{' '} {' '}
Limited time offer
SAVE % on your upgrade.

Page 1

Volume 31 / Number 1 / 2019

Volume 31 / Number 1 / 2019

Journal of

Media Psychology

Journal of Media Psychology

Editor-in-Chief Christoph Klimmt Associate Editors Nick D. Bowman Jesse Fox Diana Rieger Catalina Toma Ivar Vermeulen

Theories, Methods, and Applications


Clear, up-to-date guidance for professionals working with obese children “This book should find a place on the bookshelf of all pediatric practitioners.” Marsha D. Marcus, PhD, Professor of Psychiatry and Psychology, University of Pittsburgh School of Medicine, PA

Denise E. Wilfley / John R. Best / Jodi Cahill Holland / Dorothy J. Van Buren

Childhood Obesity (Series: Advances in Psychotherapy – Evidence-Based Practice - Volume 39) 2019, x + 80 pp. US $29.80 / € 24.95  ISBN 978-0-88937-406-5 Also available as eBook One in every six children, and more in some ethnic groups, are obese, which can lead to serious health problems in adulthood. Successful treatment of young patients is complex, requiring time-intensive, evidence-based care delivered by a multidisciplinary team. Help is at hand with this well written, compact book by leading experts, which gives health professionals – pediatricians, psychologists, other health workers – a clear overview of the current scientific knowledge on childhood obesity, from causality models and diagnosis to prevention and treatment.

www.hogrefe.com

In particular, the authors outline a family-based treatment method which is best supported by the evidence and meets the recommendations of the American Academy of Pediatrics and other organizations. The appendix provides the clinician with hands-on tools: a session plan, a pretreatment assessment form, self-monitoring forms, and a meal planning and physical activity worksheet. This book is essential reading for anyone who works with children and their families, equipping them to guide patients to appropriate and effective treatment.


Journal of

Media Psychology Theories, Methods, and Applications

Volume 31/ Number 1/2019


Editor-in-Chief Editorial Assistants

Associate Editors

Editorial Board

Publisher

Production Subscriptions Advertising/Inserts ISSN Copyright Information

Publication Subscription Prices

Payment

Electronic Full Text Abstracting/Indexing Services

Journal of Media Psychology (2019), 31(1)

Christoph Klimmt, Department of Journalism and Communication Research, Hanover University of Music, Drama and Media, 30539 Hanover, Germany (E-mail christoph.klimmt@ijk.hmtm-hannover.de) Sophie Bruns, Katharina Emde-Lachmund, Katharina Knop-Hülß, and Sabine Reich, Department of Journalism and Communication Research, Hanover University of Music, Drama and Media, 30539 Hanover, Germany (E-mail jmp@ijk.hmtm-hannover.de) Nick D. Bowman, West Virginia University, Morgantown, VA, USA, E-mail Nicholas.Bowman@mail.wvu.edu Jesse Fox, The Ohio State University, Columbus, OH, USA, E-mail fox775@osu.edu Diana Rieger, University of Mannheim, Germany, E-mail diana.rieger@uni-mannheim.de Catalina Toma, University of Wisconsin-Madison, Madison, WI, USA, E-mail ctoma@wisc.edu Ivar Vermeulen, Vrije Universiteit Amsterdam, The Netherlands, E-mail i.e.vermeulen@vu.nl

