay well below the level reached by this date; the goods referred to by this term at that time later proved to be grave goods at foot of B. 10 grave. In any event, it seems apparent that this sample originated in a post-layer 21 pit, possibly B. 10 or B. 9; it is in any case an unreliable date (cf. GaK-1033). (5568) 1860±140 bp (5730) 30±140 bc MASCA CORRECTION: 112±140 AD KLEIN CORRECTION (2 sigma): 165 BC-435 AD BC PEARSON/STUIVER CORRECTION (1 sigma): AD 0 (128) 333 ±GaK-1027 1966 LAYER 9 SQUARE D6, SE LEVEL: LP 2Wt. 10 g BAG NO. none8 DATE COLLECTED: 3-7/3/66 DATE SUBMITTED: 24/6/66 PROVENIENCE: NB VII 3/3/66: “12/ D6-several large pieces of charcoal clearly contemporary with body in NE corner and ‘potters’ workshop.” NB VIII 7/3: “cleaning up last layer 9, possibly into 10 in places.” Early date is unexplained, despite secure provenience. (5568) 2480±80 bp (5730) 600±80 bc MASCA CORRECTION: 662±80 BC KLEIN CORRECTION (2 sigma): 810-395 BC PEARSON/STUIVER CORRECTION (1 sigma): BC 793 (760, 684, 656, 638, 592, 586, 550) 408 GaK-1028 1966 LAYER 11 SQUARE C7 LEVEL: LP 1 Wt. 1.5 g BAG NO. none8 DATE COLLECTED: 10/3/66 DATE SUBMITTED: 24/6/66 PROVENIENCE: NB VIII 10-12/3/66: “removing layer 11 in C7.” NB VIII, 13/3: “6/ bottom of 11-11 confirmed 12/6/66.” Sample was in layer 11, but clearly upwardly intrusive from MP. (5568) 2560 100 bp (5730) 690 ±100 bp MASCA CORRECTION: 780 100 BC KLEIN CORRECTION (2 sigma): 875-415 BC PEARSON/STUIVER CORRECTION (1 sigma): BC 819 (793) 540 GaK-1029 1966 LAYER 17/18 (17?) B. 48 SQUARE E5 LEVEL: MP 4(5?) Wt. 10 g BAG NO. none8 DATE COLLECTED: 10/4/66 DATE SUBMITTED: 24/4/66 PROVENIENCE: Charcoal apparently collected by R. Tirot during lifting of this burial by Sood and students-cf. letter from RHP to WGS of 21/10/66. Provenience is obviously not reliable.


(5568) 2830±100 bp (5730) 960±100 bc MASCA CORRECTION: 1080±100 BC KLEIN CORRECTION (2 sigma): 1315-800 BC PEARSON/STUIVER CORRECTION (1 sigma): BC 1154 (998) 900 Gak-1030 1966 LAYER 18 SQUARE C5 LEVEL: MP 3Wt. 8 g BAG NO. none8 DATE COLLECTED: 1/28-15/2/66 DATE SUBMITTED: 24/6/66 PROVENIENCE: NB VI, 15/2/66: “15/ C5-sherd of white porcelain (or piece of polished shell) probably in same pit as more easterly of two pots exposed today. Both these are probably in burial (?) pits but very possibly not in same pit.” No mention of charcoal being collected, but it seems likely the sample originated in an LP disturbance. (5568) <250 bp (5730) 1693 ad-present KLEIN CORRECTION (2 sigma): ca. 1400 AD-present PEARSON/STUIVER CORRECTION (1 sigma): — Gak-1031 1966 LAYER 18 SQUARE C4 LEVEL: MP 3Wt. 2 g BAG NO. none8 DATE COLLECTED: 13-16/4/66 DATE SUBMITTED: 24/6/66 PROVENIENCE: NB VI, 13/2/66: “14/ C5-fair quantity of charcoal from 19 surface. NB VI, 14/2: “C4 layer 18 coming off exposing 19.” No mention of charcoal in C4 during time sample was collected, but charcoal was also present in adjacent squares D5 and E5 in layer 19; cf. GaK-956 and TF-651. Late date is unexplained. (5568) 2530±20 bp (5730) 660±120 bc MASCA CORRECTION: 750±120 BC KLEIN CORRECTION (2 sigma): 840-410 BC PEARSON/STUIVER CORRECTION (1 sigma): BC 810 (777) 410 GaK-1032 1966 LAYER 20 B. 68 SQUARE D4 LEVEL: MP 2Wt. 0.8 g BAG NO. none8 DATE COLLECTED: 9/4/66 DATE SUBMITTED: 24/6/66 PROVENIENCE: NB XI 9/4/66: “18 D7-mound [B. 68] in SW corner, pit in SE quadrant. Charcoal in fill.” the sample is thus not from B. 68, but a post-layer 20 pit. Obviously unreliable. (5568) 8100±250 bp (5730) 6390±bc MASCA CORRECTION: — KLEIN CORRECTION (2 sigma): — PEARSON/STUIVER CORRECTION (1 sigma): ?? Gak-1033 1966 LAYER 18 B. 9/17 B. 10? SQUARE C4 LEVEL: MP 4/5 Wt. 1.5g BAG NO. none8 DATE COLLECTED: 3-5/3/66 DATE SUBMITTED: 24/6/66 PROVENIENCE: Letter from RHP to DTB of 12/5/69: “GaK-1033 is not from [Layer 21] B. 6 . . . it is in a later disturbance cut across the end of this grave . . . It is therefore from some period post-layer 21.” Both B. 9 and B. 10 cut into B. 6 grave, but sample is clearly not well provenanced; cf. GaK-959. (5568) 2990±110 bp (5730) 1130±110 bc MASCA CORRECTION: 1316±110 BC KLEIN CORRECTION (2 sigma): 1520-900 BC PEARSON/STUIVER CORRECTION (1 sigma): BC 1410 (1261) 1040 GaK-1034


1966 LAYER 21 B.26 SQUARE D4 LEVEL: EP 3 Wt. 5 g BAG NO. none8 DATE COLLECTED: 26/4/66 DATE SUBMITTED: 24/6/66 PROVENIENCE: NB XI, 26/4, note 5/ discusses this grave, but makes no mention of charcoal being collected. In any event the sample is probably intrusive from a considerably earlier event; cf. 1968 date FSU-340 (5568) 5370±320 bp (5730) 3590±320 bc MASCA CORRECTION: 4273±320 BC KLEIN CORRECTION (2 sigma): 4730-3660 BC PEARSON/STUIVER CORRECTION (1 sigma): BC 4530 (4238) 3816 TF-651 1966 LAYER 19 B. 55 SQUARE E5 LEVEL: MP 3Wt. 8 g BAG NO. 1810e DATE COLLECTED: 20-26/4/66 DATE SUBMITTED: 28/6/66 PROVENIENCE: Collected by R. Tirot while clearing and recording lower part of B. 55. NB XI, 24/4/66, note 8/ mentions carbonized seeds. Unfortunately, photos and any drawings made of this burial were lost. In view of Higham’s comments pointing out the obvious problems in dating charcoal in graves, it is worthwhile to quote at some length from a letter of 21/10/66 from RHP to WGS, if only to show that the excavators were well aware of the problem at that time: Naturally charcoal from the fill of a burial can never be wholly secure . . . the making of the [B. 55] grave could have picked up charcoal from the surface of either Layer 20 or Layer 21. It weighs however somewhat against this that no charcoal or patches of blackened soil were detected on the surface of either layer in Square E5. On the other hand, the surface of Layer 19 right across the western side of this square did have a large patch of blackened soil which contained a few flecks of charcoal and a number of small irregular nodules of bronze. Burial 55 was cut through this and the black patch is therefore a very likely source of the charcoal contained in the fill. If one had to guess at the odds I would say they were at least 3 to 1 in favor of a Layer 19 origin for the sample. Also cf. date GaK-956 on layer 19 charcoal in adjacent square D5, and layer 10 charcoal in adjacent squares C5 and C4 (GaK- 1031), although this last produced a considerably later date. (5568) 4155±200 bp (5730) 2325±200 bc MASCA CORRECTION: 2807±200 BC KLEIN CORRECTION (2 sigma): 3350-2310 BC PEARSON/STUIVER CORRECTION (1 sigma): BC 3015 (2868, 2807, 2773, 721, 2700) 2470 PUBLISHED Radiocarbon 10(1), 1968 Y-1851 1966 LAYER 17/18 B. 78 SQUARE: G4 LEVEL: MP 4 WT. ? BAG NO. 1810e DATE COLLECTED:20-26/4/66 DATE SUBMITTED: 28/6/66 PROVENIENCE: Probably collected sometime in above date range, but I can find no reference to the sample in RHP’s field notebook entries for these dates. Nor was WGS able to retrieve information on weight and submission date of the sample from the now closed Yale laboratory. (5568) 3170±200 bp (5730) 1310±200 bc MASCA CORRECTION: 1595±200 BC KLEIN CORRECTION (2-sigma): 1875-910 BC


PEARSON/STUIVER CORRECTION (1-sigma): BC 1680 (1439) 1221 PUBLISHED Radiocarbon 11 (4), 1969 PT-? (TL) 1966 LAYER 17 SQUARE: G4 LEVEL: MP 4 WT. ? BAG NOs. 868.696 DATE COLLECTED:4/3/66 DATE SUBMITTED:late 1973 PROVENIENCE: Two matching crucible fragments (NP-807); one recorded at layer 17 surface (NB VII 4/3/66, notes 3/ and 14/), the second found during sherd washing the same day. Sherds are heavily tempered with rice chaff. Date supplied in letter of 7/3/74 from Carriveau to WGS; no PT-number assigned. Wide standard error makes it of quite limited quality. 3070±520 bp 1120±520 bc 2-sigma range 2160-80 BC PUBLISHED Carriveau 11 (4), 1969 FSU-339 1968 LAYER 5 firepit SQUARE 3E/3F LEVEL: LP 3 WT. 16.2 g BAG NO. 1479e DATE COLLECTED: 21/5/68 DATE SUBMITTED: 4/10/69 PROVENIENCE: from lower layer 5 firepit, charcoal, bamboo ash lens shown on F/S section; also see NB II, 19-20. Cf. GX-1609 on same material. MODERN + 0.07 counts/min/g FSU-340 1968 LAYER 11 SQUARE 4G LEVEL: pre-EP 1? WT. 24.5 g BAG NO. 517e DATE COLLECTED: 27/3/68 DATE SUBMITTED: 4/10/69 PROVENIENCE: Large lump located 12 cm deep in layer 11; not a root, and not in any detectable disturbance. Measured 6 x 10-12 cm. “This seems to be the best, most secure early sample we’ve gotten at NP 7. In fact, the chief worry is that it’s too early, as 11 is sterile and 10 almost so . . . However, this should make a much better sample for establishing an ante quem than that mentioned on p. 145 [post quem was of course intended].” NB I, 150. Given the absence of any associated cultural material, this sample may predate the EP occupation; cf. 1966 date GaK-1034. (5568) 4435 130 bp (5730) 2620 130 bc MASCA CORRECTION: 3195-130 BC KLEIN CORRECTION (2-sigma): 3510-2865 BC PEARSON/STUIVER CORRECTION (1-sigma): BC 3350 (3092, 3062, 3044) 2920 PUBLISHED Radiocarbon 13 (1), 1971 FSU-341 1968 LAYER 7 SQUARE 4F, SE LEVEL: MP 1 WT. 8.8 g BAG NO. 325e DATE COLLECTED: 11/3/68 DATE SUBMITTED: 4/10/69 PROVENIENCE: “Down to clear level 7 in SE corner.” NB I, 107. Sample from concentration of charcoal in layer. (5568) 2470±bp (5730) 590±140 bc MASCA CORRECTION: 632±140 BC KLEIN CORRECTION (2-sigma): 850-210 BC PEARSON/STUIVER CORRECTION (1-sigma): BC 800 (757, 689, 651, 648, 543) 400 PUBLISHED Radiocarbon 13 (1), 1971 FSU-342 1968 LAYER 8U SQUARE 1E, NW LEVEL: EP 1 WT. 18.6 g BAG NO. 1173e DATE COLLECTED: 7/5/68 DATE SUBMITTED: 4/10/69


PROVENIENCE: “A good charcoal sample right on 8 surface ca. 10 cm below 6A B. 60.” NB I, 229. This could conceivably date the 7/8 interface, but layer 8 was reconfirmed in notebook after final sections were drawn and rechecked against excavation levels. (5568) 3055±130 bp (5730) 1200±130 bc MASCA CORRECTION: 1403±130 BC KLEIN CORRECTION (2-sigma): 1670-900 BC PEARSON/STUIVER CORRECTION (1-sigma): BC 1450 (1379, 1343, 1319) 1127 PUBLISHED Radiocarbon 13 (1), 1971 FSU-343 1968 LAYER 4 lime kiln SQUARE 1E, NE LEVEL: LP 4 WT. 32.4 g BAG NO. 930e DATE COLLECTED: 24/4/68 DATE SUBMITTED: 4/10/69 PROVENIENCE: “Half sectioned and photographed lime kiln in NE corner. The pit, with lime and charcoal, shows nicely.” NB I, 198 MODERN + 0.2 counts/min/g PUBLISHED Radiocarbon 13 (1), 1971 FSU-345 1968 LAYER 9 B. 125 SQUARE 3E mid LEVEL: EP 1 WT. 13.7 g BAG NO. 1344e DATE COLLECTED: 14/5/68 DATE SUBMITTED: 4/10/69 PROVENIENCE: Charcoal from immediately west of Pot 4, NB I, 241-2; pot V on Drawing 42. In EP 1 grave pit dug into sterile layer 11; sample layer hence cannot derive from earlier occupation. (5568) 3560±130 bp (5730) 1720±130 bc MASCA CORRECTION: 2063±30 BC KLEIN CORRECTION (2-sigma): 2315-1655 BC PEARSON/STUIVER CORRECTION (1-sigma): BC 2131 (1908) 1971 PUBLISHED Radiocarbon 13 (1), 1971 GX-1609 1968 LAYER 5 firepit SQUARE 3F LEVEL: MP 4WT. 2.2 g BAG NO. 187e DATE COLLECTED: 1/3/68 DATE SUBMITTED: 3/6/69 PROVENIENCE: See drawing, NB I, 70, and NB II, 19-20. Bamboo charcoal, same material as dated by FSU-339 (5568) 500 ±85 bp (5730) 1430 ±85 ad MASCA CORRECTION: 1417 ±85 AD KLEIN CORRECTION (2-sigma): 1310-1515 AD PEARSON/STUIVER CORRECTION (1-sigma): AD 1625 (1422) 1450 GX-1611 1968 LAYER 6B B. 15 pit SQUARE 4H LEVEL: MP 4WT. 2.2 g BAG NO. 343e DATE COLLECTED: 12/3/68 DATE SUBMITTED: 3/6/69 PROVENIENCE: NB I, 111, makes it apparent that sample came from pit above head region, but charcoal not mentioned there or shown on Drawing 8; material was relatively large (0.5 cm) fragments of dense, shiny charcoal. As this burial was dug into an MP 1 pit, charcoal probably dates from MP 1-4, but no earlier, since layers 8 and 9 were very thin here; see 4/E section. (5568) 3685 ±110 bp (5730) 1840 ±110 BC MASCA CORRECTION: 2179±110 BC KLEIN CORRECTION (2-sigma): 2510-1865 BC PEARSON/STUIVER CORRECTION (1-sigma): BC 2276 (2127, 2077, 2043) GX-1612


1968 LAYER 9, bottom of pit L SQUARE 4F, SE LEVEL: EP 1 WT. 2.6 g BAG NO. 764e DATE COLLECTED: 18/4/68 DATE SUBMITTED: 3/6/69 PROVENIENCE: NB I, 171 shows a shell-filled pit dug from ca. 5 cm below layer 9 surface. (5568) 2750±130 bp (5730) 880±130 bc MASCA CORRECTION: 964±130 BC KLEIN CORRECTION (2-sigma): 1265-610 BC PEARSON/STUIVER CORRECTION (1-sigma): BC 1040 (905) 810 Y-2485 1968 LAYER 6 B. 95 SQUARE 2E/3E LEVEL: MP 5WT. 38.5 g BAG NO. 1619e DATE COLLECTED: 25/5/68 DATE SUBMITTED: 27/8/69 PROVENIENCE: Charcoal under bones, Drawing 32. “A good carbon sample from under the ribs and vertebrae of this burial: at least 40-50 g in solid layer 6 fill. Only right arm and right 1.2 of chest of this burial remain, but carbon is from under this.” NB II, 26. Sample would seem to be too large a piece to have been dug up and redeposited whole; in any event, surrounding layer 6 matrix indicates it dates 10 either MP 5 or MP 4. (5568) 3090±120 bp (5730) 1230±120 bc MASCA CORRECTION: 1476±120 BC KLEIN CORRECTION (2-sigma): 1650-1105 BC PEARSON/STUIVER CORRECTION (1-sigma): BC 1510 (1400) 1220 OxA-2383 1968 LAYER 9 pit “B. 65” SQUARE 4F, SE LEVEL: EP 1 WT. >1 mg CAT. NO. NNT-476 DATE COLLECTED: 22/4/68 DATE SUBMITTED: 4/10/ PROVENIENCE: “Single sherd with rice husk imprint”9 with ground clay and chaff temper from identified pit apparently dug from layer 9 surface (NB I, 192, 195). This was later (1/70) identified as being a NW extension of the empty pit of EP 1 B. 65 in the Non Nok Tha catalog. Although I did not submit the sherd, it was selected from bag 835A and was one of a group of several similarly decorated sherds found about two-thirds of the way down in the fill of this pit, which contained no skeletal remains. In late 1988 I wrote to Harris that it should provide a fairly secure date for initial use of site, although neither from a provenanced burial per se nor from a substantially complete vessel. However, although ground clay temper is relatively common from EP 1 through MP 1-2, only two substantially whole vessels with rouletted spirals were recovered from Non Nok Tha (NP-523 and NP-1723); these were both in late EP or early MP burials. If the pit originated from layer 8 (a shallow layer of 12-20 cm in this location) the sherd dated could thus be later than EP 1. Nevertheless, this is one of only two samples in the OxA-series presented here whose provenience is relatively secure. (5568) 3650 ±90 bp (5730) 1810 ±90 bc MASCA CORRECTION: 2144 ±90 BC KLEIN CORRECTION (2-sigma): 2325-1760 BC PEARSON/STUIVER CORRECTION (1-sigma): BC 2183 (2034) 1907 PUBLISHED Archaeometry 33 (2), 1991 OxA-2384 1968 LAYER 9 SQUARE 4G LEVEL: EP 1 WT. >1 mg CAT. NO. NNT-471 DATE COLLECTED: 7/3/68 DATE SUBMITTED: 4/10/88 PROVENIENCE: To layer only, in bag 247A: “Clay lump w/ charred bone fragment. 3 sherd pieces w/ poss. rice chaff.”9 No vessels with a fabric resembling the fine chaff and sand temper of these sherds were recovered from the EP; as I noted in the catalog entry written


in 1969, “chaff in Level IV [MP 1] type 71134 sherd.” The number refers to the fine chaff and sand temper, and it is highly likely that the sample was downwardly displaced from the MP. (5568) 3250±100 bp (5730) 1400±100 bc MASCA CORRECTION: 1600±100 BC KLEIN CORRECTION (2-sigma): 1760-1365 BC PEARSON/STUIVER CORRECTION (1-sigma): BC 1670 (1521) 1991 PUBLISHED Archaeometry 33 (2), 1991 OxA-2385 1968 LAYER 8X SQUARE 1F/2F LEVEL: EP 1-2? WT. >1 mg CAT. NO. NNT-453 DATE COLLECTED: 10/5/68 DATE SUBMITTED: 4/10/88 PROVENIENCE: To layer only, in bag 1210A; recovered from layer 8/9 interface. “Sherd w/ husk imprint. Stored in cotton packing at least 10 years.” 9 As with the dates OxA-2386 and OxA2391 below, obviously neither provenience nor first millennium AD date is reliable. To quote letter (7/9/90) from Oxford Laboratory to David Harris (Harris pers. comm.), “the young dates may be due to intrusive carbon but whether it is from glue, or from packing material, burial environment, etc, I don’t know.” (5568)±1110 bp (5730) 800±70 bc MASCA CORRECTION: 858±70 AD KLEIN CORRECTION (2-sigma): 650-1135 AD PEARSON/STUIVER CORRECTION (1-sigma): AD 880 (900, 902, 953) 1000 PUBLISHED Archaeometry 33 (2), 1991 OxA-2386 1968 LAYER 8 SQUARE 1E LEVEL: EP 2/3? WT. >1 mg CAT. NO. NNT-443 DATE COLLECTED: 14/5/68 DATE SUBMITTED: 4/10/88 PROVENIENCE: To layer only, in bag 1339A; “Broken sherd reveals some carbonized chaff”.9 The tempering material (which the catalog suggests was rice chaff and stalks, plus some sand, although I had no opportunity to examine the sherd to dating) is atypical of either the EP or early MP, and the provenience and date are clearly not reliable. (5568) 1110±70 bp (5730) 800±70 bc MASCA CORRECTION: 671±80 AD KLEIN CORRECTION (2-sigma): 580-895 AD PEARSON/STUIVER CORRECTION (1-sigma): AD 654 (681) 786 PUBLISHED Archaeometry 33 (2), 1991 OxA-2387 1968 LAYER 8 SQUARE 2F LEVEL: EP 2/3? WT. >1 mg CAT. NO. NNT-438 DATE COLLECTED: 1/4/68 DATE SUBMITTED: 4/10/88 PROVENIENCE: Provenanced to layer only, in bag 664A; “5 sherd fragments contain rice chaff and other plant fibers.”9 As I noted in 1969 in the catalog, “two very good chaff samples in 71134 [fine chaff and sand temper] sherds (these could be intrusive from layer 7 [MP 1])”. Like OxA-2384 above, I believe this sherd originates from MP rather than EP levels. (5568) 2950±80 bp (5730) 1090±80 bc MASCA CORRECTION: 1244±80 BC KLEIN CORRECTION (2-sigma): 1415-880 BC PEARSON/STUIVER CORRECTION (1-sigma): AD 1310 (1211, 1180, 1165) 1034 PUBLISHED Archaeometry 33 (2), 1991 OxA-2388 1968 LAYER 7 SQUARE 2F LEVEL: MP 1? WT. >1 mg


CAT. NO. NNT-321 DATE COLLECTED: 29/3/68 DATE SUBMITTED: 4/10/88 PROVENIENCE: To layer only, from bag 560A; “Single sherd w/ rice husk temper”9 Fine chaff and sand temper is characteristic of the MP, but provenience is obviously unreliable. (5568) 2880±80 bp (5730) 1020±80 bc MASCA CORRECTION: 1144±80 BC KLEIN CORRECTION (2-sigma): 1370-820 BC PEARSON/STUIVER CORRECTION (1-sigma): BC 1253 (1044) 932 PUBLISHED Archaeometry 33 (2), 1991 OxA-2389 1968 LAYER 7 SQUARE 3F LEVEL: MP 1? WT. >1 mg CAT. NO. NNT-324 DATE COLLECTED: 29/3/68 DATE SUBMITTED: 4/10/88 PROVENIENCE: To layer only, from bag 497A; “Single sherd w/ rice husk temper”9 This date clearly suffers from the same problems as OxA-2388. (5568) 2920±80 bp (5730) 1060±80 bc MASCA CORRECTION: 1214±80 BC KLEIN CORRECTION (2-sigma): 1395-845 BC PEARSON/STUIVER CORRECTION (1-sigma): BC 1264 (1124, 1113, 1108) 1003 PUBLISHED Archaeometry 33 (2), 1991 OxA-2390 1968 LAYER 6B pit SQUARE 4F, E LEVEL: MP 4WT. >1 CAT. NO. NNT-294 DATE COLLECTED: 10/3/68 DATE SUBMITTED: 4/10/88 PROVENIENCE: Recovered from Pit 1, NB I, 100, in bag 309A; “Broken sherd contains rice husk or chaff.”9 I would rank this with the two relatively secure dates OxA-2383 and OxA-2392, except that it is on a single sherd rather than a whole or partial vessel; hence the possibility of displacement is increased. (5568) 2920±80 bp (5730) 1060±80 bc MASCA CORRECTION: 1214±80 BC KLEIN CORRECTION (2-sigma): 1395-845 BC PEARSON/STUIVER CORRECTION (1-sigma): BC 1264 (1124, 1113, 1108) 1003 PUBLISHED Archaeometry 33 (2), 1991 OxA-2391 1968 LAYER 6X SQUARE 2E LEVEL: MP 1-4? WT. >1 mg CAT. NO. NNT-303 DATE COLLECTED: 22/3/68 DATE SUBMITTED: 4/10/88 PROVENIENCE: To the layer 6/7 interface only, in bag 393A; “3 sherd fragments with rice husk or other plant remains.”9 Obviously not a reliable date; see OxA-2385 and OxA-2386 above. (5568) 1320±100 bp (5730) 590±100 ad MASCA CORRECTION: 646±80 BC KLEIN CORRECTION (2-sigma): 575-885 BC PEARSON/STUIVER CORRECTION (1-sigma): AD 640 (673) 790 PUBLISHED Archaeometry 33 (2), 1991 OxA-2392 1968 LAYER 6B B.3 SQUARE 4H, E LEVEL: MP 4WT. >1 mg CAT. NO. NNT-375 DATE COLLECTED: 7/3/68 DATE SUBMITTED: 4/10/88 PROVENIENCE: Pot in bag 265A. This is the most securely provenanced date in the OxAseries, originating from “carbonized rice grain from the whole pot. Stored w/ cotton packing; 10 yrs.”9 The pot in question (NNT-375, a type 2L vessel) was clearly placed beyond the head of MP 4 B.3, along with a second vessel and a clamshell beneath both


(NB I, 78); however, the rice grain was apparently discovered during the vessel’s reconstruction, and “may not be from sherd temper.” Nonetheless, it is the only sample the provenience of which fully satisfies the criteria I desired in this series of dates. The effect of storage of the sherd in cotton within a plastic vial for over ten years is of course an open question. (5568) 3065±70 bp (5730)1210±70 bc MASCA CORRECTION: 1451±70 BC KLEIN CORRECTION (2-sigma): 1585-110 BC PEARSON/STUIVER CORRECTION (1-sigma): BC 1422 (1385, 1339, 1323) 1262 PUBLISHED Archaeometry 33 (2), 1991 OxA-2393 1968 LAYER 6U SQUARE 3F LEVEL: MP 4/5? WT. >1 mg CAT. NO. NNT-249 DATE COLLECTED: 13/3/68 DATE SUBMITTED: 4/10/88 PROVENIENCE: To upper layer 6 only, in bag 356A. “Best example of carbonized rice chaff in sherd temper.”9 Unfortunately the fabric type of this sherd was not recorded prior to being sent for dating; according to the catalog, it could have been one of three types (fine sand and chaff, prepared clay and chaff, or chaff and fiber). However, none of these three are at all common during the EP, and the sherd probably originated from the MP. Obviously its provenience is not secure. (5568) 3065±70 bp (5730) 1210±70 bc MASCA CORRECTION: 1451±70 BC KLEIN CORRECTION (2-sigma): 1585-1100 BC PEARSON/STUIVER CORRECTION (1-sigma): BC 1422 (1385, 1339, 1323) 1262 PUBLISHED Archaeometry 33 (2), 1991 AA-1321 1968 LAYER 9 B.14 SQUARE 4H LEVEL: EP 1 WT. ca. 20 g BAG NO. none DATE COLLECTED: 23/5/68 DATE SUBMITTED: Brooks 6/5/85 PROVENIENCE: BONE COLLAGEN (long bone shafts) (5568)6660±90 bp (5730) 4903±90 bc MASCA CORRECTION: ?? KLEIN CORRECTION (2-sigma): 5870-5200 BC PEARSON/STUIVER CORRECTION (1-sigma): BC 5640 (5553) 5480 PUBLISHED Archaeometry 33 (2), 1991 AA-1322 1968 LAYER 8U B. 90 SQUARE 2E LEVEL: EP 3 WT. ca. 20 g BAG NO.none DATE COLLECTED: 23/5/68 DATE SUBMITTED: Brooks 6/5/85 PROVENIENCE: BONE COLLAGEN (long bone shafts) (5568) 2794±88 bp (5730) 925±88 bc MASCA CORRECTION: 1000±95 BC KLEIN CORRECTION (2-sigma): 1015-430 BC PEARSON/STUIVER CORRECTION (1-sigma): BC 1047 (970, 928) 841 I-5324 1968 LAYER 8U B. 90 SQUARE 2E LEVEL: EP 3 WT. 50 g BAG NO. none DATE COLLECTED: 23/5/68 DATE SUBMITTED: Brooks, 1970 PROVENIENCE: BONE COLLAGEN (vertebrae) (5568) 2595±95 bp (5730) 720±95 bc ASCA CORRECTION: 838±95 BC KLEIN CORRECTION (2-sigma): 1015-430 BC


PEARSON/STUIVER CORRECTION (1-sigma): BC 834 (800) 608 N-1324 1968 LAYER 8U B. 8 SQUARE 2F LEVEL: EP 1 WT. 510 g BAG NO. none DATE COLLECTED: 7/3/68 DATE SUBMITTED: Brooks 6/72 PROVENIENCE: BONE APATITE (rib and vertebral fragments) (5568) 1010±85 bp (5730) 910±90 ad MASCA CORRECTION: 940±90 AD KLEIN CORRECTION (2-sigma): 865-1230 AD PEARSON/STUIVER CORRECTION (1-sigma): AD 964 (1015) 1152 PUBLISHED Radiocarbon 19(1), 1977 N-1325 1968 LAYER 8U B. 14 SQUARE 4H LEVEL: EP 1 WT. 510 g AG NO. none DATE COLLECTED: 31/3/68 DATE SUBMITTED: Brooks 6/72 PROVENIENCE: BONE COLLAGEN (rib and vertebral fragments) (5568) 2160±195 bp (5730) 210±200 bc MASCA CORRECTION: 268±200 BC KLEIN CORRECTION (2-sigma): 755 BC-220 AD PEARSON/STUIVER CORRECTION (1-sigma): BC 400 (196) AD 30 PUBLISHED Radiocarbon 19(1), 1977 N-1326 1968 LAYER 8U B. 52 SQUARE 4E/4F LEVEL: EP 3 WT. 510 g BAG NO. none DATE COLLECTED: 5/5/68 DATE SUBMITTED: Brooks 6/72 PROVENIENCE: BONE COLLAGEN (rib and vertebral fragments) (5568) 2160±195 bp (5730) 210±200 ad MASCA CORRECTION: 268±200 AD KLEIN CORRECTION (2-sigma): 755 BC-220 AD PEARSON/STUIVER CORRECTION (1-sigma): BC 400 (196) AD 30 PUBLISHED Radiocarbon 19(1), 1977 N-1327 1968 LAYER 8L B. 78 SQUARE 3E LEVEL: EP 2 WT. 510 g BAG NO. none DATE COLLECTED: 19/5/68 DATE SUBMITTED: Brooks 6/72 PROVENIENCE: BONE COLLAGEN (rib and vertebral fragments) (5568) 2440±125 bp (5730) 560±130 bc MASCA CORRECTION: 602±130 BC KLEIN CORRECTION (2-sigma): 825-190 BC PEARSON/STUIVER CORRECTION (1-sigma): BC 790 (523) 390 PUBLISHED Radiocarbon 19(1), 1977 N-1328 1968 LAYER 8L B. 88 SQUARE 2E LEVEL: EP 2 WT. 510 g BAG NO. none DATE COLLECTED: 23/5/68 DATE SUBMITTED: Brooks 6/72 PROVENIENCE: BONE COLLAGEN (rib and vertebral fragments) (5568) 2420±75 bp (5730) 540±80 bc MASCA CORRECTION: 582±80 BC KLEIN CORRECTION (2-sigma): 790-275 BC PEARSON/STUIVER CORRECTION (1-sigma): BC 764 (484, 437, 424) 399 PUBLISHED Radiocarbon 19(1), 1977 N-1362 1968 LAYER 8U B. 90 SQUARE 2E LEVEL: EP 3 WT. 500 g


BAG NO. none DATE COLLECTED: 23/5/68 DATE SUBMITTED: Brooks 7/72 PROVENIENCE: BONE COLLAGEN (rib and vertebral fragments, with some centra). (5568) 1820±220 bp (5730) 70±230 ad MASCA CORRECTION: 143±130 AD KLEIN CORRECTION (2-sigma): 180 BC-580 AD PEARSON/STUIVER CORRECTION (1-sigma): BC 50 (AD 214) AD 430 PUBLISHED Radiocarbon 19(1), 1977 PT-276 (TL) 1968 AYER 8U B. 90 SQUARE 2E LEVEL: EP 3 WT. 25 g BAG NO. 1553a DATE COLLECTED: 23/5/68 DATE SUBMITTED: 15/5/69 PROVENIENCE: Vessel no. NP-1635; sand temper. Pot A at head of burial, Drawing 38. 4370±200 BP 2420±200 BC 2-sigma range 2820-2020 BC PT-277 (TL) 1968 LAYER 9 B. 14 SQUARE 4H LEVEL: EP 1 WT. 25 g BAG NO.595a DATE COLLECTED: 31/3/68 DATE SUBMITTED: 15/5/69 PROVENIENCE: Vessel no. NNT-658; clay plus sand temper. From sherd sheet (C on Drawing 15) formed from 1A pot over body. 4945±320 BP 2995 ±320 BC 2-sigma range 3635-2355 BC PT-276 (TL) 1968 LAYER 7 B. 73 SQUARE 3E LEVEL: MP 1 WT. 25 g BAG NO. 1373a DATE COLLECTED: 19/5/68 DATE SUBMITTED: 15/5/69 PROVENIENCE: Vessel no. NP-1939; sand temper. One of two sherds from 1G pot forming Sherd Sheet C at feet, adjoining bronze axe, Drawing 42. 4370±200 BP 2420 ±200 BC 2-sigma range 2820-2020 BC PT-279 (TL) 1968 LAYER 7 B. 73 SQUARE 3E LEVEL: MP 1 WT. 25 g BAG NO. 1373a DATE COLLECTED: 31/3/68 DATE SUBMITTED: 15/5/69 PROVENIENCE: Vessel no. NP-1939; sand temper. Second of two sherds from 1G pot forming Sherd Sheet C at feet, adjoining bronze axe, Drawing 42. 4300±150 BP 2350±150 BC 2-sigma range 2650-2050 BC Discussion Scrutiny of the dates and their proveniences will certainly provide some justification for the criticisms made by Higham, Clark, Stargardt, and others. As Higham has said, Solheim, Parker and I “have never underestimated the problems of dating this site” (1984a:231), and it has seemed that every new attempt on our part to arrive at a conclusive answer to the dating question has instead brought one or more new problems into the arena. I think Higham is in error however in stating that “None [of the dates] conforms with the stratigraphic principles set out [by H.], and there is no relationship between the dates and their stratigraphic contexts” (1984:232). Obviously the nine bone dates and four TL dates on burial pottery are impeccably secure, as are a number of the other samples. Fig. 5-1 plots the dates (Klein corrected for comparison with the charts published in Higham 1989) against the levels, and certainly at first glance the distribution of dates relative to levels appears random, at least from EP 1 to about MP 4. It is obvious, however, that while the dates do not form a neat diagonal, sloping from early to late, neither are they randomly distributed across the seventeen levels; in fact, a regression plot of B/P. age against level (with an arbitrary two points being intersected to represent the MP-LP gap) provides a


significant correlation (r = 0.47, p = .006). The apparent lack of pattern is principally caused by the large cluster of dates in the range 1000 B.C.-A.D. 1000 from levels EP 1 to MP 1, plus the <250 bp date from MP 3. It seems obvious that there are several problems involved here rather than just one: Fig. 5-1

1. As pointed out by Higham, the stratigraphic uncertainty of some of the samples dated, particularly from the 1966 excavation. 2. The suggestions made at the 1973 London Colloquy and repeated twenty years later (Higham 1996-1997), that most of not all of the pre-2000 B.C. dates from Non Nok Tha (and Ban Chiang, etc.) seemingly associated with bronze are in fact the results of natural forest fires. 3. The total unreliability of the Non Nok Tha dates on skeletal material from burials. 4. The possible contamination and uniform distortion forward in the time of the FSU series of dates. 5. The reliability of the five thermoluminescence dates from the site, given the undatability of Ban Chiang pottery. 6. What Spriggs (1996-1997) has referred to as the â&#x20AC;&#x153;Gakushin factorâ&#x20AC;?: the general unreliability of the GaK dates from the 1966 excavation (and elsewhere in Southeast Asia and the Pacific). 7. The validity of the recently obtained series of eleven accelerator dates on organic material from Non Nok Tha potsherds.


I dealt with problems 1-4 in some detail in a lengthy paper presented to the London Colloquy in 1973 (Bayard 1979), but Higham neglected to address these arguments in his Wheeler Lecture, perhaps due to time constraints. Therefore, it is clearly necessary to cover these and the three further problems again in this final report. 1. Turning first to the question of stratigraphic reliability, I believe that a breakdown of the samples gives and accurate picture of the distribution of secure and insecure dates from the site. I have included large charcoal fragments as relatively secure because it is very unlikely that the excavation of a grave in the hard clay loam of Non Nok Tha would be able to extract and redeposit pieces of charcoal of this size intact, particularly given the nature of tools in use at the time the burials were made (cf. the difficulties encountered using modern iron tools mentioned in Chapter 3). While it is certainly clear that the provenience of many of the 1966 samples has contributed to the problem of dating the site, half of the total of fifty samples are in fact of secure provenience, as indicated in the following breakdown: a. Dates insecure contexts (nonburial features, large fragments [>15 g], clear concentrations in level, sherds from burial pots, and bone samples; N = 25): 1966 (3): GaK-908, GaK-958, GaK-1027. 1968 (22): FSU-339, FSU-340, FSU-342, FSU-343, GX-1609, GX-1612, I-5324, Y-2485, OxA-2383, OxA-2382, AA-1321, AA-1322, N-1324, N-1325, N-1326, N-1327, N-1328, N-1362, PT-276, PT=277, PT-278, PT-279. b. Insecure dates (small fragments of charcoal from grave pits, sherds provenanced to level only or not from substantially complete provenanced vessels, or diffuse concentrations of small fragments in level; N = 25): 1966 (13): GaK-956, GaK-957, GaK-959, GaK-1028, GaK-1029, GaK-1030, GaK-1031, GaK-1032, GaK-1033, TF-651, Y-1851, PT-?. 1968 (12): FSU-341, FSU-345, GX-1611, OxA-2384, OxA-2385, OxA-2386, OxA-2387, OxA-2388, OxA-2389, OxA-2391, OxA-2393. Hence I feel that half of the samples can be considered secure, if in some cases not as impeccably provenanced as Higham’s series of thirteen dates from Ban Na Di. However, even some of these secure samples (e.g., the one in an LP 2 posthole that must clearly date an earlier event) produce contradictory results, a phenomenon encountered by Higham as well at Ban Na Di (Higham and Kijngam 1984:31-32). This was also the case at Higham’s Khok Phanom Di excavation, where an initial series of secure dates tentatively suggested the beginning of burial and occupational use at ca. 4000 B.C. (Higham et al. 1986-1987:154-155), but in the light of a further extensive series of dates this was modified downward by 2,000 years (Higham 1996-1997). The dating of burials and bronze at Nong No discussed in the first section of this chapter provides another example. 2. There is also the wider question of a natural origin of the charcoal producing the pre-2000 B.C. dates; as I said in my London Colloquy paper in 1973, “It is certainly conceivable that some of the charcoal could have been the result of nonhuman action, such as forest fires, which occurred prior to the occupation of the site: (1979:22); this could be a possible explanation for the very early date GaK-1032, and for the other two pre-3000 B.C. dates. Higham (1996-1997) has raised the question again by recently completing a review of dates from bronze and pre-bronze dates in Thailand and Viet Nam and concluding that there is little hard


evidence for dates clearly associated with bronze prior to the mid-first millennium B.C. With specific regard to Non Nok Tha, he has again raised the possibility that charcoal purportedly dating the early MP and EP could have been derived from natural fires prior to occupation. In the case of FSU-340 (the large lump of charcoal in 1968 layer 11), I agree (as I thought in 1968) that this serves as a possible predate for human occupation, although probably distorted upward like the other dates in the FSU series. However, in trenching off 50 cm of sterile soil along both major axes of the 1968 excavation (47 m2 in all), charcoal was on my instruction sought after by the workmen carrying out the operation, and it was very rarely encountered. When it was, it was a cause for careful consideration, as evidenced in my field notes (quoted in Bayard 1996-1997): 4F â&#x20AC;&#x201D; In trenching down to 220 cm on E side of sq. Chaloem hit an isolated chunk of charcoal at 168 A.D., ca. 10 cm below the [layer] 12 [upper layer 12 was later redefined as overall layer 11] surface here. This location is almost directly opposite the place in the 4G/4F baulk where we encountered the charcoal patch B [FSU-340] (p.149-150), and I was worried about this turning out to be a possibly recent burned root. However, the piece was isolated, and appeared to be a burned chip of wood rather than a root fragment. As it was too small to be an effective sample, we discarded it. (1968 field notebook I, p. 200[4/24/68]). Only two pieces of charcoal were encountered, one a large lump (25 g) and the second so small (by my recollection ca. 7 x 7 x 2 mm) as to be worthless for dating by technology available at that time, even at a site where charcoal was more eagerly sought after than bronze, seem unlikely relics of even a minor forest fire. As I have said above and in 1973, I think it is highly implausible that natural fires were responsible for the vast majority of the early charcoal dated at the site (1979:22), or for that matter at Ban Chiang and other Thai and Vietnamese sites. After all, even scrub fires produce large quantities of charcoal, and this was certainly not in evidence in the basal layers of the site or in the two long trenches excavated into sterile substratum. Support for this argument is afforded by analyses of soil from Non Nok Tha: 1968 layers 1-2 contained 1.74% organic carbon (including charcoal) by weight; 1968 layer 6 contained 0.628%; layer 9 had 0.216%. But 1968 layers 11 and 12 had 0.018% and 0.020%, respectively. It is worthwhile to compare these figures with the results of analyses of samples contained in a soil profile I took from a swidden field at the foot of Phu Wiang in 1968, last used at least seven years previously: surface, 1.77% organic carbon; layer 6, at a.d. 85 cm, 0.503%. This is twenty-five times the amount present in basal Non Nok Tha (data from Nelson, to appear in the volume of specialist reports). There appears to be little support in these figures for large amounts of charcoal in the basal sterile levels of the site. Small preoccupation fires, whether natural of caused by earlier transient humans, may account for the extremely aberrant 8100 bp date (GaK-1032; but see discussion of the â&#x20AC;&#x153;Gakushin factorâ&#x20AC;? below). Similarly small fires may have also have produced the two pieces of charcoal mentioned above, which in both cases lay at the same depth below surface (170 cm), 10-12 cm below the layer 10-11 interface; two more small pieces lay above them at the interface itself. These occurrences, limited to an area of less than a square meter, were the only ones encountered in the 47 m2 of natural deposits excavated. It is simply expecting too much of


a malevolent Providence to have caused third- and fourth-millennium forest fires to occur neatly on Non Nok Tha and other sites like Ban Phak Top, Ban Chiang, Ban Tong, and Ma Dong prior to the arrival of their second millennium B.C. occupants, leaving behind them not the massive charcoal deposits produced by such fires but only enough charcoal to deceive their second millennium A.D. excavators. 3. The nine bone dates present a more fundamental problem: the reliability of apatite and collagen dating in general. Some if this series may be seen as offering support to the arguments of a few workers who still find it difficult to envisage bronze in the region prior to 1000 B.C. that is the 800 B.C. and 950 B.C. dates on 1968 EP 3 B. and the earliest bronze artifact encountered at the site; but cf. the 200 A.D. apatite date (N1362) on the same skeletal material. However, the obviously aberrant dates produced by others are clearly nonsense by any conceivable chronology. It is as difficult to conceive of the initial occupation of Non Nok Tha in 5550 B.C. (AA-1321, 1968 EP 1 B. 14) as it is to assume that the site was first occupied during the reign of Suryavarman I (N-1324,1968 EP 1 B. 8). The problems seem to be twofold: the very small amount of collagen in the skeletal material, and the fact that some of it appears to be fossilized. Thus the seven samples dated by Rikagaku Kenkyusho (the N-series) all contained less than 0.8% collagen (cf. ca. 25% in modern dry bone; John Dennison, pers. comm). Two of the seven had so little collagen that the apatite fraction had to be dated instead, producing obviously erroneous dates. On the other hand, the two samples dated in 1987 by the Arizona laboratory (AA-1321 and 1322) were “very rich indeed” in organics; the problem here appears to be fossil-derived collagen, which “is the subject of current research in bone dating. Hard data is difficult here as each bone can become its own research project owing to its particular environmental history” (pers. comm. from B. Gore to Solheim, 9/3/87). Given such circumstances, I think it is prudent simply to ignore the Non Nok Tha bone dates, at least until they have been the subject of several theses! Higham has encountered similar problems with bone dates from earlier excavations at Khok Phanom Di that suggested a date of ca. 4000 bc for initial occupation of the site; in fact, as mentioned above, a well provenanced series of charcoal dates now indicates an initial settlement of the mound ca. 2000 B.C. (Higham 1996-1997). 4. Two other problems are the six FSU dates, all apparently distorted upward in time, and the question of the reliability of the five TL dates; I addressed both at some length on the 1973 London Colloquy paper (Bayard 1979:21-24). To summarize the first, I believe that the FSU samples had been subjected to some form of fairly uniform contamination between excavation and dating for the following reasons. Firstly, the two LP samples contained more radiocarbon than modern material; secondly, the GX-1609 date on the identical material (bamboo charcoal) used to produce date FSU-339 produced a 500 B.P. date perfectly consistent with the celadon sand porcelain present in LP 3. The six Florida State University dates are consistently some 0.6 to 1.4 counts per minute per gram higher than the values would have been for the estimated ages. As a letter (dated 4/29/64) from J.R. Martin of the Florida State University laboratory to me remarked, “It is interesting how similar the differences are between the actual counts per minute per gram, CPM/g,


and what the CPM/g would have been had the samples been the ages you estimated.” I need only add that, as Spriggs pointed out in his note on the Non Nok Tha Gakushuin dates 1996-1997; see 6 below) dating technology has undergone considerable refinement since the 1960s. 5. Turning to the five TL dates, I would agree that Ban Chiang ceramics “are usually quite undatable because of a lack of significant TL output” (Fleming 1979:10) and hence should be disregarded (cf. Carriveau and Harbottle 1983). However, the Non Nok Tha TL dates have also been criticized alongside those from Ban Chiang by Stargardt; in her opinion their excessive age was due in both cases to the use of “grit” (grog) temper (1981:333). But Satrgardt is in error in equating the fabrics from the two sites; only a very small fraction on Non Nok Tha wares (<5%) are in fact grog-tempered (see Chapter 6), and only one of the Non Nok Tha sherds TL dated (NNT-658; PT-277 above) had clay as a component of the temper along with sand. The remaining sherds had only quartz for sand temper (or rice chaff in the case of the crucible fragments used for the unnumbered PT date above). Nor are the Non Nok Tha wares low-fired and hence bad subjects for TL dating, as Stargardt implies. Far from being “fired at less than 400ºC” (Stargardt 1981:334),10 their general firing temperature was apparently between 800 and 950ºC (Meacham and Solheim 1979:115). Hence the dates are internally consistent, reliable within the general range of error of TL dating, and the provenience of four of the five sherds dated is if course very secure. I can only conclude that the dates are as accurate as the limitations of thermoluminescence dating allow. 6. The “Gakushin factor.” In the course of the debate over Non Nok Tha chronology in which Higham and I participated at the 1988 Kioloa conference (Higham 1996-1997; Bayard 1996-1997??), Spriggs also became interested in a possible link to radiometric problems detected by himself and others in Oceania. As he pointed out in a brief note on Non Nok Tha chronology and what he called the “Gakushin factor” (Spriggs 1996-1997), radiometric dating technology has advanced considerably since the 1960s, and it is advisable to be cautious in accepting dates run prior to 197-—or even 1975. Fig. 5-2 After outlining some prior problems encountered with dates on carbon samples from the Pacific run by the Gakushin Laboratory, he examined the thirteen dates from Non Nok Tha that were processed by the same laboratory along side the other radiocarbon and thermoluminescence dates (then totaling thirty-nine) and concluded that similar problems were apparent with the thirteen dates from the 1966 excavation. These comprised the bulk of the ones whose lack of accurate provenience and/or aberrant values had also concerned me, and may explain why consciously or not I had none of the 1968 samples dated by this laboratory.


A glance at Fig. 5-2 will illustrate Spriggs’ point: three of the LP dates (in levels containing porcelain and celadon) are obviously too early by anyone’s standards, as is the 8100 bp date for MP 1 (GaK-1032) and the pre-4000 B.C. date for EP 3 (GaK-1034). Other Gakushin dates are clearly too late: the first centuries of the present era for the badly provenanced EP or MP date (GaK-1030). Some others (GaK-956, GaK-1033, and GaK-908, although I would discard GaK-956 and include GaK-1029) seem to be approximately correct. This spread approximates what Spriggs encountered in his examination of Pacific radiocarbon dates run by Gakushin: “Some dates are too old, some too young and, given such a spread, one would expect that some of them are correct. But which ones?” (1996-1997, in press). Fig. 5-3

For Non Nok Tha Spriggs proposed discarding all GaK dates, the six FSU dates and all nine bone dates from the site, leaving six radiocarbon and five thermoluminescence dates, which—-taken in conjunction with four dates from the nearby site of Non Nong Chik (see below)-—suggest initial occupation at ca. 3000 B.C., bronze by 2500 B.C., and a hiatus between about 200 B.C. and A.D. 1300; that is, a chronology close to the proposed here. Although I retain some of these discarded dates for initial consideration in the section to follow, Spriggs’ points are clearly valid ones, and it is obvious that the Gakushin series as a whole is not reliable. 7. The Oxford accelerator dates. In 1988 Solheim and I gratefully accepted the kind offer by David Harris of the Institute of Archaeology to have a series of twelve potsherds from Non Nok Tha dated by the then relatively new technique of accelerator mass spectrometry (AMS) at the Oxford Laboratory. This technique has progressed from its beginnings in 1982 to become one that will “have a profound effect on many aspects of archaeological research” (Harris 1987:23), as it requires


only 1 mg or more of carbon for dating (as opposed to the 1-2 g or more necessary in the late 1960s); hence tiny fragments of organic inclusions in sherds (rice chaff and other fibrous tempering materials) can be dated. I wrote to Solheim to select only sherds from substantially complete vessels securely provenanced to a specific level. Unfortunately, this proved to be impossible to accomplish; all of the over eight hundred reconstructible vessels had already been returned to the Thai Fine Arts Department, except the 10% that the Universities of Hawai`i and Otago were permitted by agreement with the Thai authorities to retain for teaching purposes.11 Also, most of the burial pottery from the site was sand or (largely silicified) fine chaff and sand tempered (see following chapter) rather than fiber tempered, and it is likely that only a minority would have proved to have inclusions suitable for dating. As a result, Solheim and Virgil Meeker had to turn to the collection of sherds that had been set aside during the main analysis because of their significant organic content for photography or further analyses, and many of these had uncertain proveniences. I had no chance to observe the sherds submitted myself, although in some cases I had made notes in the catalog on them (see date list above). In fact, I would rate only two or perhaps three of the sherds as securely provenanced: OxA-2392, the only date derived from a securely provenanced burial vessel (albeit from a carbonized rice grain rather than a sherd inclusion); OxA-2383, from a clay-and-chaff-tempered sherd found in a pit apparently dug from the surface of EP 1; and possibly OxA-2390 on one of a group of sherds from different vessels recovered from a pit provenanced of MP 4 (and because the possibility of stratigraphic displacement, I have not classed it as secure in the figures in this chapter). The remainder of the dates are on sherds that are provenanced by excavation layer only, and are hence as open to criticism as similarly unprovenanced charcoal samples. This is an unfortunate state of events, as I wrote to Professor Harris once I learned of the samples that had been submitted from Hawai`i. Nonetheless, I believe there is something to be gained from the dates. First, the AMS technique is obviously far more reliable than the collagen or apatite analysis used on skeletal material from the site; the three clearly aberrant dates in the later first millennium A.D. are perhaps 1000â&#x20AC;&#x201C;1500 years too late, rather that several thousand (as in the case of the apatite date N-1324). Secondly, the remaining eight dates fall between 1000 and slightly before 2000 B.C., with those provenanced to the Middle Period agreeing with the bulk of earlier radiocarbon determinations. But there would appear to be two problems in interpreting this series. The first of these is the possibility of downward intrusion of MP sherds in undetected disturbances into EP contexts; given the fabric/temper type of the sherds dated, I feel this is a very likely possibility for OxA-2384 and OxA-2387. Secondly, there is the possibility of contamination during the ten or more yearsâ&#x20AC;&#x2122; storage of the sherds (in cotton inside plastic vials, sometimes with traces of glue; see date list above). I feel that this factor may well have contributed to three post-500 A.D. dates; there is no evidence of any sort suggesting the occupation of the site during this time (Lop Buri or other Khmer-style wares, for example). Harris has said with regard to accelerator dating that â&#x20AC;&#x153;field or context contamination


originating outside the laboratory is less easily eliminated because it arises in many different ways during the excavation, handling and storage of samples before they reach the laboratory. . . . Much contamination can, however, be eliminated by chemically meticulous pretreatment of samples” (1987:28). However, the Oxford Laboratory stated in a letter to Harris that the young dates may be due to contamination (see commentary on OxA-2385 above). We can perhaps conclude that if the eight dates that are not obviously too late are accurate, and if the pit that produced OxA-2383 is in fact from EP 1 and not the EP/MP transition, then these dates would support a dating of all EP levels plus the first half of the MP between the terminal third millennium and 1000 B.C., as Higham has postulated (1996-1997). But, of course, six of these eight dates are as badly provenanced as the charcoal samples Higham has criticized and thus cannot be considered reliable witnesses by the stringent criteria he proposes. Given the above problems, the Non Nok Tha dates certainly cannot be described as ideal, but the lack of pattern in the dates is very largely caused by the large cluster of dates in the range 1000 B.C.- A.D. 1000 from levels EP 1 to MP 1, plus the <250 bp date from MP 3; eight of these are bone dates that as stated are best discounted. Once they are, along with the seven charcoal and accelerator dates that appear to agree with them and the <250 bp date it is difficult not to discern a pattern, particularly if the two impossibly early pre-5000 B.C. dates (one of them a bone date as well) are also eliminated. Now it is obviously true, as Higham put it to me personally, that it is always easy to form a pattern by getting rid of unwanted dates. However, the nine bone dates are among the most securely provenanced from the site, and are certainly not “unwanted”; the problem is that they simply make no sense by anyone’s chronology and are obvious erroneous. In reviewing the available dating evidence from Non Nok Tha and related sites, Higham agreed that “eliminating the clearly erroneous dates from Non Nok Tha” is a “reasonable step” to “make it possible to reach a reasonable consensus” (1989:126). Once these and the other dates rendered improbable by relative ceramic chronologies are eliminated (Fig. 5-3), we are left with a more orderly series of thirty dates that clearly suggest initial settlement of the site in the third millennium B.C. Even with the two “future” dates and the three clearly erroneous pre-A.D. dates for the LP (I know of no one who would postulate porcelain in Thailand at this date) left in consideration, a regression analysis produces a correlation of 0.84 (p = .0000). Naturally the assumption of an even spacing in duration of the levels is arbitrary, but at the least suggestive, which is all I wish to claim. Clearly there are still problems with the consistency of even those dates in Fig. 5-3, and given the above-mentioned problem of the “Gakushuin factor”, it might be wise to eliminate all of the Gakushuin dates. This would leave us with only eighteen dates (Fig. 5-4), but ten of them are secure.


Fig. 5-3

Fig. 5-4

I think it is possible to make at least two interpretations of this last sequence, small as it is (although larger than the number of dates Higham presents for Ban Na Di, Ban Chiang Hian, and Ban Kho Noi; Higham 1989:128â&#x20AC;&#x201C;129): A. The four TL dates could support an initial occupation of the site at the beginning of the third millennium B.C., with the MP middle levels dated to the mid-second millennium.


B. The accelerator date OxA-2383 and a charcoal date of about 2000 B.C. from EP 1 B. 125 (FSU-345) could suggest an initial occupation of the site in the terminal third millennium, with the Middle Period following on during the mid-second millennium (and perhaps later, given the two accelerator dates on MP 1 and 2). Interpretation B would approximate Higham’s most recently presented scenario (1996-1997), but because of my concern about the possibility of the FSU series of dates as a whole and the decoration of the sherd that produced the accelerator date OxA-2383, I would currently tend toward interpretation A. But as I said earlier in this chapter, I do not believe that Non Nok Tha can stand alone as a witness for the dating of early agriculture, bronze, or any other trait; I hope I have amply demonstrated that despite a large number of determinations, the dating of the site is still open to question, although over a much narrower range than in the past. In summary, I still believe that I am justified in saying that “there is an overall linear pattern present which dates in general the earlier portions of both Non Nok Tha and Ban Chiang sequences” (Bayard 1986–1987:122). Spriggs’ 1988 chronological revision as outlined above is also consistent with interpretation A, both internally and with the dates from nearby Non Nong Chik. It is time now to turn briefly to other witnesses and to look at the dating of the region as a whole. Regional Comparisons: Toward Synthesis and Consensus? If debate on the absolute chronology of northeast Thailand continues, at least the past decade or so has seen the development of a generally accepted relative sequence. Since the late 1970s Thai archaeologists including Charoenwongsa, Rutnin, Kijngam, Bannanurag, Vallibhotama, Natapintu, and many others, working on their own or in collaboration with westerners such as Higham, Pigott, Wilen, Welch, and Glover, have greatly added to our knowledge of northeast and central Thai prehistory. This increase in data allowed me in 1984 to postulate a four-stage developmental sequence from Neolithic to state-level societies that I hoped would be applicable to Viet Nam, Cambodia, and Laos as well as Thailand (Bayard 1984b). This scheme has subsequently been adopted by Higham (1989), and others, and seems to work well as a general chronologically nonspecific terminology. The four stages or “General Periods” [GP] are designated A, B, C, and D; they could also be simplistically labeled “Neolithic”, “Bronze”, “Iron”, and “Indianized”. In fact, although Higham used the GP system in his 1989 review of Southeast Asian archaeology in general, in his most recent work in Southeast Asian Bronze Age (1996) he reverts to the use of neolithic, Bronze Age, and Iron Age (1996:7). GP A societies, as evidenced at Khok Phanom Di and the basal levels of Non Nok Tha and Ban Chiang, seem to have had an economy based on rice—-presumably cultivated—-and well-developed stone and ceramic technologies, as well as already fairly extensive trading networks in these and other commodities. Settlements seem to have been small, but social ranking was already clearly in evidence (see Chapter 8). GP B saw the introduction of bronze metallurgy and some increase in social ranking and trade in exotic materials already present in GP A. GP C represents an increasingly well documented major shift: the rise of considerably larger nucleated villages, the introduction of iron, glass beads, and a variety of new ceramic styles, and almost certainly the beginning of intensive wet-rice agriculture and


supravillage political organization—-entities that I have termed mu’ang (Bayard 1992; see Chapter 2). GP D arose in a complex and as yet ill-understood fashion (aside from the obvious Han Sinicization of Viet Nam) at least partly from the impact of Indian politico-religious concepts, and witnessed the amalgamation of mu’ang into Indianized “states”—-what Wolters (1982) and Higham (1989) have called mandala—-the Sanskrit term--and what I refer to as monthon, using the modern Thai loan word (Bayard 1992). I should add that in a review of this article, Higham criticizes me for using specifically Thai terms like monthon and mu’ang rather than more general ones like autonomous village and chiefdom, and asks “would he wish to substitute Vietnamese, Khmer, or Mon terms in other parts of Southeast Asia?” (1994:363). My feeling is “yes”, if they offer the utility of mu’ang and monthon in the Thai context (See Chapters 2 and 9). The GP labels provide a convenient shorthand, and will be used in the remainder of this work. However, this is not the case for a detailed regional phase chronology scheme, which I presented in the same article (1984b). As I feared then, this has so far proved too cumbersome for general use, although Higham employs it for his discussion of Non Nok Tha (1989:98–106), using “Phu Wiang phase” for the site’s Early Period and “Non Nok Tha phase” for the Middle Period. However, to spare the reader any further confusion I will continue to use “EP”, “MP”, and “LP” for the three Non Nok Tha periods throughout the rest of this report. The dates for the four General Periods obviously vary from place to place, and as seen above they are still very much a matter for debate. I had hoped that by this time further research in the Phu Wiang region would have produced sufficient securely provenanced dates to either confirm or reject the tentative Non Nok Tha sequence. Unfortunately this has not proven to be the case. Only three other small excavations in the region have yielded ceramics obviously cognate with the Non Nok Tha Middle Period, and unfortunately the few dates available from them are-–like those from Non Nok Tha—suggestive but far from securely provenanced. The site physically closest to Non Nok Tha is Non Nong Chik, inside the ringwall of Phu Wiang itself about 7–8 km south of Non Nok Tha in a direct line (Higham and Parker 1970:31–36; Bayard 1977a:83 – 89). Six dates were obtained from Higham and Parker’s test excavation at Non Nong Chik, which predictably had ceramics very similar to those at Non Nok Tha. Two of these, from iron-period layers that obviously postdate the end of the Non Nok Tha Middle Period, are clearly too early and are probably intrusive from earlier layers. Three of the remaining four agree well with the dating for Non Nok Tha presented here, if we accept that the equation of the two ceramic sequences that Buchan and I arrived at (Bayard, loc. cit.) is approximately correct. Non Nong Chik layer 7, which equates with MP 5 and/or 6, is dated at BC 1010 (905) 810 (Pearson/Stuiver 1-sigma correction); a second date of BC 1932–1687 for layer 7 is seemingly too early. A sample from upper layer 6, roughly equivalent to Non Nok Tha MP 7, produced a date of BC 401 (385) 232, while a date from layer 5 (intermediate between MP 7 and MP 8, or more probably post MP 8) of B.C.182 (96) A.D. 12 also seems reasonable, although of course the internal dating of MP 7 and MP 8 at Non Nok Tha itself is very inferential; neither MP 7 nor 8 contained any of the distinctive iron period GP C pottery encountered in Non Nong Chik layer 5.


