A note on the projection of nominal appositives Rick Nouwen (Utrecht University) March 3, 2011
The projection problem for appositives
This squib 1 concerns the projection behaviour of appositives. It is often descriptively noted that appositives are scopeless. For instance, both the appositive relative clause in (1-a) and the nominal appositive in (1-b) are interpreted outside the scope of the negation. (1)
It is not the case that Jake, who is a famous boxer, lives in Utrecht. It is not the case that Jake, a famous boxer, lives in Utrecht.
It was noted by Wang et al. 2005, however, that there is a contrast between appositive relative clauses (ARCs) and nominal appositives (NAs). They observed cases where the scopal behaviour of NAs differs in interesting respects from that of ARCs. For instance, in (2) the NA restricts the set of professors the conditional is about, yielding an interpretating similar to if a famous professor
publishes a book, he will make a lot of money. A similar reading is absent from (3). 2 (2)
If a professor, a famous one, publishes a book, he will make a lot of money.
If a professor, who is famous, publishes a book, he will make a lot of money.
There are two questions that need to be answered. The hard one is what exactly is responsible for the contrast between (2) and (3). I will have some speculations on this towards the end of this work, but will not embark on a serious attempt at answering this question in this squib. The second question, the one I want to focus on here, is what accounts for the contrast between the wide-scope interpretation of (1-b) and the narrow-scope interpretation of (2). I will sketch a rather straightforward way of explaining this contrast, which is partly based on the proposal in Schlenker 2010a and Schlenker 2010b.
Nominal appositives and e-type anaphora
It has been observed before that there is a striking similarity between nominal supplements and certain e-type anaphoric phenomena (del Gobbo 2003; Nouwen 2007). For instance, notice that there is a parallel between the
data on discourse anaphora in (4) and the data on nominal appositives in (5). Just like distributive quantifiers licence plural but not singular discourse anaphora, they licence plural but not singular nominal appositives. (4)
a. Jake lives in Utrecht. He is a famous boxer. b. #Every boxer took part in the event. He is famous. c. Every climber made it to the summit. They were all experienced adventurers.
a. b. c.
Jake, a famous boxer, lives in Utrecht. #Every Dutch boxer, a famous one, took part in the event. Every climber, all experienced adventurers, made it to the summit.
Schlenker (2010a) makes the link with anaphora explicit in his semantic analysis of ARCs. I will adapt this, here, for the case of nominal appositives: 3 NA semantics â€” Semantically, nominal appositives are interpreted conjunctively, where the empty subject is interpreted as an e-type pronoun (cf. Schlenker 2010a). Given this assumption, we can account for (5-b) by claiming that it is interpreted as Every Dutch boxer took part in the event AND he is a famous boxer. Given what we
know about e-type pronouns (cf. (4)), such an analysis is expected to yield an infelicitous result. Now note the following. Schlenkerâ€™s semantics provides a handle on the examples discussed by Wang et al. (2005). That is, (6-a) can be analysed as (6-b), which yields the desired interpretation. (6)
If a professor, a famous one, publishes a book, he will make a lot of money. If [ a professor publishes a book AND he is a famous professor ], he will make a lot of money.
As such, the mere semantic assumption that NAs are interpreted conjunctively and involving e-type anaphora is prone to overgeneration. When I used this assumption to explain (5-b), I took the conjunctive interpretation to involve the matrix sentence and the appositive, while when I explained (6-a), I took the conjunctive interpretation to involve just the if-clause and the appositive. In fact, from a semantic point of view, it would be perfectly plausible to analyse (5-b) as in (7), in full parallel to (6-b). Yet no such interpretation is attested for (5-b). (7)
For every x such that [x is a Dutch boxer AND x is famous] it is the case that x took part in the event.
The generalisation is that nominal appositives can have narrow scope in the restrictor (i.e. the if-clause) of a conditional, but not in the restrictor of a quantifier. This generalisation can be accounted for using the syntactic dimension of Schlenkerâ€™s analysis for ARCs (again adapted here for NAs): NA syntax â€” Syntactically, nominal appositives have flexible attachment. They can be attached to any node of propositional type dominating the associated NP. (Schlenker 2010a) Now consider the following minimal pair: (8)
If a professor, a famous one, publishes a book, he will make a lot of money. #Every professor, a famous one, published a book.
In (8-a) there are two nodes of type t dominating the noun phrase that comes with the NA: the if-clause and the matrix sentence. We thus have two possible logical forms: (9)
If [[ a professor publishes a book ] AND [ he is a famous professor ]], he will make a lot of money. [ If a professor publishes a book, he will make a lot of money ] AND [ he? is a famous professor ].
We can discard the analysis in (9-b) since it is not in accordance to e-type pronominal behaviour: conditionals do not set up discourse referents, witness (10). (Here, and in what follows, I will indicate the lack of a possible pronoun resolution in a logical form by subscripting the pronoun with a question mark.) (10)
If a professor publishes a book, he will make a lot of money. #He is famous.