Eun-Ju Lee (Seoul, South Korea) Markus Appel (Koblenz-Landau, Germany) Jörg Matthes (Vienna, Austria) Florian Arendt (München, Germany) Peter Nauroth (Marburg, Germany) Omotayo Banjo (Cincinnati, OH, USA) German Neubaum (Duisburg-Essen, Germany) Anne Bartsch (München, Germany) Anne Oeldorf-Hirsch (Mansfield, CT, USA) Gary Bente (Cologne, Germany) Jochen Peter (Amsterdam, Netherlands) Paul Bolls (Columbia, MO, USA) Daniel Pietschmann (Chemnitz, Germany) Johannes Breuer (Cologne, Germany) Robert F. Potter (Bloomington, IN, USA) Jennings Bryant (Tuscaloosa, AL, USA) Arthur A. Raney (Tallahassee, USA) Caleb Carr (Normal, IL, USA) Leonard Reinecke (Mainz, Germany) Elizabeth Cohen (Morgantown, WV, USA) Meghan Sanders (Baton Rouge, LA, USA) Enny Das (Nijmegen, The Netherlands) Frank Schneider (Mannheim, Germany) Kevin Durkin (Glasgow, UK) Frank Schwab (Würzburg, Germany) Allison Eden (Amsterdam, The Netherlands) Stephan Schwan (Tübingen, Germany) Nicole Ellison (Ann Arbor, MI, USA) Michael Slater (Columbus, OH, USA) Malte Elson (Bochum, Germany) Kaveri Subrahmanyam (Los Angeles, CA, USA) David Ewoldsen (Columbus, OH, USA) Ron Tamborini (East Lansing, MI, USA) Christopher Ferguson (DeLand, FL, USA) Catalina Toma (Madison, WI, USA) Jesse Fox (Columbus, OH, USA) Sabine Trepte (Hohenheim, Germany) Sabine Glock (Wuppertal, Germany) Mina Tsay-Vogel (Boston, MA, USA) Melanie Green (Chapel Hill, NC, USA) Dagmar Unz (Würzburg-Schweinfurt, Germany) Matthew Grizzard (Buffalo, NY, USA) Sonja Utz (Tübingen, Germany) Dorothée Hefner (Hannover, Germany) Sebastián Valenzuela (Santiago de Chile, Chile) Shirley S. Ho (Singapore) Brandon van der Heide (East Lansing, MI, USA) Matthias Hofer (Zurich, Switzerland) Christian von Sikorski (Vienna, Austria) Juan José Igartua (Salamanca, Spain) Peter Vorderer (Mannheim, Germany) Jimmy Ivory (Blacksburg, VA, USA) Patrick Weber (Hohenheim, Germany) Jeroen Jansz (Rotterdam, The Netherlands) René Weber (Santa Barbara, CA, USA) Julia Kneer (Rotterdam, The Netherlands) Stephan Winter (Duisburg-Essen, Germany) Elly Konijn (Amsterdam, The Netherlands) Mike Yao (Hong Kong, ROC) Maja Krakowiak (Colorado Springs, CO, USA) Annie Lang (Bloomington, IN, USA) Hogrefe Publishing, Merkelstr. 3, 37085 Göttingen, Germany, Tel. +49 551 99950-0, Fax +49 551 99950-425, E-mail publishing@hogrefe.com, Web www.hogrefe.com North America: Hogrefe Publishing, 7 Bulfinch Place, 2nd floor, Boston, MA 02114, USA, Tel. (866) 823-4726, Fax (617) 354-6875, E-mail publishing@hogrefe.com Regina Pinks-Freybott, Hogrefe Publishing, Merkelstr. 3, 37085 Göttingen, Germany, Tel. +49 551 99950-0, Fax +49 551 99950-425, E-mail production@hogrefe.com Hogrefe Publishing, Herbert-Quandt-Str. 4, 37081 Göttingen, Germany, Tel. +49 551 99950-999, Fax +49 551 99950-998, E-mail zeitschriftenvertrieb@hogrefe.de Melanie Beck, Hogrefe Publishing, Merkelstr. 3, 37085 Göttingen, Germany, Tel. +49 551 99950-423, Fax +49 551 99950-425, E-mail marketing@hogrefe.com ISSN-L 1864-1105, ISSN-Print 1864-1105, ISSN-Online 2151-2388 Ó 2019 Hogrefe Publishing. This journal as well as the individual contributions and illustrations contained within it are protected under international copyright law. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without prior written permission from the publisher. All rights, including translation rights, reserved. Published in 4 issues per annual volume. The Journal of Media Psychology is the continuation of Zeitschrift für Medienpsychologie (ISSN 1617-6383), the last annual volume of which (Volume 19) was published in 2007. Calendar year subscriptions only. Rates for 2019: Institutions – from US $353.00 / 272.00 (print only; pricing for online access can be found in the journals catalog at hgf.io/journals2019); Individuals – US $195.00 / 139.00 (print & online; all plus US $16.00 / 12.00 postage & handling). Single copies – US $93.00 / 73.00 (plus postage & handling) Payment may be made by check, international money order, or credit card, to Hogrefe Publishing, Merkelstr. 3, 37085 Göttingen, Germany. US and Canadian subscriptions can also be ordered from Hogrefe Publishing, 7 Bulfinch Place, 2nd floor, Boston, MA 02114, USA. The full text of Journal of Media Psychology is available online at http://econtent.hogrefe.com and in PsycARTICLES. Abstracted/indexed in Current Contents/Social and Behavioral Sciences (CC/S&BS), Social Sciences Citation Index (SSCI), IBR, IBZ, PsycINFO, PsycLit, PSYNDEX, and Scopus. 2017 Impact Factor 1.118,5-year Impact Factor 1.651, Journal Citation Reports (Clarivate Analytics, 2018)

Ó 2019 Hogrefe Publishing


Contents Editorial

Announcing Changes in JMP’s Editorial Team: Farewell to Nicole Krämer and Welcome to Catalina Toma Christoph Klimmt