The second closely related site is Non Pa Kluai (or Kluay), excavated by Wilen, in 1984– 1985 (Wilen 1986-1987, 1989). The site is very similar to Non Nok Tha in both size and location and like it was used primarily as a cemetery rather than an occupation site during the bronze period or GP B (Wilen 1986–1987:114). The ceramics recovered from the lower phase of the small excavation (15m2) are clearly very similar to Non Nok Tha: 1D, 1R, or 6R, 4B, 5B, and 6A vessels were recovered from the excavation itself, and villagers had earlier recovered 2D/E vessels from the site (Wilen 1986– 1987:106; 1989:65,139-140). Non-ceramic artifacts from the lower phase of the site (Level V) also parallel closely the Non Nok Tha Middle Period assemblage. However, unlike Non Nok Tha, Non Pa Kluai was not abandoned with the onset of GP C; the upper phase at the site (Levels III and IV) evidenced a great elaboration of ceramic decorative styles, iron tools, glass beads, and evidence for habitation as well as funerary use. As with Non Nok Tha, charcoal was not abundant at Non Pa Kluai, and the proveniences suffer from the same problems; nonetheless, the dates appear to match at least generally those from Non Nok Tha. Three dates from the lower phase suggest funerary use from the early second into the first millennium B.C.: Level V B, 13, BC 2193 (1687) 1267; Level V B. 5, BC 1258 (1035) 910 and BC 1251 (1021) 912. The latter burial contained a “horned” 1R or 6R vessels similar to those occurring mainly in Non Nok Tha MP 2 and MP 3, but there was also a type 6A vessel somewhat more characteristic of MP 4 – 6. Unfortunately, no burials or ceramics characteristic of the Non Nok Tha Early Period were present at the site. But the site does at least generally corroborate the second millennium date postulated for most of the Middle Period above. The upper phase at Non Chai (Bayard et al. 1982–1983); some of the vessel types have very specific parallels with material from the GP C levels of Non Nong Chik and upper Don Klang (Wilen 1989:143–147), the third site with clear affiliations to Non Nok Tha in its middle and lower levels. Like Non Pa Kluai, Don Klang is similar in situation to Non Nok Tha; it is located some 25 km northeast of Non Nok Tha, on the eastern side of Phu Kao, a second smaller sandstone monadnock in the region. Excavating a test square in 1975, Schauffler similarly encountered what seemed to be two phases at the site: an upper iron period one, containing clear GP C material (iron tools, glass beads, and black-on-buff decorated ceramics reminiscent of upper Non Pa Kluai); and a lower presumably bronze period one with ceramics related to the Non Nok Tha Middle Period. Schauffler (1976:33) also mentions finds of sherds with incised curvilinear designs in a burial from a lower layer at the site, which he felt should crossdate with those from Non Nok Tha EP 1. It is unfortunate that no detailed description of the Don Klang pottery is yet published, but certainly some of the type 2C vessels from the site shown to me by Schauffler in 1980 could equally well have been excavated from Non Nok Tha. A single date from one of these layers of B.C. 762 (523) 406 is available (from the layer 8/9 interface, and this would seem to correspond well with the late Middle Period at Non Nok Tha; however, the 2C and other specifically Non Nok Tha-type vessels appear to postdate the presumed lower burial phase (White, pers. comm.). They may date from a period paralleling the terminal MP at Non Nok Tha, but clearly more analysis of the Don Klang materials will be necessary to


draw any firm conclusions. The remaining five dates from the site (Hurst and Lawn 1984) all derive from the upper phase (layer 4–6) and indicate a span of about 200 B.C.—-A.D. 100, which again correlates well with dates for the iron period upper phase at Non Pa Kluai, Non Chai, and other GP C sites. Turning farther afield, we next encounter the second major bronze period (GP B) cluster of sites in northeast Thailand: Ban Chiang and its associated test excavations at Ban Phak Top and Ban Tong (Gorman and Charoenwongsa 1976, Schauffler 1976, White 1982a, 1986); and Higham and Kijngam’s excavations at the nearby site of Ban Na Di (Higham and Kijngam 1984). As outlined above, the Ban Chiang chronology is as thorny and controversial as that of Non Nok Tha; White has devoted an entire doctoral dissertation to the question (1986; see also White 1988a), and we will not delve into it in any great detail. By both her reckoning (1986:265) and mine (1986–1987:121), the basal layers of Ban Na Di equate with the end of her Early Period (EP V) and beginning of her Middle Period (MP VI) at Ban Chiang, well after the first bronze tool encountered at the latter site in an EP III burial. However, if any ceramic parallels at all are present between Ban Chiang and the more distant Non Nok Tha, they are quite clearly confined to EP I and II at Ban Chiang and the Non Nok Tha Early Period; specifically, the curvilinear incised vessels limited to Non Nok Tha EP 1 and the rouletted “spittoons” and globular vessels with perforations in rim and ringfoot found only in EP 3/MP 1 transition contexts (see Chapters 6 and 8). These are paralleled at least in general at Ban Chiang EP II (White’s Provisional Types 3 and 4; 1986:91, 125). This level has six dates (two nonburial and four from graves) with corrected midpoints ranging from 3300 to 1300 B.C. (White 1986:206), although as White says the 1300 B.C. date seems well out of line with the other five, the latest two of which have midpoints of 1945 B.C. White concludes that a third millennium date is most probable for this level and also that “the possibility cannot be discounted that EP II ceramic types may prove to date into the fourth millennium B.C.” (1986:209). Given the possible association of bronze with densely incised ware mentioned previously, one might suggest (but certainly not demonstrate) that bronze arrived at Ban Chiang some time in the third millennium, prior to EP III. Again, I emphasize that—-like the Non Nok Tha data-—the evidence is suggestive only. A further suggestion that bronze may be associated with EP II arises from a 9 m2 test excavation at Ban Phak Top carried out in 1975 by Schauffler. Although no burials were encountered in the excavation itself, his workmen “emphatically stated that when they had dug for pots on their own, they had often found bronze objects in burials containing the black incised pottery” (Schauffler 1976:29). Here again, we cannot assume an intimate knowledge of Ban Chiang ceramic typology on the part of the workmen (although given the importance of pot “mining” to the economy of modern Ban Chiang, they would certainly be well-informed laymen), and the evidence is clearly not of the sort that would be admissible in court. A date from basal Ban Phak Top (pre-layer 12?) of BC 2484 (2457) 2298 suggests initial settlement shortly after the mid-third millennium; layer 12 itself is dated at BC 2140 (1846) 1543 (dates from Hurst and Lawn 1984). Finally, we have already mentioned the five internally consistent AMS dates on securely provenanced burial pottery from Ban Chiang; these strongly suggest an early second millennium B.C. date for the introduction of bronze (White 1997:104). Conclusions


To summarize, the Non Nok Tha sequence of dates is obviously not reliable enough to place more or less precise chronological boundaries on each of the seventeen archaeological levels; even the dating of the Early, Middle, and Late Period boundaries is contentious. I think on present evidence that we can talk of an Early Period (equivalent to GP A in the overall scheme presented above), with hazy beginnings in the earlier third or possibly even terminal fourth millennium extending to the EP/MP transition to GP B sometime after 2500 B.C. Middle Period levels 4 and 5 seem fairly well dated in the midto late second millennium, and by intercalation with the Non Nong Chik, Non Pa Kluai, and Don Klang dates (such as they are), MP 6, 7, and 8 date from ca. 1000 B.C. to the abandonment of the site in the late first millennium B.C. The date of initial reoccupation after the MP/LP hiatus is again hazy; I would estimate A.D. 1000 – 1200. Late Period 2 and 3 seem to be reasonably dated to 1300–1500, and LP 4 and 5 presumably extended up to the abandonment of the site in the previous century before the present population reoccupied the immediate area in the 1890s. How does this chronological framework agree with our still very incomplete picture of wider regional prehistory? Turning farther afield, we need only note that bronze may have arrived in central and western Thailand at a date later than it did in the northeast (Glover 1982–1983:106); however, Pigott and Surapol’s revision of the Non Pa Wai dating 1996-1997 mentioned in the first part of this chapter may indicate that bronze also arrived in the northeast only in the mid-second millennium B.C. To search for evidence supporting an earlier appearance in the region as a whole, we could look east to Viet Nam, where a number of sites of the late Phung Nguyen and Dong Dau cultures have produced dates in the range 1500–1700 B.C. associated with bronze (Ha Van Tan 1980). A single isolated date from Ma Dong has a 1-sigma Pearson/Stuiver range of BC 2882–2613, and is said to date the transition to the early bronze age (Kohl and Quitta 1978:393). Farther south, in the Thanh Hoa region, sites of the Hoa Loc culture have produced “a few pieces of bronze” at a date apparently equivalent to the late Phung Nguyen cultural period (Higham 1989:182); this is still poorly dated but would appear to be between 1500 and 2000 B.C. To the south, the site of Doc Chua has produced molds and axes that are substantially identical to those at Non Nok Tha (see Chapter 7), with an associated date in the midsecond millennium B.C. In the early 1980s, Vietnamese archaeologists dated the beginning of the Middle Bronze Age at 1800-2000 B.C. (ca. 2200–2400 B.C.; Pham Minh Huyen and Diep Hoa 1981:33, Pham Minh Huyen 1984:175, Hoang Xuan Chinh and Bui Van Tan 1980:63), but obviously the chronology requires further support. Hence the third millennium date for bronze at Non Nok Tha suggested here could be considered plausible in this wider context, but nothing more. What we can say with some degree of certainty is that by the mid-second millennium B.C., essentially similar bronze technologies were widely disseminated throughout the middle Mekong, Ma, and Hong drainages and along the Vietnamese littoral. Contrary to Higham’s belief in an unusually rapid spread of bronze technology only after 1500–1400 B.C., I find it unlikely that this phenomenon took only decades or at most two or so centuries to accomplish (see Higham 1996–1997; Bayard 1996–1997 for detailed arguments).


Despite all this, I think it is fair to say that a general consensus is beginning to emerge. Almost all workers would now agree that a case for the widespread distribution of bronze by the mid-second millennium B.C. is firmly established. As seen in the first part of this chapter, the case for bronze prior to that date is still the subject of debate, but it is now over a considerably narrower range. My own view at present is that bronze first appeared in northeast Thailand sometime between 2300 and 1800 B.C., with the Ban Chiang and Non Nok Tha evidence suggesting–-but by no means proving--a beginning in the late third millennium. Much more work will be needed in Thailand, Viet Nam, and South China (not to mention the archaeologically unexplored areas of Cambodia and Laos) before we can arrive at an unequivocal answer on the dating and, more importantly, the origin and social context of Southeast Asian bronze. I think it is very unlikely that–-after only three decades of modern archaeology in the region-–we have been fortunate enough to stumble on the earliest bronze period sites. In the original edition of his Archaeology of Ancient China, Chang quoted Max Loehr, writing in 1949 on the question of the origins of Shang metallurgy: An-yang represents, according to our present knowledge, the oldest Chinese metal age site, taking us back to ca. 1300 B.C. It displays no signs of a primitive stage of metal working but utter refinement. Primitive stages have, in fact, nowhere been discovered in China up to the present moment. Metallurgy seems to have been brought to China from outside. (Quoted in Chang 1963:140) Zhengzhou was discovered in the following year; Erlitou in 1958. I suspect that our Erlitous–-establishing a conclusive case for the area of origin and dating of early bronze, be it 1700 B.C. or 2500 B.C.-–may still patiently await discovery; certainly Pigott and his colleagues’ work at Non Pa Wai provides a possible candidate-–but is it the earliest in Southeast Asia? Or will we find still earlier sites to the east and north? No one who has followed the incredible advances in Chinese archaeology as documented in successive editions of Chang’s text can fail to be astounded by the sheer volume of work and the even more startling new insights into the unexpected antiquity of metallurgy, domestication, and advanced Neolithic societies. As a result, East Asia is no longer “dominated for millennia by the unwavering influence of the Huang Ho Valley”, as I put it in a paper written almost 30 years ago (1980b:204). At that time no one dreamed of the sixth millennium Neolithic sophistication revealed by Hemudu, or the sixth and seventh millennium B.C. farmers of Cishan and Peiligang (Chang 1986:90–-95). It could well be that we will encounter such new discoveries in our own region as our volume of data begins to approach that accumulated by our Chinese colleagues. But at present we have a long way to go to catch up. We should remember that our database of excavations from Thailand (or elsewhere in Mainland Southeast Asia, for that matter) is quite small compared to the immense amounts of excavation carried out in China over the past few decades. For example, combined areas excavated at Non Nok Tha, Ban Chiang, Ban Na Di, and Khok Phanom Di total only about 725 m2– roughly 3.6% of the area excavated at the single Neolithic village site of Jiangzhai (Chang 1986:118). That percentage is approximately the amount of Non Nok Tha sampled by the excavations described in this volume–-still a relatively large percentage compared to the 0.5% of Khok Phanom Di sampled there. There is clearly much more work to be done!


In the meantime, the lengthy debate on the chronology of Non Nok Tha and related sites, if not providing definitive answers, has at least forced us to give very careful consideration to our dates and their implications. As White has remarked, “I challenge those working in other areas to examine their chronological data with the detail and rigor that we who have worked in Northeast Thailand have been forced to do” (1988b:175–-176). But we have at least demonstrated the presence in the Southeast Asian mainland of a sophisticated bronze technology in village contexts spanning at least a millennium–-a concept undreamed of prior to the excavation of Non Nok Tha. Notes: 1. An earlier version of portions of this chapter was presented at the Kioloa Conference on Ancient Chinese and Southeast Asian Bronze Age Cultures in 1988 (Bayard 1996–1997). 2. As the result of a mistaken editorial decision to add secular corrections to the clearly already corrected dates supplied to me, I am also in print with an even more “radical” estimation for the age of bronze and iron at Ban Chiang: 4500 and 2000 B.C. respectively (1980d:172). 3. In fact I went on to complete the sentence “and samples combining fragments from different squares or even different portions of squares were never used” (1979:21); what I had in mind was a circular area with a radius of a meter, which does not seem excessive for a heart-cum-rakeoff zone. 4. References are as follows: Excavation and stratigraphy (136 pp.): Bayard 1970, 1971b, 1980b Mainly a chronology (20 pp.): Bayard 1975, 1979 Fauna (231 pp.): Higham 1975a, 1975b Skeletal material (165 pp.): Pietrusewsky 1974a, b Bronze analyses (37 pp.): Pittioni 1970, Bayard 1972, Smith 1973, Stech, Wheeler and Maddin 1976, Selimkhanov 1979 Pottery and burial analyses (89 pp.): Bayard 1977a, 1984c (External relationships of Non Nok Tha and related sites are discussed in Solheim 1967b, Bayard 1975, 1980b, etc.) 5. Solheim wrote to Time about an error on the dating of rice at Non Nok Tha in an article of 9 February 1970, and I similarly complained about a second Time account of 19 November 1984 which outlined my repeated verbal comments to their reporter on the jocularity of the “WOST—World’s Oldest Socketed Tool” nickname for the socketed bronze axe recovered from NNT 1968 EP 3 P 90, and also the still quite controversial nature of the dates. In both cases we received abject written apologies, but of course no printed corrections; to the best of my recollection Gorman had similar difficulties with coverage of Ban Chiang. In these and other cases, neither Solheim nor I was given the chance to correct texts before they appeared. Thus Raymond’s popular account of metallurgy (1984), which described the impact of the new dates for bronze as an “archaeological landmine”, contains a number of inaccuracies and exaggerations, but responsibility for this can hardly be laid at our feet. Although I have never written contributions for semi popular journals, Solheim has encountered similar problems even with articles he has authored (errors in charts in his 1971 and 1972 National Geographic and Scientific


American articles and even the “Earlier Agricultural Revolution” title of the latter, which was applied by the editors without Solheim’s prior knowledge or approval). 6. The northern Vietnamese site of Dong Dau was also being excavated between 1965 and 1968 and revealed a clear transition from early bronze (late Phung Nguyen) to more advanced bronze technology prior to the appearance of iron (Higham 1989:178); however, the American-Viet Nam War delayed publication of any results for several years. 7. Gakushuin weights are estimates of pure carbon only. 8. These samples were given directly to Solheim for submission, and hence have no associated bag number. 9. Comments in quotes are those on forms sent by Solheim to me and David Harris for forwarding to the Oxford laboratory 4 October 1988. 10. I am uncertain whether the ceramic analyses Stargardt states she had carried out (1981:334) actually included Non Nok Tha wares; her 1972 survey of Non Nok Tha, Ban Chiang, and related sites is unpublished and unknown either to me or to the then Director of Archaeological Research for the Thai Fine Arts Department (Pisit Charoenwongsa, personal communication). The credibility of her argument is not boosted by statements like “all the artifacts discussed [from Non Nok Tha and Ban Chiang] do come from burials” which is of course patently false for both sites. 11. In the case of the fifteen or so vessels I selected for Otago, I attempted in general to choose representative vessel types that were of doubtful or unknown provenience (e.g., from Green’s 1965 unit-level test pits): useful for teaching purposes, but not for accelerator dating. All vessels that would hopefully prove attractive for museum displays were of course returned to Thailand. [Most vessels kept in Hawai’i were of unknown provenience WS]. 12. Higham briefly describes a small Bronze Age cemetery site 4 at Non Praw (Phraw?), some thirty km northeast of Non Nok Tha (1996:195-196), but no radiocarbon dates were obtained, and I have not had a chance to examine the pottery recovered. Higham estimates a date of 1000 to 500 B.C.



Chapter 6 Ceramics By Donn Bayard Introduction While excavating Non Nok Tha, we had the impression that it was a site relatively rich in ceramics. The upper layers contained large numbers of small, fairly worn sherds similar to those observed on paths and walkways in modern villages, while the lower levels yielded larger, less worn sherds, most of which seemed to derive from vessels in burials that had subsequently been disturbed. It was not until the excavation of sites like Ban Chiang (Gorman and Charoenwongsa 1976, White 1982) and Non Chai (Bayard, Charoenwongsa, and Rutnin 1982-1983) some eight to twelve years later that the relative poverty of the Non Nok Tha ceramic deposits became apparent. The 476 m3 excavated at Non Nok Tha produced only some 12,000 sherds (excluding burial-associated material)-â&#x20AC;&#x201D;only a little over 20 sherds per m3. By way of comparison, Non Chai produced some 2,500 sherds/m3, or about 900,000 in a volume only three-quarters that of Non Nok Tha. The sherd total for Ban Chiang, with a volume approaching that of Non Nok Tha (420 m3), is about 1.25 million: ca. 3,600 sherds per m3 (including burial as well as nonburial sherds; White, pers. comm). Despite the smaller sherd numbers however, and the fact that only portions of nonburial ceramics from the 1966 excavation were analyzed, the sample is sufficient to provide a clear picture of changing trends in Non Nok Tha ceramics, a general seriation of 1966 and 1968 area excavation layers, and comparisons with other sites in the Phu Wiang region and beyond. More importantly, the sample of 849 reconstructable vessels recovered from the site is currently the largest in the region. Once reconstruction, analysis and quantification of this large body of data had been completed in the early 1980s, a relatively large number of analyses were carried out. These were obviously mandatory for a number of reasons. First, given the difficult stratigraphy described in Chapters 3 and 4 plus the controversial dating of the site (see previous chapter), analyses of waste sherds recovered from levels can serve as a confirmation or rejection of the stratigraphic decisions made during excavation. If a large amount of correlation between layers and sherd typology is not present throughout the sequence, then clearly the stratigraphy was not correctly interpreted. As will be seen below, the analysis of the nonburial sherds from 1966 layer 21 (a mixture of EP 1-EP3 plus MP 1 levels) provides such an example. Secondly, a typology and analysis of the very large sample of reconstructable vessels can also serve as a check on our interpretation of the various chronological levels used in this report and on the stratigraphical ordering of the burials themselves. Both of these are needed not only because of the difficult stratigraphy but also to allow those interested to test alternative interpretations against the data as presented here, as at least two have already done without access to full data (Macdonald 1980, Higham 1994). Both sets of analyses should allow us to arrive at some understanding of the relationship between the rim form, temper/fabric, and decoration of the waste sherds and those whole


vessels presumably derived from the same level. Irregularities in the proportions of wastesherd and whole-vessel remains should of course suggest possibilities about specifically funerary vessel types as opposed to domestic ones, and analyses of form, size and contents recovered from reconstructable vessels should also provide inferences as to their function. Finally, of course, all of this should hopefully tell as much about relationships to other sites in the region, both in terms if shared form, fabric, temper, and decorative categorized, and in the identification of apparently exotic vessels and their place of origin. Such at any rate were our hopes, and I think we have been successful to some extent, although hopefully the data presented here will allow others to carry out further comparisons with the everincreasing database from Northeast Thailand and the rest of Southeast Asia. Nonburial and Occupational Sherds On completing the first excavation at the site in 1966, it was our impression that the great majority of the potsherds recovered from all layers belonged to a single ware that had persisted through most of the siteâ&#x20AC;&#x2122;s history: Sand tempered, cord-marked earthenware. Other types of finish and temper, such as plain or smooth sherds tempered with rice chaff or ground clay, were certainly present, but it was our subjective view that these minor categories were probably from imported vessels and for the most part appeared late in the sequence. Certainly the few sherds of stoneware, somewhat crudely made celadon, and porcelain that were recovered seemed strictly limited to the middle and upper levels of the Late Period occupation (with the exception of a single piece of celadon in a doubtful 1966 layer 11 context). The vast majority of sherds from these upper layers, however, seemed essentially similar to those in the lower levels, both in terms of temper (sand) and overall resemblance in form, with the dominant type round-bottomed, cord-marked, globular vessel with everted rim. We were, of course, aware of distinctive categories of vessel form and decoration peculiar to one or more of the burial levels, but in general the pottery throughout the prehistoric sequence at Non Nok Tha seemed to form a coherent tradition quite unrelated to that found in nearby villages at present. Although some formal resemblances as well as apparent similarities to manufacturing techniques used in contemporary villages (Solheim 1964b) were observed, the modern pottery nonetheless seemed quite different as far as temper, surface finish, and specific formal categories were concerned (Solheim, Parker and Bayard 1966:61-62). Subsequent examination of the pottery from both excavations has modified this overall view considerably. While the general lack of resemblance of the Non Nok Tha ceramics to those in use by modern villagers is still quite apparent, it is also clear that the prehistoric pottery was more variable than we had at first believed. It became apparent that our impressions of the overwhelming predominance of sandtempered, cord-marked ware were largely stimulated by our field examinations of the numerous funerary vessels of this sort. A qualitative examination of sherds rather than vessels, as well as detailed examination of the vessels themselves in the laboratory, revealed a reasonable variety of different tempers and elements of form and decoration. As will be seen below, quantitative analysis of the pottery from both seasonsâ&#x20AC;&#x2122; excavations has upheld this subsequent impression of relative variability; it has as well produced results that were not anticipated by our earlier subjective examination of the material.


Procedure Analyses of the pottery from the 1968 excavation began in 1968, after preliminary examination and classification within a hierarchical taxonomy of material adapted by Bayard from a scheme devised by Newman (n.d.) for Hawaiian archaeological materials. This taxonomic code, with some modifications, was used for all subsequent analyses through the mid-1980s. Although certainly rather primitive in the light of computer programming technology of the 1990s, it functioned well in the era of the slow-working mainframes, before the arrival of PCs with large hard disk storage capacities and spreadsheet programs. As a generalised coding system designed to allow the entry of all materials, ceramic and non-ceramic, in a standard eighty-column computer format, to facilitate sorting, catalogue compiling, and intra- and intersite comparison, it was not intended for the entry of specialised data necessary for detailed ceramic, metallographic, or faunal studies. It was basically a material classification indicated by a series of digits, with each succeeding digit representing a finer subdivision of the category indicated by the digit preceding it. The first division made was that between organic and inorganic remains, indicated by a 3——— and 7——— respectively. Following the basic division, 71- was assigned to ceramics, 73- to metal, 74- to stone, and so on. Thus a piece of sandtempered earthenware as designated as 71 (ceramic) 1 (earthenware) 2 (sand temper) 1 (medium to large amount of sand); similarly, a piece of low-fired, incompletely vitrified porcelain would be designated 714221. Non-ceramic material was classified in a similar fashion: bronze was designated 73313, while a piece of Phu Kradung formation sandstone (see Chapter 2) became 741123 (sedimentary-sandstone-local-Phu Kradung). The above data, plus site number, catalogue number, provenience, and reliability of provenience (definite, probable, doubtful, unknown) were entered on keypunch forms (again reflecting late 1960s technology), along with the weight and number of items described. These forms were of two sorts: a non-ceramic form, on which a generalized functional/descriptive code and some metrical data were also entered; and ceramic form, which contained considerably more information on the size, form, and surface finish of the item(s). Following the twenty-five columns devoted to provenience and material code, columns 26-33 were used to record rim data, including twenty-three form categories that accommodated all rims from the site. I must add that this has since proven totally inadequate for the range of rims from other Thai sites-—in particular General Period C sites like Non Chai (Bayard, Charoenwongsa, and Rutnin 1982-1983) and Ban Chiang Hian (Chantaratiyakarn 1983). Surface treatment and decoration of inner and outer rim surfaces and the measurements of rim diameter at lip, rim-neck diameter (external), and rime height were also coded (see Fig. 6-11). Shoulders of vessels that were finished or decorated in a fashion different from the body were described in columns 34-44, and the width of the shoulder area was entered in columns 45-46. A description of the vessel or sherds was entered in columns 47-56 (including a form description as well as finish for whole or partial vessels), as were measurements pf maximum width and total height of the vessel and average thickness of body wall. Vessels with flat bottoms, ringfeet, or pedestaled bases (and sherds from these) had form and finish variables entered in columns 57-70, along with measurements of foot basal diameter, foot-body junction diameter, and height of foot. Columns 71 and 72 were used to indicate the presence of striations and/or perforations on the rim or foot. Finally, quantity and weight of the sherds described were entered in columns 73-76 and 77-80, respectively. The full outline of coding instructions and definitions was devised with a view


toward incorporation of material from sites other than Non Nok Tha, and with slight modifications it proved adaptable in the 1970s to a number of other Thai sites (Bayard 1977a, Higham 1977b), although its limitations for the entry of specialized or detailed ceramic, metallurgical, or morphological data subsequently became apparent with the advent of more advanced computer technology. Nonetheless, it made possible relatively easy preparation of matrices for analyses using mainframe package programs during the 1970s. Using this code, 388 bags of nonburial refuse sherds from the 1968 excavation levels were processed by Bayard in Hawai`i in 1969. Some forty-eight restored vessels were also processed, but the amount of time required to assemble these meant that only this small fraction (about 6%) was finished by the time of my departure for Otago in early 1970. Still, this initial period of analysis made it apparent that one of the chief advantages of the generalized coding format lay in its ability to allow simultaneous analysis and comparison of both isolated refuse sherds and more or less complete vessels on an equal basis. Details for as much of the vessel as is represented by one or more sherds were simply entered in the appropriate places; thus entire vessels were fully described in a single eighty-column entry, using most of the eighty columns, while a single body sherd received only a brief entry describing temper, finish, and thickness in the relevant columns. Groups of body sherds of identical surface treatment, temper and thickness from a single bag were recorded as a single entry, as were shoulder, rim, and base sherds if these were identical in all respects (finish, temper, thickness, form, etc.). Sherds were processed by bag lots, each of which was assigned a catalogue number; most non-ceramic artifacts and restorable vessels in bags with refuse sherds were assigned a different number. All sherds were classified by temper, using a low-powered binocular microscope; fresh breaks were examined in all doubtful cases. Sherds were further subdivided on the basis of vessel part, finish, and dimensions, where these could be obtained by means of a concentric circle sheet and callipers. Each subdivision was then entered on a separate coding form for subsequent punching on cards. An average of about ten forms was required for each bag of refuse sherds, and total of 3,385 ceramic data cards were punched (including 116 nonpottery ceramic artifacts) in this initial 1969 phase of analysis, which processed some 7,500 sherds. Analysis of materials from both excavations continued under Solheimâ&#x20AC;&#x2122;s direction in the years following 1969, as described in Chapter 1. During a five-month stay in Hawai`i in 1980, I was able to complete the assembly and coding of the 1966 and particularly the 1968 restorable vessels, as well as coding of analyses of the 1966 nonburial sherds carried out some years earlier by B. Davis. Less than half of the nonburial sherd bags from this excavation were in fact processed; however, this discrepancy is not as glaring as it would first appear, since a very large number of the 1966 refuse sherd bags were of uncertain provenience or contained a very small number of sherds. Davisâ&#x20AC;&#x2122; analyses resulted in the coding of some 4,800 sherds; when combined with the 6,000 nonburial sherds from the 1968 excavation analyzed in 1969, plus additional sherds I analyzed in 1980, these total about 11,000--over 90%-â&#x20AC;&#x201D;of the approximate site total. The vast majority of the remaining sherds were of doubtful or mixed provenience (e.g., resulting from baulk trim, from disturbances, etc.), and would have been discarded from the subsequent analyses in any case, as were all coded sherds from doubtful contexts (including layer interfaces).


Nonburial Level Sherds: The Data In coding proveniences of nonburial sherds from both excavations, the system of local layers and sublayers peculiar to each area was used; however, to minimize complexity here, the overall system of three periods and seventeen levels presented in Chapter 4 will be used. As mentioned above, all sherds of doubtful provenience were eliminated from the analyses presented here, with the exception of a few from layer interfaces included in rim and base finish analysis. Our database is thus 4,223 sherds from the 1966 area and 5,349 from the 1968 excavation, or about 80% of the total of 12,000 sherds. The samples from the two areas are not directly comparable, in that: 1. The lowest levels of the 1966 area were not completely excavated due to time pressure (see Chapters 1 and 3); 2. The Late Period levels in the 1966 area were richer in sherds than their 1968 equivalents; and 3. As a probable result, the sample of 1966 sherds processed by Davis is strongly biased toward the Late Period (83% by number as opposed to 56% for the 1968 sample; see Table 6-1). In addition, it is evident that the criteria used by Davis in coding rim form, body finish, and particularly temper significantly differed in some cases from those used by me in my analyses. Accordingly, the two samples will be dealt with separately here.

The characteristics of Non Nok Tha ceramics have already been fairly fully described (Bayard 1977a). Using the 1968 sample alone. The description presented here is similar to this earlier publication but includes the 1966 sample as well and presents slightly different figures based on a somewhat reduced and more accurately provenanced body of data. This results from a few changes in provenience made after careful editing and correcting of the catalogue, plus the omission of all sherds from layer interfaces.


For purpose of formal and decorative analysis, only those sherds not associated with burials (i.e., in grave fills, etc.) were used; however, it is obvious that a large majority of these non-burial sherds are in fact derived from vessels in burials that were subsequently disturbed. This is true for levels of all three periods, but particularly so far the Middle Period. It also seems apparent, however, that some nonfunerary (i.e., presumed domestic) wares--or at least wares distance from those encountered in burials-—were encountered, in particular during 1968 Middle Period 5. Nonburial sherds were divided into four classes based on vessel part: body sherd (total for both excavations 7,317 or 76.4%); sherds from distinctively decorated shoulder area (1,714 or 17.9%); rim sherds (1,571 or 16.4%); and base sherds (292 or 3.1%). The rim sherds were treated integrally for purposes of formal analysis, but were subdivided into inner and outer surfaces in describing decoration and finish. Fragments from flat-bottomed vessels and those with ringfeet were also treated integrally in formal description, but the latter were also divided into inner and outer surfaces, although the number of inner ringfoot surfaces available--even the larger 1968 sample-—was so low as to be of limited value (see Table 6-10). Obviously, many sherds were encountered that fit into more than one of the four formal classes-—for example a sherd from the upper portion of a pot that retained not only a portion of the rim but areas of differentially decorated shoulder and body area as well. As a result, the number of "sherds" dealt with here is 10,894 or some 13.8% greater than the actual total of 9,572 sherds. Table 6-2 gives details on the numbers and weights of the various "sherd" categories in the 1966 and 1968 samples. It is clear that the 1966 sample contains a greater proportion of body sherds of considerably lighter average weight (5.1 g per sherd as opposed to 12.0 g in the 1968 sample); fewer shoulder and foot sherds are present in the earlier excavation's sample as well. All this reflects the Late Period bias of this sample vis-à-vis that from the later excavation, which includes a larger proportion of flat-bottomed and ringfooted vessels obviously derived from the type 5 and type 2 vessels common in the Middle Period (see vessel typology below), as well as a higher proportion of distinctively finished shoulder sherds (also more common in Middle rather than Late Period levels). In the tables that follow, sherd numbers are used in preference to weights in order to avoid overemphasizing the less common thicker wares (tempered with chaff and laterite or laterite alone)


Temper As mentioned above, sand was the dominant tempering agent throughout the history of the site, but it was by no means as overwhelmingly predominant as our observations in the field led us to believe; over 40% of the combined sample is in fact made up of other temper types. A total of eleven relatively distinct types of earthenware were distinguished, although only four of these comprised more than 5% of the sample each. It is important to emphasize that temper here used in a general sense rather than strictly applied to aplastic inclusions alone, as these in their own are not sufficient to distinguish otherwise quite distinctive wares (e.g., sand and chaff and fine chaff and sand have identical inclusions, although in very different proportions). Temper as used here is thus equivalent to "fabric" or "paste" rather than temper in the strict sense (see Vincent 1991); temper is employed primarily for ease of use in phrases like "laterite and chaff-tempered ware". A description of each type in order of decreasing frequency follows (Pl. 7); percentages of frequency by number in 1966 and 1968 samples are also given. Sand: (58% of 1966 sample, 57% of 1968 sample; P1. 7a). The temper consists of moderate to large amounts of medium to coarse quartz sand, in all probability of local origin, although this cannot be objectively verified without further research on clay sources in the immediate vicinity of the site. The majority of the sherds from the Early and Middle Periods have a gritty feel but are hard-fired (Meacham and Solheim 1979) and nonfriable. On the other hand, many Late Period sand-tempered sherds are noticeably softer, more friable, and contained more fine grained sand. But as this distinction was difficult to sustain objectively--particularly as several people were involved in the analysis--the two types were grouped together, as was the "crumbly sand" category distinguished in the earlier analysis (Bayard 1977a:76). The color of the sherds ranges from reddish buff through dark brownish grey, reflecting the range of colors found on the Class 1 (see vessel typology below) pots from which the vast bulk of these sherds derive; thickness averages between 4 and 8 mm. Sand and Chaff: (20% both 1966 and 1968 samples; P1. 7b). This ware contains less sand than the above, but still enough to be decidedly gritty; it also contains medium to fairly large amounts of rice chaff (almost always wholly carbonized) and occasional pieces of straw as well. The ware is usually blackish and thick, ranging between 5 and 10 mm. Chaff: (15% 1966, 8% 1968; P1. 7c). These sherds contain moderate to large amounts of rice chaff in large pieces; the ware has a soft feel and is variable in thickness from 4 to 10 mm. Surface colors vary from grey or pinkish grey through reddish buff to almost white, depending on firing temperature and amount of oxidation; sherd cores similarly range from black to light grey. Fine chaff and sand: (only 0.7% of the 1966 sample, but 6% of 1968; P1. 7d). Temper is composed of moderate amounts of finely divided rice chaff, plus a small amount of fine-tomedium grain sand; the latter may be a natural inclusion, as may be the case with the occasional small nodules of laterite that are found in the paste. Oxidation of this ware during firing was sufficient in almost all cases to produce a buff to reddish color throughout the thickness of the sherd and to completely oxidize the chaff particles, so that only the white siliceous skeleton remains. As with chaff temper sherds have a soft feel and are often slightly friable; they average 4-7 mm in thickness. Most of these sherds originate from the very common 2C, 2L and 4D vessels described below, and it seems apparent that


most post-MP 1 sherds of this type from the 1966 sample were incorrectly attributed to the chaff or sand-and-chaff categories by Davis in his analysis (cf. P1. 7c with 7b). Laterite and sand: (3.3% 1966,2.4% 1968; P1. 7e). This is a quite distinctive ware of a buff to light orange color, tempered with small fragments of laterite and a small quantity of fine-to-medium-grained sand; the sherd cores are often more reddish than the inner or outer surfaces. Rim and foot sherds often give signs of being thrown on a fast wheel in the form of vertical striations in the clay fabric, visible in cross-section, where it has been pulled up during throwing; however, body sherds are frequently roughly finished. Sherds average 5-8 mm in thickness. The ware sensu stricto is limited to the Late Period, but a few sherds of apparently similar fabric were found in 1966 Layer 21 (Plates 7c-d). Two additional vessels (NP-1780 and NP-1857) from MP 5 and MP 6 had a similar temper, but as might be expected without any sign of fast-wheel manufacture. Eight sherds of an apparently related temper but with the addition of chaff as well were described in Davis' 1966 analysis, but as I was unable to locate and examine these they have been omitted. Clay, clay/prepared: (combined in 1966 samples [0.5%]; 1.9 and 1.7 % respectively of the 1968 sample' P1. 7f). The tempering agent in both these categories is mainly clay, but in the clay-only type this is derived from ground potsherds; in the clay/prepared type it appears to have been produced from the firing and grinding specially prepared clay-andchaff balls, as is presently done in the region (Solheim 1964b;156-157). It is obviously difficult to distinguish between this latter type and ware tempered with identical constituents derived from the crushing of chaff-tempered potsherds; hence their merging in the 1966 sample. Yet some chronological significance seems to be present in the distribution of the two varieties in the 1968 sample (see Table 6-3), and for that reason they are treated separately here. Sherds of both varieties range in color from dark grey through light greyish buff, with usually darker cores; thickness is quite variable, ranging from 3 to 10 mm. Chaff and laterite: (only 0.1% of 1966 sample, but 2.2% of 1968 P1. 7g). This ware is similar in color to chaff-tempered ware but is in general more crudely made and always quite thick (8+ mm). Temper consists of nodules of laterite and some sand as well as a moderate amount of chaff. The low percentage in the 1966 sample is apparently due to the assigning of some of these sherds to sand-and-chaff or chaff categories. No temper: (1.4% of 1966 sample, but only 0.2% of 1968; P1. 7h). This is a greyish to whitish ware with no significant amount of inclusions of any sort; it is quite variable in thickness, ranging from 4 to 10 mm. The ware is almost wholly limited to the Late Period, and the bias of the 1966 sample toward this period explains its higher percentage in this sample. Sand and red pigment: (0.2% 1966, 0.6% 1968). This bright red ware is fairly heavy tempered with sand and perhaps some amount of red coloring matter (presumably the ferruginous siltstone that was common at the site, which may have been finely powdered before addition to the clay); alternatively, a clay source naturally high in iron may have been selected for use. Sherds average about 4-8 mm in thickness and appear to derive from the red-slipped 2D, 4E, and 5B vessels, some of which (but by no means all) have this temper. Predictably, most of these sherds are concentrated in the Middle Period levels in the 1968 sample; their rarity in the 1966 sample a in reflects the bias toward the Late Period Levels.


Laterite: (0.1% 1966, 0.5% 1968; P1. 7i). Another highly distinctive ware, with large amounts of laterite nodules used as temper; sherds are dark to bright red in color and very thick (10-15 mm). Like no-temper and laterite-and-sand wares, it is characteristic of the Late Period only. Its relative scarcity in the 1966 sample is thus puzzling; this might have resulted from the analyst's lumping of many sherds of this type with the laterite-and-sand category, as sand grains are sometimes found interspersed among the laterite nodules. Again, the senior author was unable to examine the 1966 sherds and resolve the question. Although this type is rare at Non Nok Tha, it considerably more common at neighbouring sites like Don Sawan and Non Nong Chik (Bayard 1977a:85). Judging from the one or two base sherds recovered, this ware composed mainly quite large, flat-bottomed storage jars. Tables 6-3 and 6-4 present the distribution by number and percent for these eleven temper types, which represent 98.8% and 99.3% of the total secure samples from the 1966 and 1968 excavations respectively. The remaining sherds comprise minor--almost unique-earthenware temper types (eighteen sherds) and stoneware and porcelain sherds from the Late Period levels (sixty-nine sherds); the latter will be discussed below. In both tables, temper types are presented in order of chronological predominance rather than in order of the overall frequency as above. Time-frequency histograms of the temper types described below (by number and percent per level) are shown in Figs. 6-1 to 6-4. The 1968 data are presented first in both cases in order to emphasize their greater reliability due to the larger number of sherds; the wider temporal coverage (i.e., the Early Period levels); and the fact that the criteria used by Davis in his analysis of the 1966 sample were clearly somewhat different from those employed by me. Despite this last there is a good measure of agreement between the two samples. The last four temper types (chaff and laterite to no temper) serve to set apart the Late Period levels, as does the rise in sand-and-chaff and chaff tempers; however, this last is not clearly reflected in the 1966 sample. The Early Period levels are of course only represented as such in the 1968 sample, and are distinguished by the predominance of sand temper (over 80%) and presence of clay rather than clay/prepared temper. In summary, the temper distribution provides good general support for the tripartite EP-MP-LP chronological division set forth in Chapter 4.


Fig. 6-1

Fig. 6-2


Fig. 6-4



Surface Treatment Despite our impressions in the field of relative lack of variability in the ceramic material and the predominance of cord-marking in the surface treatment of body sherds, analysis of the surface finish and decoration of shoulder and body sherds (for definition of vessel parts, see whole vessel typology below) was carried out using a large number of categories to represent fully whatever variation was present in the material. Hence over thirty categories of finish/decoration were distinguished by the time analysis was complete; however, only twelve of these were numerically important, with the remainder represented by only one to ten sherds each. The results of the analysis, as with temper/fabric, indicated that our initial impressions of uniformity were again unjustified; while the various cord-marked categories do indeed contain a slight plurality of the body sherds (45.6% overall), they are by no means as dominant as we believed in the field. The twelve major categories of dominant finish are given below in order of overall predominance at the site; five of these categories were used to describe rim and base finish as well. Plain: (42.4% of body and 55.0% of shoulder sherds overall). Plain-finished pottery is presumably made or at least finished with a smooth wooden paddle in most cases. Some crude examples appear to be smoothed by hand, but facets resulting from plain paddle are often visible on sherds. Cross-hatched cord-marking: (C/M; 41.8% of body of shoulder sherds overall). The surface of the mentioned vessel was finished while still not dry by beating in two directions with a cord-wrapped paddle. The angle of intersection of the two sets of imprints ranges between about 45째 and 90째, with an average of about 70째; marks on the base of roundbottomed vessels are predictably often confused and blurred. Cords appear to have been tightly spaced together on the paddle in most cases, but evidence if loose spacing is occasionally noticeable. Cord diameter generally ranges between 0.5 and 2.0 m. This and the preceding category contain the vast bulk of Non Nok Tha sherds. Red-slipped: (3.8% of body and 15.5 of shoulder sherds). Surface was covered uniformly with a (usually) thin red slip prior to firing. The 1966 total for this category is much higher than that for 1968 (36.9% as opposed to 5.5%), presumably due to Davis' inclusion of significant amounts of reddish-fired plain ware in this category rather than in its proper place. Unidirectional cord-marking: (3.6% body, 1.6% shoulder). In most cased it was not possible to distinguish between vertical and horizontal patterns due to the small size of many of the body sherds. Discrimination is possible with most shoulder sherds, however, and vertically cord-marked sherds are entered separately in Tables 6-7 and 6-8. Smoothed cord-marking: (2.6% body, 2.3% shoulder). This category is usually distinct from any of the various cord-marled categories and is hence not included in the totals for the latter. After manufacture using a cord-wrapped paddle, the cord prints are deliberately smoothed over, but not completely. Traces of the lines and intersections of the prints remain quite visible, but in most cases the twist print of the cord is obscured. Polished: (1.3% body, 3.4% shoulder). On completion of manufacture, the vessel has been allowed to dry to leather hardness then polished with a pebble, smooth stick or other tool before firing.


Diagonal unidirectional cord-marking: (1.3% body, 1.8 shoulder). It was possible to recognize this as a distinctive variety of unidirectional cord-marking due to the small size of the vessels to which it was usually applied; sherds in this category were mostly derived from the very common type 2C and 4D vessels. Cord marks were carefully applied in a series of imprints spiralling upward and almost always to the right; this would of course be expected if a right-handed potter held the vessel's mouth in her left hand while applying the paddle with her right. Red-slipped and polished: (1.1% body, 3.0% shoulder). The surface of the vessel had been polished after application and partial frying of the slip, but prior to firing. White slipped: (1.0% body, 1.3% shoulder). A light buff to white slip, often relatively thick (>1 mm), was applied to the surface before firing. This finish is mostly found in the Late Period levels. Organic coat: (0.6% body, 0.9% shoulder). The surface of the vessel was coated with an organic wax or resin of some type after firing. White-slipped over cord marking: (0.3% of body, nil of shoulder). As with white-slipped above, but applied over a cord-marked rather than plain surface. Judging from the more reliable 1968 sample, it also appeared to be characteristic of the Late Period, but numbers are so small as to be inclusive. Red-slipped plus organic coat: (0.2 body, nil of shoulder). Organic wax or resin applied to the red-slipped surface after firing. Minor types: The categorized above include all but fifty-one and fifty-five sherds respectively from the 1966 and 1968 secure samples. Most of these are minor variants of cord-marking (vertically and horizontally cord-marked body sherds, for example); others are examples of red slip applied over cord-marking or smoothed cord-marking (1966, twelve; 1968, seven). Two 1968 sherds showed signs of impression with a woven fabric, and a small number of sherds exhibited impressed designs made with a simple tool (hollow reed or pointed object-â&#x20AC;&#x201D;1966, five; 1968, seven); a compound tool (comb or dentate stamp-â&#x20AC;&#x201D;1968, one); and one sherd featured a rouletted stamped design similar to that on vessels NP-523 and NP-1723. Tables 6-5 through 6-8 give number and percent of each finish category by overall level; again, the 1968 data are presented first to emphasize their greater reliability. The 106 sherds not included in the twelve major categories are shown in brackets below the respective level totals. As with the earlier temper tables, types are listed in order of chronological dominance rather than overall dominance as listed above. Total percentages of all varieties of unsmoothed cord-marking are also shown.




The range of surface treatments used in rims and bases is predictably narrower, due mainly to the almost total absence of cord-marking. But while plain finish accounts for a large majority of both rim and base sherd percentages, there are significant amounts of polished, red-slipped, and slipped/polished rims present. Tables 6-9 and 6-10 present these data only for the 1968 samples, as it is more representative of the sequence as a whole--both numerically and in terms of temporal spread--than the notably smaller 1966 sample. Rim and base sherd totals differ slightly from those presented under form analysis below due to separate classification for inner and outer surfaces and to the inclusion of a number of sherds from level interfaces (as well as a few unique sherds) that were not omitted from the site totals. Due to the relatively small number of base sherds, actual numbers rather than percentages are given.


Rim and Base Form Rim sherds were classified by shape into one of the nineteen categories established by preliminary observations of material from Non Nok Tha and nearby sites carried out at the field laboratory at Khon Kaen University over the summer of 1968; three additional types were distinguished by Solheim and his students while restoring whole vessels in Hawai`i during the 1970, but they are not present in the non-burial refuse sherd samples. Small rim fragments whose form could not be ascertained were coded as type 99 ("unclassifiable"); these made up 2.02% and 7.10% of the 1966 and 1968 samples respectively. Thirteen of the nineteen categories (plus the unclassifiable sherds) account for 97.4% of the 99.4% of the 1966 sample, leaving only about 3% of the classifiable sherds in the remaining six categories; some of these rare types will, however, prove of use in establishing ties with other sites outside the Phu Wiang area. The thirteen major rim types are shown in Fig. 6-5, arranged in order of chronological predominance. Table 6-11 and 6-12 list the number and percent of these major types (and the unclassifiable sherds) by level. Time-frequency diagrams of the occurrence of these types by level percent are shown in Figs. 6-6a and 6-6b for the 1966 and 1968 excavations by percent only, as the number of sherds is not great and the samples sizes for the 1966 Middle Period levels in particular are quite small. As with sherd temper, the 1968 histogram presents a clearer picture. Type 13 rims are dominant during the Early Period but continue as common forms throughout the rest of the sequence; types 14, 03 and 05 are the only other common Early Period forms. The Middle saw the introduction of the new rim forms (06, 01, 09, 16, 07), and types 02, 11, 12, and perhaps 10 are more characteristic of the Late Period. Fig. 6-5


Fig. 6-6a

Fig. 6-6b



Throughout history of the site, the majority of pottery consisted of round-bottomed vessels; this is still the case with contemporary earthenware vessels with the exception of flowerpots and spittoons (Stoneware vessels are usually flat bottomed as well. See Calder 1972.) Despite this dominance of round-bottomed forms, vessels with flat bottoms ad ringfeet certainly are present; they are particularly common in Middle Period levels, where they were obviously viewed as desirable (if not absolutely necessary) offerings in graves (see Chapter 8). In the Late Period levels, flat-bottomed forms seem to be primarily nonlocal imports, judging from their relative scarcity. Considerable varieties of base forms are present however showing significant chronological change; for this reason the data are presented here, although it should be borne in mind that the sample is a relatively small one, particularly in the case of the 1966 samples; only MP 1, 4, and 5 and LP 1-5 have over ten sherds per level in both samples. In classifying base sherd forms, it was found that four varieties of ringfeet, one type of pedestal, and four types of flat bottom were present (See Fig. 6-5). Low vertical ringfoot: (abbreviated LRF in Tales 6-13 and 6-14): as illustrated for vessel types 2A and 2F in Fig. 6-2. See also individual vessel type figures and descriptions. High vertical ringfoot: (HRF): as illustrated by vessel type 2L in Fig. 6-12; also more or less vertical rather than conical in profile but considerably taller than above. Conical ringfoot: (CRF): a flaring ringfoot very common at the site, as illustrated by vessel types 2C, 2D, 2E, 2J, 2K, and 4C-4G. Filleted ringfoot: (FRF): similar to the low vertical ringfoot, except that clay from the sides of the vessel has been smoothed down over the top of the foot to make a smooth curve and eliminate any distinct angle between the bottom of the vessel proper and the ringfoot; illustrated by vessel type 2B. Pedestal: (PED): a sort of giant conical foot, comprising over one-third of the vessel's total height; limited in occurrence to vessel type 4B. Plain flat bottom: (PFB): as illustrated by vessel types 5 and 6; base of the bowl or pot is flat rather than concave. Concave flat bottom: (CFB): the center of the flat base been pressed in to make it slightly concave; occurs on some type 5 and 6 vessels. Ridged flat bottom: (RFB): a ridge of about 1-3 mm protrudes horizontally at the sidebase junction. This type is almost wholly limited to the Late Period (the one sherd recovered from 1968 MP 5 is probably intrusive), and most if not all of these sherds appear to be from vessels thrown on a fast wheel. No whole or even partial vessels having this base type were recovered. Multiple-ridged flat bottom: (MFB): two or more concentric ridges, the wider on the bottom, make a slightly flaring flat base; as with the above, these sherds derive from thrown vessels and are limited to the Late Period.


The number and percent of eight of these nine base types recovered in 1968 and 1966 is presented below; since only one pedestal sherd was recovered from a nonburial context, this type is omitted. As with all other tables aside from 6-8 and 6-10, only sherds of secure proveniences are included, and those from layer interfaces are excluded. The types are again presented in what seems to be the best order of chronological dominance, with unclassifiable sherds presented last.


Stoneware and Porcelain Finds of pottery other than earthenware were very rare and predictably limited to the Late Period levels. Only twenty-one stoneware and forty-eight porcelain sherds were recovered from both excavations, almost all of them quite small fragments (weight 480 g and 414 g respectively). The stoneware sherds were for the most part of a brownish to greyish color and tempered with small quantities of sand but some apparently untempered sherds were found as well; most were unglazed (Table 6-15), but several bore traces of a brownish glaze. Most of the porcelain recovered had a fabric tinged with brown rather than pure white and appeared to be incompletely vitrified; it resembled that recovered by the Pa Mong Survey Programme in 1974-1975 from sites in the Loei region, and it may also have had its origins in the Kalong or related kilns of northern Thailand, reaching the Phu Wiang region by trade down the Mekong (Bayard 1980c:74). These sherds appeared mainly to derive from celadon cups and plates.

Only nine sherds of high quality export porcelain were found, including part of a spoon and portions of a small cup, both which a blue-on-white pattern. Unfortunately, the small sample of stoneware and porcelain recovered allows no firm inferences on varying amounts of outside contact during the Late Period; the high numbers of finds in levels LP 2 and 5 represent more intensive occupational use of the site or simply sampling error. Nonburial Level Sherds: Analysis The data have been presented in relatively full form above for three main reasons: a) to allow correlations between the nonburial refuse sherds and the whole vessel typology presented below, and to attempt to detect any wares found in the former that are not represented in the funerary vessels (i.e., presumed domestic wares) during the Early and Middle Periods; b) to demonstrate that the relative ceramic sequence supports the stratigraphy detected and followed during excavation and is hence secure, despite the site's still-controversial absolute chronology; and c) to allow others to perform manipulations of the data using techniques different from those employed here. We are fortunate to have one such study already available for comparison: a detailed analysis by most of the 1966 and 1968 ceramic material carried out by Macdonald (1980). During a visit to Hawai`i in 1978, Macdonald was able to examine the first hand many of the 1966


and 1968 restored funerary vessels; ha was also able to make use of the 1968 level sherd data that I coded in 1969, carry out his own coding of Davis' 1966 data, and gain access to all available whole vessel data (about 80% had been coded at that time). Regrettably, much of this last body of data had been coded following descriptions carried out by Solheim's undergraduate students during the years 1971-1977, and subsequent rechecking of the coded data against the original vessels carried out by me in 1980 revealed over one hundred errors of provenience and description, many of them quite serious. Moreover, Macdonald's time in Hawai`i was necessarily limited, giving him only a brief opportunity to gain a knowledge of the ceramics in depth. As a result, some of his finish and temper classifications, like Davis', do not agree with those made by the present authors, who have many years' familiarity with the material; for example, several of the rims classed as plain by Macdonald proved to bear traces of fugitive red slip when examined by us; similarly, some rims described as red-slipped had in our estimation no slip, but were simply the outcome of a well-oxidized firing of high-iron clays. However, this is by no means an attempt to downgrade the overall value of Macdonald's study, which was methodologically sophisticated and highly innovative to our knowledge the first of its kind to be carried out in the Mainland Southeast Asian region. In fact, despite a very different methodological and epistemological approach, a large measure of agreement exists between Macdonald's results and those arrived at in this work, although it will be apparent that differences are obviously present. Many of these differences are due to coding errors in the data, which I subsequently corrected; Macdonald was aware of this problem and tried to minimize it through the use of presence/absence coding rather than absolute frequencies. Macdonald also suggests (pers. comm.) that some of his problematical results may derive from limited firsthand experience with the data, although the reasons for such problems are not necessarily the ones we discuss here. Nonetheless, we know of few sites where a complex body of data has been independently analyzed by researchers using markedly different approaches (Macdonald's deductive framework as contrasted with the largely inductive one employed here), and for this reason alone we feel that extended discussion of his methods and results is of considerable value. We will return to his study at several points in our discussion of ceramics and in Chapter 8 (burials) as well. If we turn first to the data as presented here, an examination of the tables in the preceding section makes it apparent that correlations are clearly present between certain of the form, finish, and temper categories. Such correlations can of course result from a number of factors: actual co-occurrence of the attributes on vessels of a particular type; coexistence of the attributes on some of the different vessel types (described later in this chapter) during the same period; or merely coincidence due to sampling error. However, if we examine the correlation coefficients of the more common values of the rimform, finish, and temper variables with each other through the various Early, Middle and Late Period levels, it is clear that a general chronological pattern is present. Table 6-16 presents the correlation coefficients r of the values of the common categories (> 1% of the total) for each of the three variables as derived from a matrix of occurrence by number of sherds for each of the 1966 and 1968 levels and of both excavations combined.




Given the deficiencies in the 1966 samples described above, it is obvious that major reliance must be placed on the 1968 sample; this is reflected by the reduction in number of variables available for analysis in Tables 6-16b and 6-16c, as well as by the Late Period bias in temper types that can be included. If we then concentrate on the 1968 data in Table 6-16a, it is apparent that the attributes of each of the three variables may be loosely grouped into two categories: those more characteristic of the prehistoric levels at the site (Early and Middle Periods) and those that are more typical of the parahistoric period (i.e., the Late Period levels deposited during the Dvaravati, Vientiane, and Rattanakosin eras,


which may be roughly dated to A.D. 900-1500, 1500-1830 and 1830-present in the Phu Wiang area). Within these general groups, some associations obviously reflect actual co-occurrence of the attributes on the vessel types described below, this allowing some degree of correlation of waste sherds and whole vessels. For example, the high positive association between sand temper, type 03, 13, and 14 rims, and cross-hatched cord-marking in an obvious consequence of the frequent occurrence of vessel types 1A, 1G, and 1C in burials of the Early Period and early Middle Period. Equally clear is the association of fine chaff and sand temper, diagonal cord-marking, and type 06 or 05 rim on the very common 4D vessels of the Middle Period (the relatively high association of type 13 and 03 rims with this temper probably reflects the frequency of type 2C and 2E vessels as well as temporal coincidence). Similarly, the relatively high positive association of red slip and type 05 and 06 rimforms probably results from their co-occurrence on 4D and 5B bowls. Many of the associations are merely chronological, however, arising from the cooccurrence of two attributes in different vessel types during the same period; thus the plain-finished laterite-and-sand are shows a high positive association with organic as well as plain body finish. But and examination of the actual sherd data makes it clear that the first of these associations is the result of chronological coincidence; while sherds of this ware are indeed plain finished, organic coating is totally lacking and the major rimforms of the ware are type 13 and 15, plus some 05 and 16 bowl forms. Type 05 rimform is even more positively correlated with sand temper than it is with fine sand and chaff, reflecting both temporal coincidence and the actual co-occurrence of the two attributes in type 5 bowls. Although domestic pottery during the Late Period appears to have consisted of sand, sand-and-chaff, and chaff-tempered wares in descending order of frequency (See Table 6-3), there is some slight evidence that at least the second of these was also employed as a domestic (or at least nonfunerary) ware during Middle Period 4 and 5 times (the latter level being the only one to yield substantial occupational evidence as well as funerary features). Sand-and-chaff ware is certainly far more prevalent in the nonburial sherd sample than it is in the vessels from graves of these levels (9% and 11% respectively, as compared to 1% and 0% of grave vessels by weight). However, the slight positive association between this temper type and Late Period rimforms is due to their cooccurrence in the Late Period, as the MP 4 and 5 sherds of this ware feature a mix of 13, 10, 11, 12 and 17 type rims. The very high association of plain finish with this temper type does seem to reflect an actual co-occurrence in the Middle as well as the Late Period levels, as an examination of the sherd data reveals that most of this ware was plain finished. In general, the 1968 data clearly support a major division of the sequence at the Middle/ Late Period boundary, with the latter clearly set off by rimforms like 02, 12, and 10 (although these are of course not exclusively limited to the Late Period); the frequency of chaff and chaff-and-sand tempers and the appearance of the laterite-and-sand, chaff-andlaterite, laterite alone, and no-temper categories; and the increased proportion of plain and red-slipped finished, plus the appearance of a number of new finished (organic coat, white slip, etc.; see Table 6-5). Certain of the attributes (sand temper, red slip) appear to be common throughout the sequence, although sand temper is strongly associated with the EP/MP finish and rimforn variables and only very slightly so with the LP variables.


The 1996 data appear to offer a similar picture, but they cannot be used as strong corroboration because they are so heavily biased toward the Late Period (e.g., 83% of total sherds, 85% of body sherds, etc.). This is apparent in Table 6-16b, where the EP/MP temper types and several of the rimform and finish categories included in the 1968 sample were so rarely detected in the 1966 sample as to preclude their inclusion. The correlation patterns of the combined sample (Table 6-16c) show general similarity to the 1968 sample but are apparently blurred by the bias of the 1966 sampling procedure and by the limitation of a lesser number of variables having more than 1% in each sample; hence the data are heavily weighted toward the EP/MP levels in the rimform and finish categories and toward the LP in the temper categories. While the 1966 and combined samples do provide general confirmation of the patterns evidenced by the 1968 data, the limitations are obvious. This sampling bias is unfortunate, as the later MP levels 7 and 8 are represented only in the 1966 sequence (albeit by only 5% and 2% of the total 1966 sample respectively). The question of the nature of these two MP levels is an important one, as Macdonald's study implies that the break in the sequence is not to be found between MP 8 and LP 1 but rather between 1966 LP 2 (i.e., 1996 layer 9) and 1968 LP 2 (lower layer 5 in the 1968 sequence). His approach to the data differed from that employed here; rather than treat attributes of the nonburial sherds separately, as we have presented them above, Macdonald adopted an approach that combined nonburial level sherds, sherds from features, and the restorable funerary vessels into a number of ware groups, which were further subdivided into wares (1980:141-197). His basic criterion for establishing these was treatment of the rim; this resulted in eight groups (plain, red-slipped, polished, slipped/polished, whiteware, crude blackware, red pigment ware, and miscellaneous). Three of these groups were than explicitly divided into a series of wares on the basis of temper categories, using most of the temper categories employed here but adding some distinctions I did not make while merging some categories that I kept distinct (e.g., it is apparent that his "Plain LS Ware" is in fact a subdivision of what I referred to as fine-chaff-and-sand, rather than laterite-and-sand [Macdonald 1980:165]). Following the ceramic classification, Macdonald then constructed a matrix of frequencies by layer, using presence/absence rather than raw numbers, percent, or other measures of relative abundance due to the coding problems mentioned above. A matrix of phi coefficients was then constructed (for presence/absence data, phi = r). These were then divided by a maximum value for phi based on sample size to ensure that all values for all phi lay between +1.0 and -1.0 and to eliminate variation in sample size. Macdonald then seriated the resultant matrix using the Lingoes MINISSA multidimensional scaling routine (Lingoes 1973, 1979). As he emphasizes, this routine has the valuable attribute of possessing the capacity to fail if the stress or distortion required to derive an ordering of the data is too great; it will not-â&#x20AC;&#x201D;as will cluster analysis, taxonomy, and other routines-â&#x20AC;&#x201D;impose order on a data set when such order is not present (Macdonald 1980:200). The 1996 and 1968 samples were first seriated separately followed by a combined seriation using both samples. The results produced "clear and consistent ordering" (1980:201). The layers seriated almost exactly in the stratigraphic sequence given in Chapters 3 and 4; all of the 1968 layers were in their exact physical order. However, the less secure 1966 sequence


produced two discrepancies. One of these was the seriation of layer 21 above 1968 layers 7, 8, and 9; this led Macdonald to conclude that the Early Period/Middle Period distinction lacked substantiation. Owing to this, he placed 1968 layer 9 alone in his Phase I, and grouped 1966 layer 21 along with layer 20 and 1968 layers 7 and 8 into his Phase IIa (1980:209). But as we outlined in Chapter 4, 1966 layer 21 was a mixture of materials resulting from its incomplete excavation at the end of the season; most of the whole vessels are of MP 1 origin, with Early Period graves reached only in the eastern portion of the site. It is hence no surprise to find this layer seriating with 1968 layer 7. The justification for grouping 1968 layer 8 (EP 2 and 3) with layer 9 will become apparent when the whole vessel typology and frequency distribution is discussed below; however, it should be noted that even in Macdonald's seriation of the 1968 ceramics, layer 8 is slightly closer to 9 than to 7 (1980:fig.4.11). The second discrepancy between the two sequences is more difficult to resolve. Macdonald's combined seriation clearly grouped 1966 layers 9 and 11, along with 1968 upper and middle layer 6, as earlier than 1968 lower and middle layer 5; as mentioned above, Macdonald accordingly placed the former group in his Phase VI. This of course extends the Middle Period forward in time and implies a greatly reduced gap between the Middle and Late Periods. Unfortunately, the 1966 level sherd sample is inconclusive here; the figures for LP 2 in both samples are in fairly close agreement as far as temper frequencies are concerned, which would imply that at least 1966 layer 9 should be grouped with the Late Period; however the high proportion of sand temper in LP 1 suggests a greater affinity with the Middle Period. The fact that 32% of all our stoneware and porcelain originated from LP 2 (over half of it from 1966 layer 9) strongly suggests to us that this level at least must be grouped with the parahistoric Late Period; the situation with LP 1 is less clear as the one piece of porcelain recovered may be intrusive from above. But our view at present is that Macdonald's grouping of 1966 layers 9 and 11 with the Middle Period may well result from the omission of stoneware and porcelain in his matrix; he also included many restorable vessels from layer 9 and 11 features, in addition to the level sherds analyzed earlier on by Davis. The former were largely from cremation burials, and all our evidence to date suggests that these vessels were "mined" from Middle Period burials specifically for use as cinerary urns; many of them from layer 9 (only two or three restorable vessels were recovered from layer 11) are common Middle Period types, such as 2C (5 vessels), 5A (3), 5B (2), 2L (1), and 2A (1). Hence the grouping of layer 9 with the late Middle Period is not surprising. Macdonald questions the "mining" explanation for this similarity because of the supposed absence of stratigraphic evidence (1980:200 fn.3). However, the section drawings presented in the present report offer ample evidence of empty Late Period pits, some of which disturb Middle Period burials; more over, five cremation burials with type 2C vessels are also cut from 1966 layers 4-6, which Macdonald agrees are clearly historic layers (see also level sherd analyses below). Layer 11 remains more problematical, but Parker's description of the stratigraphy presented in Chapter 3 makes it clear that a gap of considerable proportions is represented by a largely sterile layer 12; in addition, seven cremation burials (albeit not all complete) give evidence of considerable change in funerary ritual, allying layer 11 with layer 9 rather than 13. Despite this, we would not wish to rule out completely the results if Macdonald's seriation; the possibility does exist that layer 11 is somehow transitional


between the Middle and Late Periods. We will be returning to Macdonald's study later in this chapter and in the concluding chapter as well. Further Statistical Analyses Our analyses of the Non Nok Tha level sherd date were carried out using the SPSS Factor subprogram (cf. O'Shea 1984:66). Like Macdonald's MINISSA program, factor analysis has the advantage of not imposing a preset program order on a body of data in terms of deriving clusters or a taxonomic tree, but rather it does so by simplifying or reducing a complex data set to a much smaller number of measures accounting for observed interrelationships in the data, or factors, which represent the best fit of the correlation coefficients of the different variables. As Vierra and Carlson (1981) have pointed out, factor analysis can be misleading in that it may produce clusters from a correlation matrix with low measures of association that could have arisen merely by chance (e.g., where the matrix contains few or no value for r of >+0.3). However, we have examined the correlation matrices of all factor analyses presented in this study, and in all cases the number of values of r >+0.6 (probability die to chance= .001) exceeds the number expected by chance by factors ranging from 67 to 1000. Hence there is little likelihood that the resultant clusters are due to random association. In some cases it is obvious from the variable scored on the different factors (three were usually derived in our analysis, which was sufficient in most cases to subsume 90% of the variance) which variables are positively or negatively associated with a particular factor; in other cases, of course, the variance expressed by a particular factor is made up of a combination of different variables. Most of the analyses used here are Q-mode, examining the differential distribution of variables from case to case (i.e., from level to level or pot to pot). As stated at the beginning of this chapter, one of our major goals in level sherd analysis was similar to that of Macdonald's MINISSA runs: We expected that sufficient chronological variation in the level sherds would produce a seriation that would serve as a check on the site's complex stratigraphy as well as on the proposed correlation between the 1966 and 1968 excavations put forth in Chapter 4. However, R-mode analyses of the interrelationships of the variables themselves were also carried out in some cases, and the matrices of correlation coefficients produced as a necessary step in the factoring procedure were also consulted (e.g., Table 6-16 above). Most variables were analyzed by both number and weight of sherds; in most cases, date were standardized before running and occasionally a percent matrix was employed, but in all those cases where one or the other or both of these procedures were used, the results were essentially identical with earlier runs using raw data. By and large, the results agree with those of Macdonald in producing a consistent pattern, at least as far as the three gross divisions of the sequence are concerned; inconsistencies are certainly present (in particular within the 1966 sample and between it and the 1968 sample), and some indices, like body finish and temper, are more successful in producing a coherent ordering than others (like rim-and-baseform). It is unfortunate that the 1966 sample-â&#x20AC;&#x201D;which according to the correlation proposed above contains several Middle Period levels not represented in the 1968 sequence-â&#x20AC;&#x201D;is so under-represented in these levels (only about 17% of the 1966 sample derives from pre-LP levels). However, the temper variables produced a clear clustering of layers 20 through 10, comprising all of the Middle Period levels plus LP1. The remaining LP levels were rather scattered, and later 21 was set well apart, probably representing the large mixture of sherds derived from


disturbed EP burials. The analyses of 1966 rimform, shoulder finish, and baseform produced no satisfactory results, perhaps as a result of small sample sizes (in the case of the last two variables, up to eight layers had to be omitted from the analysis, as they contained no sherds in these categories). The 1966 body sherd finish gave a general separation between LP and MP levels, eight layers 11 and 12 in an intermediate position and layer 21 again set well apart from all others. With its larger size and better representation of the Early Period, the 1968 sample produced more satisfactory results. Temper distribution produced a very tight cluster of all three EP levels with MP 1 (Figs. 6-7a and 6-7b) with MP 4 and 5 closely associated; the LP levels are not associated and are widely dispersed, probably reflecting the greater variety of temper types present. Analyses of rimform by number and weight produced a general separation between MP and LP levels, save that LP 5 was fairly closely associated with the EP and MP group. Finish of shoulder shreds (Figs. 6-8a and 6-8b) yielded a general division of the 1968 layers into EP, MP, and LP groups with the exception of the lower layer 5, which was more closely allied to the MP (hence providing some support for Macdonald's hypothesis discussed above). Fig. 6-7a

Fig. 6-7b

Fig. 6-8a

Fig. 6-8b


The body finish results (Figs. 6-7a-b, 6-8a-b and 6-9b) were very similar to those of the temper analysis: a very tight cluster of the three EP levels plus MP 1, with MP 5 and 5 somewhat removed; LP 2 in a medial position; and a second tight cluster made up of LP 3-6. As the major break in the sequence of funerary vessel types occurs between EP 3 and MP 1 (see whole vessel typology below), it is obvious that the close clustering of the latter level with the EP levels reflects no the pottery in MP 1 graves but rather sherds derived from the digging of these graves and the subsequent upward displacement of EP vessel fragments. MP 1 had by far the largest number of burials of any of the EP or MP levels in the 1968 area (See Table 6-24) and at a rough estimate about 0.25 m2 of excavated soil—including upwardly intrusive sherds-—would remain mounded above an adult grave or scattered around it. A similar but by no means identical process was probably the cause of the intermediate position of LP 2 between the two main clusters; as mentioned above, "mining" of the MP burials for cinerary vessels was common during LP times, and such mining would of course have displaced MP sherds upward along with the sought-after vessels themselves. In this case, however, the pits excavated were smaller than graves (see 1966 sections in particular) and largely refilled with the excavated soil, resulting in some upward displacement of earlier sherds-—but less than in the case of grave digging. The two processes are discussed in more detail in Chapter 8. Fig. 6-9a

Fig. 6-9b

The results of the 1968 base sherds analyses were less satisfactory; levels EP 2 and LP 6 had to be omitted due to small sample size, and the first three factors extracted accounted for only 78% of the variance. Only two loose clusters are apparent: MP 4 and 5 with LP 2; and the remainder of the LP levels. The earlier levels are widely scattered. As the samples from the two excavations are not equally weighted by level, analyses based on both samples must be viewed with caution. However, several runs using combined sample from both excavations were carried out. In the combined temper results, two factors were sufficient to account for all but 2% of the variance; however, it is clear that much of the variance results from the application of different classification criteria by myself and Davis (Fig. 6-10a). Thus we have two separate EP/MP clusters (probably resulting at least in part from Davis' use of sand-and-chaff temper for my fine-chaff-and-


sand), with LP levels 1, 8, 10, and 11 also included in Davis' cluster; the remaining LP levels, plus 1966 layer 21, are dispersed. The outcome of the combined body finish sample (Fig. 6-10b) is far more satisfactory, reflecting the more clear-cut criteria involved in the classification; again, two factors accounted for all but 3% of the variance, and both of these factors are obviously chronologically significant. As predicted by the stratigraphic correlation presented in Chapter 4, 1966 layer 21 clusters with the 1968 EP levels and 1968 MP 1, presumably due to the large amount of EP material incorporated into these deposits by grave digging in the early Middle Period; hence 1966 MP 2 is also close to this cluster. Most of the later Middle Period levels form an equally well defined group, which contains 1966 LP 1. This would seem to indicate either a large amount of disturbance from 1966 layer 11 down into the MP levels, with resultant upward displacement of the MP sherds; or, alternatively, that Macdonald's hypothesis of grouping LP 1 with the Middle Period is correct. The Late Period levels from LP 3 upward form another quite well defined cluster clearly distinct from the two earlier periods, with both years' LP 2 levels again in an intermediate position, accompanied by the 1966 MP 8 sample. This anomaly could be the result of the small sample (only 2% of the 1966 total) from MP 8, but this is about the same sample size as the other pre-LP layers; the anomaly cannot be explained by the factors that produced the aberrant seriation of LP 1 and 2, and we can only conclude that the position of the topmost level of the Middle Period is still unclear as far as this particular index is concerned. Fig. 6-10a

Fig. 6-10b

R-mode analyses of the twenty-seven major (>1% f total) temper, rimform, and body surface variables were also carried out for the 1968 material, using raw number data. The results were clear, with two clusters obviously representing distinct LP and EP/MP variables. The former includes the temper types and rimforms shown as representative of the LP in Table 6-16a (not surprisingly, as the factor analysis was derived from this correlation matrix), with red slip included along with organic and plain finishes. The remaining attributes from an equally well-defined EP/MP cluster, including sand temper. Within this cluster, fine chaff and sand temper, 06 rimform, and diagonal cord-marking form


a clear sub cluster representing the co-occurrence of these attributes on type 4D and 2C pots. As the results are obviously a reflection of the matrix already presented above, they are not figured here. A similar run using the 1966 matrix (See Table 6-16b) produced no clear clustering, due to the differing criteria employed by the analyzer of that sample as explained above. However, a combined sample using the matrix in Table 6-16c also yielded two distinct groups, with an LP cluster containing all those attributes listed in the table as typical of the LP save 01 rimform, but including red-slipped finish as well. The remainder of the attributes form a loose EP/MP cluster, with sand temper the closest of these to the LP cluster. The results in general thus provide support of the ordering of the various attributes as presented in Table 6-16, although the 1966 analysis is of course inconclusive. We have presented the nonburial sherd data in considerable detail, not only to support our general contention that during most of the Early and Middle Periods the site was used chiefly as a cemetery (with most nonburial sherds thus deriving from earlier burials distributed by grave digging-â&#x20AC;&#x201D;MP 5 being a probable exception), but also as a test of the overall stratigraphy. This is of obvious importance given the continuing debate on the chronology of the site and of bronze metallurgy in the region as a whole (see Chapters 7 and 9 for full discussion). Taken in conjunction with the evidence provided by whole vessel typology and distribution, we conclude that-â&#x20AC;&#x201D;although the 1966 sample of nonburial sherds in particular poses problems-â&#x20AC;&#x201D;the overall picture is quite consistent with the stratigraphy presented in the earlier chapters once the probable EP origin of most of these sherds in the early MP layers is taken into account. Given the pioneer nature of the Non Nok Tha excavations and the stratigraphic difficulties encountered in both seasons' work, the results of both our analysis and Macdonald's are in many ways better than we would have predicted, and they support the three basic divisions of the site's sequence, although doubts about the exact affiliation of MP 8 and LP 1 must remain. Whole Vessels and Non Pottery Ceramics By far the most time-consuming task we faced in analyzing the Non Nok Tha materials was the reconstruction of the large number of restorable vessels, almost all of them from burials. A few vessels were either wholly complete or almost so, but most were in 30 to 60 pieces (and in one case some 365 pieces), entailing anywhere from a day to two or three weeks to reassemble successfully. I began the work in 1969 with the help of Daisy Bayard and other graduate students. It continued under Solheim's direction through the 1970s, first with the support of NSF funding and, when this was exhausted, by the voluntary efforts of several individuals. We are particularly grateful to Virgil Meeker and Ludy Solheim in this regard. Finally, again helped by volunteers, I was able to complete the reconstruction of the remaining two hundred or so vessels while in Hawai`i on sabbatical leave from Otago during the summer of 1980. Both Solheim and I feel that the time and effort spent on this aspect of the analysis was worthwhile, as it resulted in a corpus of some 847 more or less complete vessels; this, to our knowledge, is the largest sample of prehistoric pottery vessels available from any site in Southeast Asia. Whole Vessels: Procedure and Typology


A number of techniques were tried in reconstruction. As the vessels had to be returned to the Thai Fine Arts Department and as shipping funds were limited, it was decided to carry out temporary reconstruction using a soluble glue so that the vessels could be disassembled after measurement, drawing, and photographing, then shipped back to Thailand in a compact form, where they could be permanently restored by the experienced staff of the Fine Arts Department. All sherds were first thoroughly washed (except in the case of those having fugitive slip or paint, when extreme care was taken); if necessary, they were then immersed for ten to fifteen minutes in a 10% hydrochloric acid solution to remove calcareous accretions. A hot glue gun was used earlier on, as the glue hardened immediately and remained flexible to some extent after cooling and this allowed the joining of two sections that had been assembled with slightly different radii of curvature; however, the hot glue proved rather messy to use, as well as not permanent enough even for our temporary purposes. Accordingly, we turned to standard PVA glue, using sand tables and masking tape to support the segments during drying. Sagging of joined pieces was prevented by the use of wooden splints, and this technique proved adequate to support all but one of the heaviest vessels until they could be recorded and then soaked in water for disassembly and shipment to Thailand. After assembly, all basic measurements were taken (Fig. 6-11) and recorded on a description sheet, which also contained details of provenience, vessel type, rim type, fabric, and a written description of specific details on constriction techniques; size, twist, and spacing of any cord marking; and any other item of significance. In most cases a scaled sketch drawing of the vessel was also made. Computer coding was either carried out by direct observation of the vessel itself or by observation plus use of the description sheets. The photographs of each vessel also included close-ups of any interesting or unique details of construction or decoration.