The only remaining analysis is (9-a), which, as was mentioned above, yields the interpretation observed by Wang et al. (2005). For (8-b) only one analysis is available, namely (11), since the matrix sentence itself is the only propositional node dominating the NP to which the appositive is associated. (11)
[ Every professor publishes a book ] AND [ he? is a famous professor ]
As explained above, the form in (11) is infelicitous because distributive quantifiers do not licence singular discourse anaphora (cf. (4-b)). Since there is no other logical form available, the example is uninterpretable. If we add more structure to the quantificational subject, more logical forms are derived, since there are more propositional nodes. For instance, (12) does receive an interpretation,
and it is indeed one in which the appositive is interpreted as part of the quantifier restrictor. 4 (12)
Every professor who wrote a book, one on linguistics, is eligible for a sabbatical.
Refining the analysis
To recap, I borrowed one semantic and one syntactic assumption from Schlenkerâ€™s analysis of ARCs and applied them to the case of nominal appositives. This appeared to successfully explain why NAs in if-clauses can restrict the conditional, while NAs in (non-propositional parts of) the restrictor of a quantifier are simply infelicitous. Given this analysis, we now come to expect non-wide scope NAs in several other configurations, and, in fact, this is what we observe. For instance, NAs anchored to an indefinite NP in the scope of negation receive an interpretation within the scope of negation. (13)
It is not the case that a boxer, a famous one, lives in this street.
This is explained readily by our assumptions. The two logical forms we derive are in (14).
It is not the case that [ a boxer lives in this street AND he is famous ] [ It is not the case that a boxer lives in this street ] AND [ he? is famous ]
In (14-b) the indefinite is not accessible to the pronoun and so (14-b) does not yield a felicitous interpretation. In contrast, (14-a) is interpretable and yields the observed interpretation. Note that what is crucial in the examples under discussion here is that the NA is associated to noun phrases whose accessibility for pronominal anaphora is subject to scopal constraints. That is, the lack of a reading corresponding to the form in (14-b) is due to the fact that the referential reach of indefinites is limited to the scope of the negation. If we change the examples to include an NA anchored to, say, a proper name, then the predictions and indeed the data change. (15)
It is not the case that Jake, a famous boxer, lives in this street.
Here, the wide scope interpretation, as in (16-b), is available. This is predicted too, since the referent of a proper name is globally accessible. (16)
It is not the case that [ Jake lives in this street
AND he is famous ] [ It is not the case that Jake lives in this street ] AND [ he is famous ]
Problematically, however, we now predict (15) to be ambiguous between the attested reading in (16-b) and the non-attested one in (16-a), which says that either Jake doesnâ€™t live in this street or he is not famous. This problem pops up in other cases too. Consistently, however, whenever a wide-scope interpretation is available, it blocks a possible competing narrow-scope one. For instance, (17) has the interpretation in (18-b), but not the one in (18-a). (17)
If Jake, a famous boxer, writes a book, he will make a lot of money.
If [ [ Jake writes a book ] AND [ he is a famous boxer ] ] he will make a lot of money [ If Jake writes a book, he will make a lot of money ] AND [ he is a famous boxer ]
We can descriptively account for this behaviour using the following generalisation.
NA attachment generalisation â€” A logical form in which an NA is attached in a high position blocks competing logical forms in which the NA is attached in a lower position. Unfortunately, I am at a loss how to explain this. Perhaps this generalisation should be seen as part of a bigger observation, namely that like other assertorically inert information (where notions like presupposition or conventional implicature have been applied), the information in an NA tends to be interpreted as scopally independent.
Definites and presuppositions
So far, I looked at indefinite and proper name anchors for nominal appositives. More support for the approach I outlined above comes from definite descriptions. Consider (19). (19)
If I ever get another son, I will call this son, my 5th one, Horatio.
The nominal appositive my 5th one cannot get widest scope here. The reason is that the existence of a referent for its anchor is a presupposition, which is plugged by the if-clause. That is, in the global context, there is no in-
formation about the existence of the son in question. The appositive receives a conditional interpretation. That is, (19) entails both (20-a) and (20-b): (20)
If I ever get another son, I will call this son Horatio. If I ever get another son, this son will be my 5th one.
According to the flexible attachment assumption, there are two LFs for (19): (21)
[ If I ever get another son, I will call this son Horatio ] AND [ he? is my 5th son ] [ If I ever get another son, [ I will call this son Horatio AND he is my 5th son ] ]
As explained before (21-a) is rejected since the pronoun in the appositive cannot be resolved. There is no such resolution problem for (21-b). However, (21-b) is not really the reading we get for (19). If I have only one son right now, this will not make the conditional in (19) false. It seems that while (20-a) is an at issue entailment, (20-b) is not at issue. The problem of (21-b) becomes even more apparent by studying (22). (22)
It is not the case that if I ever get another son, I
will call this son, my 5th one, Horatio. This example has the same conditional entailment (20-b). That is, the information that my next son will be my 5th one escapes the scope of negation. I believe the reason for this is independent of the behaviour of appositives, but is a consequence of the fact that the appositive is itself presuppositional. Notice, first of all, that (23-a) and (23-b) have exactly the same conditional inference (20-b) as (19) and (22). (23)
If I ever get another son, I will call this fifth son of mine Horatio. It is not the case that if I ever get another son, I will call this fifth son of mine Horatio.