1

Original Articles

Rational Versus Intuitive Processing: The Impact of Cognitive Load and Moral Salience on In-Game Aggression and Feelings of Guilt Marina Krcmar and Allison Eden

2

They Are All Armed and Dangerous!: Biased Language Use in Crime News With Ingroup and Outgroup Perpetrators Jeroen Vaes, Marcella Latrofa, Caterina Suitner, and Luciano Arcuri

12

Why User Comments Affect the Perceived Quality of Journalistic Content: The Role of Judgment Processes Patrick Weber, Fabian Prochazka, and Wolfgang Schweiger

24

Effects of Human vs. Computer-Controlled Characters and Social Identity Cues on Enjoyment: Mediation Effects of Presence, Similarity, and Group Identification Jorge Peña, Jannath Ghaznavi, Nicholas Brody, Rui Prada, Carlos Martinho, Pedro A. Santos, Hugo Damas, and Joana Dimas

35

Investigating the Negation of Media Stereotypes: Ability and Motivation as Moderators Florian Arendt

48

Ó 2019 Hogrefe Publishing

Journal of Media Psychology (2019), 31(1)


Editorial Announcing Changes in JMP’s Editorial Team Farewell to Nicole Krämer and Welcome to Catalina Toma Christoph Klimmt Department of Journalism and Communication Research (IJK), Hanover University of Music, Drama, and Media, Hannover, Germany

With the beginning of the year 2019, the circle of editors who run the Journal of Media Psychology (JMP) has once again undergone important changes. After altogether 9 years of editorial service, Nicole Krämer has rotated out of the group, and at the same time, Catalina Toma has joined the journal team as new associate editor. Nicole Krämer followed Gary Bente as Editor-in-Chief in 2015. Prior to holding the main editorial responsibility, she had served as associate editor for 5 years, and she has done so again since she handed over the duties of the Editor-inChief to me at the beginning of 2018. With admiration and appreciation we are looking back at Nicole Krämer’s years of outstanding engagement with the journal. Under her leadership, JMP’s operations have expanded to an annual volume of around 200 submissions. Together with her editorial assistant, German Neubaum, Nicole Krämer has achieved to manage this remarkably dynamic growth while preserving the journal’s culture of excellence in peer reviewing and editorial feedback. What is even more noteworthy, Nicole’s editorial leadership has helped to strengthen the ties within the international media psychology community and to fortify JMP’s anchor position among its members. During her office term, JMP has seen a remarkable extension of its international base with regard to the circle of associate editors, editorial board members, and reviewers. For those who have been working with her to run the journal, she has always created a positive working spirit and acknowledged everybody’s contribution. So there are many reasons to thank Nicole for her dedicated and highly successful service as editor-in-chief and as associate editor of JMP. We know she will always stay connected with the journal and trust that we may ask for her advice also in the future. This time of farewell is, fortunately, also a time of welcome. The academic executives of JMP are delighted that Catalina Toma has accepted their cordial invitation to join the team as new associate editor. Catalina Toma is Associate Professor of Communication Science at the University of WisconsinMadison (USA). She is well-known for outstanding research, particularly in the psychology of social media. She holds a long history of affiliation with and service for JMP, and thus steps well-experienced into the new office of an associate editor. We are very happy about her engagement, as it will Ó 2019 Hogrefe Publishing

make a profound difference in preserving JMP’s mission of academic excellence for the years to come. We also celebrate the continuity of great diversity among the journal’s executives, as the circle of editors is now composed of 3 women (Jesse Fox, Diana Rieger, and Catalina Toma) and 3 men (Nick Bowman, Ivar Vermeulen, and myself) and includes members from the United States (3), Germany (2), and the Netherlands (1). So waving “good-bye” to Nicole connects to waving “welcome” to Catalina, and while we sadly see Nicole turning towards new academic challenges, we are optimistic and confident about the journal’s future, given Catalina’s arrival and the renewed team composition. This announcement of change is also a good opportunity to thank all of our supporters – authors, reviewers, editorial board members, associate editors, our publisher’s team (Regina Pinks-Freybott and Juliane Munson), and our editorial assistants Sophie Bruns, Katharina Emde-Lachmund, Katharina Knop-Hülß, and Sabine Reich. You all have been and are still contributing tremendously to the success story of JMP, and I hope you will continue todo so in thefuture. Therefore, JMP is in good shape for a new year of excellent, theory-driven, methodologically ambitious research and publishing! We hope that you, dear colleagues, are feeling the same optimistic spirit and are already planning interesting and innovative scholarship that you – of course! – will submit to JMP. Christoph Klimmt Department of Journalism and Communication Research (IJK) Hanover University of Music, Drama, and Media Expo Plaza 12 30539 Hannover christoph.klimmt@ijk.hmtm-hannover.de

Christoph Klimmt (PhD, 2004) studied media management at the Department of Journalism and Communication Research (IJK) of Hannover University. From 2007 to 2010, he served as an assistant professor in the Department of Communication, University of Mainz. Since 2010, he has been a professor of communication science at IJK Hannover. His research interests include media effects and processes, entertainment, and digital games.