Fig. 6-11

An initial vessel typology had been formulated during and immediately after the 1968 excavation, and a somewhat elaborated form of this was subsequently published (Bayard 1971b, 1977a); this was necessarily somewhat subjective, as it was based on observations during excavation-â&#x20AC;&#x201D;before the vessels had been cleaned and assembled-â&#x20AC;&#x201D; and on a sample of less than one hundred vessels reconstructed in 1969. However, the typology presented here is based on the measurements and analysis of the full sample of well over eight hundred vessels and is hence considerably more rigourous.


Obviously, any vessel typology must be based on a number of criteria, including form size, temper or paste, and surfaces finish and decoration. A morphological classification by overall form was decided on as the first criterion in order to avoid any functional inferences in the first instance (although these will be made later). The vessels fall quite neatly into six morphological classes, based on two criteria: openness versus closedness (i.e., more or less bowl-like as opposed to more or less globular); and the type of base form (roundbottomed, ringfooted, or flat-bottomed). The two criteria are thus quite similar to those of access and stability as employed by Macdonald (1980:130-131) in his analysis, although they were arrived at quite independently. The six classes are as follows, together with the loose descriptive terms we will occasionally employ here: CLASS 1: Round-bottomed globular vessels with rim-neck junction ("pots") CLASS 2: Ringfooted vessels with rim-neck junction narrower than the vessel's maximum width ("ringfooted pots or jars") CLASS 3: More or less round-bottomed vessels with rim-neck junction equal to the vessel's maximum width ("cups") CLASS 4: Ringfooted vessels with rim-neck junction equal to vessel's maximum width ("goblets") CLASS 5: Flat-bottomed vessels with rim-neck junction equal to vessel’s maximum width ("bowls") CLASS 6: Flat-bottomed vessels with rim-neck junction narrower than the vessel's maximum width ("flat-bottomed pots or jars") The six classes were mutually exclusive; however, within each it was possible to distinguish from two to eleven stylistic variants or what are referred to here as types; these are based for the most part on nonmetrical differences in fabric/temper, surface treatment, decoration, shape, and rimform, although mean size and some metrical ratios were also employed (see Fig. 6-11 and the first part of this section for the metrical and nonmetrical variables used). After merging of some of the less clear-cut types within each class and exclusion of unique or ambiguous vessels, the thirty-eight types shown in Fig. 6-12 resulted. These types contain 799-—or 94%-—of the total of 847 restorable vessels; the remainder comprise 27 unique specimens and 21 class 1 vessels (referred to as type 1Z) that lack the basal portion and hence cannot be placed into one of the class 1 types; most appear to be either type 1D or type 1I. It is obvious that some of the stylistic types differ only slightly in shape or proportion, and could be perhaps be combined (e.g., 2C and 2E, or 3A and 3B); however, as some other equally subtle distinctions have proved very significant (cf. types 2C and 2L, as discussed in Chapter 8), we prefer to retain them. Only two vessels represented apparent blends of two types, and only one fell between two of the six morphological classes; all three were placed in the unique category. Bronze-casting crucibles are of course an additional type; these will be described in full under bronze technology in Chapter 7.


Fig. 6-12

The thirty eight types described below; mean dimensions with their standard deviations (s) and coefficients of variation (v = 1000s/X) for each of the seventeen most common types (i.e., with over fifteen examples) are presented in Table 6-17. Overall mean s and v for each type and each variable are given, along with an overall ranking of type means from low to high values of v (i.e., order of increasing variability of type mean measurements). Ranking by v is used in preference to the standard deviation, as the latter obviously reflects variation in the mean measurements (RD, RW, MW, and TH in particular) of the various vessel types, as well as the amount of overall variability around the means themselves. Some interesting points emerge from the table. The two most variable measurements are those with low functional implications (RH and SH); these have mean v values notably higher than the remaining eight measurements, which cluster around 25. Certain of the vessel types are obviously quite standardized in their dimensions (2E, 4D, 2L, 2C); type 1G vessels are also quite standardized. The low mean v value for type 1Z may reflect the absence of the TH dimension. The most variable vessel types are the bowls (5A and 5B), reflecting the relatively wide range of their overall size. Types 1A, 1I, and 1D receive high mean v values due to a bimodal tendency in overall size; 1F's high value is due to the presence of four vessels larger than the norm, whole the 6A figure reflects a wide range of sizes rather than a bimodal distribution.



TYPE 1A: Medium to large globular round-bottomed vessels with mouths less than twothirds the maximum vessel width; body completely covered in cross-hatched cord-marking (Fig. 6-13a). Rim usually straight everted (type 13) and plain finished, although a few specimens have red-slipped or polished exterior rims. Sand temper. Twenty-seven specimens. Fig. 6-13

TYPE 1B: Large to very large globular round-bottomed vessels (average dims. 31 cm wide by 28 cm high). Body covered with cross hatched cord-marking, shoulder smoothed, polished, and decorated with elaborate and skilfully executed incised and punctate designs (Fig. 6-13b). Rims usually massive, straight everted (type 13), polished inside and out, and decorated with additional incised and filled designs. Sand temper. Only five specimens, all from EP 1. TYPE 1C: Medium-sized round-bottomed vessels with squat ovoid profile (Fig. 6-14a). Body wholly cord-marked, shoulder often demarcated with vertical unidirectional cordmarking. Shoulder also simply and fairly crudely incised with two or three straight or wavy horizontal lines circling neck. Rimform variable, but usually straight everted, recurved everted, or recurved inverted (types 13, 21, 14); rims usually plain but sometimes polished on inner surface. Sand temper. Fifty-four specimens.

Fig. 6-14


TYPE 1D: Medium-large, squat, elliptical round-bottomed vessels Fig. 6-13c). Body cross-hatched and cord-marked, shoulder carefully smoothed and often separated from body zone by a thin appliquĂŠ ridge. Rims either straight everted or sharply curved everted (types 13 and 03); usually plain, but occasionally polished on inner surface. Sand temper. Seventy-four specimens. TYPE 1E: Medium-large, tall ovoid vessels with almost pointed bottom. Body usually wholly covered in diagonal unidirectional cord-marking; shoulder often decorated with four or more small appliquĂŠ nubs (Fig. 6-13d). Rimform usually straight everted (type 13), often red-slipped on outer or both surfaces. Fine chaff/sand or sand temper. Eleven specimens. TYPE 1F: Medium-large vessels similar in form to type 1D, but with smoothed shoulder area red-slipped and sometimes polished as well (Fig. 6-14b). Rim usually straight everted, and usually slipped, polished, or both on outer surface. Sand temper. Seventeen specimens, all but one from 1966 excavation. TYPE 1G: Small to medium, squat, ovoid, or globular round-bottomed pots with body wholly covered in cross-hatched cord-marking (Fig. 6-13e). Rim usually straight everted, sometimes slightly curved everted (types 13, 19, 20); rims usually plain, but sometimes slipped and/or polished on inner, outer, or both surfaces. Sand temper. Eighty-two specimens (after type 2C, this is the largest single category). TYPE 1H: Smallish (27 cm wide by 17 cm tall, squat, ovoid round-bottomed vessels somewhat similar to type 1D but with a marked carination at the smoothed shoulder/cordmarked body junction (Fig. 6-15a); half the specimens have a small appliquĂŠ ridge separating the two zones. Rimform either straight everted or sharply curved everted (13, 03); rims plain, save for two cases where inner surface is polished and one with outer surface polished. Sand temper. Seven specimens.


Fig. 6-15

TYPE 1I: Medium-large, tall ovoid vessels similar in profile to type 1E, but with distinctively finished shoulder zone. Body usually covered with cross-hatched cord-marking, but several specimens have unidirectional diagonal marking (Fig. 6-13f). Smoothed shoulders usually plain, but occasionally red-slipped or both slipped and polished. Rims usually straight everted, but short vertical, long vertical, and sharply curved everted rims also


occur (types 01, 02, 03). Rims usually slipped or slipped and polished on one or both surfaces. A particularly impressive example of this type (NP-1929), with almost rimless redslipped shoulder and appliquĂŠ spiral motif below it, was recovered from Pot Nest C (late MP), and could equally well have been placed among the unique vessels described below (Fig. 6-15b). Sand temper. Twenty-one specimens. TYPE 1L: Smallish ovoid vessels (ca. 15 by 13 cm) similar in profile to types 1E and 1I; whole body finished in unidirectional cord-marking, usually diagonal, but with some examples of vertical and horizontal (Fig. 6-16a). Rims straight everted, with two examples of slightly curved everted (types 13, 19); usually plain finished on both surfaces, but several specimens have cord-marking on the rim itself. Sand temper. Twelve specimens. Fig. 6-16


TYPE 1R: Medium-sized, globular round-bottomed vessels similar in form to type 1A; distinguished from it by smoothed cord-marked finish and presence of two "horns" applied to the shoulder on opposite sides of the vessel (Fig. 6-16b). Rims straight everted or tall vertical, plain finished. Temper sand, with one example each of chaff and fine chaff/sand. Only five specimens, all from 1966 excavation. TYPE 1Z: A catchall category to include ovoid vessels with smoothed shoulders but with missing basal portions, making it impossible to determine whether the vessel is a type 1D or 1I; one specimen has a slipped and polished shoulder and probably represents a type 1F. Rim descriptions as for 1D, 1I. Sand temper. Twenty-one specimens. TYPE 2A: Smallish (ca. 14 by 12 cm) globular vessels with low vertical or filleted ringfeet (Fig. 6-16c); body finished with cross-hatched cord-marking. Rims almost always recurved inverted (type 14), usually red-slipped on outer surface and ringfeet; cord-marking occasionally present on vessel base inside ringfoot. Temper usually fine chaff/sand, sometimes sand alone. Thirteen specimens. TYPE 2B: Medium-sized, tall ovoid vessels, almost always with smoothed-over cordmarked finish and usually with ringfoot smoothly joined to body of pot with no distinct angle at the junction, although low and high vertical and conical ringfeet sometimes occur with this type (Fig. 6-16d). Rims usually straight everted and plain, although two examples of red-slipped recurved inverted rims are present. Ringfeet almost always plain. Temper variable, but usually sand. Twenty-five specimens. TYPE 2C: Small vessels with tall ovoid profile and conical ringfeet (Fig. 6-16e). Body is invariably finished with unidirectional diagonal cord-marking, running upward from left to right. Rims usually small straight everted, but short vertical and vertical with internal flange (types 01, 09) also occur. Rims almost always finished on the outer surface with red slip, as are outer and often inner surfaces with of the conical ringfeet. Tamper fine chaff/sand in the overwhelming majority of cases; a few vessels tempered mainly with sand plus a little chaff are also present. Ninety-two specimens; the largest single category. Many of these vessels show signs of standardized manufacture, using a mold to form the main part of the body, with shoulder, rim and ringfoot added later (Fig. 6-16f). The mold in many cases seems to have been a cord-marked vessel itself, as negative impressions are sometimes observable on the lower inner walls of these vessels. Standardization is also suggested by the low overall variation in dimensions of this type, a trait shared by types 2E, 2L, and 4D (see Table 6-17). TYPE 2D: Vessels somewhat larger (ca. 15 by 15 cm) than type 2C, also with conical ringfoot, but with marked carination at the shoulder. Entire outer surface coated with red slip, polished with smooth pebble or other tool (Fig. 6-17a). Rims usually straight everted, but types 07, 01, and 03 also occur; sometimes on inside as well as slipped and polished on outside; inner ringfeet are plain. Temper either sand or sand mixed with red pigment to make a bright red fabric. Ten specimens.


Fig. 6-17


TYPE 2E: Vessels very similar to type 2C but with marked carination at the shoulder and slightly smaller in size (Fig. 6-17b). Rim and ringfoot from and finish otherwise as for type 2C; temper likewise fine chaff/sand. Eighteen specimens. TYPE 2F: Similar in form to type 2A but slightly larger, with upper half of body smoothed and in four cases red-slipped as well (Fig. 6-17c). Rim and foot form and decoration as for type 2A; temper is likewise fine chaff/sand, with one chaff-tempered example. Eight specimens. TYPE 2J: A large, tall elliptical variant of type 2C (ca. 14 cm wide by 20 cm tall; Fig. 6-17d). Rims straight everted and red-slipped, or in one case recurved everted and plain (this last example also has plain rather than red-slipped ringfeet, and is sand-tempered rather than fine chaff/sand). Only three specimens, all from 1968 excavation. TYPE 2K: Another variant of type 2A or 2C, slightly smaller (10 by 11 cm), and finished with cross-hatched cord-marking (Fig. 6-17e). Straight everted or low vertical rims and conical ringfeet are plain finished. Temper sand, contrasting with types 2A, 2C, etc. Six specimens, again all from 1968 excavation. TYPE 2L: Tall, ovoid, footed vessels, slightly larger than type 2C, with body finished in fine cross-hatched cord-marking (Fig. 6-17f). Rims more variable in form than in type 2C, with short vertical predominating; vertical with internal flange, recurved inverted, and slightly curved everted with lip (types 09, 14, 20) are also found. High vertical ringfeet are most common, but low vertical and occasionally conical feet are encountered. Body is smoothed up well above junction with foot, and entire smoothed area is usually red-slipped, as is inner surface of ringfoot and outer rim. No clear signs of mold manufacture were encountered, although a few pots showed evidence of a shoulder/body join. Temper uniformly fine chaff/sand. Fifty-nine specimens. TYPE 3A: Smallish (ca. 10 by 10 cm) round-bottomed cups with height and width approximately equal; body finished in rather crudely applied cross-hatched cord-marking (Fig. 6-18a). Rims high vertical or slightly everted, ad plain finished. Temper sand. Only two specimens, one from each excavation.


Fig. 6-18

TYPE 3B: The more common 3B cups are squatter in profile but otherwise quite similar (Fig. 6-18b). Several crudely shaped with fingers, with paddle used only to apply cord marks, again rather crudely. Rimforms more variable, slightly inverted and simple vertical or slightly everted lip with no distant rim treatment (types 05, 16). Some of these vessels had a slightly flattened base, but none had clearly demarcated flat bottoms of class 5 and 6 vessels; the mean figure for base diameter given for this type in Table 6-17 is based on the approximate measurement of this area as it occurred on six of the vessels. Temper again always sand. Sixteen specimens.


TYPE 3C: Very small to medium plain-finished cups (with range 6 to 19 cm; height 3 to 8 cm), similar to the cord-marked type 3B (Fig. 6-18c). Rimform varies from straight everted through slightly inverted to simple termination of wall of vessel with no distinct rim zone (types 13, 05, 16). These vessels, like some of the 3Bs, often appear to have been crudely made using fingers alone. Temper usually sand, although one chaff- and one sand-pluschaff-tempered vessel each were encountered. Fourteen specimens. TYPE 4A: Medium-sized, wide-mouthed, shallow bowls with low conical ringfoot, rather like footed 3B cups but somewhat larger (ca. 17 cm by 13 cm; fig. 6-18d). Usually cordmarked, but one smoothed over cord-marked specimen was recovered. Rims plain and slightly to moderately inverted (types 05, 06). Six of eight specimens recovered (all from the 1968 excavation, as this type is mainly characteristic of the Early Period) were sand tempered, but two were tempered with chaff and sand. TYPE 4B: Rare but distinctive medium-sized bowls on tall conical pedestals (ca. 19 cm wide by 17 cm high); both body and conical pedestal (ca. 7 cm high) either red-slipped or both slipped and polished (Fig. 6-18e). All five samples had slightly to moderately inverted rims and sand temper. TYPE 4C: Rare "giant" versions of the following type; similar to them, but much larger. Large goblets with conical ringfeet, averaging ca. 25 by 17 cm (Fig. 6-19a). Rims always inverted (type 05 or 06) and almost always red-slipped on outside. Body decorated with unidirectional diagonal cord-marking in most cases, but two of six specimens had crosshatched cord-marking, and one was smoothed. Outer ringfoot also usually red-slipped and sometimes inner as well. Temper fine chaff/sand. Fig. 6-19a TYPE 4D: Small ringfooted goblets, very common at the site (Fig. 6-18f). Rims either slightly markedly inverted, and red-slipped on outer surface in majority of cases (in many of the remaining cases we assume the fugitive slip was probably originally present, but has been lost through groundwater action). Body usually finished with unidirectional diagonal cord-marking, but some instances of horizontal cord-marking occur as well. Conical ringfoot very often red-slipped on outside and often on inner surface and on base inside ringfoot as well. As with 2C vessels, traces of "negative" cord prints are often present on the inner lower halves of 4D vessels, strongly suggesting standardized manufacture of this portion by means of a mold made from another cord-marked pot; this vessel type also ranks second lowest in overall variability of mean dimensions (table 6-17). Temper fine chaff/ sand in the overwhelming majority of cases. Fifty-eight specimens. TYPE 4E: Red-slipped ringfooted goblets, slightly larger than type 4D; they appear to parallel the latter, as type 2D does type 2C (Fig. 6-19b). Only two examples recovered (both from the 1968 excavation), one with standard inverted rim (type 06) and another with


slightly curved everted rim (type 19). The former vessel is polished as well as slipped and has sand/chaff temper, whole the second has sand temper. Fig. 6-19b TYPE 4F: This type of ringfooted goblet appears to stand in relation to the common type 4D as the type 2L footed pots do to have 2C (Fig. 6-20a). They are approximately the same size and also have slightly inverted rims, but they have high vertical ringfeet and a body finished with cross-hatched rather than diagonal unidicretional cord-marking; as with type 2L vessels, outer surfaces of both ringfeet and rims are red-slipped. Temper fine chaff/ sand. Only 3 specimens, all from the 1968 excavation. Fig. 6-20


TYPE 4G: Another minor type of ringfooted goblet, very small (ca. 8 by 5 cm) and plain finished, with no slipping on rim or foot (Fig. 6-20b). One of the two specimens has simple bowl rim (type 16) and sand/chaff temper, the second has slightly inverted rim and sand temper. TYPE 5A: Medium-sized, shallow, flat-bottomed bowls with either plain or polished finish (Fig, 6-20c); rims either slightly or markedly inverted (types 05, 06). Temper usually sand, but specimens also found with chaff, sand/chaff, and fine chaff/sand temper. Forty-three specimens. TYPE 5B: Bowls similar to above but usually somewhat larger, with entire exterior surface (save base) red-slipped; about two-thirds of specimens are also polished (Fig. 6-20d). Rims as for type 5A. Temper almost always sand, but a few specimens have sand and red pigment fabric. Thirty-one specimens. TYPE 6A: Small to medium-small flat-bottomed globular jars, with entire exterior surface (except base) wither red-slipped or both slipped and polished (Fig. 6-20e). Rimform either straight or everted, short vertical, or short curved everted (types 13, 01, 03); inner surface of rim sometimes slipped, polished, or both. Temper usually sand, with a few specimens having sand and red pigment and one with laterite and sand. Twenty-five specimens. TYPE 6B: Properly speaking, the two specimens in this category should be assigned to the unique vessels, but as they were given a type classification earlier on, they are retained here. These are both animal effigy vessels some 9 cm long, one in the form of an elephant (14 cm high) and the second a (?)frog (10 cm high), as illustrated in Figs. 6-21a and 6-21b; each has a class 6 vessel on its back, and the frog and its vessel are wholly red-slipped. The elephant vessel has a plain-finished body bearing a red-slipped jar; this vessel also has small holes on each side of its trunk in which fragments of bone representing the remains of "tusks" were present. Both vessels have sand temper. TYPE 6C: Medium to medium-sized flat-bottomed jars, with exterior cord-marked; usually cross-hatched, but examples of unidirectional diagonal and horizontal marking also occur (Fig. 6-20f). Vessels range from 9 to 20 cm ion width, and from 7 to 15 cm in height. Rimforms as for type 6A; temper almost always sand (one chaff-tempered specimen). Ten specimens. TYPE 6D: Flat-bottomed jars, slightly smaller than type 6A and plain-finished, with marked carination at shoulder not found on other types in this class (Fig. 6-22a); the three examples recovered had slightly inverted, slightly curved everted, and short vertical rimforms (05, 19, 01). Temper sand, with some amount of chaff present in one vessel as well. TYPE 6E: Plain-finished flat-bottomed jars with no carination at shoulder; larger than type 6D but still slightly smaller than type 6A (Fig. 6-22b). Two specimens had polished exteriors. Rimform varies as for type 6A; outer rims usually plain but polished in two cases and red-slipped in one. Temper sand, with one vessel having some chaff as well and a second having sand and red pigment fabric. One vessel technically of this type is distinguished by clear signs of ringbuilt manufacture (NP-443), a technique noted very rarely at the site. Seventeen specimens.


Fig. 6-21


Fig. 6-22


TYPE 6R: Flat-bottomed versions of round-bottomed "horned" type 1R vessels; like them, found only in the 1966 excavation area, and seem top have been most common during levels MP 2 and 3; not present in the 1968 area. Two of the three vessels have straight everted rims, and the third a vertical lipped rim (types 13, 07); this one also has an appliquĂŠ "cord" motif circling the shoulder, in addition to the two horns (Fog. 6-22c). Temper sand in two cases and chaff the third. Unique Vessels The following vessels could not be accommodated into any of the type classifications, although all but one could be clearly placed in one of the six morphological classes. Brief descriptions of these vessels follow, with those from the 1966 excavation listed first within each class. CLASS 1 1966: NP-985: Upper portion of a large sand-tempered vessel with high everted rim recovered from LP 2 occupation layers; shoulder smoothed and painted with simple red linear motif. 1968: NNT-384: Moderately large globular vessel with high everted collar rim found with EP 1 B. 125; cord-marked overall, but with appliquĂŠ "cord" motif encircling shoulder. Sand temper (Fig. 6-23a). NNT-646: A tiny (5 cm high by 4 cm wide), round-bottomed, roughly cord-marked vessel with a smoothed shoulder. found with MP 6 B. 61; traces of red slip on rim and shoulder. A seeming example of "paddle and finger" manufacture, possibly by a child. Sand temper with bright red paste due either to oxidation or addition of red pigment (Fig. 6-22d). NNT-693: Very tall (61 mm), vertical collar rim and part of holder from a massive, probably round-bottomed vessel found in MP 4 B. 63. Shoulder cord-marked and decorated with wavy incised pattern made using notched flat stick. Sand temper (Fig. 6-23b). NNT-734: Fairly large, round-bottomed, carinated vessel with inverted shoulder and no rim; carefully smoothed and probably red-slipped over entire outer surface (Fig. 6-23c). Found in late MP "pot nest" F. Soft, friable sand-tempered fabric characteristic of late and post-MP times in the area. (Buchan 1973).


Fig. 6-23


CLASS 2 1966: NP-552: The frequently-illustrated red-on-buff footed vessel from EP 3 B. 26. Spiral designs painted on shoulder and triangular motifs on outer rim; two perforations each on opposite sides of rim and ringfoot. clay temper (Fig. 6-24a). NP-523: A "spittoon-shaped" vessel from late EP or early MP B. 83. Very high everted rim with vertical extension; low filleted ringfoot. Lower outer rim, entire body, and upper ringfoot covered with impressed rouletted designs in spirals and waves. Chaff temper (Fig. 6-24b). NP-802: A pot similar to NP-522 recovered during initial cleaning of soil matrix from bones of EP 3/MP 1? B. 80. The globular, ringfooted vessel had a red-painted rim and body, with red-on-white arclike designs on shoulder and vertical stripes running from redpainted body over outer ringfoot. Coarse clay temper (Fig. 6-24c). NP-804: This puzzling vessel was found in MP 7 B. 43 It has high collar rim decorated with bold incised "batwing" and zigzag motifs reminiscent of those found on EP 1 type 1B vessels; however, the punctate infilling as much cruder and not done with a dentate stamp. Sand temper. (This vessel was left in Bangkok; no drawing was made of it. P1. 8.) NP-1198: A squat, globular vessel with everted rim decorated with a band of dentate stamps, with a conical ringfoot; found in disturbed B. 18, doubtfully assigned to MP 4. Lower half of body cord-marked, upper half smoothed and decorated with three bands of vertical dentate stamps; a plain band with horizontal slashes separates the two bands, and the middle and lower band are separated by an incised zigzag decoration with punctuations in the center. Both design and execution appear to be quite distinct form the EP incised and filled decoration. Sand temper (Fig. 6-24d). Fig. 6-24a


NP-1200: This vessel, found along with NP-522 in EP 3 B. 26, superficially resembles a type 2C or 2E pot, with unidirectional cord-marking and an apparently conical ringfoot (mostly missing). However, the very high everted rim and clay temper make it unique. Its appearance in this EP grave, paralleling the find of NP-1737 in a 1968 burial of the same level, has a considerable significance in our interpretation of social changes that may have occurred at this time (Fig. 6-25a; also see Chapter 8). Fig. 6-25


1968: NNT-393: Three sherds from a squat globular vessel apparently having a ringfoot recovered from MP 1 layers. Body cord-marked, shoulder polished with incised motif. Sand temper. NNT-621: Globular vessel with high collar rim, vertically everted, and flaring conical ringfoot. Red slip on outer rim and shoulder; body covered with cross-hatched cordmarking. Recovered from EP 3 B. 81, this vessel is very likely that a prototype of the 2L vessels common during the MP. Sand temper rather than the fine-chaff-and-sand temper characteristic of 2C/Es and 2Ls (Fig. 6-25b). NNT-629: A large, relatively spectacular red-and-white ringfooted vessel recovered from the mound of MP 1 B. 116. Upper rim painted red, with a second red band at rimshoulder junction; body red, shoulder plain, with thin red band encircling it decorated with upward-pointed solid triangles. Two perforated lugs were applies below this band. Foot plain, with perforations. Clay temper (Fig. 6-25c). NNT-636: An apparent blend of types 2L and 2B found in MP 5 B. 23. Low vertical rim, plain shoulder, body covered with cross-hatched cord-marks, and high vertical ringfoot. Temper chaff rather than fine-chaff-and-sand (Fig. 6-25d). NP-1686: Fairly large (20 cm diam.) globular vessel with tall, slightly everted rim and very short conical ringfoot applied to round-bottomed base found in EP 2 B. 36. Body cord marked, with possible traces of red slip; ringfoot plain with two perforations. Narrow shoulder smoothed and polished, decorated with two lines of punctuations separated by a band of vertical punctuations. A second line of these extends around rim-shoulder junction, and a final band of horizontal plus vertical punctuations decorates the upper inside of the rim. Sand temper (Fig. 6-26a). NP-1704: A globular vessel with short applied ringfoot similar to the last, but with high vertical rim and a distinct fringe protruding outward at rim-shoulder junction; it was found in EP 1 Mound 126. Body cord-marked, rim plain. Shoulder polished and decorated with incised and dentate-stamp-filled motifs similar to those on 1B vessels from this level; the motifs are separated by lines of single punctuations. Sand temper (Fig. 6-26b). NP-1723: A rouletted "spittoon" vessel similar in shape to NP-523 above, but with much shorter everted rim; recovered from EP 3 B. 81. Horizontal rouletted bands above foot and below rim; central portion of the body decorated with a complex pattern of interlocking rouletted bands. Clay temper (Fig. 6-26c).


Fig. 6-26


Intermediate between Class 2 and 4: NP-1637 (1968): An interesting cuplike vessel found in EP 3 B. 79; in shape it is intermediate between class 2 and class 4 vessels, with a slightly everted lip. Outer rim, shoulder, and upper body smoothed; lower body cord-marked, conical ringfoot plain. It is tempting to see this vessel, like NNT-621 above, as a prototype for the 2C/E/L and 4D so common in the succeeding MP. Sand temper (Fig. 6-26d). CLASS 4: NP-1300 (1996): A wholly red-slipped, chaff-and-sand-tempered bowl resembling a class 5 vessel, but with traces of a ringfoot. Recovered from LP 2 occupational layers (Fig. 6-27a). Fig. 6-27


CLASS 5: NP-1728 (1968): A unique, tall, plain bowl, oval in plan, with simple pouring spout formed by pulling out one end of slightly inverted rim; found in mound of MP 1 B. 104. Exterior red-slipped; no temper of any sort appeared to be added to the fabric (Fig. 6-27b). CLASS 6: 1966: NP-70: A small flat-bottomed pot found with B. 23, doubtfully assigned to MP 5; redslipped shoulder and outer rim, lower half of vessel cord-marked. Sand temper. NP-477: A medium-sized but squat flat-bottomed pot with high-collared everted rim possibly but doubtfully associated with MP 1 B. 69; profile of vessel is quite distinct from "standard" class 6 types. Sand temper (Fig. 6-27c). NP-919: A squat, globular, flat-bottomed vessel found in MP 2 layers not associated with any burial; rim apparently everted, but mostly missing. Entire body had smoothed-over cord-marking; appliquĂŠ ridge with punctate decoration encircles shoulder. Sand temper (Fig. 6-27d). NP-1109: A fairly large, squat, globular vessel with flat bottom and everted rim found in MP 3 B. 38. Outer surface polished and decorated with a wide band of fairly simple incised straight lines and triangles, filled with punctations made by a simple tool rather than a dentate stamp in a style resembling NP-1198. Sand temper (Fig. 6-28a). 1968: NP-1790: An extremely squat, flat-bottomed jar with plain finish and short vertical rim recovered from MP 1 nonburial layers. Chaff temper (Fig. 6-28b). NP-1873: A fairly large flat-bottomed jar with high everted rim, upper half of body redslipped and polished and lower half cord-marked; halves of body separated by appliquĂŠ ridge. Found in MP 4 B. 39. Sand temper (Fig. 6-28c). The more or less unique vessels present a fairly wide range of variants. Some of these are probably nonce forms (e.g., NNT-636, NNT-646); however, we feel that others may well have considerable chronological and even social significance (the rouletted and painted clay-tempered vessels and the forms apparently ancestral to the "C" and "L" vessels of the MP; see "Temporal Variation" below and Chapter 8). But even including these vessels, the typology arrived at here does prevent six relatively clear-cut morphological classes. Q-mode factor analyses of the mean dimensions for each type clearly shoe clustering into the morphological classes on the bases of dimensions alone, with no overlap of types between classes (Fig. 6-29; types with less than four specimens. type omitted). This would be expected, as the basis is morphological. The nonmetrical criteria (surface, finish, vessel proportions, etc.) used to distinguish types within the six classes are of course more subjective, and several of the smaller types could well be merged. However, all will be outlined below, the types to have considerable chronological significance and will be seen to delineate presumed socio-economic groups as well (see Chapter 8).


Fig. 6-28


Fig. 6-29

Functional Inferences The possible function of the various vessel types (or at least of the morphological classes in general) is of obvious interest in inferring general economic patterns and changes in these; as Macdonald had pointed out (1980:129), it is also important to attempt to determine functional variables in order to remove them from those considered to be temporarily sensitive. Macdonald's approach to the problem of functional inference followed the methods of Arnold (1975), de Boer and Lathrop (1979), and others. He took mean measurements from a sample of over 1,300 modern Northeast Thai vessels (from Calder 1972) and used these to arrive at three indices that seem correlated with function: volume; ease of access to the vessel (width of mouth divided by maximum width of vessel); and stability (width of base divided by maximum width, with round-bottomed vessels assigned an arbitrary base width of 1 cm). It is obvious that, in general, high stability is to be correlated with storage and serving functions, while ease of access is correlated with cooking and serving. Similarly, narrow access is more likely in storage vessels, and stability, is not crucial in cooking vessels that are supported in some fashion over a fire (as is the case throughout most of Southeast Asia at present). When access was plotted against stability and volume taken into account to distinguish between preparation (large volume) and serving (relatively small volume) vessels, Macdonald found that three functional clusters could be distinguished among the vessels considered in his study (1980:134-141). These clusters and their presumed functions are clearly paralleled in the six Non Nok Tha morphological classes (Table 6-18).


An access/stability plot of the morphological class mean dimensions (Fig. 6-20) illustrates the lack of overlap between classes, even at the 1-s range, and suggests a general tripartite division of the vessels into cooking (class 1), serving (classes 3, 4, 5), and storage (classes 2, 6). This may be further refined by an examination of a similar plot (Fig. 6-31) of the fourteen most common vessel types at Non Nok Tha (treating all class 3 vessels as one type and also introducing the five large 1B vessels), compared with fourteen types of modern vessels (again from Calder 1972). The fit between prehistoric and modern types is a fairly close one, save that the narrow-mouthed, flat-bottomed stoneware jars used for dry and paste storage today are missing from the Non Nok Tha inventory. This diagram also suggests a division of class 1 vessels into cooking and large liquid storage vessels. The bimodal size distribution of types 1A and 1D in particular indicates that these may have two functions depending of size; mean volume for the two peaks of small, 8 large) are ca. 2.1 and 8.9, and for type 1D (39 small, 23 large) 2.2 and 5.5. A volume histogram (Fig. 6-32) suggests that 1Bs, 1Is, and the larger 1Ds and 1As all of which range from 5 to 30 1 in capacity, may have served as liquid catchments and storage vessels, despite their low stability values (modern round-bottomed vessels are often used for liquid storage, made stable by the use of a straw ring around the base of the vessel). The smaller 1As and 1Ds, as well as the rest of the class 1 vessels, have volumes in the range of 2 to 3 1, and were probably used for cooking.


Fig. 6-31

Fig. 6-32


Additional support for this division comes from the results of factor and cluster analyses of a sample of 365 vessels with almost complete metrical data from the more secure 1968 excavation, using BMDP FACTOR and CLUSTER routines. These were run using nine of the ten metrical variables for each vessel (thickness was excluded, as this was not recorded for a number of the 1966 vessels). The factor analyses (BMDP P4M routine, with varimax rotation), with 365 vessels as variables and only nine â&#x20AC;&#x153;casesâ&#x20AC;? (the metrical variables), is obviously intended only as a heuristic portrayal of the dimensional relationships between the vessels and nothing more. It produced two linear clusters, the first consisting of class 2 and 4 vessels, and the second class 1; a small third cluster in an intermediate position included the class 5 and 6 vessels (Fig. 6-33). Obviously overall size is responsible for the first factor values, with the clusters produced by a sharp differentiation in factor 2 (i.e., base form); however, within the class 1 linear cluster, one can discern three groups: a very dense cluster made up of medium-small vessels; a more diffuse adjacent cluster of larger ones; and, adjoining that, a scattered group of very large vessels. The first two were presumably used for cooking, and the last for liquid storage. The factor analysis obviously does not discriminate between presumed storage and serving functions in the remaining morphological classes, but rather appends class 3 cups to the small-medium class 1 cluster, overlaps classes 5 and 6 almost wholly, and separates out only the larger 2B and 4C vessels from classes 2 and 4. Fig. 6-33


A number of cluster analyses (PKM routine; K clustering of means) were run on a slightly smaller sample of 353 vessels (a further 12 vessels had to be excluded due to one incomplete measurement); a run requesting nine clusters seemed to produce the most satisfactory of results (Fig. 6-34). Fig. 6-34

Clusters 5 (2C/E and 4d vessels) and 8 (the slightly larger 2L and 2D vessels) form a group of standardized vessels (cf. Table 6-17) probably intended for ritual â&#x20AC;&#x153;storageâ&#x20AC;? of food in burials; clusters 6 (small flat-bottomed bowls and storage vessels) and 9 (smaller cooking vessels, mainly 1Cs and 1Gs) group together, and cluster 7 (mainly larger 1Cs) is closely allied. Clusters 2 (mainly 1Ds) and 4 (smaller 1As and some 1Ds and 1Cs) form a third group of medium-large cooking vessels. Cluster 1 contains the large serving vessels of type 4A and 4C, as well as the very largest 2B vessel. Finally, cluster 3 contains four very large 1A and 1B vessels, as well as a single 1I vessel; these presumably functioned as liquid storage vessels. Contents Analysis A sample of about one hundred vessels from the 1968 excavation that were lifted with contents intact had these carefully screened and examined under low magnification in order to ascertain the presence of small bones, plant remains, and the like. A large majority of these vessels did in fact contain mammal bone fragments, small fish bones, and fragments of gastropod shell, wither alone or in combination with each other (Table 6-19). Carbonized rice grains and other seeds were recovered in a few instances, indicating that vessels contained vegetable as well as animal foods. Unfortunately, the extraction of the seeds from the hard soils matrix and the washing process used in preparing the samples made subsequent botanical identification of other possible seeds infeasible (D. Yen, pers. Comm.).


As considerable quantities as gastropod shell are to be found in the cultural deposits of the Early Period and early Middle Period and hence could have found their way into the vessels during of after burial, only amounts greater than 4 g were considered significant. Even so, the fact that only four vessels contained shell fragments by themselves, with no mammalian or fish bone, makes it seem likely that the shells were natural inclusions rather than remnants in the vessel at time of burial (although several species of gastropods are commonly eaten in the region today). However, the same cannot be said of the mammalian and in particular fish bones. The former were usually in such small fragments as to make identification impossible, although one vessel did contain an identifiable bovine tooth fragment. The fish fragments were in most cases small vertebrae and probably represent the smaller species utilized in the region today; the smaller ones in particular hint at the presence of the fermented fish paste (pa daek or pla ra) that is the protein staple of the region today. The small size and fragmentary condition of the bones have prevented speciation, but they probably derive in the main from several of the small species commonly used for this purpose today (Channa, Mystus, etc.). It is difficult to discern any clear pattern existing between vessel type and contents. In the Early Period most if the vessels with mammalian and fish bone are of class 1 (26 vessels), with only six class 2 vessels represented; the situation is reversed in the Middle Period, with its much greater variety of vessel types, and only nine class 1 vessels are represented alongside fourteen class 2 and six class 4 vessels. It is interesting to note that the class 1 vessels almost all fall into Cluster 9 above, with dimensions averaging about 16 cm wide by 13 cm tall. This is quite close to the mean of all vessels in Cluster 9 (15.1 cm wide by 12.3 cm tall, and it seems apparent that these vessels functioned as curry/stew cookers, similar to the modern mo kaeng and mo hung (mean dimensions 19 cm wide by 17 cm tall and 15 x 19 cm respectively; Calder 1972). The vessels that were assigned to Cluster 9 make up some 31% of the contents sample; a further 21% fall into Clusters 8 and 5 (small storage vessels). The remainder are scattered among the other clusters (excepting Cluster 3, the very large liquid storage vessels). Hence the contents analyses provide some general support for the functional inferences arising from the cluster analysis: The round-bottomed vessels functioned mainly as cooking vessels, with the smaller ones used for the preparation of dishes to accompany rice. The medium and larger round-bottomed vessels may have functioned as rice-boiling


pots, but firm evidence is lacking. The ringfooted class 2 and 4 vessels could not have functioned as cooking vessels, and a specialized role as funerary storage/serving vessels, possibly containing condiments like fish paste, is proposed here, at least for the very highly standardized 2C/E, 2L, and 4D types; a serving as well as storage function is suggested by the discovery of ladles placed inside two class 2 and two class 6 vessels. Similarly, the common, highly standardized type 4D vessels may have served as condiment containers rather than the liquid serving vessels that our informal term goblet would imply. The class 3 vessels would appear to be better candidates for the serving of liquids, and only one of them was encountered with any faunal remains inside. The class 6 vessels also seem to have been used for â&#x20AC;&#x153;ritualâ&#x20AC;? storage, but the sample that had contents analyzed is small. Finally, the class 5 vessels were apparently intended as serving vessels; their wide mouths and low profile would make them unsuitable for storage or cooking purposes. Whole Vessels: Temporal Variation The chronological distribution of the various vessel types is obviously of great importance in attempting to determine the validity of the Early/Middle Period distinction crucial to the overall stratigraphic picture presented in Chapter 4. As discussed above, Macdonaldâ&#x20AC;&#x2122;s analyses suggested that this division lacked substantiation, and in viewing the results of level sherd analyses it did seem clear that MP 1 clustered with the three EP levels rather than with the later MP levels. As already explained, however, this is almost certainly due to the incorporation of large numbers of EP sherds into MP 1 deposits through extensive grave digging. Examination of the distribution of whole vessel types, almost entirely derived from in situ burial contexts and hence unmixed, should provide a final resolution to the question. This distribution is shown in Table 6-20 and Figs. 6-35 to 6-38 and provides convincing support for the reality of the Early/Middle Period boundary; it is obvious that a number of types are either limited to or numerically dominant in the three EP levels in the 1968 sequence. Thus types 4G and 1B occur only in these three levels and in 1966 layer 21 (which as already stated is a mixture of the EP levels and MP 1). Type 1B I of course more strictly limited in distribution to EP 1 only, as is, by and large, the elaborate incised decorative technique that distinguishes it. It is interesting to note that only one other vessel was found decorated with a similar technique; a unique class 2 ringfooted vessel with filled incised bands on the shoulder and both sides of the rim (NP-1686; Fig. 6-26A) from EP 2 Burial 36. Incised decoration of this sort is thus not only limited to the Early Period, but to its earliest portions only. There are two apparent exceptions to this rule that are puzzling. NP-804 (P1. 8) is a unique specimen (see description above) with elaborate, roughly filled incised designs on a high collar rim; this was found in 1966 Burial 43 (MP 7). The second even more unique specimen is NP-1198 (Fig. 6-24D), which features zigzag incisions and horizontal rows of punctations on the upper body and rim quite unlike the EP 1 motifs and techniques; this was found in association with 1966 Burial 18 (MP-4??). This latter vessel may well by of exotic origin. NP-804, with motifs reminiscent of the EP 1 vessels, could conceivably have been recovered from a burial of that period and subsequently reburied, but its presence at this late date must ultimately remain unexplained.



Fig. 6-35

Fig. 6-36

Fig. 6-37

Fig. 6-38


Other Early period types such as 1A, 1C, and 4A do occur in the earlier Middle Period as well, but they are relatively scarce when compared to the large suite of new types that make their appearance at the onset of the Middle Period and thereafter. Table 6-21 and Figs. 6-35 and 6-37 make the same point more clearly; here percentages of the level total are shown for major vessel types only. Of the twenty-nine major vessel types portrayed in the 1968 whole vessel sample shown in Fig. 6-37, fourteen are confined to the Middle Period only. It is obvious that if any break in the sequence at all is to be recognized, it must lie between the EP and the MP levels. Certain differences are apparent between the earlier and later portions of the Middle Period as well, but these are by no means as clear and may reflect sampling error. Thus the “horned” vessel types 1R and 6R appear to by limited to the earlier MP levels only (in particular MP 2 and 3, represented only in the 1966 sequence, plus one each in MP 4 and MP 5). Type 1D vessels are also most characteristic of the earlier Middle Period but appear in smaller numbers in other levels as well. Types 6A and 6C appear to be more common in the later Middle Period, but the numbers are not sufficient to allow any conclusions to be drawn.

Finally, factor analyses of the distribution of twenty seven major types for the 1968 sample clearly show a major break between EP and MP levels (Fig. 6-39). This contrasts markedly with the evidence provided by the non-burial sherd distribution for the affiliation of MP 1 with the EP and supports our argument that this apparent relationship resulted from extensive grave digging during MP 1 times. As expected, the situation is not as clear when combined factor analysis of both 1966 and 1968 samples is performed. Separate runs using all thirty-seven types and only the major types (Fig. 6-40), however, present very similar pictures. Four distinct clusters are present, and these have obvious chronological significance: An EP cluster, which also contains 1966 layer 18 (perhaps due to its small sample size); an earlier MP cluster containing 1966 layers 21, 10 and 19; a third cluster linking 1968 layer 6B with 1966 layer 17 and, somewhat more distantly, the group of burials from the western portion of the 1966 excavation provenienced as “17/18, all of which are equivalent to the MP 5 and finally, a cluster containing the later MP levels plus 1968 layer 7> This last I certainly puzzling, but it is presumably due to the quantity of 1D, 2C, and 2L vessels—and a large variety of other vessel types—present in the 1968 sample of this level. In any event, it is apparent that the overall distribution of vessel types lends strong support to the Early/Middle Period distinction and provides some concrete evidence for the correlation of the two sequences presented in Chapter 4.


Fig. 6-39

Fig. 6-40

Decorative Motifs Non Nok Tha vessels are not distinguished by a wide variety of decorative techniques or designs, such as those found through much of the Ban Chiang sequence. However, a few vessels at Non Nok Tha do have distinctive motifs, some of them apparently quite limited in time span. The most obvious cases are the impressive type 1B vessels, with incised curvilinear and angular geometric motifs on the their polished shoulders and rims; the incised areas have been filled with wavy lines impressed by some type of dentate stamp (P1. 9A). These are largely limited to level EP 1, with only one other incised vessel recovered from the level immediately above (NP-1686). Also very limited in time span are the handful of red-on-white painted vessels and those with impressed rouletted swirl motifs (P1. 9c; NP-522, Fig. 6-24a; NP-802, Fog. 6-24c; NNT-629, Fig. 6-25c; P1. 9b; NP-523, Fig. 6-24B; NP-1723, Fig. 6-26c). These occur only in burials of the terminal Early and earliest Middle Period levels; they have atypical fabrics and are quite likely imports, perhaps from the middle Chi Basin to the east. Obviously, only a detailed ceramic analysis of the sort carried out by Vincent on a few late MP and LP sherds (to appear in the projected second volume of this report) could answer this question. However, clear parallels to the swirl-rouletted pots are present in the rocker-stamped vessels of the earlier half of the Early Period at Ban Chiang (Whiteâ&#x20AC;&#x2122;s provisional types 3 and 4, dating from her EP II, ca. 3000-1900 B.C., 1968:92, 279); clear resemblance in the form if some of these vessels are also apparent (e.g., White 1983a:60, 30). The technique of filled incised patterns characteristic of early Ban Chiang (EP I and II in Whiteâ&#x20AC;&#x2122;s revised chronology: 1986:90, 91) is also reflected in the EP 1 type 1B vessels, although no really close parallels in motifs exist. We originally thought that some relationship existed between the three red-on-white painted vessels from Non Nok Tha and the famous Ban Chiang ones (Bayard 1977a); however, now that the Ban Chiang chronology has become clearer it is obvious that no direct ties are possible between the famous Late Period Ban Chiang vessels and the much earlier Non Nok Tha ones, as White also pointed out (1982a:57). Two of the painted vessels have red bodies with a simple motif painted only on the white shoulder are; the curvilinear spirals that cover the


third vessel (NP-522; Fig. 6-24a) are in fact more reminiscent of the early impressed swirl designs than they are of the elaborate variety of Ban Chiang Late Period motifs. A further possible similarity to Ban Chiang, pointed out by White as well (loc. cit.), is the presence of matched perforations on each side of rim and foot in this and several other Non Nok Tha vessels, which appear to be foreign imports. Other motifs are more widespread in time. These included the simple incision of one, two, or sometimes three horizontal bands around the shoulder of type 1C pots. Also, the distinctive appliqué band — often itself impressed to resemble a cord, broken and curved downward at one point on the pot — which we refer to here as the “cord” motif (P1. 9d), occurs sporadically on vessels from EP 1 (NP-1688 and NN-384; Fig. 6-23a) through MP 3 (NP-831 and NP-924, both in Burial 55) to MP 5 (NNY 641 and NNT-763, both in Burial 49). The upper portion of a very large and apparently exotic pot in Pot Nest B (NNT-801; see Chapter 8) has the downward curl applied to its neck, but no encircling “rope”; a second vessel (NNT-798), also from this Pot Nest, has both rope and curl. The motif is hence one of long duration, spanning most if not all of the Early and Middle Periods. It is also fairly widespread in space; it occurs at Non Pa Kluai in GP B contexts (Wilen 1982a: 63; 1986:91, Fig. 6d) and is found as well on funerary vessels from the first burial phase at Ban Na Di (layer 7; Higham and Kijngam 1984), with additional variants such as the presence of two downward curls separated by an applied vertical line or applied oval between the curls. We have no clue a to the significance of this motif, although one could speculate that, given its duration of two to three thousand years and its occurrence in funerary contexts, it may have had some religious meaning; on the other hand, it could of course be purely decorative — but it so, its limited distribution is rather puzzling. Also puzzling, given its apparent rarity, is its occurrence on two vessels in the graves of both 1966 Burial 55 (male, 40+) and 1968 Burial 49 (female, 25); both of these burials are relatively rich, and it could be that the motif makes a special(religious?) status held by these individuals. But these speculations cannot be supported by ant concrete evidence, and the significance of the motif must remain a mystery. Non Pottery Ceramics In addition to pottery, clay was employed in the manufacture of other types of artifacts at the site, the most common of which were the objects referred to here as “spindle whorls” and “pellets”. In almost all cases the clay employed was unmixed with any significant amount of tempering material; some amount of fine sand was almost always present, but this appears to have been a natural inclusion. Other non pottery ceramic artifacts include roof tiles, fragments of partly baked clay that appear to derive from crude animal effigy figurines, a few clay stamp and cylinder fragments, and the enigmatic objects we call “tripod tables.” The distribution of the more common of these artifacts is presented in Table 6-22.