This suggests that (21-a) might be the proper analysis for (19) if we recognise that the information in the conjunct corresponding to the appositive (he is my 5th son) is completely presuppositional, and has therefore non-trivial scopal properties of its own.
The approach to the projection behaviour of nominal appositives that I have described above came in three parts: (i) semantically, NAs are interpreted as conjuncts
with an e-type pronominal subject; (ii) syntactically, NAs may attach at any propositional node that dominates the associated NP; (iii) pragmatically, a logical form in which an NA is attached high blocks alternative logical forms in which the NA is attached in a lower position. These three ingredients account successfully for the NA data discussed here, but they will not do for appositive relative clauses. As I mentioned in the beginning of this squib, ARCs resist the kind restrictive readings of if-clauses that are observed with nominal appositives (Wang, Reese, and McCready 2005). (24)
If a professor, a famous one, publishes a book, he will make a lot of money.
If a professor, who is famous, publishes a book, he will make a lot of money.
Schlenkerâ€™s approach to ARCs differs from my adaptation of that approach for NAs only with respect to the pragmatic component. According to Schlenker, ARCs are locally non-trivial, while at the same time they are translucent. Translucency means that we should be able to add unsurprising assumptions to the context that make the ARC locally trivial: that make the local context entail the ARC. Note that such a pragmatic principle cannot account for the reading in (24). If given some
additional assumption it is locally entailed that the referent for a professor is famous, then we would not get a restrictive reading, but rather one in which all professors are famous. It is moreover difficult to see how in this case this involves an unsurprising assumption (since professors are not generally famous), and so it is that Schlenkerâ€™s ARC proposal makes a prediction with respect to (25). Since it is possible to add an unsurprising assumption to the context about some specific professor, it is predicted that a professor in (25) is interpreted as a wide-scope indefinite. It appears that this is indeed the only available reading. Note the difference with an example like (26), where the narrow scope reading is available for the indefinite and the ARC. This is because here it is perfectly possible to add to the context an assumption like all students are required to fill in form B35. Given the universal nature of this assumption, the ARC does not restrict the conditional. (26)
If a student, who by the way is required to comply with all Statutory Policies, asks for legal advice, it is best practice to contact the school lawyer.
Cases where translucency gives non-trivial projection behaviour are typically cases with ARCs associated to definites or proper names. See Schlenker 2010b for ex-
amples and discussion. Here, is an example from (Amaral, Roberts, and Smith 2007) (27)
In each class, several students failed the midterm exam, which they had to retake later.
In this example, the ARC will have to be attached locally in the scope of the quantifying in each class, for otherwise the plural pronoun they will lack a proper antecedent. The information that the students who failed the midterm exam in class X had to retake the exam later is non-trivial. It is translucent since we can add to the global context the information that any student failing the midterm exam needs to retake the exam. This addition would make the ARC locally trivial, which is what translucency requires. There is much more data to look at, but these preliminary thoughts suggest that while whereas ARCs are translucent, NAs come without such a condition. Summing up, I have drawn attention to the heterogeneity of the group of nominal supplements of which NAs and ARCs are part, which adds more data to the current debate on the proper treatment of supplementary material (Potts 2005; Amaral, Roberts, and Smith 2007; Harris and Potts 2009; Schlenker 2010a; Schlenker 2010b; Sæbø 2010). I’ve suggested that certain patters of projec-
tion observed with nominal appositive can be explained quite straightforwardly using the semantic and syntactic dimension of the proposal in Schlenker 2010a. I’ve furthermore suggested that the difference between NAs and ARCs could be down to a difference in the kind of pragmatic conditions they express.
References Amaral, P., C. Roberts, and E. A. Smith (2007). Review of the logic of conventional implicatures by chris potts. Linguistics and Philosophy 30, 707–749. del Gobbo, F. (2003). Appositives at the Interface. Ph. D. thesis, University of California, Irvine. Harris, J. and C. Potts (2009). Perspective shifting with appositives and expressives. Linguistics and Philosophy 32(6). Nouwen, R. (2007). On appositives and dynamic binding. Journal of language and computation 5(1), 87–102. Potts, C. (2005). The Logic of Conventional Implicatures, Volume 7 of Oxford Studies in Theoretical Linguistics. Oxford University Press. Sæbø, K.-J. (2010). Appositives in modal contexts. In Proceedings of Sinn und Bedeutung 15, Saarland University.
Schlenker, P. (2010a). Supplements within a unidimensional semantics i: scope. In Proceedings of the 2009 Amsterdam Colloquium. Schlenker, P. (2010b). Supplements within a unidimensional semantics ii: Epistemic status and projection. In Proceedings of NELS 2009. Wang, L., B. Reese, and E. McCready (2005). The projection problem of nominal appositives. Snippets 10, 13â€“14.
I draw attention to the heterogeneity of the group of nominal supplements of which nominal appositives (NAs) and appositive relative clauses...