Journal of Media Psychology (2019), 31(1), 1 https://doi.org/10.1027/1864-1105/a000256


Original Article

Rational Versus Intuitive Processing The Impact of Cognitive Load and Moral Salience on In-Game Aggression and Feelings of Guilt Marina Krcmar1 and Allison Eden2 1

Department of Communication, Wake Forest University, Winston-Salem, NC, USA

2

Department of Communication Science, VU University Amsterdam, Amsterdam, The Netherlands Abstract: This study explored two main theoretical propositions. First, we tested Hartmann’s (2011, 2012) notion that video games are processed via two separate cognitive systems: System 1, the automatic system, and System 2, the rational system. Specifically, we used a cognitive load manipulation to test if intuitive moral responses such as guilt and anthropomorphism are processed in System 1. Second, we utilized moral foundations theory to test the effect of care salience on guilt and in-game aggression. Using an experimental design (n = 94), the results indicate that under conditions of cognitive load, players had somewhat lower in-game aggression. Effects on guilt and anthropomorphism were in the same direction, albeit with small effects. In terms of moral foundations, we found that care salience was not negatively related to in-game aggression but was directly related to guilt, indicating that greater emphasis on the moral foundation of care resulted in greater guilt. Also, anthropomorphism was positively related to experienced guilt and negatively related to in-game aggression. Keywords: video game play, moral reasoning, cognitive load, rational vs. intuitive processing

Moral reasoning in video game play has gained increasing attention in recent years (e.g., Grizzard, Tamborini, Lewis, Wang, & Prabhu, 2014; Hartmann & Vorderer, 2010). Research on moral reasoning in video game play has stressed two key findings. First, research suggests that users tend to automatically perceive virtual characters in a social way, as if they were social beings worthy of proper moral treatment (Farrar, Krcmar, & McGloin, 2013; Hartmann & Vorderer, 2010). Accordingly, experimental studies have shown that users tend to feel intuitively guilty about violating personal norms if they engage in unfair transgressions against virtual characters (Grizzard, et al., 2014; Hartmann, Toz, & Brandon, 2010). Second, studies revealed that users tend to refrain from breaking moral taboos (i.e., conducting morally unacceptable behavior) while in virtual worlds (e.g., Klimmt, Schmid, Nosper, Hartmann, & Vorderer, 2006; Whitty, Young, & Goodings, 2011). Specifically, players tend toward moral behaviors that are highly important to them. Therefore, on one hand, video game play can be conceptualized as a moral experience. On the other hand, if asked, most video game players tend to distance themselves from the negative implications of seemingly immoral virtual behavior (e.g., killing other social beings) by stressing they were constantly aware that it was just a game (Klimmt et al., 2006). A study by Journal of Media Psychology (2019), 31(1), 2–11 DOI: 10.1027/1864-1105/a000215

Hartmann and Vorderer (2010) suggests that, indeed, the more players believed the “just a game” rationale, the less they experienced guilt and negative affect after performing immoral acts in-game. This finding seems to undermine the idea that playing a game evokes moral processing overall. In an attempt to provide a cogent theory for these seemingly paradoxical findings, Bowman (2016) has argued that video game play is a demanding technology that requires cognitive and emotional resources, and that certain cognitive skills, such as spatial reasoning, actually predict game success (Green & Bavelier, 2006). This argument, as well as the empirical findings that support it, suggests that how games are processed is key to explaining divergent outcomes. More specifically, Hartmann (2011, 2012) has proposed that users may process and react to video games based on dual-process models of cognition (Evans & Stanovich, 2013; Stanovich & West, 2000). The present research examines the idea that to the extent System 1 (automatic) processing dominates over (reflective) System 2 processing during game play, users may tend to believe in the apparent reality of the game. As a consequence, users should be motivated not to violate their moral norms while playing and different kinds of processing (i.e., System 1 and System 2) would result in particular outcomes. First, we would expect greater guilt and less in-game aggression occurring with more reliance on System 1, and Ó 2017 Hogrefe Publishing


M. Krcmar & A. Eden, Cognitive Load in Video Games

perceptions of character realism (i.e., greater anthropomorphism) would be affected by more reliance on System 1. Second, we would expect anthropomorphism to influence guilt and in-game aggression. Third, we would expect these patterns to be influenced by individual differences in moral reasoning. We use an experimental design, impairing users’ System 2 processing capacity via a cognitive load induction to test the use of System 1 versus System 2 processing and game outcomes.