Spindle Whorls Among the most common clay objects recovered at Non Nok Tha were 81 generally spherical pierced items, all with a circular profile when viewed end-on but with variable profiles when viewed from the side (Fig. 6-41). By far the most common profile was biconical (thirty-three examples), followed by flattened biconical (thirteen), spherical (eleven), spherico-conical and truncated spherical (nine each), hemispherical (two), and flat (one); three of the items were so fragmentary as to make determination of the profile impossible. The objects are fairly standardized in dimensions and weight, averaging some 31 mm in width and 23 mm along the pierced axis; weight averages 22 g for complete specimens. Initially we interpreted these either as net weights — possibly used on the circular throw net (hae) that is still commonly employed in the region today — or as spindle whorls. However, a survey of the ethnographic literature carried out by Daisy Bayard in 1969 makes a strong case for accepting the second function as by far the most likely (D.N.H.L. Bayard 1969). Most of the whorls are plain finished, but four are simply decorated with lines radiating from the hole on one side of the whorl or with a ring of impressed dots or tiny circles around the hole. The various shapes occur throughout the temporal span of these items and do not seem to have any chronological significance. The whorls are unknown from the Early Period and first make a tentative appearance in the early Middle Period, but they cannot be viewed as common until MP 4. They continue in abundance up through LP 4, and one is present in LP 5, but these are probably upwardly intrusive as the numbers drop off sharply in the later Late Period; evidence from Non Chain suggests that they ceased being used there during General Period C, at about 200 B.C. (Bayard, Charoenwongsa, and Rutnin 1982-1983:51). Similar items are well known from other sites in Mainland Southeast Asia, ranging from relatively near neighbors like Ban Chiang (White 1982:76) to Mlu Prei (Levy 1943:42, P1. 25) and Samromg Sen (Mansuy 1902:P1. XIII, 16-17; he describes these as “pin heads”) to Vietnamese sites like Phung Nguyen (Hoang and Nguyen 1978:100-102) and Thieu Duong (Pham Minh Huyen and Deip Dinh Hoa 1981:28) in the north Doc Chua (Nguyen Duy Ty 1977:37, 39-40) in the south. Six Middle Period burials had spindle whorls as grave goods; although the three 1966 burials were all of young or middle-aged women and would suggest that women were mainly responsible for thread manufacture, the three 1968


Fig. 6-41


burials comprised of a child of six to eight and two thirty-year-old males, one of them (B. 124) with seven whorls in the grave. Hence there is little clear evidence for a restriction of use (at least in a funerary sense) of these items to one sex or the other. Finally, we should mention that a noticeable concentration of ten whorls was found at the layer 5/6 interface of the 1968 excavation; these were presumably concentrated by the erosion of the upper Middle Period layers mentioned in Chapter 4. The implication from the temporal distribution at Non Nok Tha is that at least this form of textile production arrived in the region at about the same time as bronze manufacture and was not known previously. The presence of woven textiles during the bronze period is strongly suggested by traces adhering to bronze artifacts from Ban Chiang or a nearby site (White 1982:90, Fig. 166). The fibers used in such textiles were presumably spun with the whorls, but the type of fiber is still in doubt; cotton is a possibility, but analysis of the fabric remains from Ban Na Di suggests that silk is the most likely choice, at least in the samples examined from that site (Higham and Kijngam 1984:126-127). The use of the whorls apparently continued in the Phu Wiang region well into the present millennium, raising the question of when and from whence the spinning wheel that is universally used in the region today was introduced; our workmen were unanimous in their lack of recognition of the function or purpose of the whorls found at the site. Although a date in the sixteenth or seventeenth century seems likely for the local introduction of the spinning wheel, evidence from the site of Non Chai, near Khon Kaen city, indicates a considerably earlier date for decline of use of spindle whorls in the region as a whole (Bayard, Charoenwongsa, and Rutnin 1982-1983:50-51). Spindle whorls were common finds in Non Chai layers 4 and 5 (ca. 500-250 B.C.), but only two were recovered from lower layer 3. Beginning in upper layer 3 and throughout layer 2 (ca. 150 BC-A.D. 200) a second type of textile-related artifact became common: the so-called thread spacer (ipae). This is a small, shallow bowl with a centrally raised, perforated lug that allows for the combination of several small filaments into a single more sturdy one. It may be associated with the use of the spinning wheel to produce silk thread, although further study of the thread spacers and their modern counterparts is necessary to confirm this. Clay Pellets Even more common than spindle whorls were the small round pellets of clay found at the site; only three of these were found in burials, with the other eighty-five recovered from the layers themselves. These are again fairly standardized in size and weight, ranging from 11 to 18 mm in diameter (mean 14.2 mm) and from 1 to 7 g in weight (mean about 3 g). Our older workmen were quickly to convince us that they were in fact ammunition for a pellet bow (krasun or kathun in Phasa Isan), an implement used for bird and small-game hunting that was popular in the area until its replacement by the Western slingshot some thirty years ago. One of our workmen was quickly able to make an example of the pellet bow and demonstrate its use. Large numbers of similar pellets were recovered at Ban Chiang, and workmen at that site also immediately recognized their function and were able to make similar bows (White 1982a:24, Figs. 26-27, p.93, Fig. 176). As at Non Nok Tha, most of the finds were in nonburial contexts; however, one grave at Ban Chiang did contain a large pile of pellets as a deliberate grave offering. Finds of pellets are common at other Northeast Thai sites such as Non Chai and Ban Na Di (Higham and Kijngam 1984); they are also


found farther afield at Vietnamese sites like Phung Nguyen, where sixty-six were recovered (Hoang Nguyen 1978:103), Ben Do (Pham Van Kinh 1977:21), and Doc Chua (Nguyen Duy Ty 1977:39-40). At Non Nok Tha, the pellet bow-â&#x20AC;&#x201D;like the spindle whorl-â&#x20AC;&#x201D; appears to have been introduced at about the same time as bronze manufacture an the wide range of Middle Period vessel types made their appearance: that is, the end of the Early Period. The large number of pellets recovered from nonburial contexts in MP 5 again supports our interrelation of this level as the only one to have seen extensive occupational as well as cemetery use of the site, at least as far as the 1968 excavation area is concerned. The slight decline in number of pellets during the later iron period levels may be correlated with the introduction of the locally made musket (presumably in the eighteenth century or later), but this is of course only speculation. Roof Tiles These are flat slabs of fired clay with a slight ridge on one end for anchorage to each other. No whole examples were found at Non Nok Tha, but hose recovered during a later survey in Loei Province measured about 10 by 15 cm (Bayard 1980c:77); like the Loei examples, the Non Nok Tha ones were ca. 5-7 mm thick and contained slight amounts of fine sand temper; this is possibly a natural inclusion rather than deliberately added temper. Tile fragments were distinguished from sherds only in the 1968 analysis; the analyzer of the 1966 material presumably grouped them with plain and sand-tempered earthenware sherds. It is apparent from the 1968 distribution that these tiles are limited to the latter half of the Late Period; the finds in levels LP 3-6 are probably the result of chance discard of only a few tiles rather than indicative of the presence of a tile-roofed structure on the site at this time. Historically, such tiles were used only for important structures like bot or other religious buildings rather than ordinary houses (which used wooden shingles and nowadays use roofing iron), and the tile fragments are in such low numbers as to make the former presence of such a structure at the site unlikely. Firebox Fragments There are simply lumps of partially fired clay with one flattened side; they are interpreted as the remains of clay-lined cooking boxes for use on the kitchen platforms of houses. Such boxes may still be seen in use nowadays; however, they have largely been replaced by the portable clay-lined buckets that are the common form of stove in the area today (these are in turn slowly being replaced by bottled gas cookers). The distribution of these fragments in both excavation suggests considerable antiquity for this type of cooking facility; the large numbers of fragments in MP 5 again provide support for considerable occupational use during this level vis-Ă -vis the other EP and MP levels. Ladles Occasional finds were made of large clay spoons or ladles; these featured a fairly large bowl (3-6 cm deep by 4-6 cm wide) and a usually upward-curving, tapering handle some 7-8 cm long (Fig. 6-42). These were usually tempered with sand and plain finished, but several specimens were made of the same fine-chaff-and-sand fabric used for the common 2C/E, 2L, and 4D vessels; one of the seven specimens was red-slipped and another polished. The ladles appeared to have some funerary significance, as a majority were recovered in funerary contexts. In the case of the four 1968 specimens, all were placed inside class 2 or class 6 vessels, presumably as servers for the stew or condiment


in the pots; the three 1966 specimens are fragmentary and were not recovered from burials. Similar ladles have been found at Ban Chang, where they also appear to be characteristic of the early and middle bronze period (White 1982a:69), and at Ban Na Di (Higham and Kijngam 1984:173-174); these and several Non Nok Tha specimens closely approximate gourds in shape, complete with upturned “stem.” Clay Rollers and Stamp Nothing really resembling the now-famous and still enigmatic clay rollers of Ban Chiang (White 1982a:74-75; van Esterik and Kress 1978) was found at Non Nok Tha; however, the two clay cylinder fragments were recovered. The first of these was a plain cylinder (NP-1626) perforated down its long axis recovered from 1966 layer 17 (MP 5). The second and more interesting was carved is a simple hemispherical meander design with dots centered in each meander; this was found beyond the head of MP 8 Burial 47, but the context was not secure. If the cylinder is in fact associated with this burial, the estimated date is approximately equivalent to the appearance of the elaborately carved Ban Chiang cylinders after the onset of the Late Period at that site (ca. 300 B.C.), but the simple incised design of the Non Nok Tha specimen contrasts sharply with the elaborate designs carved in high relief of the Ban Chiang equivalents. The function of these latter has been the subject of considerable debate (van Esterik and Kress 1978, Folan and Hyde 1980 [AP 23(2)]); whatever their function, it seems likely that if differed from that if the incised Non Nok Tha roller, which may have been simply an ornament. In the absence of any complete specimens of further fragments, any additional speculation is of little value. Crucible: One partially cylindrical object (NP-133) and one cylindrical fragment (NP-1626) both from the 1966 excavation (provenience unknown) may have been part of an unusual crucible. The top portion of NP-133 (Fig. 6-43A) is hollow with a portion of the rim indicating the top was a small, hollow container. The reason I suggest this may be a crucible is that one quite similar to this, but more of it remaining, was recovered from the black pottery site of Ham Parker excavated in Phimai. [The Phimai piece is also a question but it is suggestive of a very small crucible on a long, solid cylindrical handle that could be used to hold the crucible with liquid metal in it and pour directly into a small mold to make an ornament. NP-1626 (Fig. 43B) could be a portion of the long, solid handle WS]. A single fragment from an incised clay stamp (NNT-778) was recovered from upper layer 6 in the 1968 excavation (i.e., MP 5); it measures about 1.5 by 3 cm and ws apparently designed to impress a raised design made up to hoked spirals (Fig. 6-43c). Similar stamps were found at Ban Na Di and also occur in the Vietnamese site of Hoa Loc (Ha 1980:Fig. 4c-f); this site is unfortunately ill dated but would appear to be of the third millennium B.C. according to one Vietnamese authority, bronze fragments and a few bronze artifacts were present (Pham Minh Huyen, pers. Comm.). “Tripod Tables” Three of these puzzling artifacts of unknown function were recovered. The first and largest was recovered from the pit of EP 1 Burial 6 during the 1966 excavation (NP-546; Fig. 6-43e). It is a 2 cm thick clay disc measuring some 10 cm in diameter, with three thick legs some 3 cm in length; a small hole is pierced through the disc near the edge. The disc is


rather roughly smoothed and plain finished. The remaining two “tables” come from the 1968 burials. MP 6 Burial 61 produced a similar but smaller (7 cm diameter) and unpierced disc, again with three legs of 2 cm length (NNT-376; Fig.6-43d), which was found lying well beyond the feet of the burial. The third specimen was located next to a standard nest of 2C and 4D pots beyond the head of MP 4 Burial 33; it is quadrangular rather than round (NNT-514; Fig. 6-43f) and measures 9 cm on each side, with three short legs of 1 cm length and a hole of about 1 cm in diameter pierced through one corner. This last object had clear traces of red pigment on its upper (?) surface, and the only functional speculation we can offer at present is that the objects were used for the preparation of cosmetics of some sort. White has also suggested (pers. Comm.) that they might function as stands for round-bottomed pots, which should be possible if the pot were a small one, although in this case accounting for the traces of red pigment would be problematic. Solheim has reported seeing very similar objects at a Vietnamese ceramics factory, where they were used as supports for round-bottomed vessels during firing, with the three legs being upward and the somewhat convex bottom set in sand. All three burials were relatively rich, and all were male. To the best of our knowledge similar artifacts have not been recovered from any other site in the region. “Effigies” Two possible fragments of animal effigy figures were recovered from the 1966 excavation. The first of these may be a portion of a bovine figurine (NP-1087; Fig. 6-43g), but it so it is a very crude representation; it was recovered from layer 13 (MP 8). The second, from layer 17 (MP 5), may represent an equally crude horse (NP-1066; Fig. 6-43H), but again considerable imagination must be used. If these fragments are in fact effigies, it is possible to make ties with similar crude figurines recovered from Ban Chiang and dated ca. 1000 B.C. to A.D. 2000 (White 1982a:70); and to the large number of more elegant bovine and other animal figurines found at Ban Na Di and dated 1000-600 B.C. (Higham and Kijngam 1984:158-172). “Pestle” A single conical clay object measuring 10 cm in height by 5 cm across the base was recovered from 1968 MP 4 Burial 39; it was found lying to the left of the skull (NNT-511; Fig. 6-43i). Its shape and relatively heavy weight (138 g) suggest use as a pestle of some sort, but it could not have been employed as such on anything but fairly soft material without severed abrasion; hence the functional attribution is quite speculative. Although it could have served as an anvil in pottery manufacture, its shape is quite different from the obvious anvils recovered from many Thai sites. Conclusions The results of the ceramic analyses cannot be said to be entirely consistent and clear, which is not surprising given the great bulk of data from the site, the pioneer nature of the excavations, and the number of different workers involved in the analyses. However, general points can be made. A. Ceramic conformity with stratigraphy: There is an overall conformity and agreement in the temporal variation of nonburial sherds and funerary vessels, reflecting a close


parallel to the general stratigraphy outlined by Parker and Bayard in Chapters 3 and 4; chronological variation in the whole vessel types is by and large closely echoed by variation in vessel attributes like temper, rimform, and surface decoration, although often with a obvious lag and some degree of blurring caused by upward displacement of earlier sherds as a result of grave digging. Macdonald’s independent analyses (1980) are also in relatively close agreement with the results arrived at here, with two relatively minor points of difference: the reality of the Early/Middle Period boundary and the chronological affinities of LP 1 and 2. I believe that the results of both non burial sherd and burial vessel analyses make a conclusive case for the presence of a marked cultural boundary at the EP 3/MP 1 transition and that the data cannot support Macdonald’s Phase I/IIa boundary placing only 1968 layer 9 in the former and lumping EP 3 and 3 with MP 1 and 2 in the latter (1980:209-210). The discrepancy almost certainly arises from Macdonald’s use of presence/absence rather than frequency data in his treatment of wares. Similarly, I believe on the evidence of a number of stoneware and porcelain sherds that LP 2 (or at least its 1966 manifestation as layer 9) cannot be grouped with MP 5-8 in Macdonald’s Phase IV; the level’s affinities are clearly with parahistoric Late Period. LP1 remains problematical; lacking further data from this rather ill-defined level, I accept that Macdonald’s grouping of it with the later Middle Period may be correct, although this seems stratigraphically unlikely given the presence of 1966 layer 12 (Chapter 3). Incorporation of quantities of Middle Period sherds into 1966 layer 11 by extensive pit digging would still seem a more likely explanation. With exception of this last problem, I submit that the stratigraphic and ceramic evidence presented in the above chapters demonstrates support for the security and reliability of the Non Nok Tha stratigraphic sequence in the relative sense. In spite of the considerable debate over the site’s absolute chronology. As mentioned in the previous chapter, we have never attempted to gloss over the stratigraphic difficulties encountered in both years’ excavations (Solheim 1968, Bayard 1971b:5-6); this has led some commentators to dismiss the evidence from the site as unreliable, since the site “was so heavily disturbed by burials, root holes, termites and washouts” (Clark 1977:346). However, such cultural and natural disturbance is the norm rather than the exception in open tropical sites of this type. Four of the five major (>50 m2) area excavations of burial sites carried out to date in Thailand have featured signs of just such intensive and continuing disturbance; the stratigraphy at Ban Kao and Ban Chiang was in fact more refractory than that at Non Nok Ta, and only the lower layers (7 and 8) of Ban Na Di present clearer stratigraphic contrasts than those of Non Nok Tha (they give ample evidence of intensive human disturbance as well). Khok Phanom Di (Higham and Bannanurag 1990), with its intricate but clear stratigraphy built up of ash, midden, and sherds, is the exception. Hence if we were to follow Clark’ dictum, the evidence from most Thai sites (not to mention many others) must be discarded; this is hardly reasonable, to say the least. In any event, the Non Nok Tha relative sequence, as defined during excavation and validated by the ceramic analyses presented here, can now be viewed as reliable. B. Site function: The analyses also shed considerable light on a second difference between our interpretation and that of Macdonald: the function of the site during the Early and Middle Periods. We have always held that the site (or at least that portion of it excavated) functioned primarily (but not exclusively) as a cemetery, with sporadic industrial (e.g., bronze casting) and occasional occupational use; only level MP 5 was viewed as representing fairly intensive occupation (Bayard 1971b:20). Macdonald, on the other hand,


follows Peterson (1977) in interpreting the site as basically a habitation, with interments being made adjacent to or under houses (1980:219; pers. comm.): that is, a situation similar to the convincing case he makes for the Ban Kao burials (Macdonald 1978). Obviously, the structural evidence presented in Chapters 3 and 4 makes it clear that some structures were indeed present on the site during many of the other EP and MP levels; this is also the case at other sites during periods when they functioned primarily as cemeteries (i.e., layer 7 at Ban Na Di). At both site, the structures themselves may have had some funerary function. However, our arguments for the mainly (but by no means entirely) funerary function of the site are based on the following points. Firstly, the site is a very small one (1.1 ha) compared to modern village sites in the region, which average about 25 ha in area; this argument is, however, not a particularly strong one, as many prehistoric habitations would have been similar than modern villages if our assumptions of an economy based on some form of wet swidden are correct. For example, the modern village of Ban Chiang is some 40-50 ha in area; however, archaeological deposits appear to extend over only some 8-15 ha (White 1986:29). Secondly, we have the more convincing stratigraphic/depositional arguments: the extreme density of the Non Nok Tha burials when compared to sites like Ban Kao (S贸rensen 1967) or even Ban Chiang. The scarcity of extensive middens, firepits, hearths, or concentrated structural evidence (except for MP 5); the relatively shallow depth of cultural deposits (1.4 m); and finally, the very low overall sherd density at the site. A portion of this low figure may be explained by our inability to screen or wash deposits, due to the soil and water supply problems mentioned in Chapter 2; however, this cannot account for the striking contrast with other Thai sites used for occupation or industrial activities as well as burial (Table 6-23). Industrial/occupational sites like Non Chai (Bayard, Charoenwongsa, and Rutnin 1982-1983) and occupational/burial sites like Ban Na Di (Higham and Kijngam 1984) feature sherd densities ranging from about 6.5 to 13 kg per m3. Ban Chiang (Gorman and Charoenwongsa 1976; White, pers. comm.) has an overall density of about 18.7 kg/m3; all these sites have deposits more than 3 m deep. Khok Phanom Di has a 6.8 m build up of deposits and an overall sherd density of about 14.8 kg/m3; most of this (5.3 m) was accumulated during five-hundred years of burial and apparent mortuary ritual use, producing a sherd density for this portion of the sequence 10.1 kg/m3 (figures preliminary only, courtesy of Brian Vincent). By way of contrast, Non Nok Tha averaged 0.45 kg/m3, with only 1.4 m of deposit build up over a time span greater than Ban Na Di even by the most conservative reckonings.


These data strongly support the contention that the siteâ&#x20AC;&#x2122;s primary purpose throughout most of the Early and Middle Periods was a place of burial; it is otherwise difficult to rationalize the low number of m3 per burial and the very low density of nonburial sherds. Moreover, the only level viewed here as giving evidence for fairly intensive occupational use (MP 5) stands out from the other EP and MP levels in a number of ways: the higher proportion of apparent domestic (or at least nonfunerary) sherds (See Table 6-3); the greater proportion of clay pellets and firebox fragments; and the markedly greater number of nodules of bronze apparently resulting from casting activities (Bayard 1981 and Chapter 7 below). These and a number of other indices for each of the 1968 levels are presented in Table 6-24.


This table provides the approximate thickness and volume of each of the levels as represented in the 1968 excavation, as well as densities of sherds and some other artifacts within them; a number of conclusions may be drawn. First, the density of sherds in MP 5 in terms of both number and weight per m3 is markedly higher than in the EP levels and MP 4 and resembles the pattern of the LP levels instead. However, the density index for MP 1 is even higher. It is our contention that the concentrations of sherds in MP 1 and MP 5 arose in different ways. When the number of burials (either secure, complete ones or the total figure) of each level is divided into the sherds/m3 figure, it is clear that MP 1 much more closely approximates the other EP levels and MP 4; our implication is, or course, that the high sherd densities of MP 1 result from the extensive grave digging and disturbance of earlier EP burials that produced MP 1’s close ties to the EP un the nonburial sherd analyses. This MP 1 has the highest figure for number of burials per m3; MP 5 has the lowest. MP 5 also has the highest figures for number and density of bronze scraps, firebox fragments, and clay pellets. The one weak point in our argument is the average weight of nonburial sherds in MP 5; we would expect that his figure would be closer to those of the LP layers (5-15 g/sherd) rather than those of the other MP and EP levels (17-20 g/sherd), but this is clearly not the case. Aside from this, we think the internal evidence strongly supports our case for occupational use of the site being largely limited to MP 5 prior to the beginning of the Late Period, although we wish to emphasize again that some sporadic occupational or industrial use of the site may well have occurred during the other EP and MP levels. Turning briefly to one final argument, a survey of the ethnographic literature also strongly suggests that Mainland Southeast Asian societies that have the degree of differential wealth apparently present at Non Nok Tha (see Chapter 8) very rarely inter their dead inside the village. Burials take place either in the forest (e.g, Lamet, Meo; Izkowitz 1979, Lebar et al. 1964) or in a discrete cemetery outside or alongside the village (Akha, Palaung, Shan, etc.; Lebar et al.). The Northern Chin (Lehman 1963) seem to form one of the few exceptions to this rule in that burial can take place wither in the house compound or in a discrete cemetery outside the village. Hence we feel it is safe to assume that Non Nok Tha indeed functioned mainly as a cemetery for most of its pre-Late Period existence — a burial place possibly utilized by the occupants of the other site in the area discussed at the end of Chapter 2. These sites appear to range in area from 10 to 20 ha and are far more likely candidates for extensive habitation than Non Nok Tha. C. Vessel function: The use if Calder’s ethnographic pottery data and Macdonald’s analytical scheme has allowed us to make a number of functional inferences. While these must remain only inferential hypotheses pending the development of more sophisticated methods of testing, they are at the least plausible. Our biggest need at present is for a representative sample of domestic wares from a clear, nonfunerary context. It seems safe to assume that the common undecorated 1A, 1G, and perhaps 1C vessels are representative of ordinary domestic cooking vessels during the Early and most of the Middle Period; similarly, type 3, 4A, and 5A vessels may represent domestic serving ware, although this is perhaps less supportable. But we believe that many of the highly standardized class 3 and 4 vessels (See Table 6-17, Fig. 6-31)--and very likely class 6 as well-—were manufactured especially for funerary use and are not simply domestic wares utilized as grave goods. Our views thus differ from those of Macdonald, who believes that the marked increase in storage vessels-—be they ritual or domestic-—during the Middle Period (roughly, his Phase IIa-Phase IV) reflects on overall increase in emphasis on


storage technology in the economy as a whole (1980:256). Macdonald also feels that the apparent decline in average size of cooking vessels (i.e., our class 1) from EP 1 to the late MP serves as “an index of [reduction in] the size of the domestic unit within which food was prepared.” On the basis of other evidence, Macdonald then proposes that “there seems to have been a sharp increase in population pressure to which a response was a decrease in the size of the average domestic unit” (258). While both hypotheses are ingenious ones, we cannot accept them on present evidence. Our belief is simply that the increase in storage vessels and decrease in average size of class 1 cooking vessels are due to changes in funerary ritual; these or these may not reflect socioeconomic changes is society as a whole, but it is unlikely that they reflect the changes Macdonald postulates so directly. We feel that the marked increase in storage vessels is primarily a result of the introduction of ritual making the inclusion of type 2C/L and 4D vessels a desirable if not mandatory part of the burial custom, rather than solely an increased economic emphasis on storage technology. Similarly, the decline in average size of cooking vessels is due to the replacement of the Early Period ritual, which is emphasized large type 1A and 1B vessels as well as the smaller 1Cs and 1Gs, by what might be called the Middle Period “ritual storage” complex; during the Middle Period, types 1C and 1G also declined in number (see Table 6-21). Hence the overall reduction in size is due more to replacement of types than to a reduction of average size within each type of category; in fact, many of the later Middle Period type 1D, 1E and 1I vessels are quite large (see Table 6-17 for mean type dimensions). Our argument for the introduction of new burial ritual as a reason for these changes in vessel typology of course begs the question of the origins and reasons for such changes in ritual (but see the concluding chapter); however, the same might be said about arguments relying on increased population pressure. In fact, a strong argument against such pressure serving to reduce average domestic unit size exists, in that analyses of the skeletal remains from both excavations provide no evidence of nutritional stress; the population appeared to have been as well fed during the Middle as during the Early Period--if not more so (see reports by Pietrusewsky and Brooks in the projected volume of specialists reports). In short, we feel that while both our functional attributions and Macdonald’s are reasonable, we do not think that changes in funerary wares can be used to infer the side socioeconomic changes that Macdonald postulates. Such changes may well have taken place, particularly during the later Middle Period, as Macdonald proposes; however, we feel that a convincing demonstration will require a representative sample of domestic wares and domestic structural evidence as well as our large sample of funerary vessels; in other words, the excavation of a representative village with well-defined structures and refuse areas. D. The Early/Middle Period transition: The ceramic analyses have emphasized the importance of the Early/Middle Period boundary. While it seems clear that an overall replacement of one cultural tradition by another did not take place, as evidenced by clear continuities in ceramic technology and vessel types like 1A, 1C, 1D, 1G, and 2B, it is equally obvious that rapid cultural and technological change was taking place. In addition to the appearance of bronze in relative abundance and the techniques of casting the metal as well, the first occurrence of spindle whorls and clay pellets at about this time hint at other technological innovations arriving in the region. The sudden appearance if a whole suite of new “ritual storage/serving” types (i.e., most of the class 2, 4, and 6 vessels) also strikingly suggests considerable cultural change, most likely as the result of groups or


influences entering the Phu Wiang region at this time. These new vessel types did not all remain static during the Middle Period; on the contrary, elaboration of new types and varieties continued (2A, 2F, 6C, etc.), and changes took place in burial ritual as well. These changes in ritual and the distribution of the various vessel types occurring in burials form the basis for several hypotheses on the nature of social structure at Non Nok Tha during the Early and Middle Periods that will be discussed at some length in Chapter 8.


Chapter 7 Non-Ceramic Artifacts By Donn Bayard and Wilhelm Solheim II The two years of excavation at Non Nok Tha yielded a relatively large number of artifacts aside from the ceramics discussed in the previous chapter. These included a fair quantity of bronze and much smaller number of iron objects. In addition to numerous artifacts of both local and imported stone. Also found were smaller number of shell items and a few artifacts of bone, antler, and horn. In view of the interest generated by the evidence for early bronze technology at the site, we present the data on bronzes and bronze-related artifacts (molds, crucibles) in the first section of this chapter, along with the iron materials recovered; the remainder of the non-ceramic technology is described in the second section of the chapter (“Stone, Bone, Shell, Etc.”). BRONZE AND IRON · BAYARD INTRODUCTION One of the most important results of the excavations at Non Nok Tha was the unequivocal stratigraphic documentation of a bronze period at the site—-a lengthy period during which an obviously sophisticated bronze technology was practiced but lacking any evidence at all for the use of or familiarity with iron. While earlier excavations in Mainland Southeast Asia had of course been yielding bronze artifacts-—albeit in many cases of uncertain provenience (e.g., Mansuy 1902)-—Non Nok Tha provided the first clear evidence that a bronze period of some length preceded the first use of iron in the region. Although meager by comparison with later sites of the first millennium B.C. in Viet Nan, Non Nok Tha did produce a significant body of bronze artifacts, casting spillage, and bronze-associated artifacts such as sandstone molds and crucibles. The relevant data for the 1966 and 1968 excavations are presented in tables 7-1 and 7-2. TABLE 7-­1 BRONZE  AND  RELATED  ARTIFACTS  FROM  THE  1966  EXCAVATION AXE


















1 -­

-­ -­

-­ 1

5(1) -­

-­ -­

3 2

19 5

-­ -­

-­ 3

-­ -­



















































(Definite and  probable  provenience  only;;  including  burial  goods)




















1 1 -­

-­ 1 -­

-­ 1 1

-­ -­ 1

(1) 2 1(2)

-­ -­ -­

-­ 2 -­

9 14 23

12 10 91

4(1) -­ 1

-­ -­ -­

2(1) 2(4) -­
































Definite and  probable  provenience  only;;  including  burial  goods

These tables clearly show the continuous presence of bronze at the site from the end of the Early Period onward; the frequent finds of bronze lumps and droplets that in most cases may be interpreted as casting spillage also indicate that bronze casting was carried out at the site throughout the Middle and Late Periods. This appears to have been concentrated particularly in MP 5 and LP 2, but MP 3 and LP 1 also feature fairly heavy concentrations of nodules. As with subsequently excavated bronze period sites in the region, the evidence for bronze technology consists of molds and crucibles as well as the bronze remains themselves. Because a considerable number of the metal items and related artifacts such as molds and crucibles were recovered in fragmentary form from nonburial contexts, provenience in the descriptions that follow is given by the rather more precise 1966 and 1968 layer systems as well as the overall EP/MP/LP level systems. MOLDS, MOLD FRAGMENTS, AND MOLD BLANKS Non Nok Tha produced a relatively large number of bivalve molds and mold fragments, as well as unfinished sandstone blanks. In all cases these were made of locally occurring Phra Wihan and Phu Kradu’ng sandstone (see Chapter 2 for lithic descriptions); not surprisingly clay molds for lost-wax casting more characteristic of late first millennium sites like Non Chai (Bayard at al. 1982-1983:49-50) and the upper layers of Ban Na Di (Higham and Kijngam 1984:128) were not found at Non Nok Tha. Although the latest one or two levels of the Middle Period very likely overlapped the beginning of economic “takeoff” in the region (see Chapter 9), clay mold casting-—like iron tools—-seemed not to have reached the site prior to its abandonment. But it is interesting and perplexing to note that at the earliest dated metalworking site in Central Thailand-—Non Pa Wai in the Khao Wong Prachan Valley (Pigott and Natapintu 1994—-only bivalve clay molds (with no clear sign of the use of the lost-wax process) were recovered; no sandstone molds were found. Copper working (with no tin added) dates from the mid-second millennium B.C. at Non Pa Wai, some five hundred years or so later than the chronology proposed here; see Chapter 5 for a full discussion. MP 1: NNT-529a/b 3E layer 7; B. 73 grave pit (Fig. 7-1). This pair of axe molds was made of grayish, fine-grained Phra Wihan sandstone; total weight of the pair was 1,160 g (side A


578 g; side B 582 g). The mold pair was rectangular in elevation rather than the more usual trapezoidal shape and measured 12 by 8.6 cm. Each mold half Was 3 cm thick, with the assembled mold having a casting vent aperture measuring some 2.3 by 7 cm. Three incised grooves were carved into the walls of the vent on each half just above the bladesocket junction. The mold cast an axe with a diamond-shaped blade below the socket, with a hemispherical cutting edge not as pointed as that produced by the NP-82 mold (see below). The axe would have measured about 6.7 cm from tip of the blade to socket and about 7.3 cm across; the socket would have extended another 4 cm. It should have had a thickness of 2.4 cm at socket mouth, 1.6 cm at socket-blade junction, and 1.4 cm at the middle portion of the blade, tapering to zero at the cutting edge. The bronze axe NNT-528 that was found lying between the two mold halves at the foot of the grave in fact agrees quite closely with these measurements (see below); only the blade is some 6-7 mm shorter than expected, and this could be explained by wear and sharpening. Fig. 7-1


Yet, the axe has three hafting ridges on one side but only two on the other. Unless the lowest of the three grooves was deliberately filled with clay at the time of casting, it is hence unlikely that the axe was actually cast in this mold pair. Heat blackening appears to extend fairly evenly over the entire hollowed portions of the molds, with no clear demarcation between blackened and nonblackened portions at the top of the vent. NNT-530a/b 2F layer 7 â&#x20AC;&#x153;Mold Complexâ&#x20AC;? (Fig. 7-2a). A second much smaller pair of molds, also of Phra Wihan sandstone, weighing 324 g (158 and 156 g). Trapezoidal in shape, the assembled mold would have been 8.7 cm long and 6.2 cm wide at the bottom, tapering to 4.8 cm at the top; it should have been about 4 cm thick, with a pouring aperture of about 1.5 by 3.6 cm. Three grooves were carved into the bottom of the vent on one valve, but only two on the other. The mold would have cast a small axe with slightly flaring blade and very lightly curved cutting edge; the blade would have measured about 2.4 cm at the socket junction, flaring out to a maximum width of 5.0 cm at the edge. The socket would have extended about 3 cm above the lowermost hafting ridge. The thickness of the axe would have narrowed from 1.7 cm at socket mouth to 1 cm at socket-blade junction to about 0.7 cm in the center of the blade. As with NNT-529 above, heat blackening appears to extend over the entire socket portion of both valves. One of the valves had broken along the edge of the socket and blade portions with a deep crack on the other side of the valve, and the mold may well have been selected for burial use for this reason. Fig. 7-2


NNT-409 2F layer 7 (Fig 7-2b). A single fragment weighing 325 g, apparently from a halfmold blank of Phra Wihan sandstone. The piece measures 7.5 by 5.6 cm, and is 4 cm thick at its thickest point. It originated from one of the upper corners of a trapezoidal valve. Although the inner surface of the fragment is well smoothed, the outer side has a somewhat rougher finish, and the valve may have broken prior to hollowing of the mold and final finishing of the exterior. MP 2: NP-550a/b E3 layer 20; mound of B. 32; prov. doubtful (Fig. 7-3a). A pair of axe molds of Phu Krad’ung sandstone measuring 13.5 by 8 cm and weighing about 1000 g; they are rectangular rather than the more usual trapezoidal shape. Although both valves were broken at the upper end, only one piece is missing from a corner of one of the pair. The molds were designed to case a round-bladed socketed axe measuring some 7 cm wide by 9.5 cm along the socket axis, with the usual three ridges at the socket to facilitate hafting. The axe would have been rather similar to modern (forged rather then cast) ones observed in use in the area today. Extensive heat blackening is present over most of the socket area, and was apparently present lower in the molds as well; however, most of it in this region has flaked away. Heat blackening also covers the area above the socket, but the extreme upper left and right corners of the molds are free of discoloration. The fairly even line between the two zones suggests the use if some sort of a fitted plug during casting. The molds were recovered from the edge of a presumed burial mound; as the burial was never cleared and tied into the overall stratigraphy, provenience must remain doubtful. NP-486 layer 20 (Fig. 7-3b). A small axe mold fragment of Phu Kradu’ng sandstone weighing 71 g; it came from the edge of the socket area of the valve. The original mold half would have been 3 cm thick, and the fragment measures about 3 by 3 cm. NP-1430 E4/F4 layer 20 B. 28 grave pit (Fig 7-3c). A small fragment of Phu Krad’ung sandstone weighing 48 g that appears to come from an unfinished half-mold blank; as the inner surface of the presumed valve appears slightly convex rather than flat, the attribution must remain tentative. The piece measures approximately 4.5 by 4.0 by 2.0 cm, and the original valve (if in fact this is from one) would have had a maximum thickness of about 4.5 cm. NP-1585 E4/E5 layer 20 B. 20 mound (Fig. 7-3d). A large mold blank fragment of Phra Wihan sandstone weighing 512 g. This appears to have been shaped and partly smoothed before being split into two halves; as the bedding of the sandstone runs at an angle of about 25º off the desirable plane of cleavage, manufacture may have been abandoned when this was discovered. Only the upper half of the trapezoidal block was found; this measures about 10 by 6 by 5 cm.


Fig. 7-3

MP 4: NP-461a/b D4 layer 18 B. 22 (Fig. 7-3e). Two fragments of Phu Kraduâ&#x20AC;&#x2122;ng sandstone, possibly part of one half of a bivalve mold. These fragments are badly weathered, and can in fact be portions of grindstones rather than molds. The smaller fragment appears to come from the curved outer edge of a valve and measures about 5 by 3 by 2 cm. The larger may have come from a part of the socket vent, but this is doubtful; it measures 7 by 4 by 2 cm. Total weight is 14 g.


NP-1434 C3/C4 B. 4 (Fig 7-4a). A small fragment of Phra Wihan sandstone weighing 64 g possibly from a mold valve or blank. As it is badly chipped the functional ascription is quite doubtful. About 6 by 4 by 1.5 cm in size. Fig. 7-4

NNT-31 2F/3F lower layer 6; prov. doubtful (Fig 7-4b). A half-mold blank of Phra Wihan sandstone weighing 390 g. The top and bottom of the blank are missing, so that it presently measures about 7.5 cm square, with a thickness of 3.5 cm.


MP 5: NP-82 C5 layer 17 B. 10; prov. Doubtful (Fig. 7-4c). A complete pair of bivalve axe molds of Phra Wihan sandstone; total weight is 796 g (side A 376 g; side B 420 g). The trapezoidal molds measure 10.8 cm in height and are about 6.7 cm wide across the top and 9 cm across the bottom (the two valves are not exactly the same size). Maximum thickness of each half are 3.0 and 3.2 cm; thickness of the assembled mold is 6.2 cm at its midpoint. The mold was designed to cast a diamond-shaped axe flaring out from the socket and narrowing almost to a point at its tip. The axe would have rounded measurements of 7 cm in width and 6 cm along its socket axis to the single cast-in ridge, which may have been its end; however, the mold extends some 6 cm above the groove and an extension of the socket for part of that length is likely. Grooves in the molds running from the transverse hafting groove extend down to the shoulders of the blade; these are faint in one half but deeply carved in the other. Here blackening is present over the whole of the hollowed-out casting area, as well as around the central part of the rim of the pouring aperture on both mold halves; however, discoloration is lacking at the outer ends of the pouring hole. Several greenish bronze oxide stains were present on the exterior surface of the mold. NP-107 C5 layer 17 B. 10; prov. doubtful (Fig. 7-5a). A second complete pair of axe molds from the same provenience, also of Phra Wihan sandstone; total weight 1500 g (side A 742 g; side B 758 g). Aside from two chipped areas in the margins, this massive pair of molds was recovered intact. Each is of trapezoidal shape with the gently rounded bottom and measuring 14.9 cm in the smaller of the pair and 15.2 cm in the larger; top ends measure 9.9 and 9.6 cm respectively. The assembled mold stands 13 cm high, with the larger and heavier of the pair tapering from a thickness of 3.1 cm at the base to 1.5 cm and the other member widening from 1.8 to 2.6 cm; the two sides are thus joined on a bias rather than along a vertical join. The mold was used to cast a thin, wide-bladed axe similar to 1968 NNT-527, with a slightly curved blade measuring about 13.3 cm in length along the cutting edge, 10.6 cm along the back of the blade, and about 4.2 cm wide. Two grooves were inscribed in one half of the mold where blade joins socket, but only a single groove is present in the other half; grooves are also inscribed along the back and sides of the blade, probably to create ridges strengthening the axe. Green bronze oxide stains are clearly present on the base of valve A and back of valve B. Blackish heat discoloration is present over most of the casting area but is noticeably darker below a depth of about 4 cm in the socket area of one of the pair; this may indicated the upper limit of the cast socket, which would then have a length of about 4 cm behind the blade. On the other hand, the case socket may well have extended farther up the 8 cm length of the vent, as this would seem to give a more secure hafting to an implement that would absorb a fairly large amount of energy on impact. Cracks running from the casting area outward to the edges that are present in both valves could be the result of thermal shock; it may be that the mold was approaching the end of its working life.


Fig. 7-5

NP-310 C4/C5 layer 17 B. 10; prov. probable (Fig. 7-5b). Most of one-half of a third pair of axe molds from the same burial but of Phu Kraduâ&#x20AC;&#x2122;ng sandstone; weight 561 g. The trapezoidal mold measured 15 cm along its base by 11 cm high and was designed to cast a wide-bladed axe rather like NNT-527 but with straight rather than curved blade; the axe would have measured about 10 cm along the socket axis (with three hafting ridges), with a 14 cm blade. The probable provenience of this artifact slightly strengthens the case for the ascription of NP-82 and NP-307 to this burial rather than to the mound of MP 1 B. 8. NNT-755 3F/4F layer 6 B. 28. A large smoothed piece of Phu Kradâ&#x20AC;&#x2122;ung sandstone weighing 670 g and measuring about 8 by 10 by 3 cm; it was recovered from the foot area of the burial, ad its shape suggests that it may have been a mold valve blank.


MP 6: NNT-36 3F layer 6A B.27 (Fig. 7-6a). A mold fragment of Phra Wihan sandstone weighing 230 g. The fragment measures some 7.5 by 6 cm and is 4 cm thick; it comes from the socket and pouring aperture section of the mold half and clearly shows three grooves for casting ridges into the neck of the axe. A clearly defined heat-blackened potion above the socket area suggests that a plug of some sort was inserted into the mold pair prior to pouring of the molten bronze. The axe to be cast was apparently of a semilunar type similar to NNT-528. Fig. 7-6


MP 8: NP-607 F4 layer 13 (Fig. 7-6b). A fragment of a mold half of Phu Kradu’ng sandstone similar to NP-486 above, measuring about 4.5 by 4 cm by 3 cm thick and weighing 58 g. This piece came from the socket portion of what was apparently a fairly small mold; two groves have been carved in the neck area. NP-1141 E5 layer 13 (Fig. 7-6c). A massive mold-half black of Phu Kradu’ng sandstone weighing 1395 g. The blank is in a rough, unfinished state, and is rectangular rather than trapezoidal. It measures 11 by 10 by 6 cm and may well have been intended to be split into two halves for construction of a pair of molds. Although the piece has been trimmed parallel to the bedding plane of the sandstone, it was apparently judged unsatisfactory and seems to have finished its career as a sharpening stone, resulting is a shallow, saddleshaped concavity covering part of one side. NP-1564 E3 layer 13 (Fig. 7-6d). A piece of half-mold blank of Phu Kradu’ng sandstone; weight 115 g. The piece was originally on the edge of a central part of the mold blank and measures 5 by 6 by 2.7 cm; breakage seems to have occurred prior to smoothing of the inner surface of the blank, as it is in a fairly rough state. LP 2: NP- 1486 D7 layer 10 (Fig. 7-6e). A small fragment from the edge of a bivalve mold or blank made of Phra Wihan sandstone; it comes from the curved edge of the valve and measures ca. 5 by 5 cm, with a thickness of 2.7 cm and weight of 95 g. The inner surface is weathered, with no trace of any carving or hollowing. NP-1500 D6 layer 9 (Fig.7-6f). A small (4.5 by 3 by 2 cm) possible valve fragment of Phu Kradu’ng sandstone; weight 95 g. This may have come from a side of a thin mold half; no trace of a casting indentation is present. It more likely came from a partly finished blank. CRUCIBLES The excavations produced additional evidence of in situ bronze casting in the form of several whole crucibles included as offerings in burials, as well as a small number (twelve) of crucible fragments recovered in layer removal. The relatively scanty finds of fragments in occupational/industrial deposits contrast sharply with sites like Ban Na Di, where some 158 crucible fragments were recovered (Higham and Kijngam 1984:84-91). This as well as the lack of any clear signs of bronze casting furnaces suggest that bronze working was carried out to a far less intensive degree at Non Nok Tha; although very similar numbers of bronze lumps and detritus were recovered from Ban Na Di and Non Nok Tha (approaching five hundred in both sites), the excavation of the former site encompassed a considerably smaller area (70 m2). Unlike Ban Na Di, all crucibles and fragments recovered at Non Nok Tha were tempered with rice chaff and a small amount of sand (possibly a natural inclusion); no examples of the grog-tempered crucibles characteristic of layers 7 and 8 at Ban Na Di (Higham and Kijngam 1984:85-87) were found at Non Nok Tha. Although size and volume estimates are given below for the more complete crucibles, it should be


emphasized that the latter are tentative, based as they are on average measurements and the application of the ideal formula for a prolate spheroid (4/3 ab2) after wall thickness is subtracted. The total volume is divided by 3 to reflect the subhemispherical shape of most of the crucibles; weight of bronze capacity is based on a specific gravity of 8.8 g/ml EP 3(?): NP-1359 D4 layer 21 B. 26; prov. doubtful. Fragment from rim of crucible; weight 13 g, thickness of body 13 mm. Sherd obtained from disturbed area in grave fill, and may well be MP rather than EP 3 in origin. Contains rather more sand and less rice chaff than other specimens. MP 1: NNT-513 3E/S layer 7 B. 73 (Fig. 7-7a). Complete crucible recovered in one piece from lower portion of B. 73 grave; although slightly disturbed by later grave digging, it seems likely that the crucible was placed over the lower legs of the corpse, along with the bronze axe and pair of molds recovered from this grave. The crucible measured 9 by 7 cm, and had a total height of 6 cm; its weight was 154 g. Subtracting the 10 mm wall thickness, it is estimated that it would have held about 120 ml, or about a kilogram of molten bronze. Fig. 7-7


NNT-531 2F layer 7 “Mold Complex” (Fig. 7-7b). Substantially complete crucible recovered in eight pieces from the “Mold Complex,” which also contained a small pair of axe molds (see above) and portions of a human cranium as well as several vessels. It measured 13 by 10 cm including the spout and 10 by 10 cm excluding spout; total height was 7 cm and weight 101 g. Wall thickness averaged 15 mm and estimated volume is about 150 ml of bronze, weighing about 1.3 kg. NNT-532 1F layer 7 mound of B. 114. Fragment from base of a relatively flat-bottomed crucible; thickness 10 mm, weight of sherd 24 g. Bronze nodule adhering to inner surface. MP 4: NNT-525 2F/3F layer 6B B. 86 (Fig. 7-7c). Almost complete crucible found beyond feet of B. 86 in fifteen pieces. Total length and width at rim was 15 by 10 cm; excluding spout the body measured 11 by 10 cm. Total height 7 cm, weight 370 g, and mean wall thickness 12 mm; estimated volume of bronze 200 ml, weighing about 1.7 kg. NP-1984 2F/3F layer 6B B. 86 (Fig 7-7d). A second substantially complete crucible recovered with NNT-525above. Total length including estimation for ca. 2 cm of missing spout is 18 by 10 cm at rim; body measured 15 by 11 cm. Total height 8 cm, mean thickness 11 mm, weight 370 g. Prominent bronze stain on inner lip. Estimated volume of contents 250 ml, or about 2.2 kg of molten bronze. NNT-270 4F mid-layer 6. Fairly massive fragment from below spout area; mean thickness 14 mm, weight 31 g. NNT-271 1F mid-layer 6. (Fig. 7-7e). About one-half of a uniquely small and thin-walled crucible recovered in two pieces. Of standard shape and chaff-and-sand temper but oxidized to a reddish color, with walls averaging only 6 mm thick. Estimated original dimensions 9 by 6 cm, height 3 cm; approximate volume only 20 ml, holding less than 200 g of molten bronze. The crucible hence may have been used for some sort of specialpurpose work. Weight of the two surviving sherds is 27 g. NNT-278 mid-layer 6. Small crucible fragment weighing 10 g. NNT-488 4D layer 6B B. 97. A fairly large fragment from below spout area; thickness of sherd 12 mm, weight 31 g. Original crucible may have been about 5 cm high. MP 5: NP-521 C4/C5 layer 17. Smallish crucible recovered in one piece from area between Bs. 7 and 10 (both originating from layer 17). Rim measurements 10.5 by 8 cm; body 9 by 8.5 cm. Total height 5 cm, average thickness 10 mm, weight 163 g. Several nodules of bronze adhering to interior. It would have contained an estimated 75 ml of bronze, weighing about 660 g according to the formula adopted here, although Smith in his examination of this object estimates a capacity or about 800 g (1973:30). Smith concluded that heat was applied to the crucible by heaping fuel over it and supplying oxygen with a bellows or


blowpipe (more likely the former, if ethnographic analogies are any guide); the crucible was not laid in or on an open fire. See discussion below. NP-807 C3 layer 17. Two matching crucible fragments (total weight 25 g) recovered from layer 17 contexts and sent to the University of Pennsylvania for thermoluminescence dating in 1973. The resulting date of 3070 BP (PT-? in Chapter 5) generally agrees with the chronology presented in Chapter 5; however, the side margin of error (520 years) makes the date very unreliable. MP 7: NP-212 C5 layer 15. A very small crucible fragment, apparently from a rim; thickness 12 mm, weight 3 g. Reddish slag (?) nodules adhering to exterior. NP-1594 e$ layer 15. A single fairly large fragment from rim; thickness 14 mm, weight of sherd 28 g. Thin reddish slag (?) deposit on inner side of rim. LP 2: NP-1526 D5 layer 9. Two non-fitting crucible sherds, the smaller from a rim. Thickness 13 mm, total weight 24 g. Slight slag traces on interior of larger sherd. NP-1545 D7 layer 10. Crucible sherd, thickness 12 mm, weight 10 g. Many bronze nodules adhering to lip and inside of rim. NP-1595 E4 layer 9. Small body sherd; thickness 9 mm, weight 7 g. LP 5: NP-1593 D4 layer 2. Small rim sherd; thickness 14 mm, weight 5 g. Thin slag adhesion inside rim. PROVENIENCE UNKNOWN: NP-1527 recovered from dump. A fairly large body sherd from a crucible probably 4 cm high; thickness 13 mm, weight 25 g. BRONZE ARTIFACTS The 1966 and 1968 excavations yielded some 79 classifiable bronze artifacts, plus another 116 possible fragments of objects and about 340 nodules from casting spillage; the total weight of the bronze recovered is about 2.2 kg. When one recalls the richness and quantity of bronze finds of Dian and Dong Son sites--or even those of the Go Mun period in northern Viet Nam-â&#x20AC;&#x201D;the Non Nok Tha finds seen petty by comparison and, indeed, Higham has commented on the relative rarity of bronze at sites like Non Nok Tha and Ban Na Di (1989:147). On viewing the Non Nok Tha bronze finds, however, at least one archaeologist specializing in the Near East, Lamberg-Karlovsky, was surprised by the quantity of bronze represented by the few pieces he saw (Solheim, personal


communication). Given the relative abundance of copper and tin cores in the region as a whole (Charoenwongsa 1978) and the obvious presence of established trade networks in Northeast Thailand well before the first appearance of metal (as evidenced at Non Nok Tha by exotic shell beads, adzes of nonlocal stone, etc.; Bayard 1971a, Kennedy 1977), it is not surprising that bronze rapidly became a commodity in this network (Higham 1984a, 1988b:143-144, 1989:187). As the Non Nok Tha sequence spans the period of transition to bronze technology, the bronze and bronze-related artifacts can provide some insight into the nature of the technology, if not its origins. Well over one hundred analyses of the Non Nok Tha bronzes have been carried out by a half dozen different metallurgical specialists (Pittioni 1970, Smith 1973, Stech Wheeler and Maddin 1976, Carriveau 1978, Selimkhanov 1979, Rajpital 1983); these included both qualitative and quantitative analyses of most of the bracelets and bracelet fragments, a number of possible object fragments, and lumps interpreted as casting spillage (summaries of quantitative analyses will be presented in the projected volume of specialist reports). However, only two of the five bronze axes were quantitatively analyzed, as the Thai Department of Fine Arts was reluctant to have the remaining pieces sectioned for sampling. The relevant analyses are referred to in the descriptive section below, and a general discussion of the technology follows. Axes: EP 3: NNT-152 2E layer 8U B. 90 (Fig. 7-8a). The earliest and probably best known on the Non Nok Tha finds, this was the only metal artifact recovered form an Early Period context at the site; the only other definitely confirmed finds of bronze in a pre-Middle Period context were two tiny fragments in the pit fill of the empty grave of EP 3 B. 93, adjoining B. 90. The tool was found in a very secure context, among the ribs of B. 90, which is provenienced to EP 3 both by stratigraphy and typology. The tool weighs 149.8 g and is trapezoidal in shape, narrowing from a width of 5.3 cm at the socket to an estimated original width of 4.3 cm at the blade end (the lumpy mass of corrosion that now covers the blade is slightly wider, at 4.6 cm); the total length of the tool is 9.0 cm and its thickness is 2.2 cm at the socket and 1.1 cm at the corroded blade end. The socket measures 4.6 by 1.7 cm. The shape is thus quite distinct from the later bronze axes and molds recovered from the site, all of which exhibit widening of the axe from socket to blade. Also distinctive is the asymmetry present along the long axis of the tool: Although evenly curved from side to side on one side, the other is faceted into two narrow side planes and a wide central plane. Given the early and secure context of this tool, I jocularly nicknamed it WOST (“World’s Oldest Socketed [Metal] Tool”). an acronym that has unfortunately gained more publicity—including taking the nickname seriously—-and longevity than I ever intended! Initial qualitative analysis carried out by the Thai Department of Mineral Resources in 1968 suggested that the tool was made of hammered native copper; however, the presence of phosphorus argues for a smelted source for the copper. Subsequent analyses of the only other two metal fragments from this level using an electron microprobe also revealed that these two samples (and, by implication WOST as well) were in fact tin bronze, with about 5% Sn (Bayard 1972). Quantitative analysis carried out by Maddin (personal


communication) suggests that WOST is in fact 17% Sn, a percentage much higher than would seem desirable in a cutting or digging tool of this sort. Fig. 7-8

MP 1: NNT-528 3E layer 7 B. 73 (Fig. 7-8b). A semi-lunar axe recovered from the foot region of the burial, along with axe molds NNT-529a/b and crucible NNT-513 (q.v. above). The axe is a relatively massive one, weighing 241 g, and measures 9.8 cm from socket to midblade. Width across the shoulders of the blade is 7.9 cm, 3.1 cm at the blade-socket junction, and 6.2 cm at the socket mouth. The socket mouth itself measures 3.8 by 1.8 cm. The tool is 2.8 cm thick at the socket mouth, narrowing to 1.8 at neck and 1.3 cm at the bladeâ&#x20AC;&#x2122;s widest extent. X-ray photography of the tool shows the socket extending 7 cm deep into it, well below the broadest part of the blade. The socket narrows to 1.3 cm at the neck but then broadens to a maximum width of 3.7 cm shortly before ending. Obviously no sort of permanent plug was used to shape this socket, as intact removal after casting


would have been impossible; a specially shaped plug of fired clay may have been used and broken out after the casting cooled. Hafting ridges or flanges measuring about 3 mm wide and 2 mm high were cast across and above the neck of the tool, three on one side and two in the other; they are uniformly spaced about 3 mm apart. MP 3: NP-549 C6/C7 layer 19 B. 69 (Fig 7-8 c). This was the only bronze axe recovered from the 1966 excavation and is quite similar in shape to NNT-630 from the second excavation described below; it was found lying on the left thigh of the burial. It weighs 100 g and measures a maximum of 6.6 cm from the center of the blade to the center of the socket mouth; the top ends of the socket rise 3 and 5 mm at each side. The splayed “horned” blade measures 7.1 cm across the horns, narrowing to 2.7 cm at the blade-socket junction and widening again to 3.9 cm at the socket mouth, which is 2.7 by 1.3 cm. From a thickness of 1.9 cm at mouth, the tool narrows to 1.4 cm at blade socket and 0.6 cm at the blade’s maximum width. X-rays reveal a socket 5.2 cm deep, tapering from 2.7 to 2.1 cm in width. Smith’s analysis of this tool (1973:27-28) indicates it is made of a relatively low tin bronze (Sn <8%), with a “fair amount” of lead. While obviously cast, the cutting edge of the blade (probably including the horns as well) “has a totally different microstructure” (p. 27) and has been shaped by annealing and cold working of what was originally a thicker part of the casting. Smith suggests the original casting was a simple tapered piece; the axe could possibly have been produced from a mold such as NNT-530 described above. No hafting ridges were cast into this axe, however. Pittioni’s spectrographic analysis supports Smith’s rather surprising finding of lead in a tool such as this, indicating approximately equal amounts of tin and lead (1970:159); however, at least some Dong Son bronze weapons also contain from 4% to as much as 21% lead (Murowchick 1988). Together with the five bronze bracelets (NP-553a.e below), this is the earliest example of leaded bronze recovered from the site. MP 4: NNT-630 2E layer 6B B. 85 (Fig. 7-9a). An axe very similar to the above, but larger; found lying on the right chest of the burial. Weight 135 g, maximum length along socket axis 9.4 cm; width across horns of blade 7.2 cm, 2.9 cm at neck, 6.6 cm at socket mouth. Ends of socket mouth are about 3 mm higher than middle of socket. Axe narrows from 1.8 cm at mouth to 1,1 cm at socket-blade junction to 0.45 cm at widest part of the blade. Socket mouth measures 3.1 by 1.1 cm; X-ray photographs indicate the socket extends some 7 cm into the body of the axe, tapering from 3.1 to 2.0 cm in a fashion very similar to NP-549 above. Three faint hafting grooves measuring ca. 1.5 mm wide by 1 mm high are cast into one side of the axe and above blade-shoulder junction, spaced about 2.5 mm apart. Only one such groove is present on the other side of the axe, paralleling the lowest groove on the obverse; however, two faint ridges appear to have been cast, running down the flared sides of this face of the blade (as in mold NP-107). This makes is seem unlikely that the flared shape of the blade has been obtained by extensive annealing and cold working; rather; such a process may well have been responsible only for the horns and the blade edge.


Fig. 7-9

MP 6: NNT-527 4H layer 6A B. 1 (Fig. 7-9b). A broad-bladed axe resembling the casting product of mold NP-107 described above, with side, flaring blade; weight 141 g. Total length along the socket axis is 11.5 cm, maximum width at corners of out-curved blade 14.9 cm; allowing for broken portion of one side, maximum original width would have been 16.5 cm. Width across â&#x20AC;&#x153;hornsâ&#x20AC;? on upper side of blade 9.5 cm, width at blade-socket junction 3.1 cm, and at socket mouth 4.2 cm. The socket is 1.7 cm thick at the mouth, narrowing to 0.7 cm at blade junction; the blade itself is fairly uniform 3 mm in thickness. X-ray photography reveals that the blade is 1 mm or so thicker along its side and upper margin, almost certainly as a result of cast ridges such as those present in mold NP-107; the socket is by no means as clearly shown in x-ray as in the axes already described, but it


appears to fill most of the area between socket mouth and blade junction, ending in a rather abrupt blunt curve. The x-ray shows a clearly thicker ridge of metal near the inner base of the socket, which is quite likely a strengthening element formed by means of a groove carved in the core. Of greater interest is what appears to be an area below this where the base of the socket casting overlaps the blade itself, which may suggest that the socket was cast on in a separate operation, perhaps through accident or mischance. LP 3: NNT-536 4H layer 5U (fragment; Fig 7-9c). A 17.3 g blade fragment, not likely from an axe; 4.7 cm by 1.7 cm. The alloy used was about 12% tin, with almost no lead added (the <1% present is probably not intentional). Rajpitak’s analyses indicate that the tool was annealed and worked after casting. NNT-540 3E layer 5M (fragment Fig. 7-9d). A second axe fragment, also from LP 3 and also probably intrusive from MP layers. This one is slightly larger, measuring 4.5 by 3.1 by 1.4 cm; it weighs 35 g. It is derived from the shoulder of a semilunar axe similar to NNT-528 but noticeably smaller. The fragment is large enough and contains enough of the socket to make a tentative reconstruction of the original axe, although with a wide margin of error: socket to blade midpoint ca. 6.3 cm; maximum blade width ca. 6 cm; width at socket mouth ca. 3.5-4.5 cm; socket ca. 4.5 cm; walls of socket about 4 mm thick. Arrows: MP 4: NNT-307 2F layer 6.7 interface. An arrowhead or similar projectile point, very badly corroded and broken into three pieces; dimensions 3.7 by 2.5 cm, weight 6.6 g. Rajpitak’s analysis was limited by the severe corrosion (presumably producing the low Cu percentage), but it suggests a standard tin bronze: 39% Cu, 5% Sn, 0.5% Fe; no Pb detected. MP 5: NNT-41 4F layer 6U (Fig 7-8e). Rajpitak describes this piece (or rather three fragments) as “a very heavily corroded chisel-headed arrowhead, lacking the tang”; it measures 2.4 by 2 cm and weighs 7.2 g. The analysis again produced only suggestive results due to the heavy corrosion: Cu 44%, Sn 5%. As 0.02%, Fe 0.14%. Pins or Nails? MP 5: NNT-236 1E layer 6U (Fig. 7-9f). This 3.1 g fragment has a 2 mm square cross-section, and Rajpitak’s analysis shows clear signs of annealing and heavy working after it was cast; it could also be a tool fragment. LP 1:


NP-1547 C4/D4 layer 11/12 interface (Fig. 7-9g). As this piece comes from a layer interface, it could derive from the erosion of MP 8 layers that formed layer 12 in the 1966 sequence; alternatively, it could have been brought up into LP 1 by disturbance or manufactured at that date. The piece, described as “a conical bronze fragment” by Rajpitak, weighs only 1.8 g and is 2.0 cm long, tapering from 6 mm at one end to 2 mm at the other; its function remains doubtful. Wire? MP 4: NP-271 E6 layer 17/18 interface. Two small, thin bronze fragments weighing only 0/8 g; may be wire or pieces of a very thin bracelet. MP 5: NNT-267 1F layer 6M. Another tiny fragment of bronze with circular cross-section weighing les than a gram. Bracelets: Level totals are shown in parentheses. The first number indicates “substantially complete” bracelets; the second is for “fragmentary” bracelets, including fragments or doubtful provenience not shown in Table 7-2. MP 1 (0-2): NNT-328 2F layer 7. A 2.7 g fragment of a bracelet with a rectangular cross-section measuring 4 by 6 mm; Sn 10%, no Pb detected. NNT-522 2E layer 7; prov. doubtful. A heavily corroded 0.5 g fragment of a bracelet with 3 by 3.1 mm square cross-section; Sn 6% Pb 0.5%. MP 3 (5-1): NP-553a/e D5 layer 19 B. 63 (Fig. 7-10). Five bronze bracelets recovered from the skeleton of a young adult female; she wore two of these on her right wrist and three on the left. All are of rectangular cross-section with slightly convex sides-—what Smith described as “barrel-shaped” (1973:29) in his description. The three bracelets on the left arm (Fig. 7-10a) have been fused together by corrosion and weigh a total of 103 g. The two remaining bracelets (Fig. 7-10 b and c) weigh 20.5 and 24.5 g respectively. These bracelets have an inner diameter of 5.5 cm, and a thickness of 4 mm, and width perpendicular to the diameter of 4.5 mm. Smith’s examination of the bracelet from the right wrist showed that they were cast in their present shape, with no sign of subsequent working; the alloy employed was a medium-tin bronze with some amount of lead (1973:29).


Fig. 7-10

After subsequent cleaning, the two bracelets were also analyzed by Rajpitak in London. Electron probe microanalysis indicated a tin component of 2 to 10% and 0.3 to 0.6% lead; however, analysis of the corrosion product gave percentages of 5% for tin and 6% for lead. Rajpitak also commented on the large amounts of black lead globules present and concluded that the bracelets were cast and rapidly cooled. At an estimated date in the midsecond millennium B.C., these (along with the axe NP-549 above) are the earliest leaded bronzes recovered from the site. The fused set of three bracelets from the left wrist was not cleaned and analyzed, as the outer surface of the artifacts had over one cm2 of mineralized woven fabric adhering to it similar to the fabrics described by Higham and Kijngam from Ban Na Di (1984:126). These and the other bronze artifacts from the site are held by the Thai Fine Arts Department, and is hoped that further analyses will be carried out on the nature of the fabric.


NP-1552 C6 layer 19. A corroded bracelet fragment weighing 1.1 g with circular crosssection of 1.5 mm; estimated diameter of original bracelet is 5.5 cm. Corrosion product (CP) analysis shows a relatively high percentage of tin (16%). MP 4 (3-1): NNT-310 3F layer 6/7 interface. A small (0.7 g) bracelet fragment with 2 mm square crosssection; CP analysis shows 9% tin. NNT-546s (Fig. 7-11a), NNT-547a/b 4H/E layer 6 B. 15. Three bronze bracelets recovered from the wrists of a child of about five: originally each wrist bore two bronze bracelets with a bracelet made of marine shell (Tridacna sp. 7; G.M. Mason, personal communication) between them, but as the burial was partially in the baulk, one bronze bracelet and most of a second were not lifted. The two intact bracelets, one of them the shell bracelet (see Solheim section following for description, weigh 34 g each and have a D-shaped crosssection 4 mm thick and 12 mm wide perpendicular to the 3.6 cm inside diameter. Fig. 7-11


MP 5 (18-3): NP-551/1-8; NP-551/a-i E3 layer 17 B. 31. Perhaps the most visually spectacular bronze find of both excavations was the skeleton of a young woman wearing sixteen bronze bracelets on her lower left arm and an additional one on her right wrist (Pl. 10 a and b). This last and one of the sixteen on the other arm were circular in cross-section (3 mm thick) and decorated with four groups of knobs (Fig. 7-11b), each formed of two rows of three circular bosses 5 mm thick, set at 90º intervals around the bracelet; the knobs are bordered by parallel bands. The remaining fifteen bracelets all had a D-shaped crosssection measuring 9 mm wide by 4 mm through the thickest part of the D (Fig. 7-11 c to f). Although these were of the same width and thickness, their diameter varied from 5.3 to 7.0 cm (inside diameter), and they appear to have been a deliberately graduated set: 5.3 cm, 1; 5.4 cm, 2; 5.5 cm, 1; 5.7 cm, 2; 6.0 cm, 3; 6.2 cm, 1; 6.5 cm, 1; 6.7 cm, 1; 6.8, 1; 7.0 cm, 1 (one bracelet was not analyzed by Rajpitak). Their weight varied accordingly from 20.5 to 31.7 g. Five of the D-shaped bracelets were examined by Smith, and Rajpitak analyzed all but two of the seventeen. Smith arrived at the conclusion that casting in a bivalve mold would have been difficult; lost wax manufacture was a more likely method, although “neither certain nor necessary” (loc. cit.). If such were the case, Smith offers a very plausible method for their casting: “The D-shaped section may have been formed [in wax] over a performed cylindrical core of clay; then the whole was embedded in clay, the wax melted out, and the metal poured into the space formerly occupied by it” (loc. cit.). If the cylindrical core were tapered and multiple wax models applied around it, the result could have been a graduated set of bracelets like these. Rajpitak’s analyses of the bracelets reveal an interesting pattern: the fourteen D-shaped ones contain between 8 and 12% tin (mean 10%) and from 0.5 to 1.2% lead (mean 1%); hence it is debatable whether or not the lead was a deliberate addition. However, the single decorated bracelet with circular cross-section that she analyzed contained 8.6% tin and 4.5% lead, resembling NP-553 above (but note that the analysis is based on corrosion product). One of the decorated bracelets was also reported on by Stech Wheeler and Madin (1976:43). They also suggest that it was cast as a whole (including decorative motifs) by the lost wax process and rapidly cooled. NP-1093 C6/C7 layer 17 B. 15; prov. doubtful. This 5 g, 1.3 cm segment of a decorated bracelet was also examined by Stech Wheeler and Madin (1976:44); it is 3 mm thick , with knobs 3 mm wide and 2 mm high set at approximately 10 mm intervals. It is difficult to envision a method other than lost wax for the construction of the bracelet. The curvature of the piece suggests an overall diameter of about 9 cm. NNT-233 2F layer 6U (Fig. 7-12a). A small piece of a rectangular sectioned bracelet 2 mm thick and 3.5 mm wide, weighing 1.3 g. CP analysis indicated 17% lead.


Fig. 7-12


NNT-249 3F layer 6U. Two pieces from a circular cross-sectioned bracelet weighing 7 g. NNT-544 1F layer 6 B, 23 (Fig. 7-12b). A fairly massive small bracelet or large ring (circular cross-section 4-5 mm); inside diameter 3.6 cm, outside diameter 4.5 cm, total weight 29.2 g. Rajpitak’s CP analysis indicated a lead content of about 12% and about 9% tin. The bracelet/ring was found resting over (not on) the left wrist of a male in his early thirties. MP 6 (1-1): NNT-526 4H layer 6A B. 1 (Fig. 7-12c). Complete bracelet with circular cross-section of 4-5 mm and a diameter of 7.2 cm (OD; 6.3 ID; weight 23.2 g) It was found on the right forearm of a male aged about thirty-five. Rajpitak’s CP analysis shows a very low tin content (2.2%) and a lead content of over 12%. NNT 543 3F layer 6A B. 27. A very corroded fragment from a bracelet of inderterminate section; thickness 3.8 mm, weight 4.0 g. MP 8 (0-2): NP-1536 E4 layer 13. Three possible pieces from a bracelet with thin rectangular crosssection the fragments are only 1-2 mm thick and about 7 mm wide and are heavily corroded. Weight 2.2 g. NP-1551 C4 layer 13. A very small, corroded fragment of a bracelet with semirectangular cross-section (slightly convex on outer face), 2 mm thick by 3 mm wide; weight 0.5 g. Rajpitak’s CP analysis shows 74% copper and 3% tin but 21% lead. LP 1 (0-2): NP-253 C3 layer 11. This 1.1 g bracelet fragment was too corroded to allow Rajpitak to determine its cross-section. NP-1533 C3 layer 11. This fragment has a cross-section identical to NP-1551 above but with more normal proportions of tin (12%) and lead (2%), with “large amounts of small black lead globules” (Rajpitak 1993). Estimated diameter of original bracelet is 5.9 cm. LP 2 (0-11): NP-265 C& layer 9 (Fig. 7-12d). A corroded fragment of 3 mm circular cross-section weighing 1.8 g from a bracelet with an estimated diameter of 5.6 cm. Almost equal percentages of tin and lead (7.6% and 7.1% respectively). NP-277 D5 layer 9A. Two very small fragments from a bracelet with a rectangular crosssection measuring 5 mm wide by 2 mm thick; Rajpitak mentions that these are decorated with a circular relief pattern. The two fragments together weigh only 0.6 g and contain 17% lead and only 2.6% tin.


NP-280 C7 layer 9. Three corroded fragments weighing a total of 3.7 g from a bracelet with rectangular cross-section measuring 4 mm wide by 2-2.5 mm thick; one fragment is the catch of the bracelet. Analysis again shows a high lead content (17%). NP-293 C7 layer 9. Two fragments from a bracelet with a 3 mm cross-section; weight 1.5 g. NP-820 C6 layer 9A. A fragment from a circular cross-section bracelet weighing less than 1 g. Not analyzed. NP-1509 C5 layer 9A. A single piece weighing 0.9g from a thin bracelet with a B-shaped cross-section 1.5 mm thick and 7 mm wide, with a pronounced valley running down the center of the bracelet and cast-in circular relief decoration. Contains 15% lead. NP-1550 E6 layer 9 (Fig. 7-12e). A small (0.7 g) fragment from a very thin (1 mm) circular section bracelet with an estimated diameter of 5.0 cm; 13% lead. NP-1553 D5 layer 9 (Fig. 7-12f). A second similar piece weighing 0.8 g and slightly thicker (1.5 mm); estimated diameter of original bracelet 7.1 cm; 17% lead. NNT-141 4E layer 5L. A fairly weighty (2.9 g) but heavily corroded bracelet fragment about 2-2.5 mm thick. Cross-section not ascertainable. No lead detected in CP analysis, but with the sum of elements detected amounting to only 18% results are not conclusive. NNT-185 3E layer 5L (Fig. 7-12g). Fragment from bracelet with 3 mm square crosssection, with outer side convex; weight 1.7 g. No lead detected. NNT-207a 3F layer 5L. Small bracelet fragment weighing 1.2 g; CP analysis indicated 64% copper, 36% lead, and only 2% tin. LP 3 (0-2): NP-275 C7 layer 8. A small (0.6 g) fragment from a bracelet with 1.5 mm circular crosssection; estimated diameter of original 8.2 cm, but thinness of sample may have distorted estimate. Contains 17% lead. NNT-134 4E layer 5U (Fig. 7-12h). A similar piece weighing 1.8 g from a bracelet with a 3 mm circular cross-section; 13% lead. LP 6 (0-1): NP-1092 G4 layer 1 (Fig. 7-12i). A 5.6 g piece from a bracelet with rectangular crosssection 3 mm wide by 5 mm thick; 86% copper, 11% tin, and only 0.13% lead. The piece is almost certainly upwardly intrusive from earlier LP or possibly MP deposits below; some of the other fragments described for the LP above quite likely originate from MP deposits. Rings:


LP 2: NP-1094 D7 layer 9 (Fig. 7-12j). This delicate (0.6 g) but complete ring is only 0.5 mm in thickness and 13 mm in diameter; it is rectangular in cross-section, with the outer surface convex. Due to its completeness and uniqueness, it was not metallurgically analyzed by Rajpitak. NNT-189 2F layer 5L. A very small (0.2 g) piece from 1.5 by 1.2 mm square ring, with slightly convex surface facing outward; 19% lead. LP 3: NP-264 E6 layer 8 (Fig. 7-12k). A 0.9 g fragment from a ring with ribbed relief design on outer surface. Cross-section is rectangular 1.5 mm thick by 6 mm wide, with decorated outer surface slightly convex. The decoration was formed during lost wax casting of the ring as a whole; 11% lead content and only 0.6% tin. NNT-149 4F layer 5M. A small (<1 g) fragment from a rectangular cross-section ring; not analyzed. Beads: MP 4: NNT-548 3E/3F layer 6B B. 47. A plain ovoid bead weighing less than 1 g. LP 3: NNT-192 1F layer 5M. A second plain bead, round in shape, weighing less than 1 g. DISCUSSION Technology Even when it was the sole stratigraphic witness to what I referred to in a 1971 paper as “an early indigenous bronze technology” (Bayard 1980b), Non Nok Tha alone suggested the outlines if not the flesh of the manufacturing techniques employed. Bronze was obviously being cast at the site (crucibles, nodules, molds), although not smelted (no slag present); tools such as axes were being cast in bivalve sandstone molds, while decorative items like bracelets and rings were in all probability being made by the lost wax process. By 1970 it was apparent from analyses such as Smith’s (1973) that leaded bronze was the alloy of choice for at least some of these ornaments as well as some of the weapons, and all this at estimated dates in the second millennium B.C. and earlier. Although rather insecurely provenienced specimens of bronze and stone axe molds had been recovered from much earlier excavations (e.g., Mansuy 1902, Lèvy 1943), the novelty of the stratigraphically secure finds and the startlingly early radiocarbon dates


associated with them aroused considerable interest and controversy. While Non Nok Tha stood as almost the sole evidence for this technology, such controversy was of course understandable and justified; as I said in my 1971 paper, “given the lack of data and adequately excavated sites in this huge region this [speculation] will doubtless continue to be the case for at least a decade” (1980:204). In fact, as discussed in Chapter 5, speculation and debate has perhaps predictably continued for over thirty years.1 Fortunately the range of the debate has been considerably narrowed by other excavations since then, both in Northeast Thailand and farther a field in Viet Nam. The 1974-1975 Fine Arts Department--University of Pennsylvania excavations at Ban Chiang (Gorman and Charoenwongsa 1976, 1978) provided evidence of similar bronze axes, crucibles, and mold fragments at what seemed to be even earlier (pre-3000 B.C.) dates than Non Nok Tha. Subsequently, the excavation by Higham and his student-—now our colleague-— Amphan Kijngam at Ban Na Di, near Ban Chiang, produced very important evidence for early bronze technology in the region (Higham and Kijngam 1984:81-147). Firstly, the carefully collected and provenienced charcoal samples from this site provided quite conclusive evidence for well developed bronze technology by the middle of the second millennium B.C. Even more importantly, the site produced not one but several in situ bronze-casting hearths from layers dating between 900 B.C. and the beginning of the present era. Finally, the large interdisciplinary team recruited by Higham allowed an indepth examination of artifacts and technology, including even an analysis of the wax used in lost was casting (Ibif:124-126). Farther a field, intensive archaeological research in Viet Nam beginning in the 1960s and continuing to the present has revealed a large number of parallels with the materials from Non Nok Tha; there are general technological parallels at the northern sites of the Dong Dau and Go Mun cultures, and possible parallels with some of the as yet incompletely reported early Sa Huynh sites in central Viet Nam. The most striking parallels are with the site of Doc Chua, northwest of Ho Chi Minh City; many of the numerous axe molds and axes from this site could just as well have been excavated at Non Nok Tha (Ha Van Tan 1980; Murowchick 1988; Higham 1988b, 1989:171). The technology as evidenced by the totality of this material seems generally quite straightforward. Bronze was traded into the site in metallic form, either as ingots of copper and tin or more likely of the finished alloy itself; as no ingots have been recovered, this remains to be discovered. Some of the sources of the copper used in Northeast and Central Thailand-—if not the tin-—have now been investigated through the important research carried out by Pigott, Natapintu and their associates in the Thailand Archaeometallurgical Project. Extensive copper mining and smelting was carried out at Phu Lon, on the Mekong northeast of Loei, during the latter second millennium B.C. (Pigott and Natapintu 1988; Natapintu 1988; Pigott at al 1992). Their further research in the Khao Wong Prachan Valley (Natapintu 1988, 1991; Pigott and Natapintu 1996-1998) has revealed a very extensive copper mining and smelting industry beginning at about the middle of the second millennium B.C. and extending well into the first, including the casting of copper ingots that were very likely the vehicle for trading the copper component of bronze. The sources of the tin and lead employed remain to be positively identified, along with possibly earlier copper mining sites; as Murowchick (1988) points out, the Lao side of the Mekong may be one possible location.


This plus some evidence from Non Nok Tha suggests that the constituents of bronze usually arrived there as separate metals rather than as ingots of bronze or a finished product. If Selimkhanov’s twenty-eight spectrographic analyses of eleven bronze-casting spills from Non Nok Tha are any guide, it appears that lead may have been traded in separately and added to the alloy at the time of casting; none of the spillage samples, which ranged from EP 3 to LP 2, had more than 0.04% of lead (1979). The bronze was melted in crucibles made of clay tempered with large quantities of rice chaff and straw, and in the case of the earlier layers at Ban Na Di with grog; chaff temper came to predominate at the latter site only after about 500 B.C. (Higham and Kijngam 1984:84-91). These crucibles had an average capacity of about 100-200 ml at Non Nok Tha, sufficient to hold about 0.9 to 1.8 kg of molten bronze. However, at least two smaller crucibles were found at the site (one holding only an estimated 20 ml), and both of the crucibles recovered from Ban Na Di in latter first millennium B.C. contexts were toward the low end of the Non Nok Tha range, holding about 70 to 90 ml by the formulae adopted here (75 and 80 ml respectively by the excavators’ estimate; Higham (1988;134). Much larger crucibles were of course employed by this time in Viet Nam. At the Dong Son cemetery site of Lang Ca, a crucible capable of holding an estimated 12 kg—-over five times the total weight of the bronze recovered from Non Nok Tha-—was recovered (Murowchick 1988). Although there were at least two major periods of fairly intensive use of the site for bronze casting as evidenced by abundant casting spillage (MP 5 and LP 2; see Tables 7-1 and 2), no clear traces of hearths of furnaces to melt the bronze (or copper and tin ingots) were discovered. However, we can extrapolate from those discovered at Ban Na Di: The bronze was melted in small, scooped-out furnaces lined with blocks of clay. The layer 7 example described by Higham (1988:Fig. 13.5, 133) measured about 75 by 60 cm; it had a base of compacted earth and sides of burnt earth and clay enclosing a hollow some 25 cm deep filled with charcoal. An inlet at one side presumable admitted tuyères worked by blowpipe or bellows (given the widespread ethnographic distribution of piston bellow-—still in use in the area today-—the latter is more likely). As the inlet in the Ban Na Di example described is at the top of the furnace, it seems likely that Smith is correct in assuming the crucible was placed in the furnace and fuel (charcoal) then piled on top rather then underneath it (1973:30). Once melting was complete, some sort of relatively heat-proof tongs-—perhaps of green hardwood or bamboo-—was used to remove the crucible for pouring into the (probably preheated) stone molds. The mold blanks which were found at Non Nok Tha-—ranging from rough unsplit blocks to fragments from split and smoothed halves-—hint strongly at the manufacturing process employed: roughing out and splitting of a suitably sized sandstone block; smoothing of inner and outer surfaces; incising and hollowing out the shapes of the item to be cast; and final fine smoothing of the inner surfaces of both halves to ensure a good casting with minimal flashing. Subsequent excavations at Doc Chua in 1976-1977 (Dao Lin Con 1977, Nguyen Duy Ty 1977) have produced axe molds almost identical to those recovered at Non Nok Tha (e.g., see Nguyen Duy Ty 1977, Fig.5). More importantly, as Murowchick points out in his summary of the metallurgical finds there, more than fifty molds in all stages of manufacture have been recovered, illustrating the cutting, chiseling, and polishing if the preliminary stages of manufacture:


Several of the Doc Chua molds retain faint traces of yellow clay outlines, whereas others suggest that a model of the object to be cast was held against the mold and exposed to a smoky fire, staining the outline of the model onto the mold surface. The shape was then chiseled out, a delicate process subject to errors that forced the Doc Chua mold makers to discard a number of partially finished molds. Finally, the inner surfaces of the mold were carefully polished; a few of the molds received simple decoration at this stage, usually consisting only of one or two raised ribs running around the body of the axe between socket and blade. Many of the molds show a variety of incised lines on their outside edges, undoubtedly meant to facilitate the careful alignment of the mold halves before poring the molten bronze. (Murowchick 1988:191-192). No traces of clay or smoke outlining were found on the Non Nok Tha molds, nor was there any sign of incised registration grooves for aligning the halves; nonetheless it is clear that the mold technology employed must have been nearly identical to that at Doc Chua, some 800 km southeast of Non Nok Tha. Although we have always assumed that the sandstone molds were employed to cast bronze items directly, Higham has suggested that perhaps they were used only in an intermediate stage to case models of low-melting lead alloy. These latter models would then be used to shape clay models that when fired could better resist the thermal shock of casting than sandstone, which would have been destroyed during casting (Higham and Kijngam 1984:83; Higham 1984c:140, 1989:149); this conclusion was stimulated by the discovery at Ban Na Di of a unique lead-tin alloy casting sprue (Higham Kijngam 1984:110). However, given the lack of any traces of clay axe mold fragments at Non Nok Tha, Ban Chiang, or other related sites and the obvious signs of heat blackening on many of the Non Nok Tha pieces, their use for casting of lead models only seems unlikely; it would be difficult to explain the bronze corrosion stains on molds noted above if this were the case. At least one pair of molds has demonstrably withstood the shock of casting after some three thousand years of burial; White (1982a:Fig. 56, pp. 48-49; cf. p. 57) illustrates a bronze axe cast in mold pair NP-82, flanked by the molds themselves. Murowchick also finds this argument unconvincing (1988:192), and Carriveau suggests that the apparent slow cooling of cast axes was accomplished by preheating of stone molds (1978); it would be difficult to achieve such slow cooling in a clay mold unless it were of considerable thickness. Even Higham and Kijngam conclude that it is still probable that the stone molds “were used to cast directly in bronze” (Ibid:125). As Higham says, “There is, as yet, no other supporting evidence that this intermediate lead-casting stage was used” (Higham 1989:149), and he repeats this view in his latest work (Higham 1996:233). There remains the interesting question of the significance, if any, of the differential pattern of heat blackening observed in the socket mouth area of some of the molds (particularly NP-550, NP-82, and NNT-36). This may have resulted from the use of some type of (fired clay?) plug shaped to fit the socket opening and ensure a symmetrical socket in the cast tool. No such patterns are shown on drawings of molds from Doc Chua and elsewhere, however, and the suggestion remains speculative until such plugs are found.


Typology As the bronze materials from Non Nok Tha were less variable than those from Ban Na Di (in particular, clay-shaped beads and the bronze bells characteristic of the upper Ban Na Di layers are absent), it is not possible to present a fine-grained typology of such artifacts as bronze bracelets. However, we can definitely speak of three types: plain circular section, plain D-section, and decorated circular. The majority of these are made of leaded bronze; only the fifteen plain D-sectioned bracelets worn by 1966 MP 5 B. 31 have low (but perhaps still intentional) lead content. The five bronze axes and six major axe mold finds allow a general but obviously very tentative typology of bronze exes to be postulated: A. Trapezoidal axes with sides slanting in toward blade (NNT-152); this “type,” represented by only one specimen with a blade consisting of a mass of corrosion, is obviously ill-defined but seems rare in other sites. The only general parallel that comes to mind is an undated specimen form the Mlu Prei area (O Pie Chan) pictured by Lévy (1943:P1. XXII, 3). B. Trapezoidal axes with sides sloping outward to blade, with the blade often flared and “horned” (mold NNT-530, axes NP-549, NNT-630). The three specimens range from MP 1 to MP 4 at the site and are fairly common throughout the region. C. Axes with semilunar or diamond-shaped blades (molds NNT-529, NP-82, NNT-36, axe NNT-528). The four specimens span MP 1 to MP 6 at the site; the type occurs widely throughout Southeast Asia. D. As above, but with smoothly rounded shoulders producing an ovoid axe; the only sample is the mold pair NP-550 from MP 2(?), and in fact the type appears to be quite rare on the mainland of Southeast Asia (although it occurs in the Philippines; e.g., see Solheim 1967b:Fig. 3A). Higham illustrates a vaguely similar specimen from Samrong Sen (1989:Fig. 3.541, p. 180; 1996:21), presumably after Noulet (1879:P1. VII,6); however, this axe has a slight shoulder on one side and appears form the original plate to have been more pediform than oval in shape when cast, with its asymmetry largely removed by subsequent corrosion. E. Axes with widely extended “wings” and either a straight or slightly curved blade (molds NP-107, NP-310, axe NNT-527). These appear to be limited to MP 5 and MP 6 at the site, although it is of course impossible to be conclusive with such a small sample. The type is quite distinctive, and even general parallels are difficult to find elsewhere. Given the small sample of axes and molds (large though it is by regional standards), little can be said save that types B and C, fairly widespread elsewhere, enjoyed a fairly long period of popularity at Non Nok Tha during the Middle Period, while type E is apparently limited to the later part of the period. The status of type A remains uncertain until more excavations are carried out of sites that also bracket the transition to bronze technology.


Alloys Similarly, no really clear-cut pattern appears in bronze alloy variation through the span of the Middle Period. Higham and his associates have documented a transition at Ban Na Di from a more or less uniform use of low-tin (2-12%) bronze to the use of a range of alloys, including leaded tin and very high tin, and Stech and Maddin (1988) have presented evidence for a similar transition at Ban Chiang in the latter first millennium B.C. The transition at Ban Na Di coincided with the abandonment of the cemetery at the site and the beginning of manufacture of specialty items like bronze bells (Higham and Kijngam 1984:106-107; Higham 1988b). The more or less uniform use of low tin alloys was thus the norm from about 1500 B.C. until layer 5 times at about 100 B.C. Given the relative chronologies of the two sites, we would expect a uniform use of low-tin bronze throughout the Non Nok Tha Middle Period, the end of which coincided with the abandonment of the site and the economic and technological “takeoff”-—what I have termed “General Period C” above-—evidenced at Non Chai (Bayard at al. 1982-1983), the Ban Chiang Late Period (White 1982a), and layer 5 at Ban Na Di. The Non Nok Tha Middle Period bronzes however exhibit a considerable wider range of alloy variation than those of Ban Na Di layers 8-6. Figure 7-13a presents a graphing (by MP 1-4, MP 5-8, and LP levels) of tin vs. lead percentages for forty-four artifacts analyzed by Rajpitak (1983); and forty eight lumps and nodules analyzed by Rajpitak (twenty analyses) and Selimkhanov (twenty-eight analyses of eleven pieces). As the latter were done using spectroanalysis only (Selimkhanov 1979:33), they should be viewed more cautiously than Rajpitak’s. The analyses done by Rajpitak shown here do not included badly corroded specimens with less than 70% copper present in the analysis. When compared to a similar figure for thirty-nine Ban Na Di artifacts (Higham 1988: Fig. 12, p.136), it seems clear that while overall tin content is roughly the same for the Non Nok Tha Middle Period artifacts and layers 8-6 at Ban Na Di (0-15% as compared to 2-12%), there is more variability in lead content during the latter Middle Period than exists in Ban Na Di layers 8-6; Late Period Non Nok Tha (terminal first millennium A.D. on) specimens predictably exhibit the same wide range of lead content as the earlier layer 5-3 specimens from Ban Na Di. The used of leaded bronze alloys obviously has considerable antiquity at Non Nok Tha, not surprising given the occurrence of lead bronze alloys in late Phung Nguyen contexts (Murowchick 1988:183). Rather more surprising is the uniformly low lead content in the thirteen early Ban Na Di artifacts analyzed. Factor analyses of the Non Nok Tha bronzes present a pattern similar to that found at Ban Na Di (Higham and Kijngam 1984:116-120; Higham 1988b:Fig. 13, p. 136). Rather than list lengthy tables of factor loadings for each artifact analyzed, we instead utilize a Q-mode analysis based on elements rather than objects, using the six elements detected in the vast majority or artifacts; we also limit the sample to the 87 artifacts analyzed by Rajpitak, excluding the spectrographic results obtained by Selimkhanov (which give no percentage figures for Cu). The sorted results from the SPSS× program (which extracted three factors; see Chapter 5 for a brief discussion of the method) are shown in Table 7-3, with factor loadings less than 0.3 omitted; note that similar results were obtained by including only the sixty-four artifacts with 70% or more copper.


Fig. 7-13



FACTOR 2 (22.9%)

FACTOR 3 (17.2%)









Pb Cu Fe

.687 -­-­-­ -­-­-­

-­-­-­ .897 -­.880

-­.313 -­-­-­ -­-­-­





Instead of the sharp contrast in variability of tin and lead evidenced by the Ban Na Di bronzes, the first factor in the Non Nok Tha analysis subsumes variation in arsenic, antimony, and lead (the first two as minor trace elements, of course); the second contains copper; and the third subsumes the variability in tin, The degree of variability of both tin and lead throughout the assemblage (as opposed to little or no lead admixture in the earlier Ban Na Di bronzes) may account for the placing of lead and tin in different factors. Analysis of lumps and nodules at Non Nok Tha-—most of them presumed to be casting spillage-—produces a quite different pattern (Fig.7-13b). Tin content is considerably more variable, but this is true only when the spectrographic analyses are taken into account. Lead content, however, is uniformly low: from 0/trace only to 1% in MP 1-4, 0/trace to 4% in the later Middle Period, and 0/trace to 2.5% in the Late Period. Either lead was being added separately to the molten bronze, as suggested above, or a considerable amount of lead has leached out of these usually very small and corroded specimens. However, some ring and bracelet fragments are also very small and corroded and yet contain a considerable proportion of lead. On the other hand, it is difficult to envisage bronze casters being more prone to spill before adding lead to their crucibles than afterward! Only further analysis of a much larger body of samples from a larger number of sites will provide a definitive answer. A final point about the bronze nodules and droplets: When the number and weight of nodules for each of the two excavations is plotted against the intercalated overall sequence of levels, a satisfying degree of congruence is present. Peaks in both number and weight of nodules occur at the same points in both sequences: MP 1, MP 5, and LP 2, with additional peaks at 1966 MP 3, MP 8 and LP 1, not present in the 1968 area (Bayard 1981). These presumably reflect periods of more or less intensive bronze casting at the site. The close agreement in distribution provides evidence corroborating that of the nonburial pottery presented in Chapter 6.


ORIGINS The source of the Southeast Asian bronze technology described here for Non Nok Tha and now documented at a number of other sites remains one of the most problematic and controversial questions in the regionâ&#x20AC;&#x2122;s prehistory. I think it is fair to say that almost all workers in the field will now accept 1500 B.C. as a minimum date for its appearance in Mainland Southeast Asia, as agreed at the London Colloquy in 1973 (Smith and Watson 1979); we can discard with confidence the estimates of 300-800 B.C. for the beginning of the Southeast Asian â&#x20AC;&#x153;Metal Age: that were current prior to the late 1960s. I know of only two recent authors who seem to imply a post-1000 B.C. date for the appearance of bronze in the region. (Loofs-Wissowa 1983a.b; Nitta 1985). The crucial questions now are not whether there was early bronze, but just when and how it appeared. Debate on the chronology has been intense at times, as has already been discussed in Chapter 5. Dates for the appearance of the technology are still very sketchy in Viet Nam and almost nonexistent for Laos and Cambodia, but it seems fairly safe to say on present evidence that bronze technology developed some time in the later half of the third millennium or early second millennium B.C. and was apparently widespread throughout the region prior to 1500 B.C. One important but sadly inconclusive piece of evidence from Non Nok Tha gives a tantalizing suggestion that bronze may have appeared even earlier in the Non Nok Tha sequence. In 1974, while slowly removing the contents of pot NP-1681 for analysis, the Solheims found two nodules of bronze at the very bottom of the pot, under some 15 cm of the soil it contained; the pot is a ringfooted type 2B definitely in the 1968 B. 4, provenienced to late EP 1 (it had been dug into earlier EP 1 B. 8); while there is some chance that B. 4 could be early EP 2 in origin, there is little chance it could be any later (it in turn has been dug into by an EP 3 pit). No difference in soil color or texture suggesting rodent or insect burrows were present in the pot contents, and the fragments (3-4 mm in diameter and together weighing about 5 g) appeared identical to the many others recovered at the site. There is of course no chance that the bronze was introduced from a bronze-period layer in the course of gravedigging, as EP 1 represents the earliest use of the site and is otherwise ametallic, as is EP 2. The most economical argument would seen to be that the nodules were placed in the pot at the time of the burial. If the material were to prove to be bronze, the implications are very significant, as it is difficult to see on present evidence how EP 1 could date any later than the late third millennium B.C. at the latest. Unfortunately, we are unable to present any substantive data on this early metal. Solheim recognized the importance of the find immediately and made plans to have the specimens and a sherd from the pot that contained them hand-carried to the U.S. Mainland for analysis and thermoluminescence dating. The samples were duly delivered to the laboratory by ChEt Gorman in the autumn on 1974, but they went astray thereafter; all attempts to locate them to date have been unsuccessful. It is tempting to speculate about the first tentative movements of the new and still very rare commodity from as yet unknown source along trade networks that were already bringing marine shell and fine-grained stone into the Phu Wiang region at a date possibly in the early third millennium B.C. But sadly, speculation is all it is; hard evidence is lacking. Hopefully, further research in the


region will supply us with a definitive answer to the dating of early bronze in the region within the next decade or two. Unfortunately, the picture is still unclear, although a beginning or metallurgy in Northeast Thailand in the period 1500-2000 B.C. rather than earlier now seems more likely, based on the revision of the dating of Non Pa Wai (Pigott and Natapintu 1996-1997; see Chapter 5). But equally or more important than dates is the question of how the technology appeared; was it a totally indigenous development, without outside influence, or was it somehow introduced by a form of diffusion form the Huanghe region of Northern China or-—farther afield-—the Indus or Near East? It is simply impossible to answer these questions at present, and most of one’s beliefs (for such may be a more honest term than opinions) are dependent of preconceptions of human capacity for innovation. At issue is the admittedly extreme difficulty in establishing the links between pretty rocks, very high heat, and a hitherto undreamed of but very malleable and useful metal. As the noted metallurgist Cyril Smith wrote to me many years ago, metallurgists tend to be diffusionists and anthropologists tend to be independent inventionists. Yet, Bronson, an anthropological archaeologist well-versed in metallurgy, has said he feels that Southeast Asian bronze must have an ultimate origin outside the region, if only because the great gap in time that separates the earliest Middle Eastern experiments with metals to their appearance several millennia later in Southeast Asia “makes it almost certain that diffusion of some sort occurred-—perhaps only of general concepts rather than of artifacts or specific technologies” (Bronson 1992:107). Wertime’s 1973 arguments for the necessity of a “polymetallic” phase of experimentation preceding intensive, welldeveloped metallurgy are also cogent ones against a completely indigenous origin in Southeast Asia, where no full evidence of an experimental phase has yet been recovered: Metallurgy appears as a very uniform technology over a wide area in a relatively short span of time, although there are certainly copper rather than bronze tools present at Non Pa Wai in the mid-second millennium B.C. (Pigott and Natapintu 1994). Others have implied that pyrotechnology was not well enough developed to achieve temperatures high enough for smelting and melting (Higham 1984a:248). But, this argument has been weakened by experiments suggesting that temperatures of 1,000ºC were probably attained in normal firing of pottery (Meacham and Solheim 1979; Vincent in Higham and Kijngam 1984:674). While the case for totally independent invention is thus a weak one, we must then ask the question, “If introduced, from where?” and this presents almost equal difficulties. I have discussed this question elsewhere (Bayard 1980b [written 1971], 1986-1987), and to my mind a clear, documented source outside the region at the proper date and with a related technology is still lacking. Higham (194a:248) had implied a possible source for Southeast Asian bronze in recent finds in Gansu purported to date to the earlier half of the third millennium B.C. I have questioned the likelihood of any connection (1986-1987), and after firsthand examination oF the Gansu materials, Higham has agreed that the two traditions are quite distinct (1986-1987:144). I-—like Higham-—“am not alone in being unable as yet to identify the origins of bronze-working,” but would certainly agree that “it represents a particularly Southeast Asian technological tradition quite distinct from that centered to the north in China” (Higham 1989:358). It does not appear to have arrived from the west; western Thailand apparently acquired bronze only shortly before it gained iron, in the mid-


first millennium B.C. (Glover 1991). The areas to the north ad east of the Khorat Plateau appear more likely at present, but obviously more research is required in Laos and the Vietnamese uplands, not to mention southern China, before a definite answer can be determined. What is needed is intensive clarification of the relationships between the similar regional technologies of northern, central, and southern Viet Nam, Cambodia, Laos, Northeast Thailand, and the Chao Phraya Valey; further work on the location of the ore sources utilized by these technologies is also crucial. Even more crucial is the need for a comprehensive review of all of the evidence of prehistoric metallurgy available from Fujian south and west through Guangdong and Guangxi to Yunnan, in order to establish links presently obscured by modern political divisions. Higham has now completed such a review, and concludes, “This study has not given specific consideration to the origins of bronze working and no firm conclusions are offered” (1996:311). He does, however, present two hypotheses: diffusion, ultimately from the Zhongyuan, or a completely independent invention in Southeast Asia at a date earlier than the late third millennium B.C. date for the appearance of metal in the Zhongyuan-—the central plain of the Huanghe River. Obviously to prove the second we would require dates earlier than this in Southeast Asia, and as Higham states these are not forthcoming (although the first AMS dates on Ban Chiang pottery approach it (see Chapter 5). Higham prefers the first hypothesis, but emphasises the likelihood of stimulus rather than physical diffusion of the technology (1996:312), and this I find quite plausible given the dates available at present. As he says: “There must, therefore, always be the possibility that the Southeast Asian tradition of copper-based metallurgy was initially stimulated by the spread of ides and goods down the very rivers which introduced the rice farming groups a millennium or so earlier” (1996:338). If so, “it remains fascinating that the further dissemination of this technique should have occurred so rapidly across Southeast Asia” (996:312). But it is also clear that we are dealing with two technologically distinct traditions, and it is worthwhile quoting at length Higham’s summary of his position: This possibility, which necessarily involves skepticism over the validity of the dates which have led to claims for bronze in Southeast Asia before the mid second millennium BC, should not underestimate the individuality of the local bronze tradition. The Chinese and Southeast Asian bronzes cannot be confused. The former involved a huge scale of specialist production, the use of piece moulds and the production of large sumptuary vessels of remarkable virtuosity. The latter at the same juncture could hardly have been more different. Examination of the copper mines and associated ore processing and smelting sites, as well as the villages where ingots were cast and the dead buried with metal artifacts, reveals a distinctive local tradition. But this distinctiveness does not exclude its origins in the recognition of the nature of copper and tin ore through the spread of knowledge, and I have indicated how contact between the Shang Dynasty and later Neolithic communities in Lingnan [Guangdong and Guangxi] could have facilitated this. I am not, however, prepared to reject an alternative hypothesis, that there were quire independent, local origins to this southern bronze tradition, because it is still possible that acceptable assemblages of dates before 1800 to 2000 BC for


bronze casting will one day be obtained. My preference for the former alternative is based solely on the information which is currently available. (1996:338) This is clearly a most plausible argument; at present my position is similar; that the metallurgical technology developed in Mainland Southeast Asia in the centuries preceding 2000 B.C. was largely an indigenous one, in the sense I used the term in a paper presented in 1971: ..a technology indigenous and independent in all but the very germ of its origin. We will probably never know whether the first very basic steps in metallurgy were independently discovered in the area or introduced from outside...; the point I wish to emphasize is that the technology was indigenously developed if not discovered. I would agree wholly with [Cyril] Smith (personal communication) that the presence of the inventive aborigine on the one hand or the knowledgeable traveler on the other is by and large a question of faith. (Bayard 1980b:198-199). Some fifteen years after this was written, White in a review of Southeast Asian metallurgy also proposed that “we can talk about an indigenous, ‘Southeast Asian metallurgical province’. ...Although there is site-to-site and subregion to subregion variation, in terms of typology and technology the configuration of the prehistoric metals in Southeast Asia both shows an internal coherence and, taken as a whole, considerable distinction from other major areas of early metallurgy” (1988:179). As she correctly points out, there is no need to postulate a primary origin elsewhere and diffusion of the complex as a whole into Mainland Southeast Asia simply because the region lacks signs of centralization of formalized militarism. Despite Higham’s 1984 statement that metallurgy must somehow stimulate “concomitant changes in social complexity noted under similar conditions in the Aegean, Near East or Shang China” (1984a:238; repeated but qualified in Higham 1984b:4, and subsequently rejected in Higham 1986-1987:144), there is no reason to suppose that at lest 1,000 or 1,500 years of village-level bronze casting should in some way have led directly to increased social complexity. Higham’s recent work argues merely that the time lag between the beginnings of bronze metallurgy and the development of marked social hierarchies is not abnormally longer in Southeast Asia than in the Zhongyuan, Mesopotamia, or Mediterranean (1996:316-319); marked social hierarchies and centralized societies were apparently a product of the Iron Age (or GP C, as I term it in this work). As Stech and Maddin put it, commenting on Higham’s earlier view, “technological systems, such as metallurgy, should be interpreted as integral parts of the cultural content as which they occur rather than as entities possessing their own dynamic” (1988:173). Metallurgy is surely better considered one of the servants of society than one of its masters.


IRON TECHNOLOGY INTRODUCTION The Late Period levels at Non Nok Tha contained a small amount of iron tool fragments and badly rusted unidentifiable lumps, but in nothing approaching the quantity of bronzecasting spillage from some of these levels (LP 2 in particular). The date of the onset of the Late Period at the site, after the hiatus following its abandonment at the end of the Middle Period, is still far from certain. We have assumed a date late in the first millennium A.D., but the position of LP 1 remains problematical; only one porcelain sherd (possibly intrusive) was recovered, and the earthenware seems to relate fairly closely to the Middle Period (see Chapter 5 for full discussion). The post-LP 1 layers, however, are clearly dated from the first centuries of the present millennium by the porcelain and stoneware sherds they contain: as discussed in the above chapters, we assume a date about 1829-1830 for the abandonment of the site prior to settlement by the present population. The iron subjects recovered are presented by level in Table 7-4. The seven unidentifiable lumps of rusty iron weighed between 1 and 6 g each; one piece of iron slag was also recovered from LP 5. The fifteen identifiable tool fragments are described below. TABLE 7-­4 IRON  OBJECTS  FROM  LATE  PERIOD  LEVELS  AT   NON  NOK  THA  1966  EXCAVATION: FISHHOOK































LP 2 1966: NP-268 C3 layer 9/10 interface (Fig. 7-14a). A moderately corroded nail shaft 3.5 cm long; point missing. Cross-section 4 mm square; weight 1.9 g.


Fig. 7-14

NP-1535 C3 layer 9. Nail fragment weighing 2 g; sent to University of Pennsylvania for analysis in 1970a, but no analysis received to date. NP-1588a/c E6 layer 9 (Fig. 7-14b). three heavily corroded nail or pin fragments, all apparently circular in section. The largest is 3.1 cm long by 6 mm thick and weighs 2.8 g. The second is 2.5 cm long by 3 mm thick and weighs 0.8 g. The third is 1.6 cm long and conical, tapering from 6 m to 3 mm in thickness: weight 0.8 g. LP 2 1968: NNT-251 3F layer 5/6 interface (Fig. 7-14c). Moderately corroded knife blade fragment measuring 6.0 cm long by 1.6 to 1.2 cm wide and about 2 mm thick; weight 6.0 g. No sign of a tang is present. NNT-775 3F lower layer 5 (Fig. 7-14d). An almost complete tanged knife blade or spear point in two pieces: total length 10.4 cm. Weight 11.7 g. Tang is 2.4 cm long and 8 mm


thick; shoulders of blade are about 3 mm wide, with a blade width of 4.5 mm at base tapering to a point at end. Blade width at end of larger piece 9 mm; thickness 4.5 mm at tang end of blade, tapering to 3.0 mm. LP 3 1966: NP-810 C3 layer 7/8 interface (Fig. 7-14e). This heavily corroded object could also be a staple, but relative thickness of presumed shank (6 mm) compared to â&#x20AC;&#x153;point (3 mm wide by 2 mm thick) makes a hook the more likely choice. Shank is 2.1 cm long, point side 1.6 cm, width across 1.7 cm. Thickness at bend 5 mm, weight 1.7 g. LP 3 1968: NNT-150 3F upper layer 5 (Fig. 7-14f). A knife fragment with portions of both blade and tang; total length 6.8 cm, weight 6.6 g. About 1 cm of tang remains, 7 mm wide where it joins blade; blade portion is 5.8 cm long, tapering from 1.3 cm Wide at tang to 0.7 cm at distal end of fragment. Thickness varies between 2.5 and 3 mm. NNT-539 OF upper layer 5 (Fig. 7-15a). An almost complete but badly corroded pair of scissors or shears similar to the one piece type used in the region today. About 5 or 6 mm appear to be missing from the tip of one otherwise complete blade; the other blade is only about one-sixth present. Total length of object would have been about 17.5 cm; length of blade 8.9 cm, handle 8.0 cm. Width at junction of handle 1.9 cm, width of blade 1.5 cm. Handles are 5 mm wide and 9 mm thick; blade tapers toward edge from 4 mm at back to ca. 1 mm edge. Total weight 59.5 g. NNT-537 3E/4E upper layer 5 (Fig. 7-15b). A moderately corroded tanged awl or punch (?) with tang bent at 30° angle; total length 5.5 cm, weight 3.0 g. Presumed tang is 2.6 cm long, tapering from 3 to 2.5 mm in thickness. Length of pointed blade 2.9 cm, thickness at midpoint 5 mm. NNT-128 4D upper layer 6 (Fig. 7-15c). An only slightly corroded eyelet measuring 3.8 cm long formed by forging a rod 3 mm in diameter and bending the thinned rod back against itself; the seam is clearly visible along 1.6 cm of the shaft. Eyelet is 8 mm wide OD and 7 by 3 mm wide ID; weight 2.2 g. NNT-533 2E upper layer 5 pit (Fig. 7-15d). Both the provenience and the function of this object are puzzling. It was recovered from a disturbed area between MP 4? B. 124 and MP 6 B. 70 at an absolute depth of 105 cm. Examination of the soil and sherds in the vicinity at the object showed that it lay in a Late Period pit or large posthole, but from LP 2 or more likely LP 3. The function of the object is enigmatic; it is badly corroded and in several fragments, but it appears to be a segment of about 40% of a cylinder formed of coiled (?) iron rods ranging between 3 and 5 mm in thickness; some of these rods are definitely hollow (ID ca. 1.5 mm), while others are solid. The object as recovered is about 9 or 9.5 cm long, with a height of about 3.5 cm and thickness ranging from 6 to 9 mm; it weighs 216 g. Original diameter of the cylinder would have been about 7 cm. At present no archaeological or ethnographic parallels are apparent; intuitive guesses could range from a spring to a still!


Fig. 7-15

LP 4 1968: NNT-124 3E layer 4 (Fig. 7-15e). A knife similar to one of those recovered at Ban Na Di (Higham and Kijngam 1984:146, Fig. 3-27c). Much of the tang is missing, but remains; blade portion is 3.4 cm long. Overall length 5.3 cm, weight 4.9 g. Width at midpoint of blade 1.6 cm, thickness 2.5 mm. NNT-535 1F layer 4 (Fig. 7-15f). An almost complete tanged knife blade measuring 9.0 cm in length and weighing 8.9 g. The tang is 3.0 cm long and blade 6.3 cm, with an estimated 5 mm of the tip missing; blade width tapers from 1.0 cm to 0.4 cm at broken tip. The back of the blade is 3 mm thick, tapering to 1 mm at the edge.


With the exception of the enigmatic “cylinder,” the modest assemblage of iron artifacts from the site’s Late Period presents little in the way of surprises; similar knives, scissors, hooks, and nails could probably be found in many rural houses in the region today. The almost complete lack of slag-—in contrast to the situation at Non Chai (Bayard et al. 1982-1983) and layers 5-3 at Ban Na Di (Higham and Kijngam 1984:121)-—suggests that iron smelting was never practiced in the immediate vicinity of the excavated area of Non Nok Tha and was probably never carried out to any significant extent on the site as a whole, although evidence for iron smelting is certainly abundant at other sites in the Phu Wiang region (see Chapter 2). STONE, BONE, SHELL, ETC. • Solheim STONE Artifacts of stone or worked stone recovered include: (1) adzes and axes, (2) knives, (3) grinding stones, (4) sharpening stones, (5) rubbing stones, (6) scraper, (7) tektites, (8) hammerstones, (9) pestles, (10) metates, (11) red pigment, (12) bracelets, (13) disks, (14) rings, (15) beads, (16) pendants, (17) miscellaneous, (18) flakes, and (19) polished fragments. Adzes and Axes It has been difficult to distinguish between adzes and axes in this series of artifacts, possibly because as a group they are so small. The majority of these tools are similar in shape, both in plane view and cross section. Three different classifiers, working independently, have had sufficient disagreement that I am going to treat the two together. The one area where there is major agreement is that the great majority of these tools were used as adzes. Adzes, of course, are hafted with the line of their bit, or cutting edge, perpendicular to the axis of their haft; in other words, they are hafted like a hoe. An axe is hafted with the line of its bit parallel with the haft. An adze is used in the same way as a hoe: striking towards the body with the sharp edge of the adze thinning or shaving off wood from the surface. It can be used for cutting a tree or a branch from a tree by swinging it sideways. In both cases, the force of the blow and the resulting action of cutting are perpendicular to the edge of the adze. On the other hand, the axe is used like a knife, giving a splitting or slicing action. Instead of the force being perpendicular to the sharp edge, it is at an acute angle to the edge. This should mean that minute scratches resulting from use would be perpendicular to the sharp edge of an adze and at varying acute angles to the edge of an axe. Takeshi Ueki, in a 1969 anthropology research seminar with me, made a study of the adzes and axes recovered in the 1968 excavation. All the illustrations of the 1968 axes and adzes presented here are taken from his paper (Figs. 7-16 to 7-20). In his analysis he attempted to differentiate between axes and adzes on the basis of the scratches resulting from use, but with the microscope he used he was unable to make these out. Thus he decided whether a specific tool was an adze or an axe by inspection, basing this on lengthwise cross sectional symmetry for an axe and, for an adze, asymmetrical tool


beveling of the two sides in a lengthwise cross section. He stated that this was not satisfactory in many cases and that with these small tools this method could not be depended upon. If the beveling of this bit on both sides is symmetrical, the artifact could be used either as an axe or an adze. Ideally an adze would be unifacially beveled and thus easy to recognize as an adze. Most of the adzes from Non Nok Tha were beveled on both sides, with one bevel at a slightly greater angle than the other. The adzes and axes excavated by Bayard were all returned to Thailand before I was able to examine them, but are illustrated in Figs 7-16 to 7-20. I have been able to examine the relatively few axes and adzes recovered in 1966 at a magnification of forty times; in most cases, I was able to make out the scratches resulting from use. In one or two cases where I had considered the tool either an axe or an adze by inspection, it turned out that the direction of the use scratches did not agree with my identification from inspection. The use striations evident on NP-1510 (P1. 11a) from the unchipped portion of the bit back up the body about 5 mm are at an angle to the edge of the bit, indicating that this was used as an axe. As this axe, from Square D4-D5, was without provenience for level of layer, it is not included in the following table or in the totals. The stone used for the adzes and axes was never positively identified by mineralogists, as they indicated that a thin section of each artifact would be needed for identification. We did not feel that this datum was worth the destruction of each tool. They were able to state that all polished adzes and axes, as well as the gray to black stone bracelets and beads, were made from phylites, diorites, or slates. A brief description of the adzes and axes follows in Table 7-5, with an explanation of the abbreviations used at the end of the table. PROVENIENCE AND DESCRIPTION OF ADZES AND AXES Measurements on NNT adzes and axes are not exact in relationship to NP measurements, as without definition by Ueki of points from which he made his measurements, they probably differed somewhat from those that I used. For width of bit, and butt, I used the points of angle from bit, butt, and side; for thickness, Ueki used the maximum thickness on the line he used for his cross section, while I used the maximum thickness, regardless, of its location. The differences between our measurements usually were less than 1 mm. The common cross section of these adzes and axes perpendicular to the long axis of the tool is elliptical, with thin, flat sides. As can be seen in the vertical cross sections of most of the tools, the flat side is quite irregular both in width and length. It would appear that this flattening of the sides was to do away with sharp edges that would be liable to cut through the hafting material. I have noticed nothing on these tools to suggest how they were hafted. In all cases, for both axes and adzes, the bit is wider than the butt. The ratios of butt to bit vary from 0.45 to 0.88. All recovered axes and adzes, except the much larger NNT-353 (Fig. 7-20a) have the same general trapezoidal appearance in plain view



Fig. 7-16


Fig. 7-17


MP 2: NP-1080. The bottom surface of the adze (the surface without bevel) is largely polished with some of the original flat surface remaining unpolished, while the top surface (the face with the bevel) is polished primarily near the cutting edge of the bit over an irregular surface. Four small flake scars are present on the bottom surface at one side of the bit and two scars on the top surface close to the center of the bit. On opposite faces of the bit from the scars, the polishing extends to the edge of the scar, which forms the edge of the bit in that area. The edge is sharp and even. Polishing striations on both surfaces are roughly perpendicular to the remaining edge (Pl. 11b) and extend over the total polished surface, suggesting that this adze may have been broken not in use but either in first manufacture or when it was being resharpened, after considerable resharpening had been done (Fig. 7-18a). NP-544. At 40x magnification, small light green crystals can be seen in the stone and numerous bright red splotches on the surface, neither of these being visible to the naked eye. The red splotches probably result from finely ground red ochre worked in to roughness in the surface. The red material probably was worked into the adze surface unevenness when grinding and polishing on grinding stones, most of which also have considerable red material in irregularities in their surfaces (see description of grinding stones). The edge of the bit is almost perfect, with very slight wear scratches at an acute angle to the edge. The few very small nicks in the bit edge have smooth margins and look as though there had been slight melting from high heat at these margins. The edge of the bit appears slightly curved when looking at it edge-on, but when a straightedge is placed next to it, one finds that it is obviously straight. The curvature is an optical illusion resulting from the differential curvature of the two faces to either side. The sharp edge of the bit runs in a slight diagonal between the two sides of the axe. This is true for all the axe bits (Pl. 12b, Fig. 7-18b). NP-675. This is made from a banded light green stone. There are extremely few wear scratches on the smooth bit. In one area these are perpendicular to the edge and they extend back from the edge only a very small distance (Pl. 12c, Fig. 7-18c).


Fig. 7-18


MP 3: NP-541b. A few small flake scars on the edge of the bit have rounded edges, suggesting continued use after the flakes had been removed. Wear scratches on one face of the bit are primarily perpendicular to the edge of the bit, while on the other face the scratches are fewer and primarily at an acute angle to the edge. This suggests the possibility of hafting both as an axe and an adze. Fig. 7-19


NP-537. This was probably broken during manufacture as there are a few small but deep flake scars near the edge of the bit but not extending from the edge, and they appear to have been in the process of being polished over. Scratchesâ&#x20AC;&#x201D;-probably polishing marksâ&#x20AC;&#x201D;near the bit edge are at a very acute angle, almost parallel for some, and at one side of the bit the edge is still quite dull, as though the grinding and polishing on this side had not been completed (Fig. 7-20b). Fig. 7-20


Descriptive data on forty-six adzes and axes are presented in Table 7-5. Of these, only twenty-eight are in sufficiently complete condition that they could be used for analysis. The artifact from Late Period 5 (NNT-339) obviously came to be found in such a late context because it was curated as an heirloom or amulet or had been removed from its original context in the site through disturbance; so it cannot be used. The movement through disturbance could, of course, apply to any of those not recovered from a burial (thirteen for analysis were associated with burials). NNT-353, from Middle Period 5, is so different in size (Fig. 7-20a) from all the other complete adzes or axes that it is not included in this analysis. Thus only twenty-six are used for the analysis. With twenty-six artifacts recovered from eight layers covering a time span possibly as long as three thousand years, a sophisticated statistical analysis is not called for. Table 7-6 presents average measurements and ranges for levels for the Early and Middle Periods for two non overlapping ranges of weights and for the kind of stone. Visual inspection of these average measurements suggests no consistent changes in any measurements over time. Variation between levels within a period is greater than the variation between the Early and the Middle Period. There is a similar bimodal range in the weights for both periods. The average weight of the smaller size is 6.9 g and the larger 15.3 g. The greatest weight of the smaller size is 10 g, and the smallest weight of the larger size s 12 g. This suggests a different function for each of the two weight sets. I suspect, however, that both sizes were used for working with wood, the modest difference in size possibly meaning the working of somewhat larger wooden materials with the larger adzes. This could also be a factor of size at discard. The larger group averages slightly more rectangular in plain view than the smaller. Those axes and adzes made from metamorphic stone average heavier than those from chert, the two most common kinds of stone used. The fifteen metamorphic axes and adzes average 14.5 g while the ten of chert average 9.3 g. The one igneous axe is larger, at 17 g, but its measurements fall within the range of the metamorphic measurements. There is considerable non-consistent variation between metamorphic and chert axe and adze measurements. It must be remembered that most or all of these axes and adzes were either discarded when they were broken or when they became too small to be functional. Those placed in a burial may well have been used to the point where they were ready to be discarded. Thus it is reasonable to think that the measurements of these artifacts were about at their minimum for functional use. Their size when first used may have been considerably larger (see Roy 1981:209 for illustration of those observation). Possibly the one much larger adze (NNT-353, Fig. 7-21a) is the one example of the original size of the adzes before much reworking and resharpening and final discard, when an adze becomes too small for further use. The axes and adzes are more commonly associated with burials during the Early Period than the Middle Period. Fifteen of the seventeen whole and partial axes and adzes from the Early Period levels came from burials, while only seven of the twenty-nine from the Middle Period were from burials. Of the Middle Period ones, four came from level 4, two from level 1, and one from level 5.


I do not intend to make either an intensive or extensive comparison of the nonpottery artifacts with artifacts outside of Non Nok Tha. I do feel it of value, however, to make a few comparisons with artifacts from other sites in northeastern Thailand, Thailand as a whole, and from nearby countries.


Except for Hoabhinian sites in Southeast Asia, not much attention has been paid to artifacts other than pottery. Six adzes are illustrated from a surface survey of two sites in the proposed Pa Mong Reservoir (The dam across the main stream of the Mekong that would have backed up this huge reservoir west of Vientianne was never built). One of these is a shouldered adze and four are rather similar to the Non Nok Tha adzes, but not exactly the same (Bayard 1980a:79, Fig. 12 b-g). The sixth adze is about twice the size of the four, but if worn down and resharpened a number of times cold easily come out about the same. Five adze/axes are illustrated from the excavation at Ban Chiang (White 1982a: 83, numbers 130-132, 134, and 136). Number 134 is a small shouldered adze, while the rest are similar to those from Non Nok Tha. No stone adzes were recovered from Ban Na Di, near Ban Chiang (Higham and Kijngam 1984:187). It is curious that at Non Nok Tha there were many stone adze/axes continuing in use long after bronze artifacts started being made at the site, while at Ban Na Di, which had bronze from the beginning of the site and which overlapped close to two thousand years on its early end with Non Nok Tha, there were no stone adzes recovered. We found six stone adzes during the first year’s field survey. Three of these were shouldered, one from near Kalasin, Sahatsakhan, found in an Ayutthaya jar by the local villagers (Solheim and Gorman 1966:127, Pl. VIIe) and two were found about 1 km south of Sahatsakhan when a small drainage canal was being dug (Ibid.:163, Pl. XVIIIf-g). From the northern extension of the Ubonrat Reservoir, three adzes were shown to us that had been collected by local people. Two of these were similar to those from Non Nok Tha (Ibid.:174, Pl. XXIVk) and one was rectangular in cross section with possible low shouldering. This makes a total of seventeen adzes, of which five were shouldered and one more possible, ten were similar to those from Non Nok Tha, and one was rectangular. Seventeen adzes are not enough to consider that they are representative, but at least it is probably safe to say that shouldered adzes are relatively common in the Northeast. No shouldered adzes having been found at Non Nok Tha has meaning, but what that meaning is we can not say. We really have no idea what other artifacts are associated with the shouldered adzes. From Khok Charoen in Central Thailand, “A number of very small ground stone adzes were found, but not always clearly associated with a burial. They were mostly trapezoid in shape, with a unilateral beveled cutting edge and a rectangular to lenticular cross section. In several cases adzes made of soft limestone but having the same form as real stone adzes suggest that substitutes were given to the departed instead of real stone adzes which were presumably regarded as being too valuable to be wasted in this way” (Loofs and Watson 1970:75). While these are not illustrated, this could explain the Non Nok Tha adzes. The site of Khok Phanom Di, about 65 km southeast of Bangkok, has been dated from about 2000 to 1400 B.C. (Higham and Bannanurag 1990:19). No bronze was found at the site. Many stone adzes are recorded from the site; none are described, and only one is illustrated-—associated with a burial-—in the first volume of the final report (Ibid.:358). This one adze is like those at Non Nok Tha. Higham (1989:75, Fig. 2.25a) has illustrated one other stone adze from that site, and it is not like any of those from Non Nok Tha.


From Ban Kao, in western-central Thailand, thirty-eight axe/adzes were recovered from burials (Sørensen and Hatting 1967:85). On the basis of symmetrical form, two of these were classified axes (86-87). Many of the remaining thirty-six adzes are similar to those from Non Nok Tha, but the most similar are those Sørensen classifies as trapezoidal in plain view (Ibid.: Fig. 2, Pl. 18-12, 23A 5, 33A-8, etc.; 1211-f). His classification difference between trapezoidal and rectangular is artificial, specifically tailored to classify those adzes from Ban Kao. Most of the larger adzes classified as either trapezoidal or rectangular, when worn down to considerable smaller size, could be very similar to those from Non Nok Tha. In contrast to Ban Kao of the early Neolithic phase with well-manufactured adzes, Sorensen states that “most adzes are of a rather limited size; they are generally poor, often rehewn and resharpened, but good specimens occur as well” (Ibid.:106) From the site as a whole, including surface collection, about 1,200 whole and fragments of adzes were recovered. Only four of these were shouldered adzes. “A preliminary counting hints that adzes having a lenticular section and rounded triangular shape are about as common as those of a more quadrangular/rectangular section and varying shapes” (Sørensen 1964:81). The first form description would fit the usual adzes from Non Nok Tha. The initial survey of Kanchanaburi Province of the Thai Danish Prehistoric Expedition, in 1960-1961, discovered or were shown a number of stone adzes. From Ongbah Cave on the Menam Kwae Yai, two polished adzes were found (Sørensen 1988:16, Figs. 31-32). One of these was almost rectangular in plain view, the other triangular, and both were longer than the Non Nok Tha adzes. If worn down through use and resharpening, these could well have become like those from Non Nok Tha. In Chande hamlet, on the Menam Kwae Noi, a polished shouldered adze was purchased (not illustrated). From the second Chande Cave, nearby, two polished axes and four adzes were reported and illustrated (Heekeren 1988a:64-65, Pl. 27). Except for the one of these illustrated at the lower right (B12, Pl. 27), which was rectangular in plane view and in cross section, all of these-— when worn and resharpened-—could be like the Non Nok Tha adze/axes. One artifact that was called a “relatively broad chisel” (B27.65 and Pl. 27 lower left) looks the most like the Non Nok Tha adzes. The one axe recovered from the Sawmill Site, on the eastern bank of the Menam Kwae Noi, was a large rectangular axe (Heekeren 1988:86, Pl. 32b) unlike any at Non Nok Tha. The suggestion that I have made that many of the axe/adzes from other locations in Thailand-—when worn down and resharpened-—would be like those from Non Nok Tha suggests two things about those Non Nok artifacts: that before much reworking they were considerably larger, and that it was the practice at Non Nok Tha to continue using their adze/axes until they were too small to be used further rather than making or trading for many new, full-sized ones. I am not acquainted with any excavated sites from Burma, Laos, or Cambodia of the period covered by the Late Neolithic and Bronze Age of Thailand. All reports on artifacts from these countries for this time span are for surface finds or from private collections without reliable provenience. The one relatively recent study of stone adzes from Southeast Asia is that by Roger Duff (1970). He presented a typology of adzes based on his work with adzes in Oceania. At the time of his writing, virtually no adzes from


excavated sites in Southeast Asia had been published. The adze/axes from Non Nok Tha would be considered Late Neolithic in time, but their closest resemblance to type illustrated by Duff is to what he calls Melanesoid types of eastern Indonesia (p.40) and Bacsonian adzes (p.80) of Viet Nam that are much earlier than those from Non Nok Tha. These latter he classifies as Type 2, Variety G, and where they are found in other countries of Southeast Asia, they are the most like those of Non Nok Tha. For Island Southeast Asia including Taiwan and for peninsular Malaysia, except for the Melanesian types there are no similar adze/axes. For Thailand, the Type 2, Variety D and G (68) are generally similar. Duff was not able to examine any adzes from Burma or Laos. I saw many stone adzes from Burma when I visited the Pitt Rivers Museum at Oxford in 1969. To the best of my recollection they were generally similar to the Non Nok Tha adzes, though somewhat larger. When I visited Heine-Geldern in Vienna in 1968, he took me to the third subbasement at the Museum fĂźr VĂślkerkunde and showed me a number of adzes from Laos. As I remember them, they were rather similar. He told me that it was seeing these adzes from Laos that first gave him the idea of the rectangular adze movement from South China into Malaya and from there into Indonesia and out to the Pacific. Type 2 of Duff he calls rectangular, referring to cross section. I supposed that the Non Nok Tha adze/ axes could be considered somewhat rectangular in cross section, but to me they are more often oval to lenticular. They do stand out, however, in contrast to the great majority of those adzes illustrated in publication, as being badly worn or generally roughly made. This could be, at least in part, the result of long use resulting in secondary flaking and then resharpening. The other possibility I see is that these are close to the original size, probably for fine woodworking, and that the people of Non Nok Tha did not regard as important the complete polishing in consistent, regular forms of their adze/axes. Knives The three stone artifacts classified as knives were so classified because they have roughly parallel sides and are longer than they are broad. Only one of them might have been used as a knife, such as for cutting (NP-158). All three were recovered in the 1966 excavation. NP-1470. From Sq. E3, L. 19 (MP 3), weight 4.8 g, max. length 28 mm, max. width 21 mm, max. thickness 7 mm; a very fine grained, sedimentary stone, light tan with diagonal lines of light gray. It was made on a flake from a possibly prepared core with no retouching on the concave inner surface of the flake. Three or more flakes had been removed from the outer surface resulting in a central, lengthwise ridge. Both edges had been retouched, removing small flakes to give a steep angle on both sides. Very little edge damage is evident. This would have been used as a scraper or to shave small-diameter wooden objects (Fig. 7-21a). NP-158. From Sq. C5, L. 9X (LP 2), weight 6 g, max. length 51 mm, max. width 15 mm, max. thickness 9 mm; medium-grained, sedimentary stone, chocolate brown. Probably natural fragment from the shattering of a large rock, with no retouching. One side is naturally flat while the other side is sharp with a rather ragged edge due to removal of a few small flakes probably through some use. One end had a sharp edge and could have been used resulting in the removal of two small flakes, giving a rough edge. The other end comes to a dull point, the somewhat rough edges leading to the point suggesting that the


Fig. 7-21


point could have been used to make perforations. This stone would not have withstood much pressure before breaking. It could have been used briefly as a knife or a drill (Fig. 7-21b). NP-1599. From Sq. C3, L 3 (LP 5), weight 8 g, max. length 33 mm, max. width 24 mm, max. thickness 10 mm; a tan, fine grained sedimentary stone-—under 40 x magnification, made up totally of evenly graded crystals. Probably a natural fragment from a large rock, it may have been retouched to give a blunt, steep-angled edge on one side for possible use as a scraper. It is apparent that a knife was not a common artifact at Non Nok Tha, at least made from stone. Heekeren reports a “crude stone knife??” from the Sawmill Site in Kanchanaburi (1988b:84) but does not illustrate it. He mentions “an odd crudely flaked stone knife” (1988a:63) from a burial in Chande Cave B, with no illustration. The two artifacts called knives here could also be called blades. Blades are not common in Southeast Asia and when found, though not well dated in many cases, appear to be much earlier that these from Non Nok Tha. Grinding Stones Very few of the grinding stones are whole. Many of the fragments are so thick that it is hard to understand how they could have been broken, unless they were deliberately smashed. An unbroken grinding stone had one or two working surfaces. Surfaces may have a very fine texture, a coarse texture, or anything in between. No attempt is made to define numerically a fine or coarse surface: when the texture of the surface is mentioned it should be considered as a relative description. Some surfaces have varying remnants of a fine red powder, suggesting that the red ochre stones commonly found in the site with one or more polished surfaces were ground on such a surface to make a red powder. The amount of powder imbedded in the irregularities of the grinding surface varies greatly, from small quantities found on a very small portion of the surface to large quantities covering practically the whole surface. A moderate number of the grinding stones were not brought to Hawai’i and thus only a few measurements are available for these. Table 7-7 presents the raw data for the grinding stones. Very few of the stones are symmetrical, so the length is taken as the maximum length (in most cases), the width as maximum width, and thickness as maximum thickness. For the broken stones these measurements may have little relationship to the original size or shape of the stone. Directly following the table are remarks with additional information on specific artifacts where further data is available. No attempt is made to present the same information on each grinding stone; instead the remarks give an idea of the range of additional data but do not make it available for quantification. One important variable not included is the kind of stone of each artifact. While there are obvious differences in the kind of stone used, without mineralogical identification by an expert I do not feel our impression of this identification should be presented.


The requirements of a grinding stone are two parallel flat surfaces, one or both of which may have a moderate concave area worn down through the grinding or sharpening of stone tools.





EP 1: NNT-9. Grinding surface concave, up to 2.5 mm deep; back surface very rough. NNT-28. Red pigment on both stones not evenly distributed but considerably heavier in some areas than others, as is the case with all of the grinding stones with red pigment. NNT-474. Horizontally layered with 4 mm thick grinding surface layer, about one-third broken off from rest of stone. Grinding surface slightly concave. No red pigment apparent to naked eye but under 40 x magnification, red pigment is obvious in many of the small depressions in the surface. For most stones with red pigment, whether obvious to the naked eye or not, the pigment is in small, natural depressions in the grinding surface. EP 2: NNT-23. With slightly concave surface. EP 2-3: NNT-11. With uneven grinding surface. NNT-439. With the red pigment much thicker in one area that measures about 1 cm2 than over the rest of that surface; the other grinding surface with rare, very small areas with red pigment. EP/MP 1: NP-1011. With red pigment on both surfaces but on one face only along one edge, while on the other face it is unevenly distributed over the whole surface.


NP-1615. Has one surface considerably smoother than the other. The smoother surface is quite flat, while the rougher surface is slightly concave from edge to edge across its width and flat from side to side across length. NP-1425. Has both faces rather uneven and about equally smooth; only one surface has red pigment. NP-387. Presents a problem present in a number of the grinding stones; the question of whether both surfaces were used for grinding. There is little difference between the two surfaces and it is a matter of personal judgment as to whether only one or both surfaces were used. MP 1: NNT-327. With little pigment present and most of this on or close to projecting portions of the surface rather than in depressions, suggesting little use for powdering pigment and little or no use for grinding stone after powdering the pigment. NNT-395. With most of grinding surface missing; this surface layer is ca. 1 mm thick. NNT-401. Grinding surface concave diagonal to length of stone (Fig. 7-21c). NNT-327c. With one certain grinding surface slightly concave and some pigment but little of this in depressions; second surface questionable as a grinding surface, with no pigment. NNT-313c. With uneven surface and only a small area of this surface shows grinding so must have been used very little. NNT-316a. One of rare, unbroken grinding stones, this one triangular in plain view, concave on one face (Pl. 12g). NNT-318. Slightly concave in diagonal; a small portion of the second surface may have been used for grinding. NNT-312b1-2. These were cataloged as one artifact but are two unrelated grinding stones; b1 has a slight concave area on one face (Fig. 7-22). NNT-315. May have had its second face used for grinding.


Fig. 7-22


Fig. 7-23


MP 2: NP-1503. Has a very fine surface. NP-438. Questionable whether ever used for grinding, as both surfaces are quite course. Small amount of red pigment in lines on one surface. NP-1329. With a very fine surface. NP-919. Has some red pigment on both surfaces. NP-1428, NP-1513, NP-1429, NP-756. Have red pigment on one surface. NP-1067. Has some red pigment on one face in small area; slightly concave on one face and deeply concave (3-4 mm) on other face (Fig. 7-21e). NP-629. Such a small fragment that only one surface is certain, so it could have been bifacial and could have had red pigment elsewhere on surface. NP-1429. Has small amount of pigment near one edge (Fig. 7-23a). NP-1546. Has red pigment on both surfaces. MP 3: NP-133. Has only small portion of the possible grinding surface remaining. NP-1356. Bifacial with one face smoother, with a fine texture and a slight groove 10 to 12 mm wide down the center of this face; rougher side shows less wear (Fig. 7-23b). NP-401. Fragment, medium texture with some crystals (Fig. 7-23c). NP-220. Whole, rectangular stone, slightly concave groove (ca. 1 mm deep) on one side, groove ca. 15 mm wide, diagonal but not reaching either corner (Pl. 12h, Fig. 7-23d). NP-1480. Both surfaces slightly concave (Fig. 7-23e). NP-406. Has red pigment powder in wide area of more heavily used surface; this surface has several different slightly concave areas and one groove 40-45 mm wide, running across one side of the surface, ca. 110 mm long at centerline of groove, up to ca. 4 mm deep; areas of red pigment deposit show no relationship to concave areas. MP 4: NP-1138. Questionable grinding stone, surfaces rather uneven. NNT-310. About half of grinding surface quite smooth and slightly concave; next to this area is a rougher surface that is considerably more concave, ca. 2-3 mm deep.


MP 5: NNT-403. Has an even but relatively coarse surface. NNT-265. Includes three different grinding stones: a) one face is slightly diagonal to the long axis of the stone, while the other face is highly concave (ca 4 mm) in a different axis; b) with both faces slightly concave and both with red pigment. NNT-271a-b. Each has one slightly concave surface. NNT-272. May have been bifacial. NNT-274. Very small amount of red pigment on one face in a small area. NNT-264b. Has some light brown possible pigment on one face. NNT-270b2. One surface moderately concave, the other moderately convex. NNT-275. Both faces slightly concave. NNT-228a. Used face fine textured and very smooth and flat, other face slightly uneven (Fig. 7-23f). NNT-232. Has slightly concave surface; one spot <0.5 cm2 with heavy red pigment and a much smaller area close by with no pigment evident on 95+% of the surface. NNT-233a. Has two fitting pieces; one with a considerably concave surface (ca. 3 mm deep); the other with a nonconcave face with red pigment. NNT-238. Both surfaces highly concave (ca. 4 mm deep) with concave axes in different directions. NNT-246. One surface moderately concave (ca 2 mm); red pigment in small area on one face (Fig. 7-24b). NNT-247b1. Has red pigment on both surfaces. NNT-253. A questionable grinding stone. NNT-259. A questionable grinding stone with relatively coarse surface. NNT-230. Two pieces fitting together face-to-face having broken horizontally; both faces slightly concave. NNT-244b. Has a deep concave area on one side (ca. 5 mm) and a moderately concave surface on other side (ca. 2 mm deep); close to the complete artifact (Pl. 13a, Fig. 7-24c).


NNT-240a-b. Probably from one artifact, but they have no joining surfaces. NP-630. Has one very small spot of red pigment. Fig. 7-24


MP 7: NP-1512. One surface considerably coarser than the other. NP-1518. Has red pigment on one of the two surfaces. MP 8: NP-607. Used face somewhat uneven but very smooth and fine texture (Fig. 7-25b). Fig. 7-25


LP 1: NP-1373. Slight concavity on both surfaces; close to a complete stone. Three edges have been partially ground, with two of these having moderate grooves, suggesting that something narrow (one ca. 4 mm wide, second ca. 7 mm wide, third obvious one is at edge of break so cannot measure) and probably flexible was used for smoothing or polishing around the side, or for the former two, over one edge of the side (Pl. 13b, Fig. 7-25c). NP-1367. Two opposites sides have been used for grinding or polishing across the sides, like NP-1373; one side has two old scratches in grinding surface with a red deposit. Under 40x magnification, this is shiny as if melted, and the opposite surface is without scratches and with the same red material over much of its surface-â&#x20AC;&#x201D;and over the angles at both edges to the grinding surface-â&#x20AC;&#x201D;but not at any distance out on the main grinding surfaces. The side with scratches also has some of this at the angles between the side and grinding surface (Fig. 7-26a). Fig. 7-26


NP-13. Surface slightly concave. LP 2: NP-1502. Surface slightly concave. NP-1562. Fragment, possibly bifacial, both faces medium texture, one face slightly concave (Fig. 7â&#x20AC;&#x201C;26b). NP-159. Possibly both surfaces used. NP-518. Surface slightly concave. NP-691. Surface moderately concave. NP-1456. Possibly bifacial, with one surface considerably coarser than the other. NP-236. Used face fine textured with slightly concave surface (Fig. 7â&#x20AC;&#x201C;26c). NP-459. One surface slightly concave (Fig. 7-26d). NP-1517. Three out of the four edges are unbroken. NP-1457a. Grinding surface remaining is convex; possibly a rubbing stone. NP-1457b. Coarse surface. NP-135. One side slightly concave. NP-183. Used side medium textured, smooth, and slightly concave, other side slightly uneven (Fig. 7-26e). NP-1417. Reddish groove in grinding surface, reddish color from groove extending down into reddish crystals in stone; groove 3 mm wide and ca. 10 mm long. NP-1464. Surface moderately concave. NNT-189. Possibly bifacial, second face somewhat uneven but very smooth. NNT-210. Possibly bifacial. NNT-213. Both faces fine textured and slightly concave (Fig. 7-26f). NNT-215. Used face fine textured and slightly concave; other face rough (Fig. 7-26g).


NNT-216. Moderately concave, some red towards margins, away from concave surface, suggesting early use with red ochre but much use for sharpening, without further ochre grinding, wearing away the pigment on the concave working area (Fig. 7-27a). Fig. 7-27


NNT-219a. Moderately concave surface. NNT-221. Fragment without any of second surface. NNT-223. Slightly concave surface. NNT-250. Fragment without any of second surface. LP 3: NNT-145. Two pieces, not available for inspection. NNT-191. Pigment in very restricted areas. NNT-192. One surface probably considerably concave into missing section of stone. NNT-167. Complete stone minus two flakes; two faces and one side used, with the used side slightly concave (Fig. 7-27b). NNT-169. Possibly bifacial. NP-566. Questionable grinding stone, coarse surface. NP-1414c. Questionable grinding stone. LP 4: NP-1519. Fragment with only one grinding surface remaining. NP-1242. Questionable; uneven surfaces. NP-1499. One surface with fine texture, the other coarse. NP-1520. One surface slightly concave. NP-1443. Some pigment on one face. NNT-98a. Fragment with one side and red pigment. LP 5: NP-1264. One of two pieces missing. NP-1365. Both surfaces slightly concave. For a summary of the grinding stones, see Table 7-8.