Dual Processing of Video Game Play Several related ideas regarding cognitive processing support the notion that audiences process media content based on two distinct information processing systems that operate in parallel during exposure. These two distinct processing modes have been expressed and empirically substantiated in various psychological dual-process theories including the distinction between experiential and rational processing laid out in cognitive experiential self-theory (e.g., Epstein & Pacini, 1999), System 1 versus System 2 processing (e.g., Stanovich, 1999), rule-based versus associative processing (Sloman, 2002; Smith & DeCoster, 2000), reflective versus impulsive processing (Deutsch & Strack, 2006), and central versus peripheral processing of persuasive messages as proposed by the elaboration likelihood model (Petty & Cacioppo, 1986). In a recent theoretical approach combining dual-process theories and media psychology, Hartmann (2011, 2012) proposed a general dual-system model of media reality. According to this approach, individuals process media, including video games, based on two distinct informationprocessing systems that operate in parallel during exposure. The rational, rule-based system is likely to be engaged when players try to make sense of plot information, understand character motivations, and recall and utilize environmental and spatial information. On the other hand the experiential, associative system may be engaged because players must also make quick judgments, form impressions, and experience emotional responses. The joint outcomes of both processes determine how users experience a video game, for example, its realism and moral significance. In System 1, the intuitive processing system, information is processed quickly, effortlessly, and unconsciously. This system is thought to be the evolutionarily older (Moskowitz, 2005) system, which may in part account for its automaticity and our reliance on it as the default system. On the basis of this, Hartmann (2012) argues that when individuals are exposed to media and consequently make reality judgments, System 1 relies on correct sensory information (“seeing is believing”). In the context of video

Ó 2017 Hogrefe Publishing

3

gaming, System 1 processing may give rise to a feeling of apparent reality. Users may intuitively perceive video game characters not as pixels on the screen, but as social beings with individual thoughts and feelings, they may automatically feel like being part of the scenario (Presence; e.g., Schubert, 2009), and may feel that depicted events are really happening. In contrast to System 1, information processing in System 2 is slower, more effortful, deliberate, and conscious. System 2 operates on rule-based logical or analytical processing. In general, System 2 processing has been linked to uniquely human abilities such as hypothetical thinking, mental simulations, and detection of illusions (Evans & Stanovich, 2013, p. 235). Therefore, System 2 processing may also be related to reality judgments, simply those made at the more analytical and conscious level. Importantly, System 2 responses are derived more slowly. According to the dual-system model of media reality, both systems operate in parallel and at times may derive conflicting or seemingly paradoxical impressions of reality (e.g., something feels real while one knows it is not.) For example, in a virtual reality environment, many users refrained from walking across the edge of steep virtual cliff despite knowing, and articulating that, the cliff was virtual (Slater, 2009). In a virtual replication of the famous Milgram experiment by Slater, Antley, Davison, Swapp, and Guger (2006), participants reported that they felt like they were inflicting pain on a person, despite knowing that the in-game character was a computer simulation. Lastly, players feel intuitively guilty about unjustified virtual violence (Grizzard, et al., 2014; Hartmann & Vorderer, 2010; Hartmann, 2011) and players behave consistently with certain moral norms (e.g., not shooting women and children; Eastin, 2006). Broadly speaking, media environments are typically processed automatically, and that processing can become discounted through System 2 (e.g., Busselle & Bilandzic, 2008; Zillmann, 2006). According to Hartmann’s dual-system model of media reality, as well as other dual-processing theories, users may engage in slower and more effortful System 2 processing to contextualize or regulate their immediate impressions. For example, studies in the context of emotion regulation suggest that users rely on reflective processing to regulate fear (Hoffner & Levine, 2005) or sadness (Sheppes & Meiran, 2007) induced by movies. Likewise, gamers may rationalize immediate feelings of guilt by recalling that this is just a game. However, such System 2 operations are only likely if users are both motivated to engage and, importantly, capable of engaging in analytical reasoning (Evans & Stanovich, 2013). In the context of video gaming, users would generally engage both System 1 and System 2, the former because

Journal of Media Psychology (2019), 31(1), 2–11


4

it is automatic and the latter because it is necessary to follow a plot, make decisions, and utilize controls that have been learned previously and stored in working memory (Przybylski, Ryan, & Rigby, 2009). However, because gamers are likely motivated to feel entertained and to maintain a positive mood, it is likely that they are also motivated to suppress unwanted System 1 impressions, including annoying moral gut feelings. Thus, System 2 is necessary to override the System 1 reactions regarding hurting virtual characters.