Only four of the grinding stones were whole or sufficiently whole to be able to see the appearance of the original stone. NNT-316a, from MP 1, is triangular in plain view (Pl. 12g) and easily portable. NP-220, from MP 3, is rectangular in plain view (Pl. 12b, Fig. 7-23d) and is also small. The two others (NNT-2445 from MP 5), NP-1373 from LP 1) are very similar in shape and size (Pl. 13a and 13b, Figs. 7-24c and 7-25c), probably six sided in plain view, with rounded sides and also easily portable. The four are almost exactly the same thickness, with NNT-316a and NP-1373 at 15 mm and NNT-244b and NP-220 at 16 mm. While the small triangular stone is unifacial, the other three are bifacial, two of them with concavities on both surfaces. Any of the four would easily fit in a small carrying bag or basket. This strong suggestion of portability presents a question as to whether these were actually used as grinding stones. To me the term grinding stone suggests use in the second stage of manufacture of stone axes and adzes or any other stone tool that is finished to a smooth surface. The grinding stone would be used to complete the form of the intended artifact from a blank that had already been roughed out to the general form and size desired. Such a second-stage job would probably be done at a home base where blanks in some number were accumulated and the grinding then done in this one location. The grinding stones would probably be rather large, and portability would not be a consideration. These four stones would seem to me to have been carried when work was to be done using an axe or adze in the field and used to sharpen the tool immediately on need after accidental flaking or general dulling of the axe or adze resulted from use. Because practically all of these


grinding stones were badly broken when recovered in the excavation, their classification was not based on form or size but on smoothed surface indication of use. Little can be said about the original size and form of most of the grinding stones because they have been so badly broken. It is certain that there was much variation in original size, as can be seen by looking at the dimensions of the stones recovered. Of the five largest stones recovered, only one was brought to Hawai`i (NNT-23), and it gave no clear indication of its original shape or size. These larger fragments start from EP 1 and extend to MP 4. The only similarity that all have is that they are unifacial. This suggests that these stones were on the ground when used and were kept in the same position throughout their use until broken. Only two of these large fragments were rather similar in size: NNT-23, from EP 2, weighed 1860 g, length 190 mm, width 160 mm, and thickness 25 mm, while NNT-25, from MP 1, weighed 1,410 gm length 185 mm, width 175 mm, and thickness 20 mm; only the thickness need have any close relationship to the original size of the stones. The three larger stones measured as follows: NNT-9, from EP 1, 2,450 g, 180 x 150 x 70 mm; NNT-5, from MP 4, 3,100g 160 x 110 x 90 mm; and NNT-1, from MP 4, 2,500 g, 245 x 195 x 40 mm. The grinding stones recovered from Early Periods 1 to 3 averaged 561.3 g. The only layer above this in which the stones even approached that average weight was MP 4 at 476.5 and it was the two largest stones recovered that made this average so high. In none of the layers above MP 4 was the average weight above 91 g. The average thickness of the Early Period grinding stones is 27.5 mm. The only other layer with an average thickness above 20 mm was MP 4 (average 24.7 mm) with the two very large stones. The fact that the largest fragments of grinding stones were found in Early Period layers suggests that there were considerably fewer of the portable grinding stones from these layers than from later layers and possibly that there was somewhat less breakage on the large stones form the earlier layers. Breakage of the stones was horizontal, vertical, and at varying angles to the grinding surfaces. Often only one of the original surfaces remained, so that for many stones it was impossible to tell whether originally they had been bifacial or unifacial or to tell the extent, if any, of the grinding of the red pigment. Only four of the stones with red grinding surfaces had red powder on both surfaces, while there was a total of forty-nine stones with at least some red pigment on one or both surfaces. Undoubtedly, had all stones been complete, there would have been more with the pigment on both surfaces and more of the total with some pigment. With this inherent inaccuracy, I can only say that with what data we have, the use of the grinding stones or powdering of red pigment appears to have been more common in the earlier layers. From the 1968 excavation, five out of the six grinding stones had red pigment. The ratio for the whole site was one stone out of 4.27 with the pigment, but for the Early Period and through MP 4 for the stones from both 1966 and 1968 excavations, the ratio was one out of 2.29 while for MP 5 through LP 6 it was one out of 6.14 stones. Not only breakage affects identification of red pigment on the stones. Some may first have been used for grinding both artifacts and the pigment and then used only for grinding artifacts, grinding away all or much of the pigment that had become imbedded in the


surface of the stone (see NNT-216 of LP 2). With only a small percentage of one grinding surface remaining on many of the stones, very little can be said about the original grinding stones. This is equally true for the presence of bifacial stones. Without doubt, many of those stones marked as unifacial had been used on the missing surface. As identified, there were 158 unifacial and 61 bifacial stones, or 72% unifacial and 27% bifacial. Looking at the same subdivisions as for the pigment, for the Early Period it was 9 and 21 or 82% unifacial and 18% bifacial; for Early Period through MP 4 it was 50 and 24 or 68% and 32%; and for MP 5 to the top it was 108 and 37,or 74% unifacial and 26% bifacial. Considering that there were more of the larger pieces relative to the smaller stones for the Early Period recovery and thus somewhat more chance that at least a portion of both surfaces of the stone could be examined than for the later layers, it seems reasonable that a larger proportion of the later stones would change to bifacial if all the stone were present and that there was little, if any, change in the proportion of unifacial to bifacial use of the grinding stones through time. This could modify my earlier inference that there were more of the large stones used only on one side in the earlier layers to one saying that the largest stones were primarily unifacial and, while there were more of them in the earlier layers, size for middle-sized and small-sized stones was not necessarily a factor in unifacial or bifacial use of a stone. The one exception to the largest stones being unifacial was NP-1480 from MP-3, which weighed 1,196 g, somewhat smaller than the six noted above. There was considerable variation in the texture of the grinding surfaces, some of them being extremely fine and others being quite coarse. The texture of the two faces of the bifacial stones often was different. A very small number of the obviously used surfaces were rather uneven, while some surfaces that were questionable for use were more even. There is certainly a possibility of misclassification of some grinding stones. One almost certain indication of use as a grinding stone is the concavity of the grinding surface. Yet some with a very fine texture that had certainly been used as grinding stone show no indication of a concave surface. The patterns of concavity vary somewhat as well. Usually the concavity is longitudinal; that is considerably longer than it is wide. In some cases it is rather central and appears to be either oval or even circular. With almost all of the grinding surfaces only partially present, it is difficult to judge the shape of the concavity. The difference between a grinding and a sharpening stone is probably indicated in the form of the concave working area. Without question, there is no clear boundary between these two classifications. One stone, NP-1406 from MP 3, had both types of concavity present, so it should be considered both a grinding and a sharpening stone. Twenty-eight (14%) of the 220 stones classified as grinding stones were associated with burials. None of these were whole stones, so it is possible that the association came about because the stone was in the soil placed back in the grave to cover the body. Only in the Early Period was this common, with six out of eleven from burials. None were found with burials or cremations from MP 7 to the end. Not counting the EP/MP 1 combination of the 1966 excavation, there were only twenty out of one hundred found with burials from MP 1 through MP 6. Two-thirds of the grinding stones recovered from Early Periods 1 and 2 were from burials, this at a time when there were probably only a few broken grinding stones on the surface or in the soil, so it would seem likely that they had been placed with


the burials. Considering my earlier question of how some of the very heavy stones could have been broken by use and the fact that all of the stones found in the graves were broken, and further that at least for the earlier use of the site it appears to have been used solely, or at least primarily, as a cemetery, why were so many-—all but three-—of these stones broken? I would suggest the possibility that the grinding stones had been purposely broken at the cemetery, probably at the time of burial. An alternate possibility is that it was traditional to use these grinding stones in the cemetery, so that was where breakage took place. From a present-day, Western point of view, this does not seem likely to me. Two grinding stones from LP 1 (NP-1373 and NP-1367) had been used not only on their surfaces but also across or around their edges (Pl. 13b, Fig. 7-25c and 7-26a). Most of these edges are concave surfaces lengthwise and the wear goes completely around the edge. This suggests to me that something flexible, such as rattan or material for baskets or matting, was being smoothed or polished by working back and forth around the edge. If this were so, it would require two hands to hold the flexible material so that the stone would have to be held by a second person, held by the feet, or wedged up in some way. The second and third possibilities seem more likely to me. These also might have been used to grind down the inner surface of bracelets (see discussion of the shell and stone bracelets). I know of no standard definitions of the three categories of stone tools that are used here; grinding stones, sharpening stones, and rubbing stones. For the most part, I have kept the classification that was made in the field, although I have changed a few. Were I to start from scratch in making the classification of these three tool types, I would change a few more, but the great majority would remain as they are here presented. I have noted illustrated reference to only two grinding stones in other reports. Sørensen (1988:16 0B. 29-30, Figs. 29-30) lists two broken grinding or polishing stones as surface finds from Ongbah Cave in Kanchanaburi. From the illustrations, they could be grinding stones, as they have flat working surfaces-—but their form is not like the grinding stones from Non Nok Tha. Because they were broken, with no idea of the original size or shape, they could have come from a large, monofacial grinding stone that had been set in the ground. In a preliminary report, Sørensen (1963:82) mentions the difficulty in distinguishing “between fragments of polishing and grinding stones.” He neither illustrates nor describes these, but he does say that about two hundred pieces were recovered. Sharpening Stones The classifying element used in the field for distinguishing sharpening stones from grindings atones was that the sharpening stone had grooves rather than the fine-textured flat or slightly concave surface. The grooves were of varying forms and sizes. Data on the sharpening stones are presented in Table 7-9.



EP 2: NNT-454. Distinct edge to broad groove, width of groove 15-16 mm. Two narrow grooves ca. 1 mm wide and 26 and 35 mm long, one end of each not distinct (Pl. 13c, Fig. 7-27c). NNT-460a. Two grooves, at two levels and different angles, one at lower level cutting across the other, upper groove straight, lower groove slightly curved, upper groove 1.7 mm deep, lower groove as much as 2.2 mm deep. Single deep groove on the other side of stone, of slanting depth 4.6 to 7.4 m deep (Fig. 7-27d). NNT-460. One curved groove at broken edge of stone, depth of groove 3.6 to 4.0 mm (Fig. 7-27e). EP 2-3: NNT-447. Uneven surface except for shallow groove with a distinct curved edge (Pl. 7-13d, Fig. 7-27f). NNT-448. One groove ca. 8.2 mm wide, distinct at one end, widening out and indistinct at other end, depth ca 1.8 mm, distinct for ca. 27 mm then 23 mm indistinct. EP/MP 1: NP-1424. On one surface there is one straight, deep groove, 4.6 mm deep, with relatively even curvature, and one shallower groove with distinct edge and probably a relatively flat bottom. 1.6 mm deep. Deep groove ca. 12.3 mm wide at surface of stone. A third apparent groove at one edge of stone, varying in depth 3.5 to 5.5 mm (Pl. 13e, Fig. 7-21g). NP-1421. Six grooves on different faces, at different angles, some deep, some shallow, curved lengthwise; deepest 6.0 mm and 12.7 mm maximum width, shallowest 1.4 mm deep and 5.7 mm maximum width. Their forms suggest smoothing something circular in cross-section and flexible lengthwise (spiny rattan or split bamboo?). All these stones are a coarse, light tan sandstone (Pl. 12a, Fig. 7-28b). NP-1426. One relatively shallow groove along edge, ca. 20.5 mm wide and 2.0 to 2.5 mm deep, and a deep somewhat u-shaped groove ca. 6 mm deep, over 25 mm wide, and slightly curved lengthwise (Fig. 7-29a). NP-1358. Broad, shallow groove, at least 35 mm wide, ca. 2.5 mm deep, on limestone (Pl. 14b, Fig. 7-29b).


Fig. 7-28


Fig. 7-29


MP 1: NNT-492. A bifacial sharpening stone with portions of two narrow, V-shaped grooves, one practically all present, 12.0 mm long, maximum width 2.1 mm, maximum depth less than 1 mm; only tip of second groove remaining (Pl. 14c). NP-1485. A broad, rounded bottom groove with a narrow groove running the full length in the center of the larger groove; maximum width of larger groove 53.7 mm, maximum length 91.2 mm, maximum depth ca. 5.5 mm, width of smaller groove ca. 2.2 mm; along broken (?) edge of large groove is a relatively flat grinding surface 11.0 to 14.0 mm wide (Pl. 14d, Fig. 7-30a). Fig. 7-30


MP 3: NP-953. Deep grooves on both faces; on one face, deepest groove up to 5.0 mm deep, all groves asymmetrical, narrow-â&#x20AC;&#x201D;ca. 8 mm-â&#x20AC;&#x201D;in the middle and widening at both ends (9.0 and 11.3 mm); two other grooves crossing over the deepest one, curving and running together; grooves at one edge of stone probably led to relatively straight edge break. Other surface with one very distinct curved groove with sharp edge slightly undercut in one area, ca. 5.8 mm deep on sharp side, very asymmetrical, a second much less distinct groove cutting into end of sharp groove with a third sharp, straight groove along broken edge (Pl. 14e, Fig. 7-29c). NP-1340. Varying grooves on four of six faces, several of them very indistinct because of wearing away of edge of groove, with one edge remaining and the other edge of groove gone; in some cases two or more groves running together (Fig. 7-30b). MP 4: NP-1062. Oval depression, smoothly rounded, maximum remaining length 97.5 mm, maximum remaining width 47.0 mm; small portion of a second sharpening area at corner of stone with ridge ca. 9 mm high between the two areas (Pl. 14f, Fig. 7-30c). NP-540. Deep, curved, worked surfaces on both faces makes this look a bit like it was being made into a mold for casting bronze and broken, so not completed. One face with one curved groove, sharp on one side and gradually reducing in depth on other side, maximum depth ca. 3.5 mm; two curving grooves on other surface, one with maximum depth ca. 5.2 mm, the other 2.3 mm. Two shallow grooves across one end, one all the way across, the other ca. halfway across (Pl. 15a, fig. 7-31). MP 5: NNT-231a. Gradual, shallow, concave area ca. 3 mm maximum depth and at least 22 mm wide. NNT-243. Narrow, slightly curved groove, ca. 70 mm long, ca. 5 mm wide, and 1.3 mm deep. Other surface has a concave grinding area (Pl. 15b, Fig. 7-32b). NP-1484. One face is a sharpening stone with a shallow (2.5 mm) diagonal channel ca. 48 wide an a coarse surface; at edge of the channel is a short, round-bottomed groove 19 mm long and 12.8 mm wide. The other face appears to have been used as an anvil, as roughly in the center of this face is an area ca. 49 mm by 25 mm without clear boundaries, with a roughened surface of relatively undefined, shallow pits and gouges (Fig. 7-32a).


Fig. 7-31


Fig. 7-32


LP 1: NP-515. Has only the edge of a curved, deep, gradually concave area (Fig. 7-35a). Fig. 7-33


LP 2: NP-1463. A shallow channel ca. 1.5 mm deep on one face and on the other surface a roughly oval depression ca. 27 by 19 mm (Fig. 7-33b). NP-1353. Deep groove, widening at both ends, 9 to 13 mm wide, 3.0 mm deep (Pl. 15c, Fig. 7-33c). NP-1462. Shallow groove along straight edge and asymmetrical, deep groove on one side, ca. 6.5 mm deep; very coarse sandstone (Fig. 7-33d). NNT-137. Two asymmetrical grooves on one face, neither completely present (Fig. 7-33e). NNT-204. One face for sharpening, the other face with three grooves present (Fig. 7-33e). For a summary of the sharpening stones, see Table 7-10.

Many of these stones were broken and some were only fragments of the original artifact. With the relatively small number of stones and the great variation in size, a sophisticated analysis would probably have little meaning. There is no average weight for a layer that is even close to the average weight of the total and, looking at the stones individually, only one (NP-1484) is within 10 g of the average. The median weight of all stones is 147 g, median length 66 mm, width 5 mm, and thickness 29 mm. The median measurements probably come closer to the original size of the stones than the averages.


There is no pattern in the shape of the stones, but there is a pattern to the grooves. Most of them are curved, there is more than one groove on a surface and often grooves on more than one surface, and when there is more than one groove on a face, they are in no specific relationship to each other. More than half of the stones were recovered from the three layers EP 2 to MP 1, and nine of the eleven stones found in burials come these same three layers. Sharpening is probably not a true, functional description for this group of artifacts. Some were probably used for sharpening axes and adzes, while most were likely used for smoothing rough surfaces such as thorny rattan in preparation for use in making baskets or the inside of bracelets. These stones could well have been used for more than one purpose, or they could have been made for one purpose and when broken, as with NP-540, used for a second, different purpose. Most of the sharpening stones are distinct from the grinding stones. They are about double the weight of the grinding stones and double their thickness. The narrow grooves are clearly different from the flat or slightly concave surfaces of the grinding stones. Four of the illustrated sharpening stones (Figs. 7-30c and 7-33a, b, and d) do not have distinct groves, but their concave surfaces are either considerably deeper or more abrupt than those of the grinding stones. NNT-137 (Fig. 7-33e) has two lengthy and relatively narrow concave surfaces and on NP-515 and 1463 (Figs. 7-33 and b) the concave surfaces appear to be linear rather than circular, so these are considered to be sharpening stones. I find NP-1484 (Fig. 7-32a) questionable and would perhaps classify this as a sharpening stone were I starting over. I have been able to find little comparative material from other sites. In part, I suspect this is because there are so few final reports published, particularly on the nonburial artifacts, and this type of artifact is not spectacular or well defined and thus may be listed somewhere in the text but not illustrated. From Lam Pao 2 in Kalasin, Sahatsakhan, a local villager showed us an artifact she had found in an Ayutthia-style jar. At the time we had no idea what it might have been, but on looking at the illustration (Solheim and Gorman 1966:Pl. VIIf), it appears to be a sharpening stone rather like NP-1062 (Fig. 7-30c). Higham and Kijngam list and illustrate eleven “whetstones” from Ban Na Di (1984:187-191). All but one of these I would classify with our sharpening stones. The one illustrated as Figure 3-47a I would call a rubbing stone. Roberto Ciarla illustrates one “sandstone polisher” from Tha Khae, Changwat Lop Buri, in the eastern part of the central plains of Thailand (1992:127). Stone and shell bracelets were manufactured at this site and the author believes that the polisher he illustrates was used for polishing the bracelets in their final stage of manufacture. The stone is rather smaller than most of the sharpening stones from Non Nok Tha and the grooves are much more symmetrical and much shorter than in the great majority of stones from Non Nok Tha. The short description of the manufacture of stone bracelets appears to differ from the way stone bracelets were made at Non Nok Tha (see below).


Higham, in his plans showing features from different layers of Khok Phanom Di, locates “burnishing stones” from many layers (1990). In this volume he does not define or illustrate a burnishing stone, so I cannot say which one(s) of our types of modifying stone he means. Sørensen lists a polishing stone that-—from the illustration-—I would call a sharpening stone (1988:47, Pl. 44-8). It is oblong, with three parallel channels running the length of the stone. While none exactly like this were recovered at Non Nok Tha, it would not seem out of place if found there. I have not attempted to locate any of this type of artifact outside of Thailand. Rubbing Stones This is an ambiguous category, as are the grinding and sharpening stones. All three categories were no doubt used for grinding, sharpening, or smoothing stone, bone, shell, and possibly softer materials such as wood and vines. I feel confident that all of these were used for modifying and/or finishing the surfaces of artifacts, such as bracelets, or the materials for making artifacts such as mats and basketry. Table 7-11 presents the location and measurements of the rubbing stones.

EP 1: NNT-464. Fragment, possibly part of a sharpening stone. MP 1: NNT-337b. Possibly a complete artifact, one curved surface used, slightly concave lengthwise, total face smoothed (Pl. 15e, Fog. 7-33f). NNT-402. Possibly whole, one surface used, surface somewhat irregular and somewhat concave (Pl. 16a, Fig 7-34a).


Fig. 7-34


MP 3: NP-1469. Fragment, probably from sharpening or grinding stone, one used surface slightly concave. MP 5: NP-119. Fragment, probably from sharpening or grinding stone, one concave, used surface. NNT-266. Fragment, probably from sharpening stone, one face concave across width and one side used (Fig. 7-34b). MP 8: NP-954. Complete, all lengthwise surfaces used, several slightly concave areas (Pl. 16b, Fig. 7-34c). LP 2: NP-1369. Fragment, probably sharpening or grinding stone, used face slightly concave. LP 4: NP-1709. Fragment, probably sharpening or grinding stone, one used face. Rubbing stones are the most questionable of these three types of artifacts used for modifying other artifacts. Two of those illustrated, I feel, are a truly different artifact from the other two types. NNT-337b and NP-954 (Figs.7-33f and 7-34c) were probably rubbed against the material being modified-—probably wood or vine-—as opposed to the grinding stones and sharpening stones where the object being modified was worked against the stone. If this is a distinct artifact type, the difference in form from the other two types is that rounded surfaces were used for working instead of flat surfaces a concave surface on a flat face. I know of no other such artifacts reported from other sites in Southeast Asia. I have not made an intensive search for this type of artifact, so they may have been reported elsewhere. Scrapers None of these are definitely scrapers, but they were so classified in the field. Their recovery from adjacent layers-—and thus possibly close in time—-suggests the possibility that these are intrusive and represent a culture different from the resident culture. Table 7-12 presents data on the so-called scrapers.


EP/MP 1: NP-1228. Possible scraper on a core with original cortex on one face, no indication of use; would not be out of place in a Hoabhinian context (Pl: 16c, Fig. 7-34d). EP 3: NNT-32. Possible scraper or chopper; from fill of burial; purplish red and white banded. Probable bulb of percussion on one side. NNT-34. Possible crude scraper. MP 1: NNT-24. Possible scraper, possibly from a prepared core; concave working edge near one end. Because as explained above, I question whether these few artifacts are really scrapers and whether they are indeed part of the Non Nok Tha culture, I have made no attempt to compare them to artifacts from other sites. Tektites Tektites are very hard and difficult to work. The suggestion of flaking and retouching found on several of these is unusual. Tektites are not common and are not found in large numbers in a deposit, so they would not furnish a readily available material for artifacts. These few apparently flaked and possibly retouched tektites suggest a knowledge of flaking not otherwise indicated at Non Nok Tha. Table 7-13 presents data on the recovered tektites.


MP 1: NNT-496. Roughly oval in form; several flakes removed from one end (Pl. 16d, Fig. 7-34e). MP 3: NP-1576. A flake with some tektite cortex left; small flakes on opposite sides of one edge suggest that the flake may have been used (Pl. 16e). NP-1508. A roughly circular flake with no original cortex left; may have been retouched (P1. 16f). LP 1: NP-1530. Flake with total original cortex on one side; no use or retouch evident (Pl. 16g). NP-650. Fragment of what could have been a large tektite, with only a small portion of original, typical pitted cortex remaining. Ripple marks from percussion suggest possible flaking of original tektite (Pl. 16h). LP 2: NP-1575. Some cortex remaining on one side; use unlikely (Pl. 16i). H. H. E. Loofs has reported on tektites found at Khok Charoen in Central Thailand (1977). He suggests that many, if not most, of the tektites were naturally broken when they struck the ground at the time of their fall from the sky, producing many natural tektite flakes. Microscopic examination, however, has shown that many of the resulting flakes were used (Ibid:P1.III). This may well have been the situation at Non Nok Tha. This area is well within the wide area of the original fall of the tektites (Ibid:Fig. 1). A broken tektite was found in the first surface survey of Nam Phong 6, a bit over 1 km west of Non Nok Tha (Solheim and Gorman 1966:175, Pl. XXVb).


Hammerstones Data on hammerstones are presented in Table 7-14. Stones were classified as hammerstones based on surface indications of percussion use.

MP 1: NNT-314. Mottled gray and brown (Pl. 17a, Fig. 7-34f). MP 4: NNT-310. Pebble with several large flakes removed; original surface red, flaked surfaces light tan to light gray with reddish tints; small area on one end with pecked surface (Pl. 17b, Fig. 7-35a). The very small number of hammerstones recovered from Non Nok Tha and the moderate number of polished stone tools that could well have been pecked into blanks for manufacture suggest that the original forming of the stone artifacts to be ground and polished was not done at this site. This would suggest either a trade in finished, polished stone artifacts, a trade in artifact blanks, or that locals went to the stone source and roughed out the blanks there. Hammerstones being such a common and usually unillustrated artifact in preliminary reports, I will not attempt a search from other sites for comparative purposes.


Fig. 7-35


Pestiles The term pestle is perhaps not a good one for these artifacts, as they are not in the usual form of a pestle. As they were used with metate-like stones rather than mortars, it might be better to call them manos-â&#x20AC;&#x201D;but they are not formed like the traditional mano, nor were they used in the same way. Table 7-15 presents data on these artifacts.

MP 3: NP-1473. Has a very neat, quite smooth, rounded working surface at one end and a small (11.8 mm x 8.9 mm). slightly rounded working surface at the other end (Pl. 17c, Fig. 7-35b). MP 5: NNT-6. A questionable pestle, while with a rounded end as if used in a rocking motion, there are faint, parallel scratches parallel to the bottom edge suggesting use was parallel to that edge. These scratches could have resulted from the process of manufacture. MP 7: NP-1483. Probably primarily used as a pestle or a rocking/ grinding stone. One side and the two ends have rounded and relatively smooth working surfaces from use, but the ends have some pitting and the rounded side has semicircular indentations. These do not appear to be from use as a hammerstone but made purposely to give a better grinding/ crushing surface. There is some pitting on the original remaining cortex (Pl. 17d, Fig. 7-35c). LP 2: NP-652. A Pestle or rocking/grinding stone, pentagonal to cubical in form, with rounded, smooth working surfaces. Three faces have the original cortex and four surfaces were used for grinding. The original cortex on opposite sides of the cubical portion has a polish,


probably from fingers grasping the stone. The fifth pentagonal surface, with original cortex, is slightly concave and has some rounded, small pitting in the surface, but because of the concavity that surface was not used for grinding. The pitting was probably the original preparation of this surface for grinding, through not used (Pl. 17e, Fig. 7-35d). NP-1482. A pestle or rocking/grinding stone on a river cobble with the edges of the working surfaces rounded. Original cortex on much of the two faces and one side, but some of this was worn away or intentionally worked away to give a better surface for holding the stone when in heavy use. One edge of the cobble was naturally broken off, as the patina on the broken surface is much the same as that of the original cortex-â&#x20AC;&#x201D;indicating an old break-â&#x20AC;&#x201D; with two edges of this broken and worn smooth, through the patina cortex, giving a moderately concave edge as if something narrow and flexible had been smoothed or thinned over that edge (Pl. 17f, Fig. 7-36a). LP 3: NNT-172. A pestle with four flat sides, two of these with most of the original cortex; all surfaces with many fine scratches. One end, coming to a rounded point, probably used as a pestle; from this point a groove runs 3.4 mm wide with parallel scratches the length of the pestle and over all of this surface (Pl. 17g, Fig. 7-36b). These pestles do not appear to have been used with a deep socketed mortar but rather to grind on a less deeply concave surface like a metate, with a rocking motion rather than a rotating motion like the usual pestle, or in back and forth motion like the usual mano. Three of the six are igneous stone, two are sedimentary, and one is metamorphic. With such a small number, the distribution could well be a matter of chance, though it would seem that it did not matter what kind of stone was used to make a pestle. None were recovered from early levels, but so few were found that this could also be meaningless. I have found no reference to or illustrations of manos, metates, mortars or pestles in any of the reports on other Thailand sites.


Fig. 7-36


Metates The stones on which the grinding presumably took place using the so-called pestles are not typical metates, as they are considerably smaller than the typical metate, but they do have a rather shallow concave grinding surface. Table 7-16 presents data on the metates.

EP 2-3: NNT-22. Questionable metate; stone left in Thailand, so my impressions are from the brief description and the drawing made in Thailand. It looks similar to NP-540, which is called a sharpening stone. In the Thailand description it is called a â&#x20AC;&#x153;grooved sharpening stoneâ&#x20AC;?; broken. MP 4: NNT-10. Possible metate, stone left in Thailand and the drawing is not sufficient for a description. MP 5: NNT-280. Metate and sharpening stone; one face is a very smooth sharpening surface with faint remnants of red pigment here and there; the other surface a likely metate with a short, heavy streak of red pigment near the original edge; broken (Pl. 18a, Fig. 7-37a).


Fig. 7-37


LP 2: NP-1481. The original cortex is still on the one working surface of this metate, and small patches of cortex remain in somewhat concave areas on the side but there is no cortex left on the bottom. The cortex is 1-2 mm thick and a dark reddish brown, while the stone is tan to tanish gray. The cortex on the used, slightly concave working surface has many shallow pits, some with rounded bottom, some with pointed bottom. Remnants of these pits are found even in the smoothest of the used areas, not extending through the cortex. Almost certainly these pits are artificial as the surface of the other remaining cortex area, while uneven, is not pitted. All of these possible metates are of sedimentary stone. While they are larger than most of the recovered stone artifacts, they do not have large surfaces available for the grinding that is the purpose of the metate. On the other hand, if the pestles were used with these metates, the working areas of pestles employed in a rocking motion are not large either, so they would fit well with these stones. The relatively small size of both metates and pestles suggests that they were not used to process large quantities of material-â&#x20AC;&#x201D;such as carbohydrate staple-â&#x20AC;&#x201D;but rather something of which only a relatively small quantity was needed at a time-â&#x20AC;&#x201D;such as spices. While the deepest, questionable metate is earlier than the deepest pestle, the vertical distribution of the two is similar. This rough match in quantity, distribution vertically, and size of working surfaces suggests to me that at least some of these were used together as pestles and metates for grinding a food substance that they were not an important element of the culture, at least in volume. Red Pigment Stones Red pigment stones are either iron oxide or in some cases a sedimentary sort of stone made up of small, tightly packed nodules of laterite. They have one or more smooth surfaces that were worked against a grinding stone to produce a red powder. While some burials had a considerable amount of this red powder, there were no doubt other uses for it. One likely use would have been to make a red slip for slipping pottery, while another might have been for body decoration. Table 7-17 presents data on the red pigment stones.




EP/MP 1: NP-1491. Six faces of varying size and shape, worked surfaces primarily flat or somewhat convex; abrasion lines parallel, but sets of lines in different directions (Pl. 18c, Fig. 7-38a). NP-1010. Unused, unbroken, perhaps too small for easy use. Fig. 7-38


MP 1: NNT-406. Probably not a red pigment stone but like the heavy, red deposit in the bottom of jar NP-957 (with Burial 69 of Layer 19, MP 3). This is a relatively soft, crumbly concretion with numerous small (1-4 mm) rounded pieces of laterite inclusions, all a red color, with no working surfaces and no possibility of working like the other stones, possibly for making red slip. NNT-427. Working surfaces primarily on one side with only a little working on two other uneven faces, one of them slightly convex (Pl. 18d). MP 2: NP-645. Wedge shaped with two main faces but worked on both sides as well; one side slightly convex with one working surface, the other side with 2 working surfaces, one convex and the other flat (Pl. 18e, Fig. 7-38b). NP-1567. Roughly rectangular with 6+ faces, two main faces with the two edges between the main faces also used by rubbing lengthwise but not producing a truly flat surface side to side, the wider end with two working faces and the narrower end rounded with three small working surfaces (Pl. 18f, Fig. 7-38c). MP 3: NP-682. Two main used surfaces on wedge-shaped stone also with a small portion of the butt and one side used (Pl. 19a, Fig. 7-38d). NP-1493. One main used face with small convex face on one edge (Fig. 7-38e). NP-1578. Small fragment, one curved and one flat working face. MP 4: NNT-268. Six or more faces, all surfaces used, including a rounded edge between two faces and to both ends (Pl. 19b, Fig. 7-38f). MP 5: NNT-272. Four used faces plus a small area of a fifth. Grinding marks in any direction, though relatively consistent for any one face (Pl. 19c, Fig. 7-38g). NNT-278. Two faces, one only partially used as is one side. NNT-237. Four faces used, two main faces not totally ground through the original relatively flat cortex, edge and end face rounded; stone is a conglomerate of tiny laterite pebbles.


NNT-249. Two used faces, one an edge 2-4 mm side and 31 mm long, convex at one end; only moderate use on one large face (Pl. 19d). NNT-257. A conglomerate stone of small laterite pebbles, only one face used but nonused end shows polish possibly resulting from finger pressure when working. (Pl. 19e). NNT-261. A fragment of larger stone with one face and one edge used. NP-1370. Laterite conglomerate pebble, more compressed than NNT-237, not showing a filling between the small laterite grains making up the stone; one used face. NP-1360 and 1360a. Larger piece similar to NNT-257, one used face and polish on other original surface possibly from finger pressure during grinding; â&#x20AC;&#x153;aâ&#x20AC;? is a small fragment with three faces, not broken from 1360 but same kind of stone. MP 6: NNT-507. Irregular shape, one used face with two working surfaces at slight angle to each other. MP 7: NP-1077. Laterite conglomerate, one side used with two used surfaces at slight angle to each other. NP-1371. Laterite conglomerate, wedge shaped with two used surfaces and some use on narrow edge between faces. NP-1524. Small fragment with one worked face. NP-1572. Laterite conglomerate with some use on two faces. NP-1589. Laterite conglomerate, wedge shaped, one face used, additional two narrow facets at one edge at slight angle to each other. NP-1361. Six faces, every surface used, ends and edges convex (P1. 19f, Fig. 7-38h). MP 8: NP-642. Laterite conglomerate, one used face with many tiny pits, possibly from remnants of small laterite nodules falling out of the grinding surface. Unworked surface with polish from finger pressure or from wind polish. NP-1487. Laterite conglomerate, with one primary face but with four other partially used faces (Pl. 19 g, Fog. 7-38i). NP-683. Laterite conglomerate, two primary faces with three more small working faces.


LP 1: NP-335. Laterite conglomerate, wedge shaped, two faces; one face with a slightly concave groove ca. 5 mm wide, diagonal across at center of face; other face with a deeper and wider grove across, ca. 7 mm wide (Pl. 19h, Fig. 7-38j). LP 2: NP-648a. Wedge shaped with ground, rounded back, two flat faces with edge of wedge a narrow third used face (Fig. 7-39a). Fig. 7-39


NP-1490. Wedge shaped with two faces and small portions of the wide back used. NP-1495. Fragment of possibly considerably larger stone, not a laterite conglomerate but a crystalline, harder stone than the laterite as scratches can scarcely be discerned on used surfaces, even under high magnification; two primary faces remaining, with one end and two small faces also used (Pl. 19i). NP-1566. Laterite conglomerate, one face, one edge used plus very slight use of a second face. NNT-139. Not laterite, similar to NP-1495, one relatively flat face partly used, convex side partly used (Pl. 20a). NNT-215a. Crystalline stone, portions of two faces broken off, probably ca. six faces plus two facets at an angle to the side face probably practically all surfaces used. NNT-215b. Laterite conglomerate, two faces used, one a side and the other the end of a wedge-shaped stone. LP 4 NNT-105b. Probably laterite conglomerate but considerably harder than usual laterite; does not show scratches under hand lens, worked faces very even, laterite nodules showing only rarely in partly smoothed or unused surfaces; one large face used and use on rounded edges and side (Fig. 7-39c). NP-1600. Laterite conglomerate, one flat face with additional facet, one rounded face. There is no pattern to the form of the red pigment stones. I consider only weight as a meaningful variable to examine statistically. I do not include NP-1010 and NNT-406 in this simple analysis as they do not appear to have been used. By visual inspection I divided the stones into two weight groups: One group of 72 stones with a range of weight of <1 to 29g; the second group of five stones with a range of 53 to 460 g. This results in 72 of the smaller red pigment stones with a total weight of 692 g (if weight was <1 g I used 0.5 g for the weight), an average weight of 9.6 g and a median of 7 g. The group of large stones is so variable that its average at 188.4 g has no meaning. The group of smaller stones probably had been used so much that they had become difficult to grind for powder, and if so, these would have been discards. The larger stones are probably more representative of sizes still of easy use. No red pigment stones were recovered from the bottom two layers and only eight of the smaller stones from layers of EP 3 through MP 3. MP 4 through 8 had 24 of the smaller stones recovered and LP 1 through 5 had 36. The most stones from one layer (17) were recovered from LP 2. LP 6 had only one stone. For the layers with more than one stone, the average weight for a layer varied from 5.9 gm for LP 5 to 14.5 g for MP 2, showing no consistent change in size from early to late levels.


Eleven of the pigment stones were recovered from burials, starting in EP 3 (NNT-506), one from EP/MP 1, seven from MP burials, and two from LP burials. Descriptions of only three burials mention the pigment stone found with the burials. NNT-506 was found on the left shoulder; NNT-697, from MP 4, was found on the bottom of the pit; and NNT-507, from MP 6, was found to the right of the legs. It is possible that all but the one from EP 3 were not directly associated with the burial but found there with soil from the backfill of the grave. Red pigment stones were not a common item for burial furniture. I have found no evidence of the use of red pigment in burials at Ban Chiang nor of red pigment stones (White 1982a). It would seem likely that red pigment stones will be reported as suggested by the extensive use of red painted designs on the middle- and later-level pottery. I have not come across reference to red pigment at Ban Na Di (Higham and Kijngam 1984). Red pigment in powder form was common on burials at Khok Phanom Di (Higham and Bannanurag 1990). No red pigment stones were reported associated with the burials. No red pigment is mentioned in the detailed discussion of burials at Ban Kao (Sørensen 1967:65-74). The use of red pigment on burials appears to be random at sites in Thailand, although probably not on burials within a single cemetery, based on the present limited evidence. Stone Bracelets Fifty-four stone bracelets were recovered in the excavations, three of sedimentary stone, 22 of igneous stone, and 29 of metamorphic stone. Cross-sections of the bracelets were recorded in the field at the time of cataloging. On examination in the lab we disagreed with the recorded cross-sections in many cases and on closer inspection realized that a number of the cross-sections were between defined sections. As a result I decided not to list sections in the table but include them in the notes when the section is clear. The crosssections are: D, rectangular, and square. The major problem in recognizing the crosssection is that many of the interior surfaces of the bracelets are not smoothed or polished so they often are very rough and uneven. The makers and wearers were apparently not particular. Table 7-18 presents data on the stone bracelets.



Sedimentary Stone Bracelets EP 2-3: NNT-499. This is one of three bracelets of sedimentary stone, a white marble with a D cross-section (P1. 20b, Fig. 7-39e). Unfortunately, NNT-547 was unavailable for measurements, so it is not included in the analysis involving measurements. Igneous Stone Bracelets EP 1: NNT-476 and NNT-500. Both have rectangular cross-section (P1. 20c, Figs. 7-39f and g). EP 2: NNT-504. Rectangular cross-section (P1. 20d, Fig. 7-39h). EP 2-3: NNT-498. Outside surface and sides well polished, ca one-third of inside surface and sides were cut, the rest a rough surface probably having been broken apart (P1. 20e, Fig 7-39i). MP 1: NNT-401. All surfaces well smoothed, inner surface and sides straight while outer surface convex, D cross-section (P1. 21a, Fig. 7-39). NNT-489. Outside surface and sides well polished, inner surface rough (Fig. 7-39k). NNT-419, NNT-497, and NNT-494. Like NNT-498, NNT-489, and NNT-401 respectively (P1s. 21b-c, Figs. 7-39l and m, 7-40a).


Fig. 7-40


MP 4: NNT-502a. Square cross-section, outer surface slightly convex, inner surface rough in part (Fig. 7-40b). NNT-502b. Rectangular cross-section, inner surface slightly convex (Fig. 7-40c). NNT-495 and NNT-503. Like NNT-489 and NNT-401 respectively (P1s. 21d, 22a, Figs. 7-40d – e). MP 5: NNT-269. Outer surface polished with small flake scars on this surface extending from edges, inner one-third of surface relatively smooth and even with horizontal striations, other two-thirds very rough with diagonal, deep gouges from edge inward to ridge at base of the smooth band; smoothing of the ridges between diagonal gouges. Rough area feels smooth to touch, with no sharp edges to the gouges (P1. 22b, Fig. 7-40f). NNT-238a–b. Inner surface of “a” like NNT-269; “b” with total inner surface rough like portion of NNT-269 (Fig. 7-40g-h). MP 6: NNT-501. Outer surface has high polish, one side well smoothed while other side not as well smoothed, inner surface mostly rough (Fig. 7-40i). LP 2: NNT-203. Outer surface and sides well polished, inner surface slightly rough, cross-section rectangular (Fig. 7-40j). NNT-250. Outer face and one side with moderate polish, other side not as well polished, inner surface rough; one-third of inner surface circumference smooth from the cutting of the cylindrical drill used to cut out central disk, while two-thirds rough from breaking out the disk without cutting through. There are diagonal channels extending through the rough surface from the smooth surface suggesting that a small-diameter bit was used to drill a number of holes on the diagonal from the bottom of the circular channel made with the cylindrical drill. This might have been done to form a weakened circle from the bottom of the channel to the other face of the stone, which would allow the central disk to be broken out of the stone without requiring the no doubt long and tedious job of drilling all the way through the stone with the large, cylindrical drill (Fig. 7-40k). LP 3: NNT-198. Outer surface and two sides with moderate polish, inner surface very rough (Fig. 7-40l).


NNT-169. Outer surface not polished but with horizontal grinding striations not smoothed over, sides smooth, inner surface smooth but mostly horizontal grinding striations evident, with some striations perpendicular or diagonal, extending primarily from one edge. LP 4: NNT-121. Outer surface with high polish, sides well smoothed, inner surface rough with uneven cross-section (Fig. 7-40m). Metamorphic Stone Bracelets EP 1: NNT-772. High polish on outer surface, well-smoothed sides, half inner surface horizontally ground and half rough (Fig. 7-40n). EP 2: NNT-504Y. High polish on outer surface and sides, inner surface smooth, outer surface a smooth arc from side to side (Fig. 7-40o). EP 3: NNT-428. Outer face and sides smoothed but with some diagonal striations on outer face, inner face smooth with horizontal striations and concave from side to side. Stone is dark gray at a break but outer patination a brownish gray (Fig. 7-40p). EP/MP 1: NP-674. Sides and outer face smooth, the latter with many fine horizontal striations, inner surface smooth. On inner surface, horizontal, shallow grooves are present on about fourfifths of the width from one side with diagonal, shallow grooves in one and some areas two directions from the other side to the inner edge of the horizontal groove area; cross-section rectangular with rounded corners (P1. 22c, Fig. 7-41a). NP-1091. Outer surface and sides well polished, inner face smooth and slightly concave from side to side, outer face convex (Fig. 7-41b).


Fig. 7-41


MP 1: NNT-406. Outer face and sides well smoothed, inner face with deep horizontal striations extending ca three-quarters of the way from one edge with the remaining one-quarter much smoother. Stone a light gray with a darker gray layer (P1. 22d, Fig. 7-41c). NNT-320. Outer surface and sides well smoothed, inner face with horizontal striations (Fig. 7-41d). MP 2: NP-1088. Same as NNT-320 (Fig. 7-41e). NP-1476. Moderately polished outer face and sides, inner face rough (P1. 22e, Fig. 7-41f). NP-660. Outer face and sides smooth, inner face with deep horizontal grooves (Fig. 7-41g). NP-798. Outer face and sides smooth, three-quarters of inner surface with horizontal striations, one-quarter with varying striations, cross-section rectangular with rounded corners (P1. 22f, Fig. 7-41h). NP-799. Moderate polish on outer face and sides, horizontal striations on inner face (Fig. 7-41i). MP 3: NP-1597. Outer face with high polish, sides smooth, inner face partly smoother (Fig. 7-41j). NP-596. Outer face has moderate polish with some diagonal scratches and remnants of horizontal striations, sides and inner surface smoothed; a moderate horizontal ridge on inner face, cross-section rectangular but close to square (Fig. 7-41k). NP-657. All surfaces well smoothed (Fig. 7-41l). NP-1507. Like NP-657, cross-section rectangular with rounded corners (Fig. 7-41m). NP-1422. Outer face has high polish, sides smooth, inner face mostly rough with some smoothing on half toward the thicker edge (P1. 23a, Fig. 7-41n). MP 4: NP-1090. Like 1422 (Fig. 7-41o). NNT-287. Moderate polish on outer face, sides and inner face smooth, one side slightly rough with horizontal striations; considerable variation in thickness (Fig. 7-41p).


MP 5: NP-1478b. Outer face smooth (Fig. 7-41q). NP-1089. All surfaces finished. MP 7: MP-1478a. Outer surface moderately polished, sides and ca. half inner face smooth but latter with few diagonal rough areas; other half smooth with horizontal, shallow grooves and ridges (Fig. 7-41r). NP-681. All surfaces smooth (Fig. 7-42a). NP-1475. Outer face has polish, moderate polish on sides, inner face rough (Fig. 7-42b). Fig. 7-42


LP 1: NP-824. All faces have moderate polish (Fig. 7-42c). NP-644. Outer face smooth, other faces uneven, probably unfinished (P1. 23b, Fig 7-42d). LP 5: NP-1366. All surfaces smooth (Fig. 7-42e). The small variety of cross-section does not define the variety of the completed bracelets. The D cross-section had the flat face to the inside with a curved outer surface (see Fig. 7-39j and P1. 21a). In a few cases there is close to a D cross-section, but with the flat face to the outside and the curved face inside (see Figs. 7-40k–l). The rectangular crosssection could have the long side of the rectangle perpendicular to the arm, as in NNT-504 (see Fig. 7-39h) or more commonly, parallel with the arm (see Fig. 7-40a). The most impressive type of bracelet (P1s. 20e, 21c, 22a – b; Figs. 7-39i and m and 40e–f) could be considered rectangular in cross-section with long side parallel to the arm, except that the inner surface is, in part, very irregular, due to the method of manufacture and the inner surface not being smoothed and polished like the outer surface. Only two of the stone bracelets found (NNT-546 and 547) were complete and these were, one each, on the arms of child Burial 15. To make these even more unusual, they were each in the middle between a pair of bronze bracelets. Sixteen other bracelet fragments were found with burials, but in all cases where I have located detailed information about the burial, the fragment(s) were either in the mound above the burial or not directly associated with the skeleton. Thus Burial 15 appears to be unique in having stone bracelets directly associated with the burial, and in this case, on the arms. Bracelets made from sedimentary stone may have been early in the site. With child Burial 15 being so rich, and the only one with stone bracelets on the arms, these might well have been heirlooms. With only three of these bracelets, however, nothing definite can be said about their distribution in time. There were twenty-two fragments of igneous stone and twenty-nine of metamorphic stone. For the three types of stone, there were eight bracelets from the Early Period layers, two from EP 3/MP 1, thirty-five from the Middle Period, and nine from the Late Period. Igneous and metamorphic stone bracelets were recovered from all three periods, being clearly most common during the Middle Period. NNT-547, NNT-774 and 204 were not used in summarizing the measurements, so a total of fifty-one bracelets were used. The two Metamorphic bracelets from EP 3/MP 1 were not included in summarizing the EP or MP measurements Totals. “>1.0” and “<1.0” refers to Maximum Width equal to or greater than 1.0 cm and less than 1.0 cm respectively. See Table 7-19 for a summary of stone bracelet measurements.


With measurements of only two sedimentary stone bracelets, the several measurements do not necessarily have reliable meaning. You will note that except for thickness, the average measurements for the sedimentary stone bracelets are either much the largest or much the smallest for all the bracelets. If you divide the bracelets into two groups based on the width, with the width of one group 1.0 cm and up and the second group less than 1.0 cm (the median width of all the bracelets is 0.94 cm), you get two groups without a boundary between them. If you graph the maximum width for the fifty-two bracelets for which there are measurements, you get neither a single curve nor a bimodal curve, but there is a suggestion of a trimodal curve with maximums at 0.3, 0.9, and 1.4 cm. The ranges show no clear distinctions between the different measurements except for the two sedimentary stone bracelets, which, as mentioned above, are not enough in number to give them much weight; differences for these bracelets are also present in the averages. Looking at the average measurements, there are differences suggested. There is a small increase in size for outer and inner diameters and thickness from Middle to Late Periods, while the maximum diameter is about the same. The maximum width is much smaller in the Early Period than in the two later periods and, if you remove the one sedimentary bracelet from the Early Period, the average total width is 0.5 cm–-even smaller. Only one of the eight EP 1–3 bracelets is of the type with wide rectangular-like cross-section with the rectangle parallel to the arm. With the exception of NNT-498 (Fig. 7-39i, P1. 20e, EP 2–3), they are either the D cross-section, D with rounded corners, or rectangular–-or close to rectangular–-with the rectangle perpendicular to the arm (Figs. 7-39f–h, 7-40o). This wide rectangular-like bracelet is made from either igneous or metamorphic stone. While the examples are few in number, this suggests that the wide rectangular type started


somewhat later than the other two types. In all measurements taken as a whole, the metamorphic bracelets are smaller than the igneous bracelets. In comparing the widths when divided into the two groups, the larger widths (59%) are more common than the smaller for the igneous bracelets, and the smaller widths (67%) are double the number of the larger for the metamorphic bracelets. Also, the metamorphic bracelets are slightly thinner than the others. With only fifty-one bracelets for which there are measurements, one cannot have great confidence that these differences are real, but they are sufficiently consistent that I feel the indicated differences are likely to be real. The bracelets, being apparently for everyday wear rather than specifically for burial with the body, were in use over the total area of the site and possibly discarded when and where broken. Considerable information can be inferred on the method of manufacture of the bracelets, particularly on those of greater width and roughly rectangular cross-section. Only the finishing process can be seen on most of the bracelets. NNT-401, from MP 1, is a good example of a finished bracelet. In Plate 19a, on both the inner and outer surfaces and the side pictured, can be seen scratches left over from the grinding to even these surfaces. On the inner and outer surfaces, the scratches are parallel to the long axis, and on the side they run across the thickness. The polishing, done after the grinding was completed, did not completely obliterate the grinding scratches. The grinding of the curved outer surface could well have been done on a grinding stone with a concave grinding surface such as NP-1373 (P1. 13b). The inner surface and sides of the bracelets could have been smoothed using the edge of the same stone or a similar small grinding stone, sharpening stone (NNT-454, P1.13c and NP-1353 P1. 15c), or rubbing stone (NNT-327b P1. 15e). Many of the outer bracelet surfaces were polished after the grinding to shape so that the grinding striations were no longer apparent (see NNT-498, P1. 20e). The bracelets were probably made using a flat piece of stone selected for its finished thickness as a bracelet. The stone would have been broken, probably with a hammerstone, to roughly the outside shape, which would later be ground and polished to its final shape. This is hypothetical, as no stones at this stage of manufacture of a bracelet have been recovered. The center was probably cut out of these flat stones using a hollow, cylindrical drill, such as an open-ended piece of bamboo. A very hard, fine-grain sand, placed in the circular groove after it had been started, would have been the primary cutting agent while the bamboo was rotated back and forth by hand or using a pump bow. The few cutout disks that were recovered (see P1s. 23câ&#x20AC;&#x201C;e) show that one surface of the stone had been smoothed before starting to cut the disk, but the opposite face was not smoothed until after the disk had been removed. The cylindrical, hollow drill was used to cut only one-third to one-half of the way through the stone when a much smaller, solid drill was used to perforate, often at a diagonal, from the bottom of the groove to the opposite face. The circular disk would then be punched out, leaving a horizontally smooth surface and next to this a rough inner surface (see description for NNT-250 from LP 2 above; see P1s. 21bâ&#x20AC;&#x201C;c and 22b). This punching out of the disk after cutting partway though would no doubt save considerable time, as the cutting with the cylindrical drill must have been hard work, taking a long time. A similar, somewhat simpler procedure has been presented for stone bracelets found at Lop Buri in Central Thailand (Ciarla 1992:126). Peter Francis Jr. presents another variant of manufacture found in Indonesia (1991:221 â&#x20AC;&#x201C; 222).


From Ban Tak Det, an archaeological site on the banks of the Mekong in Laos a short distance east of Vientiane, was recovered “a green stone bracelet core” (Bayard 1980a: 57). Besides this core, a fragment of a bracelet and a broken bracelet in the early stage of manufacture were recovered from this site (78 – 79, Fig. 12j–l, 120). From the surface of a site (L69) in Loei, Thailand (47, Fig. 12m), a broken bracelet in the process of manufacture was also recovered. These are similar to bracelets from Non Nok Tha. A T-sectioned bracelet of calcite was found on the arm of a burial at Ban Chiang, with a bronze bracelet of the same form (White 1982a: 40). Nine whole or fragments of stone bracelets were recovered from Ban Na Di (Higham and Kijngam 1984:69–72). All had the T cross-section, unlike those from Non Nok Tha. Five were found associated with burials, but only one of these had the bracelet on the wrist. Fragments of or complete stone bracelets were recovered at Kok Charoen on some of the skeletons. Thirteen were found on the lower left arm. Some bracelets found were of a nonlocal stone, but others of an alabaster-like calcite limestone may have been made from a local stone (Loofs and Watson 1970:75–76). Bracelets are illustrated in an unnumbered plate that is not referred to in the text, so whether these are stone or shell—-also found on the lower arm of a burial--I cannot say. No mention was made of their cross-section. A stone ring, presumably a bracelet, was recovered from Chande Cave B (Heekeren 1988a:63, P1. 26b). It is T-shaped in cross section. Sørensen mentions about one hundred fragments of “arm-rings” but wonders whether all of these were actually to wear on the arm, as the inner diameter of many of them is so small (1964:81–82, Fig. 7). He illustrates a variety of these but says nothing about their cross sections. One of the forms appears to be T-shaped in cross section and none of them look particularly like the Non Nok Tha bracelets. He notes that not one has been found on the wrist or arm of a burial. He also noted “a limited number of broken half fabrics and plug-like discs from the refuse of the drilling out of the central hole”. Khok Phanom Di was excavated in eleven layers, with Layer 11 the deepest. Higham and Bannanurag (1990) list all artifacts recovered by layer. One of these artifacts is called “stone disc/bangle”, without definition. A total of 109 of these stone disc/bangles were reported from Layers 10 up to 2; 1 in Layer 10, 2 in Layer 9, 2 in Layer 8, 4 in Layer 7, 12 in Layer 6, 8 in Layer 5, 18 in Layer 4, and 61 in Layers 2 and 3. No stone bracelets were recovered from burials. None of these are illustrated, so their possible resemblance to those from Non Nok Tha is not known at this time. Stone bracelets are commonly found in Phung Nguyen culture sites in Viet Nam. A much greater variety of cross sections are found there than at Non Nok Tha, including variants of the T section as well as those found at Non Nok Tha. Bracelets broken in the process of manufacture before the central disk has been completely cut out are also found (Nguyen 1980:31-32). Disks


Only five stone disks were recovered from the site. These are obviously the center sections of stone bracelets that were manufactured here. With only five disks and fifty-four bracelets recovered, it may be that only a few bracelets were made at the site or that, more likely, the general location where they were made is not in the areas excavated. As the disks do not appear to have been used, they could well have been discarded in the area of manufacture. MP 1: NNT-493. Square 4H, Level 7, Weight 53 g, Burial -, max. diameter 4.6 cm, max. thickness 0.8 cm. Igneous stone, horizontal striations to narrow grooves on smooth edge, one face relatively flat, possibly the original surface of the stone, other face uneven from removal of flakes from all around the side, greenish gray (P1. 23c, Fig. 7-42f). NNT-392. Sq. OF, L. 7, Wt. 7 g, B. -, Max. Diam. 3.2 cm, Max. T. 0.5 cm; with originally a smooth, flat surface, but a flake removed from ca. 80% of face results in a somewhat irregular surface, bottom face rough, horizontal striations on edge (P1. 23d, Fig. 7-42g). NNT-491. Sq. C5, L. 7, Wt. 25 g, B. 101, Max. Diam. 4.0 cm, Max. T. 0.9 cm. Metamorphic stone, faces slightly convex, faces and side mainly smooth, probably side and faces were smoothed, light greenish gray color probably from weathering, with original stone probably dark gray (Fig. 7-42h). MP 2: NP-1327. Sq. C5, L. 20, Wt. 20 g, B. -, Max. Diam. 3.0 cm, Max. T. 1.1 cm. Metamorphic stone, diameter at smooth face 2.7 cm, side smooth from smooth face, with ridge and increasing diameter to edge of rough, uneven face; probably black stone originally, with white lime covering partially removed (P1. 23e, Fig. 7-42i). MP 5: NNT-262. Sq. 4D, L. 6U, Wt. 4 g, B. -, Max. Diam. 3.2 cm, Max. T. ?. Metamorphic stone; a flake from one face has reduced maximum thickness; remaining face well smoothed, possibly worked, grayish green stone. In spite of the possibility that NNT-491 and NNt-262 may have had their edges and one face smoothed after having been removed from the bracelet, there is no indication that these disks were used for anything. As a few potsherds were recovered that had been worked into disks, probably for use in a game, it is possible that these two smoothed disks may have been used for the same purpose. The grooves from the cylindrical cutting tools would probably have been 2 to 3 mm wide, so the maximum diameter of the disks, plus 4 to 6 mm, would have them fit the minimum inside diameter of the bracelets. (Stone identification by Belshe and L. Lepley.) The central stone disk from manufacture of a bracelet is usually found in sites where stone bracelets are found. My feeling is that there are usually a considerably smaller number of the disks than bracelets found, but I am unable to document this.


Rings These two split rings, the only ones recovered from the two excavations and both of them from one grave, would appear to be imported. As the two were found together, it represents a single event--and a very rare one. It would certainly seem that they were not being made locally. They are much too large to be worn as a ring on a finger and, as they are both split, this would suggest that they were originally made to be earrings. It does not necessarily follow that this is the way they were used by the person with whom they were buried. MP 6: NNT-552a. Square 4F4G, Level 6A, Weight 10 g, Burial 26, outer diameter 1.8 cm, max. thickness 0.8 cm. Sedimentary stone, partly crystallized limestone; split ring with flat face, gap of 2 mm, roughly circular in cross section but inner surface flat in some places with a slight ridge along one side of flat surface (P1. 23f, Fig. 7-43a). Fig. 7-43


NNT-552b. Square 4F4G, Level 6A, Weight 10 g, Burial 26, outer diameter 3.5 cm, inner diameter 1.8 cm, max. thickness 0.8 cm. Split ring very similar to 552a (P1. 23g, Fig. 7-43b). I have not been able to locate any similar stone rings from other sites in Thailand or elsewhere in Southeast Asia. I have seen brass rings just like these stone rings being worn in their extended earlobes by women living in longhouses on the Baram River in Sarawak. Somewhat similar split rings were common in Viet Nam, but instead of circular in cross section they were rectangular and usually made of jade or similar semiprecious stone (Nguyen Phuc Long 1975: Figs. 69, 73, 110, 156, 158 â&#x20AC;&#x201C; 163). Split rings similar to those of Viet Nam are common in South China. BEADS Only three stone beads were recovered during the two seasons of excavation, two of them from burials. All three are cylindrical with both ends cut on the diagonal, the diagonals not parallel to each other. Data on the beads is presented in Table 7-20.

MP 3: NP-1506. Translucent, mottled white, layered, ends diagonal to long axis but not quite parallel to each other, oval diameter 10.8 x 11.7 mm, length perpendicular to ends 19.1 mm, perforated from each end, oval perforation 4.0 x 5.4 mm at one end and 2.7 x 3.5 mm at the other measured parallel to end surface, kind of stone not identified (P1. 24a, Fig. 7-43c). MP 4: NNT-569b. Mottled light gray metamorphic stone, well polished, ends diagonal in opposite directions, diameter slightly oval 10.8 x 11.7 mm, maximum length 47.4 mm, minimum


length 35.8 mm from extreme edges of ground end surface, perforation from one end 4.4 x 5.0 mm in diameter and extends to ca. 7 mm from the other end; the shorter perforation is conical and circular, 4.4 mm in diameter, probably done with the same drill (P1. 24b, Fig. 7-43d). LP 1: NP-285. Similar to 569b but darker mottled gray, metamorphic stone, diagonal ends, diameter 9.1 x 9.5 mm, maximum length 31.3 mm, minimum 27.0 mm, maximum and minimum lengths not exactly opposite each other on circumference, perforation drilled from both ends and meeting almost exactly, closer to one end than the other perforation is 4.4 x 5.0 mm (P1. 24c, Fig 7-43e). According to Higham and Kijngam, “stone beads were extremely rare at Ban Na Di. Two were recovered. The earlier, a flat stone spacer bead, was found in association with the human skeleton. . . . It was broken in antiquity. It was rectangular in shape. . . . The second bead is fragmentary. It comprises an orange-white banded agate. . . The general appearance suggests that it is part of a fairly large spherical bead. . . . It probably originated in India” (1984:79). I have been unable to find references to stone beads from Ban Chiang or Khok Phanom Di. I have been able to find reference to only two stone beads from Ban Kao (Sørensen 1967:91, P1s. 25, 27, 126a). Their description is incomplete, not mentioning their color, though questioning whether they were of “jade or nephrite”. They are cylindrical but from the three illustrations it is uncertain whether their ends are cut square or on the bias. In the Plate 25 drawing, three of the ends appear to be cut on the bias and one square. In the Plate 27 drawing, two of the ends appear to be cut on the bias. In the Plate 126 photograph, three of the ends appear to be square and one possibly at a slight bias. These are the only other stone beads I have noted that may have their ends cut on the bias like the beads from Non Nok Tha. Glover (1990:167) recovered 330 stone beads from two seasons’ excavations at Ban Don Ta Phet in Central Thailand. These were made from carnelian, agate, crystal, or nephrite. The site of Khuan Luk Pat is on the central west coast of peninsular Thailand. It was a major center for the manufacture of ancient stone and glass beads. For the stone bead manufacture, “Several kinds of evidence . . . have been found. (1) Fairly large quantities or unshaped carnelian, onyx and quartz fragments. (2) A number of carnelian bead blanks-flaked blocks which have not yet been perforated by drilling but which have the rough form of a finished bead. (3) A few fragments of carnelian with attached areas of cortex” (Bronson 1990:216). Another source of beads was farther south: Kuala Selinsing is on the west coast of Peninsular Malaysia in the state of Perak. This was also an early bead manufacturing center. Most beads found in Kuala Selinsing were made from semi-precious gemstones. Analysis.... of thirteen beads from Kelumpang 1 indicate that the materials used included beryl, sodalite, moldavite, plasma, jasper, eventurine, quartz, cats eye and analicima. Such beads occur right through the stratigraphy and outnumber those of glass. The beads were cut, ground and polished, although much of the lapidary work was rather crude.


Many of these gemstones, apart from jasper, do not occur in the Malay Peninsula. Beryl, plasma, aventurine and sodalite have been mined for centuries in India . . . . and Iran and Iraq are well known for producing sodalite and aventurine . . . . The excavations also unearthed beads of agate, carnelian and rock crystal. (Rahman 1991:148 – 149) The most complete, both extensive and intensive, treatment of jewelry in Southeast Asia with which I am acquainted is that done by Louis Malleret (1962:145 – 271). This covers stone and glass beads and other forms of jewelry that were found at Oc-Eo, in southern Viet Nam, and comparative materials from all over Southeast Asia. Oc-Eo was a bead manufacturing center and there is a great variety of beads illustrated. None of the cylindrical beads illustrated have their ends cut on the bias like the three from Non Nok Tha. It has long been assumed that India, the major stone-bead-making center of Asia . . . was the source of semiprecious stones uncovered in most archaeological contexts in Southeast Asia. . . . On the other hand, there is growing evidence that stone-bead making also took place in other parts of Southeast Asia, including Mantai in Sri Lanka, Klong Thom in Southern Thailand, Oc-Eo in Vietnam, Kuala Selinsing in West Malaysia, and Bukit Maras in Sarawak. (Francis 1991:223) The westerly trade of Southeast Asia during the period from about 400 BC to AD 500 was not a mere “trickle of trade”, nor can it be described simply as the “drift” of a few exotic and precious items to the east of India. Rather it operated on a considerable scale at panregional, and local levels; it was developed as a commercial enterprise by Indian merchants; and there is little doubt that Southeast Asian sailors and traders were also active in the exchanges. (Basa et al. 1991:381) I have hypothesized that the trade between India and Southeast Asia was first organized as the western lobe of the Nusantao Maritime Trading Network and was started long before 400 B.C. (Solheim 1993). The first two of these beads at Non Nok Tha were from Middle Period 3 and 4 and therefore date well before 400 B.C. If the assumption is true that the earliest stone beads in Southeast Asia came from India, these beads would push back the 400 B.C. date for the beginning of bead trade between India and Southeast Asia. On the other hand, while these beads are well made and polished, their color and bias-cut ends appear to be unique at this time. We need to look to India for this type of bead and its dating there. Pendant Only one recovered stone was classified as a pendant. From the description it appears that it may not have been a pendant. LP 2: NNT-218. Square 2E, Level 5X, Weight 13 g; not from a burial. A small, water-washed pebble with a natural hole and no indication of wear, but the stone is foreign to the site so it may have been meant for a pendant (P1. 24d).


Miscellaneous MP 5: NNT-524. Square 4F, Level 6U, Weight 12 g. Classified in the field as fine-grained sandstone, with lengthwise striations (possibly well-fired earthenware with considerable fine sand temper), cylinder, both ends broken, maximum length 37.3 mm, diameter 14.7 x 16.8 mm; would be good for grinding inner surface of stone bracelets (Fig. 7-43f). Flake Fragments Data on flake fragments are presented in Table 7-21.

While these flakes are of stone foreign to the site, none of them were used, so it is likely that they resulted from the accidental breaking of a grinding stone in the cases of NP-1336 and NP-1068, or flakes removed accidentally or on purpose from a stone tool. There were so few flakes recovered that they suggest the working of stone by flaking was not done at the site.


Polished Stone These pieces of polished stone are probably fragments of artifacts broken in the process of manufacture or broken in use. They are too small to determine for certain from what kind of artifact they might have come. Data on polished stone are presented in Table 7-22.

Three of these (NNT-752, NP-779, and NP-405) are fragments of adzes or axes, such a small portion in each case that total form is not known. One of them (NP-779, P1. 24e, Fig. 44c) was in the process of manufacture when broken and the other two could have been. The other pieces listed above could have been something other than axes or adzes, also in the process of manufacture, or artifacts broken at the site when in use. Taken in combination with the very few unworked flakes, this suggests two possible interpretations. Either stone artifacts were given their general formâ&#x20AC;&#x201C;-blanks--at the source of the stone and completed into polished artifacts at the site, without any stone flaking done at the site, or stone flaking was done in a circumscribed area at the site, an area or areas not touched in the small total of the site excavated. The first alternative would be supported by the considerable number of grinding, sharpening, and rubbing stones recovered as these would suggest the finishing of polished stone artifacts at the site out of previously prepared blanks. (Fig. 7-44)


Fig. 7-44


SUMMARY Table 7-23 summarizes the stone artifacts and worked stone fragments of definite and probable recovered during the 1966 excavation; Table 7-24 does the same for the 1968 excavation.


It is obvious that there was differential use of various parts of the site over time. The three levels with the largest number of artifacts in the 1968 excavation area—-MP5, MP1, and MP4-—did not produce numerous artifacts in the area of the 1966 excavation. The levels of the 1966 excavation with the second and third (tied) largest number of artifacts—-MP3, MP2, and MP7—-were not represented in the 1968 excavation. While LP2 would appear to have had the largest population, the MP1 through MP5 stone artifact count suggests considerable activity through all five levels of the Middle Period. Table 7-25 presents data for the 1966 and 1968 excavations combined.


In Table 7-26, the first column under % is the percentage for the total number of artifacts covered in Tables 7-23 and 24, while the second column under % in the lower half of Table 7-26 is the percentage for the total of the five most common artifacts. Tables 7-23 and 24 show that the most common stone artifacts are grinding stones, red pigment, bracelets, axes and adzes, and sharpening stones.


These five most common recovered artifacts are interrelated. Grinding stones and sharpening stones used either for manufacture or maintenance of other artifacts would commonly have been used in working with the red pigment stones, bracelets, axes, and adzes. This supports the interpretation presented earlier suggesting that bracelets and axe/adzes were finished at the site. The rarity of the stone disks cut from the center of the bracelets and of axes/adzes in the process of blocking out also suggests that the preliminary manufacture of blanks for these artifacts was either done at a different location in the site or done away from the site. Assuming that there is a direct correlation between the number of artifacts recovered from a level and the population that produced that level, the largest populations for the site as a whole would appear to have been those represented in Late Period 2 and Middle Period 5, or for the major periods as a whole, for the Middle Period. The numbers and percentages for the 1966 and 1968 excavations are not directly comparable as the 1966 excavation did not reach the bottom of the site in most squares, so the Early Period is considerably underrepresented in the 1966 excavations. BONE AND SHELL Relative to pottery and stone artifacts, very few bone or shell artifacts were recovered; only bone and shell beads were at all common. These were from the earlier layers EP 1 through MP 2, plus a few from MP 5. For beads as an artifact type, there was also one of mammal bone from MP 5 and one glass bead from LP 2. Data on shell and bone beads are presented in Table 7-27.


EP 1: NNT-562. A total of 592 disk beads with a child burial, two strings on the waist, two on the forehead and five on a vessel beyond the head; many are brownish and darker at their maximum diameter than toward the center, others are white with generally greater diameters (up to 5.7 mm maximum) than the brownish (minimum diameter 3.8 mm, but with much overlap). Perforations are not proportional to diameters; in the largest white beads, the perforation has a 1.7 mm diameter--the smallest diameter noted--while in the smallest brownish bead it is 1.8 mm; thickness varies from 1.4 to 0.5 mm for the brownish beads, while the white appear to be slightly thinner, with nine beads calcified together, 7.5 mm in total thickness; both are often somewhat wedge shaped. At 40x magnification, no structure can be seen on the face of the brownish beads but on some, horizontal lamination can be seen; on the white beads the lamination is slightly more apparent. White beads tend to have more calcification than the brown and for both many are calcified together, showing that when placed with the burial they had been strung. There are about twice as many brownish as white beads by volume (P1. 24f). NT-563. Hundreds more of NNT-562.


NNT-564. Same as NNT-562, but one bead is only 3.4 mm in diameter. These beads make a string 69.5 cm long; child burial (P1. 24g). NNT-565. Two brownish disk beads, same as others; child burial. (There were twenty beads originally, but only the two were brought to Hawai’i). EP 2: NNT-566. A total of 293 disk beads at waist of a child burial; some are white but most are brownish and generally somewhat lighter in color than those of NNT-262 to 265. Many of the beads are smaller than those of 252 to 265, down to 2.0 mm, but the largest is 6.0 mm in diameter; most are about 4.4 mm with several less than 3.0 mm diameter, the largest is wedge shaped with thickness 2.2–1.0 mm, the thinnest is 0.6 mm; perforation in largest is oval 2.3 x 1.8 mm, second largest perforation is 1.8 mm, the smallest 1.6 mm. NNT-567. Eight beads below a child burial, different from the others, uneven, white, diameter 7.3–4.3 mm, perforation 3.5–2.0 mm, thickness 2.3–4.3 mm; one side is flat and the other somewhat concave (Fig. 7-45a). EP/MP 1: NP-663. A total of seventy-one white disk beads around waist of adult burial, diameter 5.5– 4.2 mm, thickness difficult to measure because of concretions, 2.0–0.3 mm; many are calcified together like the other white beads. NP-822. Total of 110 white disk beads with an adult, same as 663. NP-823. One white disk bead with an adult, probably accidental, diameter 4.7 mm, diameter of perforation 2.7 mm, thickness 0.7 mm. NP-1505. Two whitish disk beads with child burial, maximum diameters 4.9 and 4.2 mm, perforation diameters 1.7 and 1.8 mm, thickness 0.9 mm. NP-1522. Three whitish disk beads with child burial, maximum diameters 6.2, 6.6, and 5.8 mm, perforation diameters 2.0–2.2, 1.9, and 2.4 mm, thickness 1.7–2.2, 1.4–1.6, and 0.7– 0.8 mm; the white beads have more obvious horizontal layering than the whitish to brownish, perhaps resulting from the weathering, as they also look rougher. MP 1: NNT-568. Total of 756 whitish disk beads at waist of adult burial, maximum diameter 3.0– 5.0 mm, perforation diameter 1.1–2.0 mm, and thickness 0.4–1.3 mm, rather smaller on average than other sets.


Fig. 7-45


MP 2: NP-1523. Thirty-three whitish disk beads; three larger beads are 4.7–6.0 mm maximum diameter, perforation 1.2–1.9 mm (in largest 1.2), and thickness 0.8–1.0 mm; thirty smaller quite consistent in size, maximum diameter 3.2, perforation 1.2, and thickness 0.3–0.6 mm. MP 4: NNT-569a. Five beads classified in the field as made from sedimentary stone, fine striations, layering where heavy patination flaked off, strongly suggesting these are shell. Three different shapes: two roughly spherical, first bead, 11.6 mm maximum diameter, moderately flattened ends diameter 11.1 mm, conical perforation drilled from each end meeting at slight angle in middle, perforation diameter at ends 3.4 mm; second bead, maximum diameter 11.0 mm, diameter from end to end 8.6 mm, conical perforation drilled from both ends, diameters 3.6 and 2.5 mm. One bead oval in cross section 7.7 x 9.0 mm, but this had been accentuated by flaking of patination, diameter end to end 7.0 mm, perforation conical from both ends, diameter 2.1 and 2.6 mm. Two long beads roughly barrel-shaped and oval in cross section, one 4.6 x 8.0 mm, length 8.7 mm, flattened ends not quite parallel, perforation conical from both ends, 2.1 and 1.9 mm diameter; second 3.7 x 7.5 mm cross section, length 8.1 mm, perforations conical from both ends, 2.1 mm in diameter (P1. 24h, Fig. 7-45b). NNT-570. Three beads, striations suggesting these are shell, but classified in the field as sedimentary stone. Largest with oval cross section 12.2 x 14.0 mm, maximum length 17.5 mm, ends not parallel, conical perforations from both ends, diameter 6.7 mm. One almost circular in cross section 11.0 x 11.9 mm diameter, maximum thickness 7.5 mm, ends not parallel, circular perforation 3.3 mm in diameter, drilled from one end. Third like last two of 569a, 3.7 x 7.4 mm, length 7.4 mm, circular perforation 1.9 mm diameter, drilled from one end (P1. 24i, Fig. 7-45c). MP 5: NP-1525. Three to five whitish disk beads, possibly six, all calcified together, total thickness 3.3 mm, same as other small ones of 1523 except the one at end is shiny on open face, looking like mother-of-pearl. NP-1504. Sixty-one disk beads, some white and rough, others brownish, the latter slightly larger and smoother, white diameter 2.5-4.1, brownish 3.2–4.2 mm. NNT-571a. Mammal bone bead with adult burial, oval in cross section and tapering slightly at ends, maximum diameter 12.6 x 8.8 mm, length 28.0 mm lengthwise, perforation circular in cross section 3.5 mm in diameter, drilled from one end (P1. 25a. Fig. 7-45d) NNT-571b. Broken mammal bone bead, reconstruction not perfect, oval cross section ca. 13.0 x 11.9 mm, length >30 mm, probably slight tapering at ends, lengthwise perforation ca. 5.8 mm at end. At the broken end the perforation has uneven edges ca. 3.7 mm apart,


suggesting that it was broken while being perforated as the larger perforation extends evenly from unbroken end to ca. 6 mm short of the broken end (P1. 25b, Fig. 7-45e). LP 2: P-306. Dark blue glass bead, transparent with some bubbles, diameter 6.2 x 7.0 mm, thickness 4.7 mm, one narrow side flattened, perforation oval and uneven inside bead, diameter 2.5 x 3.7 mm (Fig. 7-45f). Most of the shell disk beads were with child burials and were placed, usually in strings, at the waist and/or neck. The disk beads were recovered from Early Period 1 through Middle Period 2. One set of about four disk beads and a second set of sixty-one were recovered from Middle Period 5. These would appear to have been disturbed and probably from a lower level. Two sets of five and three beads from Middle Period 4 were roughly spherical and distinct from the disk beads. The small disk beads are found throughout Southeast Asia with late neolithic and early metal age burials. Several thousand were recovered at Ban Na Di, and their method of manufacture was suggested by Higham and Kijngam: “Some misshapen beads indicate that thin pieces of shell were broken into small square shapes, which were then drilled approximately through the centre and usually from both sides with a cone shaped drill bit. Numbers of these square blanks were then threaded onto a fine stick or cord and ground down to a circular shape” (1984:58–61). There were a few larger shell beads recovered at Ban Na Di (Ibid.: Fig. 3-1), strung with the small disk beads and roughly cylindrical with varying cross sections. These are unlike the larger shell beads at Non Nok Tha. Shell disk beads appear to have been common at Khok Charoen in Central Thailand. They were found near the waist of skeleton 5 during the first season’s excavation (Watson and Loofs 1967:253–255) and in the same location on several skeletons during the second season’s excavation (Loofs and Watson 1970:76). According to the index, shell beads were found in only a few burials at Khok Phanom Di. From Burial Cluster B of Mortuary Phase 2, 327 disk beads were found in one burial and 390 barrel-shaped beads in four burials, with both types of bead in Burial 120, and 55 short barrel-shaped shell beads in one burial (Higham and Bannanurag 1990:183). From Mortuary Phase 5 a total of 189,503 small disk beads were recovered from four burials, three of these being children. The one adult female burial, however, had 120,787 beads while an infant had only 269. From three of these burials, 1585 I-shaped shell beads were recovered, a type not found at Non Nok Tha. Seven short barrel-shaped beads were recovered from one burial (Higham and Bannanurag 1990:327). From the Bang site at Ban Kao, Sørensen (1967:91) lists as ornaments “A necklace of tiny circular and flat beads with central perforation,” but does not say what they were made from. In the illustrations (P1s. 25, 27-12, 125i–j, and 126a), they certainly look like the shell disk beads. Francis (1991:220–221) has this to say about the shell disk beads: The Nautilus and Trochus shells were cut into disk beads and spacers (with multiple holes for holding


strands apart). At Gilimanuk, Bali (c. A.D. 50), the 1964 excavation uncovered 1,900 shell disk beads. These appear to have been made by the heishi technique, in which flat shell is chipped into circlets, which are perforated and then strung together on a stick or fiber and rubbed across a flat or grooved rock, producing beads of similar diameter (or tapered strand, if so desired). The technique is of great antiquity, dating to the upper Paleolithic in India. . . . It is also very widespread, found, for example, in the Philippines, Taiwan, Thailand . . . in the Americas, and in Africa. The only other bone beads I have come across are from Khok Phanom Di. They are cylindrical and in illustration look like the two beads from Non Nok Tha. They were a part of a bracelet, alternating with barrel-shaped beads (Higham and Bannanurag 1990:256 â&#x20AC;&#x201C; 257). Perhaps the reason the bone beads appear to be so rare is that being small and organic they would not last well in soil. Much new information has become available the last few years about glass beads (for example see Basa, Glover, and Henderson 1991; Francis 1991). As only one glass bead was recovered from Non Nok Tha, I donâ&#x20AC;&#x2122;t feel it is necessary to say more than that glass beads started coming into Southeast Asia from India by 400 B.C., appearing with the first use of iron, and became widespread in Thailand and all of Southeast Asia. They were being made locally soon after this in sites on the west coast of the Malay Peninsula and elsewhere. Transparent, blue glass beads with bubbles are one of the more common early types. Shell and Ivory Bracelets The shell bracelets from Early Period 1 and Middle Period 4 were badly corroded. Because shell bracelets were so rare from Non Nok Tha, the two bracelets (NP-539 and NP-677) from burials of disturbed provenience are included in the tables and following description. The portions of bracelets probably made from elephant ivory are not corroded. Data on these bracelets are presented in Table 7-28.


Shell Bracelets EP 1: NNT-561. Has an interior ridge (P1. 25c, Fig. 7-45g). EP 3: NNT-546b. Only whole bracelet recovered (Fig. 7-45h). ?? NP-539. Thickness from 0.7 to 1.0 cm; Conus shell (identification by Alan C. Ziegler; P1. 25d, Fig. 7-46i). ?? NP-677. Thickness from 0.3 to 0.6 cm; considerably weathered, cut from top side of Conus shell (identification by Ziegler). The two shell bracelets without come from Burials 61 and 35 of uncertain layers, NP-539 probably about layer 6 and NP 677 probably layer 17. As shell bracelets are so rare, they are included in the listing but not included in the layer listing (P1. 25e). Twenty-one fragments and whole shell bracelets were recovered at Ban Na Di, sixteen of these from burials, and of these bracelets from burials, thirteen were from one burial. Most of these were made on Trochus shell, a type of cone shell (Higham and Kijngam 1984:63– 68). They were considerably larger than the Non Nok Tha bracelets, the outside diameter of which averaged 5.7 cm. The smallest Ban Na Di bracelet was 7.3 cm outside diameter, while the largest bracelet from Non Nok Tha was 6.9 cm. The average outside diameter from Ban Na Di was 9.46 cm. With the little information on shell bracelets from sites in Thailand, it is interesting that the difference here appears to be in the kind of shell used to make the bracelets. The two bracelets from Non Nok Tha for which the shell has been identified were made from Conus shell. From Ban Na Di they were made from Trochus shell, a type of cone shell, so they may have been the same. From Tha Khae, to the west on the central plain, they were made from Tridacna and Trochus, the former being the shell used in local manufacture and the latter apparently imported (Ciarla 1992:126). This may well have meaning in different trade patterns, but information from many more sites is needed. As mentioned above, the two sedimentary stone bracelets with Burial 15 were the only bracelets found on the arms of the skeleton, and they were each between a pair of bronze bracelets. It is curious that the one burial found in the Pha Chang Rockshelter, in Chiang Mai Province, had one shell bracelet on each arm and nine bronze bracelets with the one shell bracelet on one arm and five bronze bracelets with the other shell bracelet on the other arm (Santoni et al. 1990:46 – 48). No stone bracelets were reported from this site. The preliminary reports on Khok Charoen do not mention how many shell bracelets were recovered, but one of them (Loofs and Watson 1970:76) mentions that twelve shell bracelets were found with a burial, “on the lower left arm with a cowrie shell near the wrist”.


The unidentified plate showing bracelets, mentioned in the section on stone bracelets, appear to illustrate shell bracelets, as suggested by the kind of breakage. The report on Khok Phanom Di does not include shell bracelets in its index, but they were present with some burials. Four shell bracelets were found on the arm of an infant burial (Higham and Bannanurag 1990:361). The one illustrated bracelet with this burial is about 5 cm in outer diameter and looks like it was made from a cone shell. The term used for these bracelets is shell disc/bangle, but its meaning is uncertain. Burial 7, of an infant (353), illustrates two shell disc/bangles, so labeled, but the caption for this illustration states “A set of four disc/bangles fashioned from fish vertebrae were found in the vicinity of the right wrist” Burials 15, 16, and 43 had each listed with one shell disc/bangle. The illustration of Burial 15 (333), an adult female, points to a disc/bangle, and this is illustrated on page 334 with a half T-like cross section, that is with a lip on only one side of the bracelet. Burial 16 (335), of an infant, has a similar bracelet around its left wrist. There were other burials with disc/bangles made from fish vertebrae. Both complete and unfinished bracelets made from the spires of Trochus niloticus shells and a few bracelets probably made from Tridacna squamosa were recovered from Kuala Selinsing mounds (Rahman 1991:149). Elephant Ivory MP 4: NP-714. Has cracks perpendicular to edges across outer and inner surfaces and almost horizontal-diagonal across sides. Concentric laminar structure is obvious (identification by Ziegler; (see note with Worked Bone — NNT-553; Pl. 25f). MP 5: NP-671. Worked the same way as some stone bracelets on inner face with striations perpendicular to edge over half the width and the other half with diagonal gouges. No concentric laminar structure is obvious, so-—while probably made from elephant ivory-— this could be dense bone of a large mammal or pig canine (identification by Ziegler; Pl. 25g). Artifacts made from elephant ivory are not common at Non Nok Tha or at any other reported site in Southeast Asia. There are a few more ivory artifacts from Non Nok Tha presented below. Fragments of four ivory bracelets are reported from Ban Na Di (Higham and Kijngam 1984:73, 75). These are larger in diameter and considerably larger in width than the bracelets from Non Nok Tha. While no ivory artifacts were reported from Khok Phanom Di, a cache of ivory fragments was found. These had been split and broken by percussion, but no other working was evident. As these fragments were only found in this one small location, it is logical to assume they were related in some way to former or future ivory artifacts (Higham and


Bannanurag 1990:45, 47). Two short cylindrical beads possibly made from ivory were reported from Ban Khao (Sørensen 1967:91). Artifacts of ivory have been reported from Kuala Selinsing but not further identified (Rahman 1991:149). Mammal Tooth Pendant MP 4: NNT-555. Square 3E-3F, Level 6B, Weight 82 g, Burial 32. Three teeth: first with perforation all the way through, drilled from both sides, circular perforation diameter 2.0, total length 37.5 mm; second tooth with perforation from both sides, but either it did not meet in center or was filled with a very hard lime; on one side a flake from tooth surface removed at side of perforation, diameter of perforation 2.0, total length 37.7 mm; third tooth perforation filled with hard, white material, two perforations on one side, side-by-side, diameter 2.0 mm (Pl. 25h, Fig. 7-46a). Fig. 7-46


Dog’s teeth and other kinds of teeth are widely found as pendants or in necklaces. Their use may have been relatively rare but, because teeth survive better in the soil than any other form of bone or other organic material besides shell, they probably appear to be more common than they were when in use. Two tiger’s teeth were found associated with an early, rich burial at Ban Chiang (White 1982a:82-83, #129). Both teeth had two perforations through them. Two dog’s canine teeth were recovered from Ban Na Di (Higham and Kijngam 1984:73-75). Both had a single perforation through the root. From Khok Phanom Di, several different kinds of teeth were found associated with burials, including crocodile, rhinoceros, and shark teeth. “Sixteen pierced Muntiacus muntjak canines were found arranged as it they had been strung from the right shoulder down to the left hip” of Burial 102 (Higham and Bannanurag 1990:274). A canine of a carnivore was found, perforated at the base, that had been associated with Burial 132 (272). Five perforated canines from a small carnivore were found on the chest of extremely rich Burial 15 (328) and one pierced mammal canine was found with Burial 43 (327). Considering that there were three burials at Ban Na Di out of 154 burials described that had perforated teeth similar to the three from one burial at Non Nok Tha, they were a little more common for a burial at Ban Na Di than they were at Non Nok Tha. Bone Point MP 1: NNT-559. Square 1E, Level 7, Weight 1 g, Burial 74. Slightly charred point, length 37.2 mm, only the point shows smoothing, probably from use, several faces extending back from the point with scratches on one face perpendicular to the length and next narrow face with diagonal scratches. Looks like an awl but differing direction of scratches makes this questionable (Fig. 7-46b). NNT-771. Square OF, Level 7, Weight 2 g, Burial 114. Length 51.1 mm, uneven widths, the few scratches are parallel to the length, somewhat biconical in shape, possibly with points at both ends and one tip broken off; could have been a large fish gorge and when tip broken off possibly used as an awl (Fig. 7-46c). MP 2: NNT-307. Square C5, Level 20, Weight 20 g, Burial 2. Biconical, mottled bone, length 73.3 mm, maximum diameter 11.2 x 10.0 mm; tips not sharp as flakes have been removed from the points, tip of points rounded showing continuing use after loss of points; possible fish gorge (Fig. 7-46d). MP 4: NNT-773. Sq. 2F, L 6B, W 1 g, Burial 33. Black bone, length 30.0 mm, one tip broken, other sharp, maximum width and thickness 2.7 x 0.5 mm, rounded cross section toward tips but rectangular through ca. 23 mm of length, possible a fish gorge (Fig. 7-46e). MP 5:


NNT-571b. Sq. 3E-4E, L 6, W 3 g, Burial 49. Bone or antler, length 53.8 mm, diameter at broken end 6.7 x 7.6 mm; for about 20 mm back from point, all striations are parallel to length, largely obliterating the scratches going around the length (Fig. 7-46f). The function of these points is questionable. Three of them were probably pointed at both ends so they could have been fish gorges. They were not perforated, however, with no wear indicating that they might have been attached to a line. The other possibility for all five of them is that they were used as awls, though NNT-773 seems rather small to have been so utilized. A long (17.8 cm), narrow (0.7 cm) tool was recovered from a rich burial at Ban Chiang. It had two perforations towards one end, and it was called a hairpin (White 1982:82-83, #126). The other pointed bone tool listed in this Ban Chiang exhibit catalog (#125) was not illustrated. From nearby Ban Na Di, twenty-two bone points were recorded and illustrated (Higham and Kijngam 1984:175-178, 181-184). Several are double pointed and most have one or both points, when indicated, broken. Several of these are very similar in size, shape, and surface abrasions to those from Non Nok Tha, except that none are as small as NNT-773. The bone point summary states: “It is noticeable that practically all the bone points were found damaged. They may have fulfilled a variety of functions. There is no evidence that any were used as projectile points. One function was probably that of an awl. . . . Notched and shaped bone is used today as a part of a composite fish or small mammal traps, the principal component being bamboo” (178). Two broken bone points, similar to some of those from Non Nok Tha were reported from Ban Kao (Sørensen 1967:90, Plate 125d-e). From Kuala Selinsing, “Bone points, some possibly identifiable as hair-pins, have been found in association with burials, particularly near the skulls (Rahman 1991:140). Bone Spatula EP 1: NNT-777. Square 4H, Level 9, Weight 82 g, Burial 14. Fossilized bone, 183.0 mm long, width at bit 42.7 mm, thickness at break 13 mm, curved working edge sharp, slightly concave on inner side. The shape and length of the bone suggest that it was made form the femur of a mammal (Pl. 25i, Fig. 7-46g). This appears to be the complete undamaged tool. The sharp edge, without nicks of damage from use, suggests that whatever this was used for, the material worked must have been relatively soft, and/or even in texture. The only other bone spatulas that I have noted are from Ban Na Di. The two were both made from a flat piece of bone, possibly a scapula, each with a handlelike projection. Both are unbroken, with a sharp edge, and the sharp edge of one of them was “worn to a fine sheen” (Higham and Kijngam 1984:179, 184 A-B).


Carved Bone Cylinder MP 2: NP-710a. Square E5, Level 20X, Weight 32 g, - . Decorated bone, length 81.1 mm, diameter 45 mm, maximum thickness 2.3 mm with some variation, smaller end diameter 37 mm; the bone at the smaller end is tapered to a sharp edge while at larger end it is flat and 1.8 mm thick. Near the larger end is a band, 9.1 mm wide, of six concentric circles around the cylinder; first circle is 3.8 mm from the end, circles not an exact distance from each other; 27.2 mm from the six circles at the end, about in the middle of the cylinder, is a band 6.7 mm wide of five concentric circles; between the central band of circles and another band of circles at the smaller end is another undecorated band 21.4 mm wide; the band of five concentric circles at the smaller end is 5.8 mm wide, with the final circle 4.6 mm from the sharp edge. The bone is a mottled brown with both inner and outer surfaces very smooth. The inner surface has grooves down to right from the larger end and vertical, less apparent grooves from the smaller end with a few grooves diagonal up to right. Concentric laminar structure seems present, so this is probably elephant ivory (identification by Ziegler; Pl. 26a, Fig. 7-47a). NP-710b. Associated with the cylinder is a small piece of fired (?) clay, maximum length 13.2 mm, maximum thickness 2.7 mm, with a very fine cloth (?) impression having fifteen of the diagonal over threads along a horizontal line in a space of 5 mm, a very even, tight weave (Pl. 26b). I have no idea what purpose this artifact might serve. Nothing else like was recovered at Non Nok Tha, nor do I know of anything else like it from another site. It was very well and carefully made. Being only a little over 2 mm thick, it was very fragile. Without question it was ornamental. Five fragments of finely decorated bone were reported form Ban Chiang. The fragments are curved, suggesting the possibility that they were bracelets. They were decorated with pairs of parallel lines very neatly incised around the circumference of the fragments bordering four or six equally well incised concentric circles (White 1982a:81-82). The bone is shiny and the workmanship reminds me of the Non Nok Tha cylinder.


Fig. 7-47


Antler Chisel MP 4: NNT-779. Square 3F, Level 6B, Weight 31 g, Burial 39. Length 115.3 mm, maximum diameter 23.2 x 18.3 mm, width of chisel tip 8.2 mm; small part of tip broken off but edges of break are smooth; no wear scratches can be seen at up to 40 x magnification but rather a shiny polish the last ca. 22 mm to the tip; possibly used as a wedge rather than a chisel (Pl. 26c, Fig 7-37b). The sheen extending back 22 mm from the working edge means that the material being worked continued in contact with the surface of the bone for that distance. If a chisel were used on wood, the wood chips of shavings would probably curl away from the sharp edge of the chisel and not produce a sheen for nearly that distance. Using the tool as a wedge on something relatively small could produce the sheen that far back from the sharp edge. Perhaps it was used to open something like a coconut. Antler artifacts and worked antler have been reported from several sites, though none of these sites produced a tool like this one from Non Nok Tha. At Ban Na Di several pieces of antler were recovered that showed some working, but no completed artifacts were found (Higham and Kijngam 1984:180, 185c). Two pieces or worked antler were included in the Ban Chiang exhibit, but with no idea as to function (White 1982:81-83, #128). Worked antler was reported from nonburial context at Khok Phanom Di (Higham and Bannanurag 1990:114-115), but neither description nor illustration of these objects was presented. A very finely worked portion of an antler was recovered from a burial at Ban Kao, with no obvious function except possibly ritual (Sørensen 1967:91, Pl. 124). It reminds me somewhat of an atlatl, but I do not know whether such an artifact was known in Southeast Asia, nor do I see why it would have been so very well made. Bone Knife EP 2: NNT-560. Sq. OF, L 8L, W 8 g, B 121, Length 81.0 mm, Max. W 18.7 mm, Max. T 5.9 mm; on a mammal rib, sharp edge of rib has small nicks presumably from use, tip somewhat broken but probably had a rounded, sharp tip, one side of rib cut diagonally to length to make sharp edge and tip (Pl. 26d, Fig. 7-47c). I do not know of another such bone artifact having been reported. Worked Bone, Elephant, Ivory or Tooth EP/MP 1: NP-1591. Sq. D5, L 21, W 2 g, B 8. Worked fragment of black material, possibly elephant ivory, with concentric laminar structure possibly present (identification by Ziegler, see note with NNT-553), L 23.7 mm, Max. W 9.1 mm, Max. T 5.2 mm, one flat face, three rounded faces, distinct diagonal scratches on all faces in same direction but different on each face (Pl. 26e, Fig. 7-47d).


MP 1: NNT-558. Sq. 1E, L 7, W 27 g, -. Worked portion of a hollow long bone, maximum length 33.8 mm, maximum outer diameter 31.7 mm, inner diameter at finished end 17.2 x 15.6 mm, irregular, unfinished end cut-tapered to smaller diameter of 22.0 x 23.7 mm with inner diameter 12.5 x 13.5 mm; inner spongy surface scraped smooth, finished and well smoothed to tapering portion at smaller end, 9.0-11.5 mm wide, not well smoothed, the smoothed outer surface 19.8 to 25.7 mm wide (Fig. 7-47e). NNT-776. Sq. 1F, L 7, W 3 g, -. Broken, curved artifact, called a tooth but probably elephant ivory or dense bone of large mammal (identification by Ziegler), with shiny, mottled surface, diameter 8.8 and 5.8 mm at the two broken ends; fine striations generally parallel to length of bone show under 40 x magnification; no concentric laminar structure obvious (Pl. 26f, Fig. 7-47f). MP 3: NP-811. Sq. C5-D5, L 19X, W 0.2 g, B 11. Fragment of possible needle, probably made from a fruit bat matacarpal (presumably Pteropus sp., commonly known as flying fox; identification by Ziegler); broken at both ends, no indication of working (Pl. 26g). MP 4: NNT-554. Sq. 3F-4F, Level 6B, W 4 g, B 63. Cut, tapering hollow bone, length 61.4 mm maximum width 10.6 mm, tip width 6.4 mm, both ends broken; while edges show cutting, broken tip does not show much use or smoothing; may have been an awl broken after little use (Pl. 26h). MP 6: NNT-553. Sq. 4F-4G, L 6A, 30 g. B 26. Elephant ivory, bone, or wood top (?) with tip broken off, maximum length 48.5 mm, diameter at top 34.9 mm (identification by Ziegler2; Pl. 27a. Fig. 7-47g). MP 8: NP-1032. Sq. C3, L 13, W 2 g, B 47. Fragment of possible elephant ivory (concentric laminar structure possibly present, see note with NNT-553; identification by Ziegler), shiny black, maximum length 21.2 mm, maximum width 10.1 mm, three flat faces parallel to length and two rounded faces; faces (except for one rounded, relatively rough surface) have fine striations, close together, slightly diagonal or perpendicular to length like NP-1591 (Pl. 27b, Fig. 7-48a).


Fig. 7-48


LP 2: NP-763. Sq. C3, L 9, W 4 g, -. Hollow mammal bone, pointed, maximum length 40.9 mm, maximum width 17.7 mm, maximum thickness 5.7 mm, sharp edges of point have battering with small flakes removed, as also the top of point; no wear to show hafting but base of point may be broken off; provenience doubtful (Pl. 27c). LP 4: NNT-110. Sq. 4F, L 4, 2 g, -. Tapered white mammal bone, maximum length 26.8 mm, maximum diameter 10.8 x 9.8 mm, worn tip, one side of tip flattened from one or two flakes broken off but smoothed and rounded over by use, broad base at diagonal (Pl. 27d, Fig. 7-48b). Four of these pieces of worked bone are identifiable as whole or broken artifacts. NNT-558 may be complete but its function is a question. It looks rather like a handle for something that would have fit into the worked hollow center. If so this must have been something organic, because I can think of nothing form Non Nok Tha that would make sense with this as a handle. I know of nothing like from other sites in Southeast Asia. NNT-553 looks very much like a top, with only its tip broken off. Nothing like this is known to my knowledge from other sites in Thailand. NNT-673a is a point that shows damage probably from use, but what that use might have been is not known. Sørensen pictures one that is quite similar (1964:Fig. 9, top row center) from Ban Kao. The form is so simple that it may well be coincidence that they look alike. The tip of NNT-110 was presumably rounded from wear. This suggests that it as used to rub something or to impress a groove in something relatively soft. What this might have been I have no idea. I have noted no other similar tool from other sites. Worked Shell MP 2: NP-308. Sq. C5, L 20, W 3 g, -. Marine bivalve (identification Ziegler), very neatly cut shell, one end 14.9 mm wide and other end 18.8 mm, both sides 21.7 thickness 3.1 mm at wide end and 4.3 mm at narrow end with beginning of hinge giving added thickness (Pl. 27e, Fig. 7-48c). MP 3: NP-846. Sq. E6, L 19, W 2 g, B 53. Fragment possibly worked, Pterida or Isognomidae (mother-of-pearl; identification by Ziegler), five sided, edges uneven, and possibly weathered, three sides concave side to side as surface harder than interior of shell. NP-1548. Sq. G4, L 19, W 47 g, B 38. Maximum length along cut side 91.0 mm, maximum width 48.2 mm; marine bivalve Veneridae or Lucinidae (identification by Ziegler), one side


cut diagonal to thickness of shell, curved edge shows wear smooth and rounded, possibly for thinning inside of plastic clay vessel; fits hand nicely, notch at one end with rough inner surface from removal of small, chunky flakes (Pl. 27f). NP-649. Sq. E3, L 19, W 2 g, -. Length 38.8 mm, maximum width 11.9 mm, maximum thickness 3.7 mm; two relatively fresh breaks have removed most of both original edges; the convex edge is totally the result of the fresh break and the concave portion of the opposite edge results from a fresh break. The face opposite the fresh break of the concave edge has the appearance of a worked edge, with small flakes out along the edge from the fresh break to the tip, and it would appear to have been very sharp and probably used for cutting (Pl. 27g, Fig. 7-48d). MP 5: NP-659. Sq. C4-C5, L 17, W 3 g, -. Waterworn fragment, possibly worked, Mytilidae (identification by Ziegler), maximum length 38.4 mm (Pl. 27h). Two of these worked shells appear to be complete. NP-1548 would be logical as a scraper to even the inner surface of unfired pottery. NP-308, though not perforated for suspension, looks to me like a pendant. With the ridge at the top of the back face, it would be simple to suspend with a cord fastened just below this ridge. I know of nothing like these two artifacts from other sites in Southeast Asia. Notes 1. It is interesting to try to guess the motivation behind this often intense debate; those on the side of an early and relatively independent origin for bronze in the region tend to have been trained in North America or Australia, while those opposed to it have by and large been trained in Europe. It is also tempting to perceive a “colonial supporter of the underdog” bias in the first group (at least in myself) and a subconscious unwillingness to accept Southeast Asian priority over Europe (at least northwestern Europe) in the second, but this may well say more about my own personal bias than anything else! In any case, the whole question of ethnocentric bias and colonialist attitudes toward Southeast Asia has received full coverage elsewhere (Solheim 1967b, Lyons 1977, Gorman and Charoenwongsa 1978, Bayard 1980a, Peterson 1982-1983). 2. Personal communication from Alan C. Ziegler on NNT-553 Elephant Tusk (“Ivory”). This artifact is made of a very fine-grained substance, with an obvious concentric laminar macrostructure. This (I am assuming it is not wood), it must be the distal end of either a horn sheath of a large bovid mammal or an elephant (presumably Elephas maximus) tusk (as opposed to say, limb-bone or other dense osseous material of a large mammal such as elephant, tapir, or bovid). My primary reasons for believing this artifact (and several others in this same lot of “NP” material) are made of ivory are explained below. It this item is of horn sheath, when a thin edge is scraped with a knife it should yield a continuous “shaving” (rather than simply crumbling); and when a tiny fragment is burned, it


should give off a characteristic “burning-hair” smell (rather than remaining essentially odorless). Neither of these two is true of this artifact’s material. It is true that if horn sheath was permineralized it would give these same results but, as far as I’m aware, horn sheath (and other keratinous material such as hooves, rhinoceros “horn”, hair, etc.) never permineralizes like bone and teeth do. The keratinous structures simply disintegrating or macerating leaving no obvious substance in an ancient deposit. Further, in regard to the type of breakage at the smaller end of this artifact, at least fresh horn sheath does not fracture in this manner, simply splintering, instead. (For at lest one of the smaller fragments of artifacts in this “NP” lot, the lack of obvious continuous laminar structure allows the possibility of material from the canines [especially the lower] of various species of pigs [genus Sus], which would be extremely difficult (if not macroscopically impossible) to distinguish from ivory or dense mammal bone, However, if the fragment is large enough in diameter, the small central pulp cavity of the pig canine would usually be apparent along much of the length of this tooth.

Chapter 8 The Burials and Their Implications By Donn Bayard Introduction Although one might argue that Non Nok Tha’s most valuable contribution to our knowledge of the region’s prehistory lay in its demonstration of a non-iron using bronze period of considerable duration, an almost equally strong argument may be made for the large sample of burials it yielded. With well over two hundred burials, it is the largest available from Northeast Thailand (see Table 8-1) and one of the largest yet obtained from Southeast Asia. While all of the burials are not of equal utility (many are incomplete or fragmentary), the wide variation in age and sex of the burials provides a fairly compete picture of changing patterns of burial practice at the site over an estimated two millennia or more. The sample enables comparisons with the three other such corpora published to date; Ban Kao (Sørensen and Hatting 1967, Sangvichien et al. 1969), Ban Na Di (Wiriyaromp 1983; Higham and Kijngam 1984), and Ban Chiang (Pietrusewsky 1978, 1982-1983, n.d.).1 I believe it also allows some tentative insights into the organization of society at Non Nok Tha when some of the methods and theoretical approaches of modern mortuary analysis techniques are employed, although I admittedly have a somewhat skeptical bias toward some of the more wide-ranging theoretical claims involved (Bayard 1969, 1978). It would thus be best to begin with a brief review of approaches to mortuary analysis. Mortuary Analysis In the early 1970s, Binford complained of the “surprising lack of literature in which attempts are made to deal with burials as a distinct class of variable phenomena” (1971:6). Nowadays there would seem to be little cause for complaint; following on the earlier work of Binford and Saxe (1970), mortuary studies have blossomed (e.g., Peebles and Kus 1977, Tainter 1978, King 1978, Chapman et al 1980, O’Shea 1984). However, Southeast Asia has been a late bloomer: Despite a wealth of mortuary data, until the early 1980s quantitative studies had been published only for the Chinese coastal Neolithic (Pearson 1981) and for the site of Ban Kao in Thailand (Macdonald 1978). Macdonald’s doctoral dissertation (1980) also contains an innovative preliminary analysis of the Non Nok Tha burials, albeit based on incomplete data. Higham’s 1983 paper at the Pacific Science Congress on the Ban Na Di burials (Higham 1984d), with my companion paper incorporating some of the analyses and material presented here (Bayard 1984c), and our 1988 contrastive analyses of the Non Nok Tha burials (Higham 1996-1997; Bayard 1996-1997) are more recent contributions. The Ban Na Di burial series has now been published (Higham and Kijngam 1984); full descriptive data are now also available form Khok Phanom Di (Higham and Bannanurag 1990). Such descriptions and analyses—-and the social, economic, and political inferences drawn form them-—are obviously of crucial importance to Southeast Asian archaeology if it is to attain a state of sophistication comparable to that of North America, Europe, and the Pacific and provide further test cases for current theories of sociopolitical evolution.


Despite this, I still feel I should include a cautionary note. Macdonald, Higham, myself and many others have obviously been influenced by the hypotheses on social ranking put forward by Peebles and Kus (1977), among others. However, they rely on ethnographic data from Hawai`i and extensive archaeological data from Moundville as tests for their propositions, and neither our ethnographic nor archeological databases are anywhere near as extensive. Hawaiian society at the time of European contact, a little over two hundred years ago, is well documented (Cordy 1981). Similarly, Peebles and Kus’ Moundville database is immense: fifty years of prior excavation, about 50,000 m2 of excavated area, and a sample of over 3,000 burials to draw on, of which 2,053 were actually used (1977:435). Although Non Nok Tha is one of the larger area excavations carried out to date in Mainland Southeast Asia and has yielded a large sample of burials, its 340 m2 and 217 burials and funerary features are clearly of a different order of magnitude from Moundville, where in any event the presence of social ranking is apparent by simple inspection. These qualms, coupled with an innate skepticism about the universality of such nomothetically oriented studies (Bayard 1968, 1978) and the existence of “whispering potsherds” (Flannery 1976), lead me to caution the reader not to expect a perfect fit of nomothetic hypothesis to archaeological data. But such perfect concordances are a rarity in archaeology (cf. King 1978); a site such as Non Nok Tha, with one-tenth the number of burials and 1% of the excavation area of the Moundville data set, may well produce a more archaeologically realistic test than the atypical abundance of Moundville. Despite its size, the Non Nok Tha corpus-—like all such burial data sets worldwide (cf. O’Shea 1984:27 28)-—has age, sex, differential recovery, and other imbalances in the samples from each level that increase the possibility of sampling error, leading to conclusions not valid for the site or population as a whole; however, I know of no site in Southeast Asia that does not share these problems. Even more important is the degree of confidence we can have in the application of our theoretical constructs. It is obvious that in digging up burials we are not excavating social systems, political organizations, trade networks, or even burial ritual. We are instead recovering a strictly limited and selectively preserved amount of material associated with burial behavior; the rest is merely inference of greater or lesser plausibility, as I have pointed out elsewhere. Nonetheless, the cautious, thoughtful application of such models (or metaphors; Bayard 1992) in Southeast Asia is crucial if we are ever to move “beyond subsistence and dating” (Redman et al. 1978) and bring Southeast Asia into “a meaningful and important place within the history and development of mankind” (Hutterer 1982:567). Despite the warnings given here, Non Nok Tha does appear-—even to my frankly skeptical mind—-to offer some insights into the social organization of prehistoric societies in Thailand. Data and Description of Burial Practices The stratigraphy and ceramic typology presented in Chapters 3, 4, and 6 clearly support the division of the Non Nok Tha sequence into prehistoric Early and Middle Periods, followed after a hiatus by a parahistoric Late occupation period. I referred to these as the Phu Wiang, Non Nok Tha, and Don Sawan Phases in some earlier publications (Bayard


1984b, c; cf. also Higham 1989), but for convenience the Early, Middle, and Late Period terminology of earlier chapters is retained here. It is necessary to say something about the function of the site during the two prehistoric periods. As mentioned in Chapter 6, Macdonald’s valuable and innovative treatment of the Non Nok Tha data (1980) assumes the site was a hamlet, with interments taking place in and around houses and house compounds in a fashion similar to that he proposed for the Ban Kao site (1978). My earlier analyses of the site suggest it was primarily a cemetery, although with some occasional occupational and industrial use (MP 5 in particular). The question is one of obvious importance in attempting to determine the social organization and economy of the two prehistoric periods (cf. Saxe and Gall 1997). I continue to hold to the cemetery argument for the reasons discussed in detail in the conclusions of Chapter 6: 1. The small size of the site (1.1 ha) compared to modern villages (10 to 40 ha), although this is perhaps the weakest of the arguments. 2. The shallowness of cultural deposits. 3. The extreme density of burials, including pits without skeletons but with goods interred (see Table 6-23); frequent intercutting with no apparent concern shown earlier burials also argues against family burials in house compounds as a regular practice (see discussion below). 4. The very low overall density at the site compared with other Thai sites, particularly those spanning both General Periods B and C (again, see Table 6-23). 5. The lack of extensive occupational evidence (with the exception of Middle Period level 5 and possibly MP 7 and 8). 6. The ethnographic parallels suggesting that Southeast Asian societies with the degree of differential wealth that I think was present at Non Nok Tha almost never bury their dead within the village. The last two points deserve some comment. First, one of the readers of the manuscript of this report as submitted for publication found it unlikely that a cemetery would have yielded so many (over 130) pieces of sandstone grinding stones; his plausible conclusion was that those pieces found in burials were accidental inclusions in the grave fill, and the site was hence used equally for occupation and burial. A segment of grindstone would certainly seem to be neither a prestigious nor sentimental grave offering, but the distribution figures on burial versus nonburial finds of grindstone pieces are interesting. In the three Early Period levels, eleven pieces were recovered, six in burials and five in nonburial contexts (1.2 burial to 1 nonburial). MP 1-4 produced sixty-three pieces, nineteen from burials and forty-four from nonburial contexts, a ratio of 1:2.3. By way of contrast, MP 5—-the only MP level that apparently saw relatively extensive occupational as well as cemetery use-—produced two pieces from burials and forty-two from nonburial contexts (ratio 1:21). Eighteen of the nineteen pieces from MP 6-8 were from nonburial contexts, but MP 6 is represented at the site by truncated graves and eroded postholes in the 1968 area and MP 7 and 8 by relatively few burials and larger amounts of possible occupational debris confined to the 1966 excavation area (see second section of Chapter 7 for distribution details). Hence I believe it is unlikely that all of the grindstone finds in EP and MP burials are accidental inclusions. It is more probable that only MP 5-—and possibly MP 7 and 8—-saw any marked occupational use of the portion of the site excavated.


Turning to the second point, as survey of the ethnographic literature suggests that societies which I believe were structured like Early and Middle Period Non Nok Tha (i.e., General Period A and B) rarely inter their dead inside the village; burials take place either in the forest (e.g., Lamet, Hmong; Izikowitz 1979, Lebar et al 1964) or in a discrete cemetery outside the village (Akha, Palaung, etc.; Lebar et al 1964). The Northern Chin (Lehman 1963) are an apparent exception in that burial can take place either within the house compound or in a discreet cemetery outside the village. Hence I feel fairly content that we can assume that Non Nok Tha functioned mainly as a burial place during the Early and Middle Periods, with occasional periods of use for occupation and other purposes. Certainly later General Period C sites in the Phu Wiang area provide apparent examples of within-village burial, at least as evidenced by potsherd densities in the surrounding matrix (e.g., Non Pa Kluai Level IV; Wilen 1989). Although Non Nok Tha was a relatively large area excavation, it covered only 3.1% of this small site; if the density of burials is generally representative of the site as a whole (as Green’s 1965 test pits suggest it is), I would estimate that the mound contains between five and ten thousand burials. The sample is thus a relatively small one, but certainly as representative as those used elsewhere (e.g., O’Shea 1984). The basic mortuary date form the site are presented in Tables 8-1 and 8-2; “empty pit” designates those pits without skeletons but often containing grave goods, while “funerary feature” indicates groupings of vessels, animal remains, and other offerings not associated with a burial but obviously deliberately placed. Given the widespread occurrence of memorial offerings in Southeast Asia today (e.g., the Lamet; Izikowitz 1979), it seems safe to assume a mortuary function for these (perhaps a commemorative one) in the absence of any more plausible explanation.


Predictably, the data are not all of equal utility; in some cases only a small portion of the burial extended into the excavation area, while in others subsequent disturbance removed portions of the burial and/or made its stratigraphic position unclear. Accordingly, the following description of burial typology will be based on a “secure” subset of eighty-nine burials and funerary features (ninety-two if multiple burials are listed separately, as they are in Tables 8-4 through 8-9). Table 8-2 presents details on the number of burials available from both excavations for each of the six chronological groups of burials presented below.2 “Secure” in the sense used here means those burials that are all or largely complete (range 60%-100%, mean 86% complete) and also of definite provenance to the three time periods in Table 8-1; the sample is thus similar to those Ban Na Di burials described as “[Complete]” by Higham and Kijngam (1984:418-420). As the age categories used in classing the 1966 and 1968 skeletons differ, and since the 1966 excavation was able to clear only four of an unknown number of Early Period burials, it is best to treat the burials initially as two separated samples, comprising sixty-one 1968 burials and twenty-eight from the 1966 excavation. The chronological subsets obviously cannot be viewed as fully representative of the Non Nok Tha population (cf. Brooks and Brooks 1983, and in the projected second volume of this report), as children are clearly over represented in the 1968 Early Period sample, while adult sex ratios vary considerably in three of the Middle Period samples. But such variation is of course the case with the overwhelming majority of such samples, and the overall ratios are more satisfactory; I know of no more reliable sample of equal size yet analyzed from Mainland Southeast Asia. 3


In addition to age, sex, and orientation, a number or metrical and nonmetrical attributes were recorded for each burial. These included grave depth and the presence or absence of a mound, to test Tainter’s (1973) hypothesis of the correlation between energy expenditure and social ranking; and the presence or absence of one or more vessels deliberately broken to form a sherd sheet, used by Macdonald as an additional possible rank indicator (1980:244). More importantly, grave goods were tabulated according to quantity and type of vessels present (using the twenty-seven most common of the thirtyeight types as listed in Table 6-21); total number of vessels of all types (in the sense defined in the preceding chapter) present; amount of non-ceramic goods, using seventeen categories of common offerings; and total amount of all goods. The variables are listed in Table 8-3. These present sex and general age ranges for the 1966 skeletons are determined by Pietrusewsky; the Brooks were unable to supply these data for the 1968 skeletons in their report to appear in the projected second volume, and the age and sex determinations used here are those of L. Kobori, a student of C. G. Turner II who corrected some of their earlier estimates of the material. Kobori’s determinations in turn were checked by Turner himself, with a high degree of agreement. Where there was not complete agreement on age, an intermediate value is given; where Turner and Kobori differ on sex, “?” is used.

General Summary of Mortuary Behavior Throughout both prehistoric periods at the site, mortuary behavior involved interment of the corpse in a supine position (only two flexed burials were recovered) in a shallow (30-90 cm) grave, together with a variable amount of goods (ranging from nothing to forty items). Child burials were most common during the Early Period (particularly EP 2), when much of the 1968 area was apparently used as a children’s cemetery. These usually featured one or more cooking or liquid storage vessels (Class 1), and the wealthiest child burial may have had the skin of a pig spread over its body (EP 1 1968 B. 14 Fig, 8-2; cf. Higham 1989:102); three other wealthy graves had low mounds over them containing additional goods. Adult burials (often disarticulated toward the end of the period) also contained cooking vessels, together with a few serving


and storage vessels. Only two funerary features are present. Although continuity is apparent in both burial practice and ceramic inventory, the onset of the Middle Period saw marked changes in ritual and the introduction of a wide range of new vessel types. Mounds topping wealthy graves were common in the early levels (1 and 2) of this period, as were a few wealthy disarticulated burials, indicating some degree of continuity from the end of the former period. Toward the middle of the period (MP 4-6), disarticulation disappeared and mounds were replaced by groups of distinctive vessels placed on top of the filled grave or in the fill itself; these almost always comprised a set of four type 2C and 4D vessels or two 2C or 2L vessels alone. Near the end of the period (MP 7-8) this custom declined, as did the average number of goods per grave. The general pattern of mortuary practice is thus roughly similar to that observed in earlier Thai excavations like Ban Kao (Sørensen and Hatting 1967), and in subsequent excavations at Ban Chiang, Ban Na Di, Khok Phanom Di, and Nong No (Nor). Non Nok Tha burials generally contain a greater number of pots than burials at these other sites but a lesser number of presumed exotic goods, as discussed below. Detailed Description In looking at mortuary behavior through the eleven levels of the Early and Middle Periods, burial practices can be most effectively divided into six chronological groups based on continuity and change. These comprise four periods of relative stability and two apparently transitional periods of relatively short duration that mark noticeable changes in burial practice. As these groups provide evidence for postulated social, economic, and political changes discussed later, this method of discussion seems preferable either to presenting the data in one massive, indigestible block or to a level-to-level account. Tables 8-4 through 8-9 give basic details on each of the six chronological groups using the variables and abbreviations set out in Table 8-3. A full description of each burial and its contents, listed by 1966 and 1968 burial number, and a master register of all 224 burials (i.e., the 217 total in addition to the five double burials and the two 1965 test pit burials) will be found in the projected second volume of this report. The data are presented in full form in order to allow others to arrive at alternative interpretations and develop new hypotheses, as Macdonald (1980) has done with the ceramic data and Higham (1994) has done with the burials themselves. See Chapters 3 and 4 for overall maps of the burials (Figs. 3-6 and 4-1); note that these portray the approximate areas occupied by each burial and its goods and not the overall size of graves (which occupy greater areas at their tops due to inwardly sloping sides).



Group 1: The Early Period (EP 1-EP 2; N secure, 18, N insecure, 18) From the initial use of the cemetery, a general pattern of funerary behavior was set that was to continue for over two millennia: The dead were interred in a supine position in fairly deep graves (45-90 cm in EP 1) and accompanied by offerings of vessels. Some of these contained small fish bones and mammal bone fragments (see Table 6-19) and presumably the remainder also contained food of some sort. Type 1G and 1C vessels were the most common offerings, but 1A and 2B vessels also occurred fairly frequently. Portions of pig skulls and whole pig limbs and segments of cattle limbs were also common offerings; somewhat less common were clamshells (actually freshwater mussel spp. Hyriopsis, Physunio, and Pseudodon; G. Mason, personal communication). Some of the richer burials (usually of children) contained one or more strings or clusters of small disk-shaped shell beads; unfortunately, because of the small size of these beads it has thus far not been possible to determine the species of shell used or whether they were freshwater or marine; given the relatively common occurrence of these beads at inland as well as coastal sites (see comparative discussion below), I would suspect a local origin in the case of Non Nok Tha-—perhaps Chamberlainia, the largest freshwater mussel species in the area. A piece of the thick portion of the lip shell from this species was recovered from an early Middle Period context that appears to have been prepared for industrial use (G. Mason, personal communication). From the onset a fairly wide degree of variation in wealth-—in terms of number of items-— was present, ranging in EP 1 from only one item up to twenty-one items in the case of 1968 B. 14 (a mean of 7.5 items per burial in the eighteen secure burials, or 6.8 if the two multiple burials are counted separately). Sheets of sherds formed by the apparently deliberate smashing of one or more pots occur in several graves; these “sherd sheets” are common features in burials throughout the rest of both prehistoric periods. The placing of additional offerings (pots, portions of animals, and occasionally other items such as stone adzes and unworked stone blocks) in a mound on top of the grave also was practiced in EP 1 (1966 B. 6, 1968 B. 125) and continues through MP 2, with an apparent survival of the custom in the form of one or more vessels placed on top of the filled grave persisting until MP 4 or 5. In some cases it possible to determine stratigraphically that these offerings were covered with an actual mound of soil; this would seem most likely even for those cases where an actual soil mound was not detected, as the animal bones show no evidence of disturbance by scavengers. All skeletons of these two levels appeared to be primary burials, with no sign of the disarticulation found in a number of EP 3 burials. At the time of excavation some of the bones presented a purplish black burned appearance, at it was thought that some individuals may have been partly cremated prior to burial. However, as the Brooks state in their report on the 1968 skeletal material (to appear in the projected volume of specialist reports), no actual charred bones were found, and the discoloration is probably the result of absorption from the surrounding clay matrix. The most distinctive (and visually impressive) features of EP 1 burials are the large, finely made type 1B vessels with elaborately incised, dentate-stamp-filled designs on rim and shoulder; these appear to be limited to the richer EP 1 graves such as 1966 B. 6 (Fig. 8-1) and 1968 B. 14 (Fig. 8-2). The two other poorer graves containing these vessels were in


fact far from complete, and the amount of goods that were recovered form them suggests they were originally rich as well. Although probably of local rather than exotic origin, these vessels represent a considerable labor investment and were probably considered prestige items. Clearly the richest burial recovered from this level is 1968 B. 14, a child of six. It featured not only two 1B vessels and an extensive sherd sheet but also had the skin of a subadult pig spread over the body and second piglet placed with legs around the head. Other child burials richer than those of most adults are present, suggesting that some degree of social ranking or ascribed status was present at the site from its first occupation. As mentioned above, however, adult burials are apparently underrepresented in our Early Period sample. Fig. 8-1

Fig. 8-2


One interesting find in this level was 1968 B. 40, which I assumed from the ribs and arm bones-â&#x20AC;&#x201D;all that remained-â&#x20AC;&#x201D;to be an infant or very young child. This was determined by Kobori to be a nonhuman primate, presumably a gibbon or langur; it was buried with a pot and stone adze. Taken in conjunction with the presence of a dog skeleton in 1968 B. 8 (a child of five), the keeping of pets seems a definite feature of Early Period society. A female on her early thirties, 1968 B. 4 included a 2B vessels from which two fairly large fragments of what was almost certainly bronze were recovered, providing the first tantalizing suggestion of the presence of a bronze-working technology somewhere in the region by EP 1 or 2 times (see discussion in Chapter 7). EP 2 represents a clear continuation of the burial practices of the earlier level. During this period most of the 1968 excavation area was apparently set apart for the burial of children. Small adzes of exotic stone were fairly common inclusions; the richest burial of the level (1968 B. 112; Fig. 8-3) featured three adzes as well as a mound and sherd sheet. Again, this burial (and 1968 B. 121, a child of three) were considerably richer than any adult burial of this level. A pot in B. 121 also contained cranial fragments from a year-old infant; a similar infant pot burial was found with EP 1 1968 B. 91, itself an infant. Fig. 8-3


Group II: The Transition to the Middle Period (EP 3; N secure 10; N insecure, 5)


Early Period level 3 is not represented by a large number of burials, but enough are present to discern marked changes as well as continuities in the funerary tradition. Six of the burials recovered featured skeletons with a marked degree of disarticulation, as if allowed to decompose prior to burial. Attempts appear to have been made to place the skull, long bones, and vertebrae in the grave in approximately correct position, but even in the more tidy burials a considerable amount of disarray is evident (cf. Figs. 8-4 and 8-5). All of the disarticulated burials are adults—-usually older adults-— and all but one are male (1968 B. 52—Fig. 8-6-—was considered male by Kobori but female by Turner). Obviously the sample is insufficient to be conclusive, but it seems possible that the custom was reserved for the burials of adult males only. Fig. 8-4 In terms of quantity of grave goods present, this level is rather impoverished compared to EP 1-2, with a mean of about 4.5 items per grave in the secure burials. Continuity is evident in the placing of portions of pigs and cattle in the grave, along with 1G and 1A pots; however, stone adzes and strings of shell beads are absent, and only one 1C vessel was recovered in contrast to the twenty-six of this type in EP 1-2. The richest burial of the level, 1968 B. 90 (Fig. 8-5), was a fortyyear-old male buried with only eleven items. However, one of these was the earliest bronze artifact recovered from the site, a socketed axe (Fig. 7-25). Consistent dating of this important burial has proved impossible; three bone dates on apatite, collagen, and amino acid fractions of collagen indicate a date of about 950 B.C.--A.D. 400 while a thermoluminescence date on a pot from the burial suggests a date of 2400 B.C. The chronology adopted here prefers the latter to the former, which appears about a millennium too late by the estimates of all but a handful of workers in the field; see Chapter 5 for a full discussion of dating problems.


Fig. 8-5

What does seem clear is that EP 3 represents a period of considerable technological and social change over a relatively short period of time. The changes herald the arrival of the Middle Period, which followed EP 3 with no noticeable break in the sequence. All other pairs of adjacent levels feature at least one case of burials from the later level intersecting a burial from the earlier; no such disturbance occurred between EP 3 and MP 1. The major postulated technological change is of course the introduction of bronze metallurgy; in addition to the bronze axe in 1968 B. 90, two other fragments of bronze were found in the otherwise almost empty pit of 1968 B. 93, and a third fragment in the grave of 1966 B. 26. Along with the appearance of bronze artifacts, spherical clay pellets—- presumably ammunition for a pellet bow (see note at end of Chapter 3-—also make their first appearance in this level.

Indirect evidence for the social changes postulated later in this chapter consists of a number of apparently exotic and unique vessels recovered from five burials of this level. Most distinctive are the clay-tempered, rouletted “spittoons” from 1968 B. 81 and 1966 B. 83 (although the latter burial more likely postdates the Middle Period transition); and the globular, clay-tempered, ringfooted vessels painted in red designs recovered form 1966 B. 26 and 80. None of these vessels have any precursors in terms of decoration of fabric at the site, and the rouletted vessels in particular seem to bear some resemblance to similar vessels recovered form Ban Chiang Early Period II (White 1982a:30; 1986:91, 279). Even more important in my view is the presence of what appear to be prototypes of the very common 2C/E/L and 4D/F vessels of the Middle Period. These occur in 1966 B. 26 (which also contained the red-on-buff painted vessel NP-522 and a small bronze fragment) and in 1968 B. 79 and 81 (see Chapter 6 for a full description of the vessels). None of these vessels have antecedents in the earlier two levels, and nothing approximating them occurs alongside the standardized 2C/E/L and 4D vessels of the Middle Period.


It thus seems plausible to speculate that the two social divisions postulated later in this chapter were coming into being at this time, perhaps stimulated to some extent by the arrival of bronze metallurgy. Enough continuities are present on burial typology to suggest that these modifications were taking place in the same general population that had utilized the site earlier, although this is of course difficult to demonstrate conclusively. What does seem clear is that a considerable amount of external influence reached the Phu Wiang region at this time. Fig. 8-6


Group III: The Early Middle Period (MP 1-MP 2); N secure, 25; N insecure, 27, including B. “X”)


Early Middle Period burials present a marked contrast with those of the Early Period, although a number of continuities are present. These include the frequent use of cattle and pigs as grave goods and the continuing occurrence of 1G pots; these were presumably everyday cooking pots (see Chapter 6). Strings of shell beads and type 1C vessels also occur but are notably less common than in the Early Period. One and perhaps two red-on-buff painted vessels are also present (in 1968 B. 116 and 1966 B. 83. As mentioned above the latter burial is more likely early MP than EP 3 in provenience). Unworked blocks of sandstone (usually Phu Phan formation) weighing about 0.5 to 5 kg are frequent inclusions in and on graves of this period, particularly in the MP 1 burials in the 1968 area, and may have functioned as markers. They were certainly not accidental inclusions, as the nearest source of the material was on the slopes of Phu Wiang, some 5 or more km distant (see Chapter 2). Small blocks of ferruginous siltstone almost certainly used as red pigment were also common inclusions in burials of this group. However, the discontinuities apparent in funerary behavior far outweigh the continuities from the Early Period and suggest a significant amount of social change. To begin with, type 1A pots disappear from burials; instead, a whole new range of vessel types appear alongside the continuing 1G, 2B, and rare 1C pots (see Figs. 6-35 to 38). The most notable among these are the 2C/E, 2L, and 4D vessels of highly standardized manufacture that appear in quantity from the onset of the Middle Period and continue as common grave offerings through MP 6. These vessel types are joined by several other types that appeared sporadically in the Early Period but now increase markedly and continue through Fig. 8-7


the period; squat, globular cord-marked vessels with smoothed shoulders (type 1D) and plain or red-slipped bowls (5A, 5B) are the most common examples. Flat-bottomed jars (types 6A and 6B) also appear for the first time, as do the rarer wholly red-slipped variants of the 2C and 4D vessels (2D and 4E respectively). The elliptical 1E vessels and “giant” 4C variants of the 4D goblets are largely limited to MP 1 and 2; it is tempting to view the latter as the expression of a “bigger is better” trend that did not last long (the single MP 4 example of this type is of doubtful provenience). The beginning of the Middle Period also saw the introduction of bronze casting as well as bronze artifacts in relative abundance (Tables 6-1 and 6-2); I stress the word relative because-—as at other sites of this period-—bronze was always a rare commodity (cf. Higham 1996:242). Hence only six of the forty-eight burials in this group contained bronze or bronze-related items (sandstone molds or mold fragments, crucibles); four of these graves also contained type 2C of 4D vessels, a marked association that continued through the period to MP 8 (see analysis later in this chapter). In addition to bronze, MP 1 saw the first appearance of the pierced biconical clay objects we assume are spindle whorls (two in 1966 B. 8); these continue to occur sporadically in MP 2 and 3 graves (1966 B. 77, 1966 B. 63) and become quite common in MP 4 burials. It seems plausible to add a new fabric technology to the list of innovations entering the region at this time. The burial typology of the early Middle Period represents a marked departure from the earlier groups. In contrast to the essentially single method of burial of EP 1/2 and the articulated/disarticulated dichotomy of EP 3, MP 1 and 2 appear to have employed four varieties of mortuary practice. The most elaborate of these consisted of fairly rich burials in graves ranging from 35 to 80 cm in depth, accompanied with a variety of grave goods. Occasionally additional vessels were placed in the fill of the grave, which was then capped with a mound containing still more pots and other goods (e.g., 1968 B. 103; Fig. 8-7). One MP 2 burial (1966 B. 59) appeared to have been laid on a mat, as fiber impressions were present in the soil underlying the bones; the head of this burial also appeared to have been propped on a cushion or pillow of some type of organic material, as the cranium had dropped back from the mandible. A second sort of burial consisted of graves of about the same depth but with only one or two items placed with the corpse (e.g., 1968 B. 64; Fig. 8-8). Six and possibly seven (counting 1965 test pit burial “X”) secondary burials were also recovered, consisting of the skull and varying amounts of the postcranial skeleton placed in a usually shallow oval pit with an also variable number of goods. These burials ranged from simple arrangements of the skull and long bones on a sherd sheet (1968 B. 111, Fig. 8-9; 1965 B. “X”?) to elaborate and richly accompanied jumbles of bones and grave goods (1968 B. 74, Fig. 8-10); one of these burials (1968 b. 71) contained not only most of the skeleton of a male in his early forties but the skull of a twenty-year-old woman as well. Finally, a large number of deposits of pots and portions of animals were recovered on the surfaces of the MP 1 and 2 layers, which had no burial below them (e.g., 1968 M. 120; Fig. 8-11). As mentioned above, it seems fairly safe to assume a commemorative or other funerary function for these and, for want of a better term, they are referred to as “funerary features.”