Dual Systems and Cognitive Load Despite the fact that both systems are likely engaged during video game play, as well as during many processing activities, there are factors that are likely to influence the ability of System 2 to override System 1. One clear factor that may do this would be the availability of working memory capacity, a central but limited source of System 2 processing. As Evans and Stanovich (2013) argue: “The inhibition of type 1 responses take cognitive capacity” (p. 235). Thus, strain on the system from external distractions, excessive information, or information complexity can overload the system. Accordingly, if working memory capacities are impaired, (e.g., under cognitive load) users should be less able to engage in System 2 processing to regulate their intuitive default impressions. Several studies have shown that higher cognitive load does enhance the relative importance of System 1 processing. For example, people under cognitive load increasingly take false information as true (Busselle and Bilandzic, 2008). Accordingly, if video gamers are artificially put under cognitive load, their System 2 processing should be impaired, and intuitive impressions of an apparent reality resulting from System 1 processing should be relatively enhanced. In terms of a dual-system approach specifically, the current study explores these theoretical propositions by manipulating cognitive load during game play in order to consider the differing effect of System 1 versus System 2 processing in these main areas: perceived anthropomorphism, in-game aggression, guilt, and moral judgment.

Anthropomorphism One important aspect affected by users’ System 1 processing may be anthropomorphism, or, “the ascription of human characteristics to nonhuman entities” (Carporael, 1986, p. 215). Numerous studies have shown that users – despite better knowledge – readily treat game characters as social actors, and feel themselves to be in social company, while they in fact encounter only pixels on the screen (e.g., Nowak, Hamilton, & Hammond, 2009). Journal of Media Psychology (2019), 31(1), 2–11

M. Krcmar & A. Eden, Cognitive Load in Video Games

For example, Nowak and Rauh (2006) found that even static image avatars are anthropomorphized by computer users. One possible reason for the human tendency to anthropomorphize is due to the automatic nature of System 1 processing. Exposure to human-appearing images, such as those in video games, increases our likelihood of feeling, intuitively, that they are human. This perception may in turn encourage identification with the pixels or even empathy for what they represent. But this quick, automatic processing resulting in potentially empathic feelings might then be overridden by System 2 processing, or the logical realization that the computer is not a social actor. However, because System 2 is later at arriving, it acts to suppress intuitive, automatic System 1 responses. As a result, when System 2 processing is suppressed, we enhance perceptions of anthropomorphism: Hypothesis 1 (H1): There will be an effect of cognitive load on anthropomorphism such that greater reliance on System 1 processing will result in perceptions of greater anthropomorphism.

Perceptions of Reality, Anthropomorphism, and Moral Intuitions Given the theoretical propositions presented thus far, the model would suggest that greater reliance on intuitive processing via System 1 is likely to increase perceptions of game characters as real or human, and gut feelings that the pixels on the screen are human. However, the process is unlikely to end there. Recall that several studies have shown that players tend to feel intuitively guilty about violating personal norms (Grizzard, et al., 2014; Hartmann et al., 2010). In addition, despite the obviously fictional nature of virtual worlds, individuals typically engage in moral behavior while in these virtual worlds (e.g., Joekel et al., 2012; Klimmt et al., 2006; Whitty, Young, & Goodings, 2011). One possible explanation for this intuitive guilt and avoidance of moral transgressions is the use of System 1 processing. Whereas dual-processing theories emphasize the dual nature of cognitive processing, the dual approach is mirrored in work in moral psychology as well. For example, Haidt (2001) has argued that moral judgments, which occur, after all, in the brain, are at least partially innate, or, more descriptively, organized in advance of experience. Haidt (2001) has suggested that owing to the innate nature of moral responses, they are indeed quick and intuitive. However, he goes on to suggest that rational processes often overtake or contribute to these gut moral intuitions, thus resulting in sometimes ambiguous or conflicting moral judgments, as evidenced by recent research suggesting Ó 2017 Hogrefe Publishing


M. Krcmar & A. Eden, Cognitive Load in Video Games

within-individual processing differences (e.g., Eden et al., 2014; Tamborini, Weber, Eden, Bowman, & Grizzard, 2010). Thus, suppression of the rational system should result in more intuitive moral responses, such as those demonstrated in previous research on guilt and in-game aggression. Although this trend toward engagement in moral behavior while in a virtual environment is moderated by the individual difference variable of moral salience (Boyan, Grizzard, & Bowman, 2015; Petitte, 2013), we would still expect a main effect of the cognitive load manipulation. Therefore: H2: There will be an effect of cognitive load on (a) guilt and (b) in-game aggression such that more reliance on System 1 processing will result in (a) greater guilt and (b) less in-game aggression.