Fig. 8-9 Fig. 8-8

Fig. 8-10

Fig. 8-11


What this group of MP 1 and 2 burials seems to represent is not only an increase in overall wealth (up to a mean of 11.0 items per grave, the highest figure of any of the six groups), but a marked increase in the range of grave wealth and funerary practices employed. Paralleling this is the firm establishment of a largely complimentary distribution of 2C/2E/ 4D/4C vessels versus 2L/4F ones, with the former containing almost all bronze and bronze-related goods; the possible significance of this is discussed in the analytical portion of this chapter. As mentioned above, the transition seems to have been a relatively sudden one, and it is possible that only a generation of so passed between 1968 EP 3 B. 90, with the first metal artifact recovered from the site, and 1968 MP 1 B. 103, whose shell beads and large number of 1C vessels give it a rather “archaic” EP appearance. It also seems likely that little time passed between MP 1 and MP 2; indeed, the minor differences between the two levels (apparently shallower graves and the presence of two “horned” pots in 1966 MP 2 B. 77; Fig. 8-12) could reflect differential funerary use of the mound by a contemporary group than any changes through time. Fig. 8-12


Group IV: A Second Transition? (N secure, 5; N insecure, 8)


Burials in this group are confined to the 1966 excavation area and comprise an even smaller sample than those of group II. However, the limited data available suggest that they can be viewed as a transitional period between the wide variety of funerary practices found in group III and the highly standardized burial behavior characteristic of MP 4-6. The “C/L” distinction is evident in this group as well, although only one “L” burial is present; this is not surprising given the tendency of “C” burials to predominate in the 1966 area (at the center of the site) throughout the Middle Period. Burials are somewhat less wealthy (mean = 9.4 items per burial) than those in the preceding group. Continuing the apparent trend in MP 2, graves are shallower (20-45 cm, with two exceeding this range). The practice of mounding goods over the top of the grave has disappeared, but in two cases two pots were placed on or just inside the grave fill instead. “Horned” pots occurred in two burials of this level (e.g., B. 55 Fig. 8-13); it also produced the earliest bronze ornaments at the site (five bracelets on 1966 B. 63). A single disarticulated and apparently secondary burial was recovered (1966 B. 24), the latest one found at the site. There is a tendency for burials to be oriented with head to the west or southwest, a feature to become much more standardized among burials of the following group. Fig. 8-13


Group V: Later Middle Period Stability? (MP 4-MP 6; N secure, 27; N insecure, 58, including B. “Y”)




This is the largest of the six burial groups defined here (total eighty-four) and generally a relatively wealthy one (mean 10.0 items per grave), although it is obvious that grave wealth declined through the three levels from a peak of fifteen items in MP 4 to only five items MP 6. The three levels share a relatively highly standardized burial method with some degree of variation between levels but little variation within each of the three levels. Graves are of medium depth (mostly in the range 40-65 cm), and it seems clear that orientation of burials had become standardized by MP 4 items. The preferred orientation for burials of this level was with head to the southwest; this changed in the succeeding level to preference for a northwest orientation, only to return to a southwest norm in MP 6. Other differences in placing of the body include an MP 4 preference for placing the hands under rather than alongside or on the pelvis and placing the feet very closely together (perhaps tied in some fashion); these traits are rare in burials of the two succeeding levels. Several examples of what may be adjacent burials of relatives are present; these include 1968 B. 32, 39, 47, and 86 during MP 4, B. 59 and 51 (a mother and child?) in MP 5, and MP 6 B. 67, 68, and 69. As mentioned above, the custom of mounding goods over the grave had disappeared by MP 3, but an apparent vestige of the custom continued through at least the first of the three levels considered here. This consisted of placing groups of pots—-in some cases nested together—- not only in the fill of the grave, but on top of the fill at one end or the other of the grave and in some cases over the center of the grave. During MP 4 these were most commonly groups of four pots: a 2C/E and 4D pot placed side by side and a second pair nested together; 2L pots occurred singly or in pairs. The most extreme example of this practice is illustrated by 1968 MP 4 B. 33 (Fig. 8-14), the richest burial recovered form the site: Groups of pots were placed in the fill of the grace and at several locations on top of the grave. An additional four-pot “nest” was placed beyond the head, along with a clamshell (Hyprosis and Physunio spp.); such shells are fairly common finds in MP 4 burials (cf. 1968 B. 39; Fig. 8-15).


Fig. 8-14

Fig. 8-15


Skulls of burials from these levels and in some cases the earlier MP levels often appear to have been originally placed on some sort of a cushion or pad of organic material and to have slipped backward as the cushion decayed. Offerings of pig and bovine limb and skull portions were also fairly common inclusions in MP 4 graves but declined markedly during the two following levels. Several MP 4 burials (1966 B. 19, 1968 B. 32, 37) and an MP 3 burial (1966 B. 63) were found with the skull either missing or disarticulated (upside down, placed to one side of the corpse, or on the corpse’s knees). In some cases (e.g., 1968 B. 32) clamshells were placed where the skull would properly have been and a “nest” of pots placed beyond them. The possibility that such burials might indicate the presence of headhunting must remain speculative, as no obvious signs of trauma were detected by the physical anthropologists examining the material (although this cannot be stated with confidence for the 1968 material, where Brooks supplied no skeleton-by-skeleton description of the material they analyzed), but I find it difficult to believe that these people were totally peaceful. We can at least state that the disarticulation of the skull in these cases was made at the time of the burial, and was not the result of postinterment disturbance. Bronze and bronze-associated artifacts continue to occur sporadically in burials of this group, particularly in the richer burials and those with 2C/E and 4D type vessels. In MP 4 a bronze axe was found with 1968 B. 85, and pairs of bronze and shell bracelets were worn by the five-year-old child in 1968 B. 15. MP 5 produced 1966 B. 31, a young adult female wearing sixteen bronze bracelets on her left arm and another decorated one on her right (Fig. 8-16). A male in his early thirties had a small bronze bangle lying on top of his left wrist (1968 B. 23 Fig. 8-17); this was the only documented occurrence of a bronze artifact in a grave containing type 2L pots. Finally, 1968 MP 6 B. 1 (a male aged thirty-five) had a bronze axe on his chest and a bangle on his right forearm. Several other burials (1968 B. 47, 48, 49, and 59) contained small lumps or fragments of bronze that in two cases appeared to have been deliberately placed in the grave. In terms of bronze-related artifacts, two crucibles were found in addition to the axe in 1968 MP 4 B. 85 (a male in his late thirties), and axe mold fragments were recovered form the fill of 1966 B. 4 and 22; these could, however, have been accidental inclusions. The assignation of the two complete pairs of axe molds to 1966 MP 4 B. 10 must unfortunately remain doubtful. In any event, it is clear that bronze metallurgy continued to flourish locally during this period, although no signs of a specialized, intensive ornament fabrication industry such as at Non Chai Phase IV (Bayard, Charoenwongsa, and Rutnin 1982-1983) or Ban Na Di layer 5 (Higham and Kijngam 1984) were evident at Non Nok Tha. Although a decline in overall wealth is clearly present from Mp 4 to Mp 6, some relatively rich burials were present in the latter level, such as 1968 B. 26 (Fig. 8-18); this burial intersected the only flexed burial encountered in the 1968 excavation, MP 5? B. 21. We are unable to explain why this latter burial-—like 1966 B. 72-—was made in a flexed position. Nor can we explain satisfactory why an apparently normative southwest orientation of burials in MP 4 shifted to northwest in MP 5, only to return to the southwest in MP 6. Given the evidence in Chapter 4 suggesting possible reuse of posts or postholes and hence rebuilding of the structure occupying the eastern portion of the 1968 excavation


Fig. 8-16

Fig. 8-17


during MP 5, it is possible that this part of the site was used sporadically by a related group somewhat distinct form the MP 4 population. It is interesting to note that of the ten MP 5 burials in the 1968 area, four have 2L vessels while only one contained a 2C pot; the distribution was quite different in the more central 1966 area, where of the fifteen MP 5 burials present, four contained 2C pots and none contained 2Ls. This scenario would then view the MP 6 burials as reflecting the reuse of the 1968 area by a population more directly descended from that of MP 4, with a much higher ratio of graves containing 2C and related pots (thirteen out of a total of twenty-one). In any case, it is clear that despite a marked decline in overall wealth during these three levels, the largely complementary distribution of 2C and 2L pots continued to be a significant feature of mortuary behavior. Fig. 8-18


Group VI: The Terminal Middle Period; Decline? (MP 7-MP 8; N secure, 4; N insecure, 8, plus six “pot nests”)


Our evidence for the terminal phase of inhumation burial at Non Nok Tha is unfortunately strictly limited; only thirteen burials (including a double burial), most of them incomplete, were recovered from the MP 7 and 8 levels present in the 1966 area alone. The only evidence from the 1968 area that probably dates from this period is the six so-called pot nests seemingly placed in shallow pits in the subsequently eroded surface of 1968 layer 6. Even this limited evidence, however, suggests that significant changes took place in mortuary behavior, chief among them an overall simplification of burials in material terms. Graves appear to be somewhat shallower (ca. 30-35 cm), but this is based on only four cases. The number and quality of grave goods declines (e.g., 1966 MP 7 B. 46, Fig. 8-19; B. 42, 43, 44, Fig. 8-20), continuing the trend of MP 4-6; the four secure burials in this group average only four items per grave. Only two burials contained 2C and 2L pots (one and two pots respectively), and only one piece of bronze was recovered in a mortuary context, a possible bronze object in the right hand of MP 8 B. 40, an elderly female (also the only burial in this group to contain a 2C pot; Fig. 8-21). What we seem to have is an acceleration of the decline in wealth seen during MP 4-6 and a probable loss of whatever social significance was marked by the 2C/2L pottery distinction. Fig. 8-19


Fig. 8-20

Fig. 8-21


The pot nests in the 1968 area seem to represent a new funerary practice quite distinct from the placing of pairs of 2C and 4D vessels on top of graves, which in any event had ceased to be common by MP 5 times. The softer, more friable sand-tempered fabric of vessels in some of these groups (particularly C, E, and F) suggests some changes in local ceramic technology, and the clearly exotic form and fabric of the vessels making up Pot Nest B suggest influences entering the Phu Wiang area, probably form the east. This speculation is strengthened by a handful of mainly red-slipped and red-on-buff painted sherds recovered from MP 7 and 8 nonburial contexts that are quite reminiscent of material from sites such as Non Chai (Bayard, Charoenwongsa, and Rutnin 1978-1983) and Ban Chang Hian (Chantaratiyakarn 1983, 1984). Even to the non-ceramicist, these appear to be foreign to the local ceramic tradition in both fabric and decoration. A ceramicist, Brian Vincent, kindly reported on a dozen of these sherds (to appear in the volume of specialist reports), and some ten or eleven of the twelve sherds are probably from areas other than the Phetchabun piedmont region: in particular the lower and middle Chi basins. Significantly, none of the pot nests contained 2C/E, 4D or 2L vessels, and shortly after their deposition and the handful of MP 8 burials in the central area of the site, use of the site for inhumation burial (or it seems any other activity) ceased for a considerable time. Late Period Cremation Burials When use of the site began again at the end of the first millennium A.D., it is clear that cremation had replaced inhumation as the dominant form of burial, with burial some of the ashes in a pottery vessel a fairly common practice during much of the Late Period. Descriptions of the Late Period cremation have already been given by Parker in Chapter 3; we need only note that the central area of the site was obviously favored for such burials, as they were rare in the 1968 area. Also, as noted in Chapter 6, mining of the MP inhumation burials to obtain vessels for cinerary used was also common and took place in both excavation areas during the LP. It may be that the discovery of the earlier MP inhumation burials and the availability of the vessels led to the initial use of the site for burial of cinerary remains. Analysis The goals of the analysis of this body of burial data are the usual straightforward ones: to ascertain variation within the mortuary data that may give insight into the social structure of those interred; and to attempt to discern changes in funerary behavior through time, again allowing plausible inferences to be made about social change. In doing this, I relied on a number of techniques: simple histograms of burial wealth vs. period, age, and sex; correlation coefficients of grave inclusions with each other as well as with sex and age of the individual interred; chi-square analysis to test the significance of spatial distributions; and factor analysis, or more properly a subtype of the technique called principal X components analysis (i.e., the PA1 option in the SPSS package routines employed here). As stated at the beginning of this chapter, I remain skeptical about the ability of such techniques to allow us to arrive at definitive answers. On the other hand, it would obviously be sheer nonsense to discard quantitative methodology in favor of intuitive interpretation, and I have in fact employed these techniques to great benefit, not only with


the Non Nok Tha analyses carried out in the early 1980s and more recently, but with sociolinguistic and accent attitudinal research that I began in 1984. The analyses presented here concentrate mainly on the Middle Period burials for several reasons. First, there is current debate about the transition to bronze, the social organization and residence pattern that followed, and the duration of the Middle Period at the site (Higham 1996-1997; Bayard 1996-1997), which is relevant to consideration of other General Period B sites in the region. Secondly, the Early Period burials at the site are relatively few in number and were not distinguished as such in the 1966 excavation. Thirdly, I believe that the ceramic analyses already presented in Chapter 6 provide sufficiently convincing evidence for the reality of the Early/Middle Period dichotomy. Finally, I believe analyses concentrating on the Middle Period burials may well tell us something about the social structure of those making the interments. As mentioned above, it was noted while completing the final listing of burial contents that the type 2C, 2E and 2L vessels so common in Middle Period burials were very largely in complementary distribution; graves with 2C or 2E pots rarely contained 2Ls, and vice verse, and I believe I can demonstrate that this pattern was not coincidental. For the purposes of detailed analysis I have usually concentrated on a sample of sixty-one secure Middle Period burials from both seasons’ excavations. The secure subset is used in preference to the full set of 164 Middle Period burials (excluding double burials, burials not lifted, and the two burials in the 1965 test pit) because of the obvious likelihood that some of the burials that were mostly outside the excavation area also contained unexcavated “C” or “L” vessels. I have also divided these into “rich” (>14 items in the grave) and “poor” (<15 items), using 14/15 items per burial as an arbitrary but plausible boundary between “rich” and “poor” burials when a histogram of total MP grave contents is examined (Table 8-10, Fig. 8-22). Such a division is probably not statistically valid (Orton and Hodson 1981) but is of some heuristic value in breaking up what appears to be a continuum from “poor” to “rich.”


Fig. 8-22

Given these criteria, the distribution of C and L vessels within this sample does indeed appear to be noncoincidential. In the combined 1966 and 1968 secure subsets, C vs. L presence produces a chi-square value of 12.74 (df = 3, Yates correction applied, p <.01). A simple two-by-two matrix incorporating secure and insecure burials from both areas (“rich”: 8 Cs, 1 L; “poor”: 10 Cs, 19 Ls) produces a phi value of -.228, x2 of 3.07 and a onetailed significance of less than .05. Also, although the distinctive vessel types comprise only four of the forty-three funerary attributes considered, they make up one-third of the vessels of secure provenance in the Middle Period sample. The very common type 4D pots are almost wholly limited to graves with 2C/E vessels. Moreover, two of the three type of 4F pots, which parallel 2Ls as 4Ds do 2C/Es, co-occur with 2L vessels. But it is of course necessary to attempt some explanation of the anomalous burials that contain both C and L vessels, although there are only eight of these out of the 164 Middle Period burials (five of the sixty-one secure MP burials). Most appear to have been rich or fairly rich burials, although in two cases only 10% of the grave was present in the excavation area. Two of the burials involved skeletal material from two individuals; the most clear-cut case is 1968 Burial 71, which contained the disarticulated skeleton of a man over forty and the skull of a woman of twenty along with seven C vessels and two L vessels. It would seem probable that this and the other burials represent special cases involving multiple interment or otherwise meriting vessels of both types. My present conclusion is that the largely complementary distribution of these vessels may reflect the presence of two distinct affiliative groups in Middle Period society, each containing individuals of differing wealth; this deliberately vague term is used in preference to clan or lineage because of my belief that kin systems cannot be excavated, they must be inferred (see discussion below). It seems obvious that inclusion of C or L vessels in the grave was not a necessary part of the burial ritual, as about a third of the poorer secure burials (nineteen out of the forty-six burials with less than fifteen inclusions) lack either; however, they were certainly viewed as desirable inclusions, since all of the fifteen secure wealthy burials contain one or more of them (1966 Burial 77 contained “giant” type 4C vessel rather than 4Ds, 2Cs, or 2Es). Examination of the histogram of grave wealth (Fig. 8-22) makes it apparent that the two groups considered as a whole were of unequal average wealth; nine of the fifteen secure rich burials contained C vessels, as against only two with L vessels and four with both types. The total amount of grave goods is significantly correlated with the presence of 2C


vessels (r = +.63) and the very similar 2Es (+.57); and 4Ds (+.61); no significant correlation is present between total wealth and 2L vessels (r = +.12; see R-mode factor analysis below). The actual means for total number of grave goods in the sample of sixty-one secure burials are: â&#x20AC;˘ C group (23 burials), 12.8 items; â&#x20AC;˘ L group (14 burials), 10.6 items; â&#x20AC;˘ Both C and L inclusions (5 burials), 17.6 items. Moreover, luxury goods (bronze implements and ornaments, beads, etc.) are largely confined to C burials; only 1968 Burial 23 and Funerary Feature 114 contained bronze or bronze-associated artifacts together with L vessels alone (1968 B. 75 had two L vessels and what appeared to be a deliberately placed bronze fragment and 1966 B. 20 and 1968 B. 97 had bronze associated artifacts and both types of vessels). Turning now to variation in grave goods as a whole, Figs. 8-23 through 25 present crosstabulations of amount of goods with age and sex of the burial for secure burials from the 1966 excavation; 1968 secure Early Period burials; and 1968 secure Middle Period burials. It seems clear that we are not dealing with an egalitarian society. Wealth distinctions clearly crosscut those of age and sex, as predicted for agricultural societies by Binford (1971:20). Differentiation is more marked in the Middle Period burials but still apparent in those of the Early Period. Adult male graves of both periods were only slightly richer than female graves (9.96 items/grave vs. 9.54); graves of children were somewhat poorer (6.95 items). Overall mean wealth increased from 6.25 items per grave in the Early Period to 10.8 items in the 1968 Middle Period sample (mean for all 1966 burials is 9.7). Fig. 8-23


Fig. 8-24

Fig. 8-25

It is apparent that there is a positive correlation between overall wealth of a grave and presence of a mound or group of pots on top of the grave, supporting Macdonaldâ&#x20AC;&#x2122;s use of this feature as a rank indicator. However, there is a lack of correlation between wealth and the presence of sherd sheet(s); the latter appear to be a part of the general burial ritual that crosscuts wealth distinctions. There is a slight positive correlation between grave depth and total wealth, but it is of insufficient significance to lend strong support to Tainterâ&#x20AC;&#x2122;s hypothesis; no correlation exists between grave depth and presence of a mound or sherd sheet or between age/sex and depth of grave.


Multivariate Analyses The multivariate methods employed in the preliminary analysis of the Non Nok Tha mortuary data (Bayard 1984c) were in some respects a simplistic version of those used by O’Shea in his 1984 book on mortuary variability, which appeared shortly after my publication. O’Shea’s work is of course a more general attempt to test a theoretical framework of mortuary analysis using six case studies (i.e., cemeteries), and he also relies in principal components analysis to attempt to ascertain patterns in grave inclusions. Two of his goals are obviously congruent with those sought here: to consider “mortuary differentiation as an expression of social distinctions in the living society”, and the “variation in a society’s funerary treatments through time.” His third goal is an attempt to discern ethnic distinctions as expressed in the mortuary record (O’Shea 1984:50). There are, however, several important differences in the databases used by O’Shea and that employed here. O’Shea is concerned with social and ethnic variability as expressed in the graves of some 460 eastern Great Plains Native Americans affiliated to three tribes (Pawnee, Omaha, and Arikara) in six cemeteries generally spanning the period A.D. 1750 to 1850 (but with one cemetery perhaps in use as early as A.D. 1680; O’Shea 1984:58, 60) during a period when the societies in question were undergoing massive social and demographic disintegration as a result of Western invaders and their diseases. Ethnographic documentation is available for all three societies during the period in question; although certainly of uneven quality, it is sufficient to provide “an independent check on the conclusions derived form archaeological analyses” (O’Shea 1984:50). By way of contrast, at Non Nok Tha we are confronted with a sample of perhaps 5% of a cemetery (if we assume that the cemetery covered most if not all of the present mound, which Green’s 1965 test pits imply is the case) with 217 burials and funerary features, spanning at least one and half millennia and possibly as many as three; the earliest ethnographic accounts of the area date from the nineteenth century, some two thousand or more years after the last interment at Non Nok Tha. Finally, O’Shea’s work is an attempt to test a body of theory; this report is not. Nonetheless, there are certainly parallels in the methodological approaches adopted here and those used by O’Shea. I rely on factor analysis (a shorter and I think more preferable term than principal components analysis, once qualified as at the beginning of this section) in my ceramic analysis in Chapter 6. I also presented a Q-mode factor analysis of secure Middle Period Non Nok Tha burials at the 1983 Pacific Science Congress (Bayard 1984:104), which I present again here. But there is one important methodological criticism of both O’Shea’s and my factor analyses that should be mentioned: It is generally agreed among most social scientists (at least among the social psychology of language colleagues with whom I have been associating since beginning research in this field in 1986) that factor analyses should ideally employ a ratio of about ten cases and in practice five cases (at least for explanatory analysis) for every one variable included in the factor analysis. Moreover, sample (i.e., case) sizes should ideally be well over fifty; one hundred to two hundred is viewed by some as the minimum for a good sample (cf. Tabachnick and Fidell 1989:603). I have certainly violated these ideal criteria in several of the analyses used in my accent attitudinal studies (Bayard 1991a, b), but in general I try for a ratio of


one variable to five or more cases (Wilson and Bayard 1992:38, fn. 7) and sample sizes exceeding fifty. Both of these goals, of course, are much to difficult to achieve in archaeology (with a sample of burials, artifacts, etc., limited by time, budget constraints, and sheer chance) than with living subjects. As might be expected, the factor analyses presented here fail to meet this ideal 1:10 criterion, as do indeed all of O’Shea’s 1984 analyses, which have variables-to-cases-ratios ranging from a maximum of 1:5.7 down to as low as 1:1.36. The R-mode analysis of grave contents presented below has a ratio of 1:2.65, higher than four of O’Shea’s six samples. Given this proviso, I believe that factor analysis and the other techniques employed above can provide a highly suggestive—-if not definitive--picture of the relationship between the variables involved, be they individual burials or the contents of graves. I also think that they provide support for the interpretation presented here and may also encourage further alternative interpretations like those already presented by Macdonald (1980) and Higham (1994; see below). Turning to the factor analyses themselves, the 1984 Q-mode factor analysis (treating the fifty-nine burials included as variables and the forty-three artifact types as cases) 4 mentioned above is reproduced here as Fig. 8-26 . This analysis confirmed the C/L dichotomy but can be taken as suggestive only because it employed fifty-nine burials as variables but only forty-three artifact “cases”: a variable-to-case ratio of only 1:0.72. Factor 1 in this analysis (43.4% of the variance subsumed by the first three factors but 16.5% of the total variance) seems clearly to be a result of chronological variation in the occurrence of type 1D vessels, which diminish on frequency during the Middle Period (see Chapter 6) and are not in complementary distribution with any other types. Fig. 8-26 For example, a chi-square test comparing presence/absence of 1D vessels with presence of 2C/ E, 2L, and 4D vessels produces a value of 0.694 (df 2, p > 0.10); an examination of presence/ absence of 1Ds vs. 4Ds in graves yields a similar result (X2 = 0.624 [df 1, p > 0.10]). The distribution of 1D vessels also fails to correlate significantly with other grave goods, save for some degree of association with 2L and 5A vessels (particularly in 1968 MP 1 burials). In other words, this factor results form diachronic variation within the Middle Period rather than the apparent long-term contrast between C and L burials; of forty-


three type 1D vessels in the sixty-one secure MP subset, twenty-nine (67.4%) originated in MP 1 and MP 2 burials. Factors 2 and 3 (as illustrated, 12.6% and 9.3% of total variance respectively) are clearly related to the presence of L and C vessels respectively. However, these factors are clearly also not what O’Shea refers to as either “bipolar,” with a set of variables loading strongly either positively or negatively on a single factor; or “grouped,” with many variables clustering strongly at either positive or negative values of a single factor (O’Shea 1984:66). Instead, some variables (i.e., individual burials) load more positively in either Factor 2 (L; e.g., B. 111, 63) or Factor 3 (C; B. 33, 69, etc.) than others, with a few burials loading fairly strongly in both; what is apparent by inspection is that the degree of positive loading is due to the number of C type or L type vessels present (as indeed is the case with Factor 1, 1D type vessels). This is of course not very enlightening; what is of interest is the relationship of these vessels to overall wealth and inclusion of nonlocal goods such as bronze. As stated above, 14/15 items per burial seems an arbitrary but plausible boundary between “rich” and “poor” burials. If we accept this and examine this subset of MP burials in terms of wealth as well as age and sex, it seems apparent from Figs. 8-23 and 8-25 that the largely mutually exclusive C and L groups do not have a dominant relationship to either age, sex, or overall wealth of the person interred; these vessel types were clearly not reserved for the well-off alone. Moreover, the first three factors generated in this analysis subsumed only 38% of the total variance present in the data (see note 4); the fact that a total of seventeen factors with eigenvalues greater than 1.0 were produced indicates the presence of many other sources of variation in the burials, most of them probably due to the co-occurrence of two or more uncommon sorts of inclusions (e.g., bronze axes or molds, shell bracelets or disc beads, type 1R, 4A, or 6E vessels, etc.) by chance alone (cf. O’Shea 1984:66). Clearly it is necessary to examine the relationships between the cooccurrence of the more common burial inclusions themselves rather than the burials containing them. This is easily done using an R-mode analysis of the burials, with the contents categories as variable and the sixty-one burials as cases. This is of course the method O’Shea employed in this 1984 study, using two types of cluster analysis to group the burials themselves; however, I generally dislike using cluster analysis (although I employ it occasionally in this report — e.g., Fig. 6-34), as it is all too easy to enable the computer to enforce a dendrogram on data that-—with the omission or addition of a few cases or variables-—could produce a very different picture (cf. discussion of Pietrusewsky’s analyses of Non Nok Tha skeletal material below). Given the problem of spurious correlation due to random co-occurrence of rare grave goods categories, an R-mode analysis should be limited to only relatively common categories of inclusions; O’Shea in most cases employed an arbitrary cutoff point, including only those artifact types that occurred in 5% or more of the burials in the sample (1984:66). I have taken a similar approach here, discarding relatively rare grave inclusions in favor of nineteen of the most common vessel types and non-ceramic inclusions. I have also totaled bronze-associated artifacts (axes, bracelets, molds, and crucibles) into a single “bronze” variable. All of these twenty variables occur in at least seven burials in the


sixty-one-burial data set (i.e., in 11.4% of the sample; cf. O’Shea’s 5% criterion) excepting bovine skulls (six occurrences, but included to test relationships with occurrence of bovine limbs); thirteen of the twenty traits occur in ten or more burials. The traits employed in this analysis cover 249 of the total of 325 occurrences of all traits in the data, or 77%. To establish any relationship between the specific grave goods categories and overall grave wealth, a total goods variable was also included, bringing to total of variables to twenty-one (variable/case ratio 1:2.95). The results of the R-mode analysis are graphed in Fig. 8-27 and provide a more coherent picture than that resulting from the Q-mode analysis. Only seven factors opposed to seventeen were generated with eigenvalues greater than 1, and these subsumed 70.4% of the total variance. Factor 1 (22.7% of variance) is clearly a “grouped” factor in O’Shea’s terms, with 4D, 2C, and 2E vessels loading very strongly at the positive end of the factor (Table 8-11). Total grave wealth also loads more strongly (+0.63) on this factor than on any other. Factor 2 (12.6%) is less strongly defined but is obviously related to the presence of 2L and 5A vessels, with 1D vessels also loading fairly high. Factor 3 might be


called a “utilitarian offering” factor, with grinding stones and type 1G (ordinary cooking?) vessels loading fairly highly; Factor 4 includes pig legs and type 1I vessels. Factor 5 suggests a clear and unsurprising association between the presence of bovine skulls and limbs, while Factor 6 contains pig skulls, type 5B and 6C vessels, and red pigment. Factor 7 is the only one approaching what O’Shea calls “bipolar,” with sandstone pieces loading fairly strongly on the positive side and total bronze-associated artifacts loading weakly on the negative size. The almost equally high negative loading for bronze-associated goods on Factor 3 suggests that bronze-associated goods were not usual inclusions in graves containing what may have been “low prestige” items like 1G vessels or pieces of sandstone. Fig. 8-27

To further investigate the interaction of these factors with variables like age, sex, bronze-associated goods, and overall grave wealth I used the technique of Manova analyses of factor scores, a method commonly employed in social psychology of language studies (e.g., Bayard 1991a). This involves submitting the scores of the individual cases (i.e., burials) on each of the seven factors to multivariate and univariate analyses of variance to test for significant effects of the more general variables like total wealth (rich/poor, using the 14/15 boundary discussed above) and presence or absence of bronze-associated goods on the cases as a whole (and on subgroups like children, females, and males within cases).

The overall effect of grave wealth on all seven factor scores proved to be highly significant (F = 9/69 [df7,51] p = 0.000); univariate analyses of scores on the seven individual factors indicate highly significant effects of wealth on Factor 1 (C vessels: F = 17.46 [df1,57] p = 0.001); and on Factor 5 (bovine remains: F = 15.81 [df1,57] p = 0.000). Amount of bronzeassociated goods also had a significant overall effect on the seven factors (F = 4.12 [df7,51] p = 0.001), and on individual Factors 3 (grindstones 1G vessels: F = 12.72 [df1,57 p = 0.001); and 4 (pig legs, 1I vessels; F = 4.51 [df1,57] p = 0.038). The Factor 3 result confirms the tendency for 1G vessels and bronze goods not to co-occur. Similarly, individual Manova analyses of Factor 1 and Factor 2 scores by rich/poor and bronzeassociated goods/no bronze goods produced significant results (but only marginally for the bronze variable) with the C Factor 1, but no significant effect with the L Factor 2. Finally, the mean C and L factor score diagrams I constructed for the two general variables for children, females and males (the last two in four age decades from 11-20 to 40+) exhibited very clear distinctions in wealth for scores on Factor 1 (C vessels included in


burial). “Rich” burials scored positively in this factor (means from +0.10 to +4.60); “poor” burials scored more negatively (+0.05 to –0.40). No such pattern was apparent in interactions in Factor 2 (L vessels). I would conclude that these analyses provide a further validation of the hypothesis that C-affiliated burials tend to be richer and contain more bronze-associated goods than L-affiliated ones; or conversely, that rich burials score considerably more positively in the C Factor 1 than they do in the L Factor 2. In summary, I think the data and analyses lend qualified support to the presence of both superordinate and subordinate dimensions of ordering (Peebles and Kus 1977:431) during at least the Middle Period. Child burials are indeed poorer than adults and female burials slightly less rich than males, providing a weak degree of subordinate ordering. However, “rich” burials include men (eight), women (six), and children (two) alike. Peebles and Kus have of course pointed out that the presence of rich child burials does not necessarily indicate the presence of ranking; such wealth may reflect parental affection rather than ascribed status. However, as Macdonald notes, the presence of rich child graves “has at least suggestive utility”; he also points out that the presence of rich adolescent burials would be more useful in deciding whether some degree of ascriptive ranking was present (1980:222). While adolescents may be underrepresented in the sample, 1968 Burial 33 (Fig. 8-14) certainly suggests the presence of some type of superordinate ranking, as do the two very rich female burials in the 1966 sample (B. 31 and 8). However, the overall ordering is about one “rich” to three “ordinary” burials in the secure subset; the apical very rich class postulated by Peebles and Kus, containing mainly adult males, does not appear to be present. Moreover, there is no evidence for ranking in the restricted sense used by Peebles and Kus in their study (422); no clearly demarcated boundaries between ranks are apparent, nor are obvious symbols or badges of rank in evidence. Hence it seems likely that-—barring surveyed area--we are dealing with a degree of social ranking considerably less than the apparent stratification exhibited at Moundville or the Dian sites (von Dewall 1967, Higham 1996:142-182). We do, however, have the additional C/L dimension during the Middle Period; this may be viewed as subordinate in Peebles and Kus’ terminology, since it crosscuts age, sex, and to some extent wealth distinctions, but it is better described by Tainter’s term horizontal, encompassing “structural components that equate on identical hierarchical levels” (1978:131). In the Middle Period context, this suggests the presence of two affiliative groups, each of which was internally variable in wealth, but one of which contained many more wealthy individuals—-that is, “ranked” in the general sense but clearly not stratified. Local Versus Exotic Goods Thus far we have been treating grave goods and total wealth as if all items included in a grave were of approximately equal value or significance. Obviously this was almost certainly not the case; if we assume the law of least effort was operational at Non Nok Tha and that Tainter’s hypothesis is at least approximately correct, it seems clear that some items were more valuable or prestigious than others. A type 1B pot or a bovine limb almost certainly represented a greater labor investment than a common 1G cooking vessel or a 1 kg block of unworked sandstone—-and bronze axes and bangles even more so. In their analysis of the Ban Na Di burials, Higham and Kijngam, following Dalton


(1977), have emphasized the presumed importance of such exotic “primitive valuables” in cementing and maintaining kin group and other alliances (1984:714-715). Of course, as they also state, this does not “rule out the possibility that artifacts made from locally available raw materials are also indicative of status” (438-439); the 1B vessels of EP 1 may well be an example. The Non Nok Tha burials in general contain a smaller number of clearly exotic goods than those at Ban Na Di. No Trochus artifacts were found and only a handful of items of marine shell were recovered from burial contexts: a cowrie shell from 1966 MP 1? B. 79 and three beads (possibly of marine shell) and two bracelets from 1968 MP 4 B. 15; pieces of worked or possibly worked marine shell identified by A. Ziegler as belonging to the families Veneridae or Lucidinae and Pteridae or Isognomidae respectively were recovered from 1966 MP 3 B. 38 and MP 3 B. 53 (see Chapter 7). One of the bracelets was examined by G. M. Mason, the Otago Anthropology Department malacologist, who believes it to be Tridachna or another marine species (Mason, personal communication). Only about a quarter of the shell disk beads at Ban Na Di were present in the much larger number of burials at Non Nok Tha: ca. 2,800 vs. 12,000 at Ban Na Di (also cf. the total of 120,000 disk beads found in a single burial at Khok Phanom Di; Bannanurag 1991:206). Not a single intact stone bracelet was found, although numerous fragments (sixty-nine) were present both in grave fills and nonburial contexts. It seems likely that Non Nok Tha was farther “down the line”-—that is, presumably upstream (Bronson 1977)-—in the system of trade links that was obviously present in the region from the time of earliest use of the site. Despite this, bronze artifacts are slightly more common at Non Nok Tha than at Ban Na Di. Still, it would be wise to consider the question of “exoticness” in a bit more detail; clearly one person’s (or culture’s) exotica may well be another’s rubbish, and we will never know exactly what criteria were employed by the population at Non Nok Tha. The blocks of unworked sandstone apparently used as grave markers in the early Middle Period were “exotic” insofar as they had to be transported several km from the slopes of Phu Wiang, but I doubt that they were viewed as “primitive valuables”! The sources of the large aquatic clamshells (Hyriopsis, Physunio, spp.) included in burials probably lay somewhat farther afield in the Phong River or possibly even the Chi. Obviously, some of the items in the burials were clearly exotic, such as stone adzes and bracelet segments (presumably from the Phetchabun Range some 40 km to the west), the pair of presumed Tridacna bracelets recovered from MP 4 B. 15, and the bronze used to cast axes and bracelets. Others, such as the axe molds and shell disk beads, could either have been fabricated locally or made outside the Phu Wiang region; the Phu Krad’ung and Phra Wihan sandstone used in mold manufacture has a widespread distribution around the margins of the Khorat Plateau (see Fig. 2-2), and it has not yet been possible to determine whether the shell used to make the disk beads found at Non Nok Tha and other sites is marine or aquatic (G. M. Mason, personal communication). Hence the definition of “exotic” grave goods must necessarily remain rather subjective. But it would obviously be desirable to test both Higham and Kijngam’s conclusions on differential areal use of the Ban Na Di cemetery by unequally ranked lineages and my hypothesis on the C and L affiliative groups by examining the distribution of exotic goods in both mortuary assemblages. To do this I have assumed the following classes of artifact to


be exotic: metal items (both bronze and iron in the case of Ban Na Di); shell disk beads and bangles; adzes and bracelets of nonlocal stone; stone molds; and pots with a high probability of exotic origin in terms of decoration and fabric. The results generally support both hypothesis (Table 8-12). The relative scarceness of exotic goods at Non Nok Tha probably accounts for the lower overall correlation between total grave wealth and exotic goods at that site relative to Ban Na Di, but even at the former site the results are highly significant. Both of the postulated affiliative systems exhibit significant correlations with total exotic goods and total wealth, although the correlation between exotic goods and “poor” (area A) and “rich” (area B) groups at Ban Na Di is only marginally significant at .05 or .01 levels even when only the “complete” ( = my “secure”) burials are used; this is of course due to the small sample of “complete” burials at the site (thirteen in area A and five in area B). I would not wish to emphasize the importance of exotic goods to the exclusion of total amount of effort employed in a burial and the goods therein, which seems to be a more general overall index of wealth. Nonetheless, it seems clear that a case for differentially wealthy burial groups can be made at both sites using both criteria; whether these groups reflect lineages, clans, residential units, or the like is of course quite another question.


Spatial Variability Fig. 8-28a

Fig. 8-28b


Fig. 8-29a

Fig. 8-29b


If two distinct social groups were present during the Middle Period-—and this is admittedly debatable (see discussion of Higham’s 1988 reinterpretation below)-—we might expect to see this mirrored in the spatial distribution of the burials as at Ban Na Di (also cf. Goldstein 1981. King 1978). There is in fact some evidence for this during the Middle Period, in that L burials appear to be mainly confined to the outer portions of the site (i.e., the 1968 area), with its center (the 1966 area) containing more mounds and significantly higher proportion of C burials: 1966 area: thirteen C burials, one L; 1968 area: 10 Cs, thirteen Ls. X2 = 7.04 (Yates corr.), df 1, p <.01 (Figs 8-28, 8-29). The northeastern portion of the 1968 area seems to have been reserved for L burials, although some degree of overlap exists. By the latter half of the period, the center of density of burials appears to have shifted to the north and less separation of C and L is evident. Instead they are intermingled as are richer and poorer burials. Yet, some degree of patterning does exist in that many of the burials are in more or less parallel rows, lying side-by-side with head to the southwest; this is particularly marked in the 1968 area, although two small groups are present in the 1966 area as well. In several cases sexes are paired, and it would be tempting to view these as family assemblages (cf. discussion under Group V burials above) and search for evidence of exogamy (e.g., male Cs, female Ls and children with either C or L); however, no clear pattern exists. Nor do any of the rows appear to be richer than others, although again the C burials are markedly richer overall. It would appear that Ls dominate the northeastoriented scatter of burials in the 1968 area, and these are also out of alignment with the predominantly C rows; however, the former date from the only level of this period (MP 5) to feature significant occupational evidence. As mentioned above, they may represent a brief habitation by an L-affiliated group between periods of intensive cemetery use by a Caffiliated group or groups. Discussion If we accept that some degree of social ranking in the general sense did exist from the first use of Non Nok Tha and that this grew more marked with the transition to the Middle Period proper, it is necessary to consider first the degree of such ranking in the Early Period and then the origins and nature of the increased ranking in the subsequent period. The presence of ranking per se in an agricultural society is hardly surprising, as egalitarian food producers are very rare; indeed, it has become apparent that many collecting/hunting societies are also ranked (King 1978). The degree of ranking in both periods at the site is of course not great when compared to Hawai`i or Moundville and probably represents something like a “simple chiefdom” in Service’s terminology, or what Cordy has termed a “simple-rank” society (1981:42). The presence of elaborate burials of both adults and children from the beginning of the Early Period (e.g., 1966 B. 6 and 1968 B. 14) could suggest the presence of ascribed status. Given the widespread occurrence in Mainland Southeast Asia of societies based on villages made of local exogamous clan segments, it is tempting to postulate a similar system here, with higher rank devolving on the most senior of the local segments and chieftainship on its leader—-roughly speaking, a society somewhat more ranked than the modern Lamet (Izikowitz 1979) but less so than the gumsa Kachin (Leach 1964).


Both Higham (1972, Higham and Kijngam 1979) and I (1971b, 1980c) have suggested that some form of swidden cultivation preceded intensive wet rice agriculture in the region; we feel this may well have initially have swamp-based rather then upland forest swidden (Bayard 1980c:132; Higham and Kijngam 1984:712), although forest swidden plots would later have been cultivated (as they are today) on the nearby Phu Wiang mountainside. Macdonald (1980:92-98) has countered this suggestion with an argument that the catchment area of sites such as Non Nok Tha or Ban Chiang is too small to support their populations if fields are long fallowed; hence some form of inundation cultivation must have been used. However, he uses larger population estimates for basal layers of such sites than I would envisage (2,000 vs. 150-500); a lower catchment radius (1.5 km) than has been observed ethnographically (e.g., 2.56 km for the Lamet; Izikowitz 1979:42); and 5 a fallowing period base on dry-field rather than swamp swidden. In any event, if ethnographic parallels are any guide (and we have little archaeological evidence at present), land tenure would have been vested in the village as a whole or in the senior clan segment, with other villagers having usufruct rights. The end of the Early Period was apparently marked by a period of rapid social change as evidenced in the mortuary remains by a noticeable increase in the range of wealth and the introduction of new vessel types (see Chapter 6); included among the latter are of course the distinctive C and L vessels. As mentioned above, their appearance is presaged by the presence in three late Early Period burials of apparent prototypes of 2C and 2L vessels; one is actually a blend of 2C and 4D types. An adjacent burial of the small fragments that appeared to be bronze were recovered from a secure EP 1 context (see Chapter 7). Immediately following the transition to the Middle Period, bronze tools, molds, crucibles, and casting spillage appear alongside the new vessel types. It is obviously unreasonable to explain this by gradual internal evolution; much more likely is the introduction of new ideological and technological concepts either by borrowing from a newly arrived group in the region or by incursion of the new group itself. The C vessels in particular, with their highly standardized temper and construction, suggest the possibility of manufacture in a single center, although this is of course highly speculative. L vessels are also standardized in temper and size but less so in technique of fabrication, perhaps reflecting imitative local manufacture. But such questions cannot be conclusively answered without a detailed analysis of the full range of Non Nok Tha mortuary vessels by a ceramic specialist; indeed, our data from Thailand have increased so markedly over the past few decades that the time is ripe for a thorough comparative study of funerary ceramics, should specialists and funding be available. I have no doubt that the results would be of great significance. Bur relying on our present knowledge, the C group appears to have been the locally dominant one in terms of wealth. Bronze may have been a significant part of this wealth given its concentration in C graves (seven of the eight graves with bronze artifacts) and rarity in other graves; however, some of these bronze-containing graves are otherwise rather poor (e.g., 1968 MP 6 B. 1). Table 8-13 shows that only four MP burials with neither C nor L vessels contain any bronze-related artifacts. Of these, two contain only small fragments of molds, which could have been accidental inclusions; one was only 40% excavated, with the possibility of undiscovered C/L vessels in the remaining 60%; and the last was not a burial but a late MP â&#x20AC;&#x153;pot nest.â&#x20AC;?



A number of plausible scenarios can be put forth to explain this apparent dominance; unfortunately the cost of testing them will be high in both time and money. Control of access to supplies of bronze (as the metal rather than ore, since no evidence of smelting was recovered) seems one important factor, but I am reluctant to seek a monocausal explanation. Control over a more basic resource—-land—-may also have played a part. The increased density of burials in the 1968 area during the later Middle Period suggests some amount of population increase and possible pressure on swamp and dry swidden lands. Several decapitated burials during this period hint at the presence of conflict perhaps related to this, but the evidence is inconclusive and no signs of trauma were present in the 1966 (and apparently the 1968) skeletal material. On the other hand, population pressure seems clearly not to have been an important factor during the succeeding wet-rice period (Welch 1985, Bayard 1992, Welch and McNeill 1991), and Chantaratiyakarn’s analysis (1983, 1984) also argues against any undue scarcity of land in the presumed swamp-swidden period. The exact nature of the two postulated affiliative groups is obviously also problematical (hence the deliberately vague label), and the following discussion is of course speculative. These distinctive vessels could have served as badges of clan membership rather than rank in the strict sense (which is highly unlikely given their occurrence in poor as well as rich burials), but the burial pattern gives no evidence of exogamy. One highly speculative possibility is that they served as markers of affiliation either to a village or to the landowning senior clan segments of different villages. The latter seems more likely, given the presence of 2E and D vessels at the site of Don Klang, some 25 km northeast of Non Nok Tha (White personal communication; Schauffler 1976). It is possible that the numerical preponderance of C burials (38% vs. 23% Ls in the secure MP sample) reflects their dominance in nearby hamlets or villages; one of these was possibly located some 500 m to the north on the present site of the village of Ban Na Di or on a second site called Don Kha adjoining it to the west (see Fig. 2-4); the two sites have areas of ca. 20 and 15 ha respectively, and villagers report burials from the latter site. The L burials may have derived from a smaller and possibly subordinate hamlet; Don Kok Pho, 1.2 km east of Non Nok Tha and about 9 ha in area (see Chapter 1), is a possible location. This was surveyed in 1965 (Solheim and Gorman 1966) and tested in 1967 by the Thai-British Expedition. Publication of this excavation and of the pottery from Don Klang may shed some light on the speculations presented here, but a proper test would involve extensive and costly area excavations on the three large sites. In addition, it is important to remember that all four of these sites, including Don Klang, show clear evidence of General Period C and D use as well, and the areas in use during the earlier General Periods A and B were in all probability considerably smaller (see Chapter 9 for a discussion of GP A and B settlement patterns in the Phu Wiang region). Alternatively, the C/L ratio could merely be the result of the clustering of C burials in the center of the site (1966 area), with L burials around the periphery of the mound. Testing this hypothesis would obviously require further extensive excavation at Non Nok Tha itself, and in view of Higham’s reinterpretation of the late EP and MP burials sequence (1996-1997; see below), I have invited any interested and qualified person to carry out such reexcavation in cooperation with the Thai authorities (Bayard 1996-1997). There are


certainly a number of further possible hypotheses that could be generated in addition to mine, Macdonald’s and Higham’s but I doubt that even the excavation of several thousand square meters at Non Nok Tha and neighboring sites would conclusively resolve these questions. As mentioned above, about two-thirds of the poorer burials from the secure Middle Period sample contain neither C nor L vessels. These include one burial with thirteen items and three more with eight to ten items; the rest have five or less. These may represent either persons of low achieved status within the two divisions or members of junior kin groups not entitled to use of the vessels or associated ritual that may have accompanied their interment. Given the prevalence of cattle/buffalo sacrifice among hill tribes today, it is interesting to note that all but two of the fifteen secure burials with bovine remains also contain C or L vessels (eight C, two L, three with both types, two with neither type). Based on the limited evidence available from MP 3, this level seems to represent a period of some amount of change and standardization in burial practice. The mean wealth of C graves, which had dropped very sharply from MP 1 to MP 2, declined still further, and overall mean wealth also declined slightly (Fig. 8-30). Mean wealth of C graves shot up sharply in the following MP 4, as did overall wealth; however, the latter figure is probably due to the preponderance of C graves in the MP 4 secure subset (seven/eleven). But it does seem possible that MP 4 represents some type of consolidation of economic or social Fig. 8-30 dominance by the C affiliative group. This presumed dominance appears to have waned during the apparently fairly stable MP 4-MP 6 period, as overall grave wealth declined and L-affiliated burials first close the gap and then slightly exceed the mean wealth of C burials. The end of the Middle Period (MP 7-8) during the late first millennium B.C. may have seen the apparent breakdown of whatever significance was associated with the C and L vessels and presumably of the affiliative groups themselves. Although our sample of burials from this period is not large (twelve, all in the 1966 area), indications are that average wealth of burial goods declined to about four items, and only two burials contain these distinctive vessels. The six “pot nests” in the 1968 area, some of which contain obviously exotic vessels, also hint at the introduction of new mortuary practices. This period coincides with or immediately precedes the first appearance of iron


tools in the Phu Wiang area (Bayard 1977a); this is also the most probably date for the beginnings of intensive wet-rice cultivation and what I have termed General Period C (Chapter 5). It is tempting to see this reflected in a shift of land tenure from village or clantype bodies to the family tenure prevalent in the region today—-as well as a decline in the importance of affiliation with larger corporate descent groups as the transition to permanent fields and family tenure progressed-— but obviously, alternative explanations are also possible. Macdonald has suggested that at this time Non Nok Tha was incorporated into an expanding state system (1980:259). I agree that state-like entitles-—or more probably complex chiefdoms-—were beginning to replace apparently autonomous village-level simple chiefdoms in Northeast Thailand at about this date, as evidenced by the amount of luxury goods and overall size of such sites as Non Chai (Bayard at al. 1982-1983); indeed, Higham has aptly titled the chapter of his 1989 book documenting this period “The End of Autonomy and Emergence of Chiefdoms.” But I feel that Non Nok Tha offers primarily negative—-albeit suggestive—-evidence for these developments. The decline in grave wealth and number of burials at the end of the Middle Period could have been due to a takeover of the local authority of the affiliative groups by a larger and more complex regional entity; but if so, this apical center has yet to be discovered and clear connections with Non Nok Tha documented. Alternatively, local group authority could have independently declined in importance with the shift to family land tenure that may have followed the adoption of wet-rice cultivation. The presence of a handful of sherds with clear affinities to wares in the lower and middle Chi Valley to the east in MP 7 and 8 (see Vincent’s ceramic report in the projected volume of specialist reports) tends to make me favor the former alternative at present, but only further investigation into the development of complex political entities in the region as a whole will provide an answer. An Alternative Interpretation There are of course, alternative interpretations of the data presented here, and at the 1988 Kioloa Conference Higham offered a plausible reinterpretation of the entire EP 2 to MP 8 burial sequence, compressing it into the space of a few generations (ca. 1500-1300 B.C.) and obviating any need to explain differences based on C and L affiliated groups (1996-1997). This employs a simple spatial patterning explanation of the burials as being based on linear interments of family or lineage members over a short period of time, particularly as evidenced by two perceived groups (Groups A and B) of parallel grave alignments in the central area of the 1968 excavation. The largely complementary distribution of C and L vessels is in Higham’s scenario due to chronological factors, with 2C/E and 4D vessels tending to replace 2Ls and 1Ds during the few generation that the cemetery was in use. This is evidenced in his view by a tendency for 2L and 1D vessels to occur more toward one end of each of his proposed alignments of burials, with 2C/Es and 4Ds more common at the other ends. I believe there are a number of weak points in Higham’s interpretation, but it its definitely worth considering—-particularly in view of the continuing uncertainly about the absolute chronology of the site and the arrival of bronze in the region (as discussed in Chapter 5). To use his words:


The coherent linear spacing, rarity of many superpositions and shallow depth of stratigraphy cause me to wonder whether we are in fact confronted by a time span of only a few generations, perhaps covering a couple of centuries. Let me post instances which seem strange given Dr. Bayard’s ascriptions of burials to phases. Consider burials 33, 67, 68, 85 and 124. These comprise a linear arrangement. Is it likely that burials 60, 67 and 68 (MP 6) could fit so neatly alongside burials 85 and 124 (MP 4) after an interval of perhaps five centuries? (Higham 1994). Examination of the overall burial plan of the 1968 excavation can provide some visual support (see Fig. 4-1) for Higham’s hypothesis; his two proposed alignments do contain a number of burials aligned in a similar direction and side-buy-side as well, although I would certainly question the “rarity of superpositions” (ca. eight out of twenty-two cases in his Group A, and ten out of twenty-two in B). Yet it is easy to find still further possible alignments in the burial plan and—-as Higham acknowledges—-I suspect some of these are indeed family burials made within a relatively short time span (see above for Middle Period 4-6 examples). It is easy to ascertain the reasons for Higham’s interest in alignments and the desire for a shortened chronology based on the shallow cultural deposits at Non Nok Tha. Higham’s first major excavation (with Kijngam) at Ban Na Di produced plausible but inconclusive spatial distinction in wealth and a conviction that “there can be no doubt that the prehistoric inhabitants of Ban Na Di knew precisely where their ancestors were buried” (Higham and Kijngam 1984:415), plus a depth of stratigraphy for the mortuary levels of 1.4 m over an estimated four hundred to eight hundred years, equaling the total stratigraphic depth of Non Nok Tha. His more recent excavation at Khok Phanom Di (with Bannanurag) had revealed clear and undeniable evidence of family clusters in burials during some five hundred years of mortuary use that produced 5.3 m of cultural deposit. Hence it was reasonable to expect Non Nok Tha to do the same. His most recent excavation at Nong No (Nor), however, has uncovered a late (1100-800 B.C.) GP B cemetery overlying a Neolithic (GP A) shell midden securely dated at 2450 B.C. (Higham, personal communication; Higham and Bannanurag 1992, Higham and Thusarat 1998) with a total depth of stratigraphic deposits not much greater than that of Non Nok Tha. I am convinced that Higham’s spatial arguments are correct for Khok Phanom Di, and perhaps at Ban Na Di as well (although further excavation would be required to provide a conclusive demonstration). I do not at present believe they are valid for Non Nok Tha. There are a number of arguments I have offered against those proposed by Higham, and details can be consulted in the relevant publications (Higham 1996-1997; Bayard 1996-1997). But I will summarize them as follows: a. Higham’s sequences of grave alignments violate the available stratigraphic evidence relevant to the burials discussed in terms of intersections, superposition, and so on. I have difficulty in resolving Higham’s arguments with the stratigraphy of the burials in these presumed linear groups and the changes in burial typology that took place from EP 3 to MP 8-—in particular, the wide-ranging changes between EP 3 and the MP. Hence it seems unlikely to me that 1968 EP 3 B. 90, lying at a depth of 160 cm below surface, could be the father or uncle of MP 4 B. 124, lying directly over it at a depth of 110 cm. Moreover, during the 400-800 year span of the Khok Phanom Di and





4. 5.

Ban Na Di cemeteries, care seems to have been taken not to disturb burials of ancestors in making fresh interments (e.g., Higham 1989:78). Such care is not apparent at Non Nok Tha, where disturbances of earlier burials by ones from the immediately following level are apparent from EP 1/EP 2 throughout the Middle Period. The sole exception is EP 3/MP 1, which I have interpreted above as a brief period of fairly rapid social change. I believe that Chapter 6 demonstrates the relative security of the chronological ordering of the whole vessel typology, and the present chapter presents what I think is a plausible typology of burial practices that fits well with this. Since Higham bases his arguments for chronological replacement of 2L/1D vessels by 2C/Es and 4Ds on my vessel and burial typologies, clearly his conclusions are erroneous if my stratigraphy is. If his linear group hypothesis is correct, then my vessel and burial typologies cannot be. While type 1D vessels do indeed decline in frequency during the Middle Period, as shown in Chapter 6 and the analyses above, type 2L vessels increase in the 1966 area not considered by Higham. Moreover, a significant negative correlation between occurrences of 1D and 4D vessels throughout the MP 1-8 sequence is lacking (r = -0.03; p = 0.416). Finally, it is difficult to postulate any functional or ritual similarity between the very different 1D and 4D vessel types. The analyses presented here also provide more than ample evidence for the high probability of the existence of the C/L distinction and its ties to overall wealth, whatever the social meaning of the distinction may have been. In short, it seems clear to me that if Higham’s alternative hypothesis is correct, most if not all of the stratigraphy described in Chapter 3 and 4 is incorrect, and almost all of the quantitative results of the analyses presented in Chapter 6 and this chapter are wrong. I thing the statistical vigor of the arguments presented here makes this very unlikely. Hence Higham’s alternative hypothesis must be rejected, at least until and equivalent body of quantitative data supporting it is produced. Higham’s view of the C/L distinction has varied from “a highly plausible interpretation” in the original draft of his 1989 book (which he kindly supplied to me for comments) through “a most interesting pattern” in the published version (Higham 1989:156) to “complicated and highly unlikely” (Higham 1996-1997). In his most recent book he has again upgraded the C/L hypothesis to “an intriguing hypothesis” (1996:193). In another paper delivered two years after the 1988 Kioloa Conference. He commented that both his model for Ban Na Di and mine for Non Nok Tha are “rather simple” (Higham 1991:324). I would agree, but I believe that the “Big Woman” model he offers for Khok Phanom Di is also rather simple and would guess that we will be very lucky to unravel more than a fraction of the details of prehistoric social structure at such sites-—although it is admittedly fun to try! For Higham’s chronological arguments about the absolute dates of the site, see Chapter 5. Finally, there is what is to me perhaps the most fascinating element: that of chance in what we happen to dig up and how we interpret it. I replied as follows to Higham’s query quoted above about the likelihood of such tidy alignments occurring after several centuries: Yes, [the neat fit of burials] is likely by chance alone, given the fact that both MP 4 and MP 6 burials employ southwest orientation. I have written a short, simple computer program which generates a 15 by 10 m square and fills it at random


with 40 male, female, and child burials, half “early” and half “late” (simulating MP 4 and MP 6). 25% of the burials are arbitrarily designated “rich”; the only variable held constant is southwest orientation [as with MP 4 and MP 6 burials]. In almost every one of the many randomly generated squares I have looked at, there are cases of “late” burials “neatly fitted” alongside “early” ones. Moreover, the computer generated a startling range of linear alignments and “rich” vs. “poor” cemetery areas, often quite clearly demarcated. I presented several examples of such “aleatory archaeology” to the [Kioloa] conference, and now have a more flexible PC-compatible Basic version of the program available should any reader be interested in trying it for him/herself6 (Bayard 1996-1997). Steven Gould pointed out in a recent book of essays that the human eye seems to perceive order in what is actually a random distribution and randomness in what is actually a nonrandom distribution. Although Gould is discussing New Zealand glowworms rather than Thai burials, the principle is clearly the same: The eye is easily fooled. As Gould says, “this nearly ubiquitous failure constitutes one of the chief, and often dangerous, dilemmas of both intellectual and everyday life” (1991:268). But obviously I, Higham, Macdonald, and the rest of humanity are all equally susceptible to the this tendency; it is for this reason that I urge readers to consider Higham’s and Macdonald’s arguments alongside mine in the relevant publications. Conclusions In conclusion, I believe that the Non Nok Tha mortuary data support the presence of some degree of ascriptive ranking in the general sense from the first use of the site; this increased noticeably with the arrival of new vessel types, local bronze technology, and the apparent development of two unequally ranked affiliative groups but appeared to decline at the end of the Middle Period, possibly because of absorption or displacement by a larger regional entity. There seem to be general parallels with Pearson’s survey of Neolithic cemeteries in coastal Eastern China (1981), which covers a longer span of three millennia. There the data indicate increasing wealth and stratification as well, culminating in the rise of more complex entities like the Shandong Longshan culture and ultimately a takeover by the Shang state. The research of the past two decades has convincingly demonstrated that similar developments leading toward statehood were in progress to the east and south of Non Nok Tha during the period from 500 B.C. to the birth of Christ; the impact of what I have called General Period C is evident at sites like Non Chai (Bayard, Charoenwongsa, and Rutnin 1982-1983), Ban Chiang (White 1982a, 1986), Ban Na Di (Higham and Kijngam 1984), Ban Chiang Hian (Chantaratiyakarn 1983, 1984; Higham 1989:210), and in the Phimai region in the transition from Tamyae to Prast and Classic Phimai phases (Welch and McNeill 1991). This transition is somewhat earlier than postulated by Higham et al. 1982, but in line with Hiham Kijngam’s 1984 estimates; as Higham et al. 1982, but in line with Higham and Kijngam’s 1984 estimates; as Higham has put it more recently, “Most workers agree that during the latter half of the first millennium B.C., and perhaps for a century or two earlier, there were profound cultural changes in the riverine and lacustrine floodplains of Southeast Asia” (1989:192). Unfortunately, Non Nok Tha bears little direct witness to them except perhaps by the negative evidence of its abandonment.


However, the site does at least testify to the presence of relatively complex, internally differentiated societies in the region as much as two millennia before such developments began, testimony that has been seconded at the important site of Khok Phanom Di (Higham et al. 1986-1987; Higham and Bannanurag 1990, 1991; Bannanurag 1991). Although further dates received late in 1987 (Higham 1989:67, 1994) now indicate that Khok Phanom Di was initially settled much later than the excavators’ tentative initial estimate of 4000 B.C. (Higham et al. 1986-1987:155), the site clearly evidences signs of social ranking equivalent to or surpassing the Non Nok Tha Early Period by the early second millennium B.C. I will return to these questions in the final chapter of this report. The Prehistoric Population: Inferences From the Skeletal Material The results of osteological examination of the skeletal material from both seasons’ excavations will hopefully be published in the projected volume of specialist reports. This section is intended only as a brief summary, much of which will rely on Pietresewsky’s analyses of the slightly smaller 1966 sample. This is because—-despite twenty years’ possession of the 1968 material—-the Brooks were unable to supply basic age and sex details for each more or less complete skeleton, nor did they compile any overall mortality, survivorship, or life expectancy curves for comparison with the growing body of osteological data form other Thai sites. Hopefully other scholars will be able to undertake these analyses. In their report, the Brooks take pains to emphasize the small number and fragmentary condition of the Nok Nok Tha skeletal material. I think it should also be emphasized, however, that the sample of about 201 individuals (about 83 from the 1966 and 118 from the 1968 excavation) is by far the largest yet available form Thailand.7 Comparable figures from other Thai sites are: Ban Kao, 38 (“most. . . in rather poor state or preservation”; Sangvichien et al. 1969); Ban Chiang, 127 (“one of the largest, best provenienced, and well preserved samples. . . from Northeast Thailand [and quite possibly Southeast Asia]”; Pietrusewsky 1982-1983:90); Ban Na Di, 63 (Houghton and Wiriyaromp 1984:391); and the large and well-analyzed sample of 154 from Khok Phanom Di (Choosiri 1988; Tayles 1992). Given this, I think the Non Nok Tha series, comprising 78 males, 60 females, 15 unsexed adults, and 48 subadults, is the largest and most representative currently available from the region. It is hence fortunate that Pietrusewsky was able to combine some data from the complete series for his paleodemographic study of the Non Nok Tha population (1974a). In terms of stature and life expectancy, the Non Nok Tha population fits well with estimates based on material from other Thai sites; most males ranged between 160 and 180 cm in height, with females ranging from 147 to 165 cm. The means for Ban Kao (male 169 cm, female 160 cm) and Ban Na Di (male 174 cm, female 161 cm) fit comfortably within this range, which in turn is quite similar to that observed with the majority of the Ban Chiang series (male 165-175 cm, female 150-158 cm). The prehistoric population of Khok Phanom Di tended to be slightly shorter, with males averaging 162 cm in height and females 154 cm (Tayles 1992:105). Squatting facets were commonly observed at Non Nok Tha as at other Thai sites (with the puzzling exception of Ban Na Di; Houghton and Wiriyaromp (1984:401). As Houghton and Wiriyaromp have noted (1984:399), life


expectancy at birth in Northeast Thai sites was quite high by the standards of tribal society: Non Nok Tha, 25.2 years, Ban Chiang, 26.1 years, and Ban Na Di 22.8 years. Survivorship curves from the three sites are also very similar (Fig. 8-31); in all three cases survival beyond the age of 50 was a very rare event. Fig. 8-31 The low mean of arrested growth lines and the other indices of possible stress discussed in the Brooks’ report suggest that the Nok Nok Tha population was a relatively healthy one; the Non Nok Tha mean of 4.6 Harris lines for a set of twenty-six femora compares quite closely to a mean of 4.3 lines for the Ban Na Di series. A low incidence of enamel hypolasia in both series also suggests a relatively healthy diet and lack of nutritional stress (Houghton and Wiriyaromp 1984:399-400). Dental evidence also suggests the chewing of betel nut was practiced, and Pietresewsky’s analyses note an unusual wear pattern in the upper incisors of females (not noted by the Brooks); whether this is due to filing, use of teeth for some type of stripping of plant material, or to another cause is not known. One of the more puzzling questions arising from the analyses of Northeast Thai skeletal material is that of porotic hyperostosis, a thickening of the cranial vault that may result from thalassemia and hence be an indication of the presence or malaria in the region. The results are sharply contradictory; Pietrusewsky identified the condition in over half of the adult crania in the 1966 sample from Non Nok Tha, as well as in the Ban Chiang material. It was also detected in the Ban Kao material by Sangvichien et al. (1969). However, Houghton and Wiriyaromp found no sign of it in the Ban Na Di series, and most puzzling of all, it is seemingly not present in the 1968 Non Nok Tha crania. Houghton and Wiriyaromp suggests that the condition may have arisen through a dietary deficiency-—for example, iron deficiency anemia-—rather than through genetically transmitted thalassemia (1984:400); however, this does not explain the surprising differences between the two Non Nok Tha samples.


At both Non Nok Tha and Ban Chiang, Pietrusewsky (1974, 1982-1983) has found evidence for a slight increase in lifespan and decrease in preadult and young adult mortality when earlier and later samples from each site are compared; hyperostosis is also more common in the later (MP 2-MP 8) Non Nok Tha sample than in the earlier one (EP plus MP 1). He has speculated that this trend may well have been due to the introduction of wet-rice agriculture in the region, leading to both a more productive economy and an increase of malaria due to large areas of standing water (1974:135). There is no evidence, however, in the archaeological record for such a shift occurring at this time (early Middle Period; the criterion for division of the Ban Chiang burials is not given). Obviously, much larger series of skeletal material examined by a larger number of scholars are necessary to answer these questions. Pietrusewsky (1974a, b, 1982, 1982-1983, 1996) has carried out a large number or multivariate comparisons of Non Nok Tha and Ban Chiang materials with samples from elsewhere in Asia and the Pacific using both metrical and nonmetrical cranial features. I must say that although a nonspecialist in the field, I find myself agreeing with Houghton and Wiriyaromp (1984:401) that such studies have their problems and often produce contradictory results. For example, the results in Pietrusewsky’s report in the projected volume of specialist reports suggest that Non Nok Tha is less closely related to Ban Chiang than Fiji is when metrical criteria are used, but that both Non Nok Tha and Ban Chiang are more distant from modern Thai material than Mongolia or Siberia based on nonmetrical features. It is difficult to reconcile such results. That being said, it can surely do no harm to apply such techniques if-—like glottochronology in comparative linguistic studies-—the limitations of the techniques and their results are recognized. Pietrusewsky’s analyses at the least demonstrate that the Non Nok Tha populations predictably had ties to both Asian and Indonesian populations; small numbers of rocker jaws in both Non Nok Tha samples hint at Oceanic connections as well. However, commentary on these wider relationships and the more general theories with which the Non Nok Tha materials have found themselves incorporated (Brace’s postulated correlation of tooth reduction with food preparation [1976] and Turner’s “Sinodont/Sundadont” dichotomy [1985]) is beyond the scope and competence of this report.


Notes: Chapter 8: The Burials an