Moral Foundations Theory Thus far, we have suggested that processing strategies influence game perceptions and game outcomes. However, emotions and perceptions such as guilt and perceived anthropomorphism have a moral component. Current theorizing suggests that moral reasoning is automatic (Haidt & Joseph, 2008) but varies between individuals as well (e.g., Weaver & Lewis, 2012). Specifically, whether or not a particular behavior during game-play violates personally relevant norms depends, of course, in part on the morality of the user (e.g., Joeckel et al., 2012; Weaver & Lewis, 2012). According to moral foundations theory (MFT; Haidt, 2001; Haidt & Joseph, 2008), individuals differ according to the salience of five innate moral foundations. One of these is the foundation of harm/care or moral concerns about the protection of children and rules against physically hurting others. Previous research has shown that the varying salience of moral foundations among users indeed affects their video game play and in-game moral decisions. Specifically, individual differences in moral domain salience among players guides in-game moral decision-making (Joeckel et al., 2012; Krcmar & Cingel, 2016; Weaver & Lewis, 2012). In the case of video game play, where players shoot at human targets, the dimension of morality that would most logically be engaged is that of care. It is clear, therefore, that those for whom this dimension is particularly important would be less likely to shoot overall and feel greater guilt after game play. Furthermore, under conditions of high cognitive load, greater reliance on System 1 processing would be likely to have an additive effect for these individuals. Specifically, greater reliance on System 1 should enhance moral (i.e., less violent) outcomes because

Ó 2017 Hogrefe Publishing

5

System 2 rational processing encourages game play with less intrusion of social norms, guilt, and moral reasoning while suppression of System 2 should have the opposite effect. Therefore: H3: Greater care salience will be related to lower in-game aggression (in-game aggression) and greater guilt if users engage in unjustified virtual violence. H4: Care salience will moderate the effect of cognitive load on in-game aggression (in-game aggression) and guilt postulated in H1.

Method Procedure Participants at a large university in The Netherlands were told they would be participating in a media multitasking experiment. All participants arrived for the experiment to a laboratory room consisting of five computers in cubicles screened for privacy. Participants first filled out a questionnaire measuring their video game experience/skills and moral salience of moral foundations. When finished with the questionnaire, they were asked to read the game instructions for the game Call of Duty: Modern Warfare 2. After reading the instructions they played a test level for 5 min. Next, they were randomly assigned to memorize a two- (low-load condition) or seven-digit (high-load condition) number, and told that it would be important for them to remember these numbers in the future. Participants then played the “No Russian” mission of Call of Duty: Modern Warfare 2 for 5 min, after which they were stopped and asked to write down the numbers they had been told to remember. After this, they filled in a questionnaire about their game play experience in general, including feelings of guilt, enjoyment, and the human nature of the nonplayer characters in game.

Participants Participants (N = 101, 46 male) were recruited through a participant pool and through the social networks of the researchers. Participants (n = 96) recruited through the participant pool received course credit for participation; the remainder were volunteers. Seven participants did not complete the study and were dropped from the analysis, leaving a final of 94 participants (44 male). Participants ranged in age from 18 to 58 years, Mage = 22.13 and SD = 3.98. In all, 22% of the participants indicated they were familiar with the specific game. Randomization

Journal of Media Psychology (2019), 31(1), 2–11


6

M. Krcmar & A. Eden, Cognitive Load in Video Games

checks show no differences were observed in age, F(1, 92) = 1.170, p = .282, gender, w2(1) = 1.042, p = .307, education level, w2(4) = 1.800, p = .773, or video game experience, F(1, 91) < .333, p = .566, between experimental conditions.

Manipulation and Game Materials Cognitive Load Cognitive load was manipulated using the Gilbert Digit Rehearsal Task (Gilbert & Hixon, 1991). In this study a seven-digit number (e.g., 2604731) was used for the highload condition, and in the low-load condition a two-digit number (e.g., 39) was used (c.f. Duffy & Smith; 2012). In the high-load condition, 90% remembered the number correctly (consistent with recommendations for this procedure [Gilbert & Hixon, 1991]; a number was coded as correct when the participant had at least four of the seven numbers right in the correct order; only 24% remembered all even numbers correctly). Game Stimulus The game used was Call of Duty: Modern Warfare 2. This game was chosen because it features violent scenarios that test players’ morality and displays highly photo-realistic graphics. In this game there is a specific mission called “No Russian,” in which players are given the choice to shoot or not shoot innocent bystanders. To ensure that it was possible for all participants (i.e., inexperienced gamers) to easily play the game, a couple of modifications were made to the game controls. For instance, players could only make use of one weapon that reloaded automatically, they could only use basic movement controls (forward, backward, left, and right), and they could not be killed during the mission.

Measures Video Game Experience Three items were used to measure video game experience – “I am a good video game player,” “I know a lot about video games,” “A lot of my free time is spent playing video games” – as well as the game-specific item “I have played Call of Duty.” Participants answered these questions on 1

a 5-point Likert scale (1 = not at all, 5 = very much). The items formed an internally consistent scale (Cronbach’s α = .94) and hence were combined into a single index (M = 3.07, SD = 1.75). Aggression In-game aggression was indicated by the amount of bullets fired during game play. While playing, the number of bullets a participant fired was recorded. The minimum number was 0, maximum 975 (M = 238.94, SD = 204.81).1 Care Salience The salience of care was measured with six items concerning the moral foundations harm/care, although in this paper we refer to it only as care salience (Graham, Nosek, Haidt, Iyer, Koleva, & Ditto, 2011). First, three items from the MFQ were used: “When you decide whether something is right or wrong, to what extent are the following considerations relevant to your thinking: ‘Whether or not someone suffered emotionally,’ ‘Whether or not someone cared for someone weak or vulnerable,’ and ‘Whether or not someone was cruel’?” Responses were recorded on a 6-point scale (0 = not at all relevant, 5 = extremely relevant). Three more items from the MFQ asked about participants’ agreement or disagreement with the statements: “Compassion for those who are suffering is the most crucial virtue,” “One of the worst things a person could do is hurt a defenseless animal,” “It can never be right to kill a human being.” Responses were recorded on a 6-point scale (0 = strongly disagree, 5 = strongly agree). Although the internal consistency of the items was not satisfactory, they were well within range for this subscale in other research (e.g., Haidt & Graham, 2007; Cronbach’s α = .62). The items were averaged to form a composite index (M = 3.44, SD = 0.70). Anthropomorphism Anthropomorphism was measured by five items modified from Hartmann, Toz, and Brandon (2010). Items included: “It felt as if the people in the game had their own feelings,” “To me the people in the game seemed as if they were living beings,” “It felt like the characters had their own personality,” “It felt like the characters had a soul,” “It was as if the characters had their own lives,” “It felt like the character was aware of me.” Responses ranged from 0 (= strongly disagree) to 4 (= strongly agree). Items formed an internally

This variable was positively skewed (skewness = 1.355), which limits the interpretation of common statistical tests assuming normally distributed data. One option is to normalize the skew by either transformation or by Winsorizing the sample (Erceg-Hurn & Mirosevich, 2008). We believe the positive skew here reflects important information about how users play games, which is why we present the untransformed data in text. However, we understand readers may wish to see an analysis using a 5% Winsorized approach given the strong positive skew. Replacing values below 10 with 10, and above 700 with 700 (a 5% replacement), results in the following descriptive data and results: Winsorized bullet count, M = 238.94, SD = 204.81, skewness = .46. H2b: The effect of the covariate game experience remained significant, F(1, 93) = 30.938, p < .001 ω2 = .192, ηp2 = .249, and positively related to bullets fired, B = 46.307, SE = 8.325, but the effect of cognitive load on in-game aggression was reduced, F(1, 93) = 3.158, p = .079, ω2 = .017, ηp2 = .033. H3: Care salience was not correlated with in game aggression, (r = .09, p = .388). H4: There was no moderation of in-game aggression by care salience.

Journal of Media Psychology (2019), 31(1), 2–11

Ó 2017 Hogrefe Publishing


M. Krcmar & A. Eden, Cognitive Load in Video Games

consistent scale (Cronbach’s α = .92) and were averaged to form a composite index (M = 2.09, SD = 0.97). Guilt Guilt was measured by three items from Hartmann and Vorderer (2010) including: “I had feelings of regret of what I had done,” “I felt like I did something wrong,” “I felt like I should be accused of something” and three items from the guilt and disgust subscale of the Differential Emotions Scale (Izard, 1993) including: “I felt like something smelled bad like a bad taste in my mouth,” I felt as if everything was so rotten it could make me sick,” “I felt resentment, as if something terrible was going on.” Responses ranged from 1 (= not at all) to 5 (= very much). All six items formed an internally consistent scale (Cronbach’s α = .92) and were hence averaged to form one composite (M = 1.85, SD = 0.91).

Results Before engaging in hypothesis testing, correlations were run between the variables of interest (see Table 1). Given the significant correlations of video game experience with aggression (r = .472, p < .001) and anthropomorphism (r = .207, p = .047), video game experience was used as a covariate in these analyses.2 To test Hypothesis 1, regarding the effects of high cognitive load (i.e., suppression of System 2) on anthropomorphism, an ANCOVA was conducted. The effect of the covariate game experience on anthropomorphism was significant, F(1, 90) = 4.601, p = .035, ω2 = .038, ηp2 = .048, such that the greater experience with gaming, the less anthropomorphism experienced, B = .120, SE = .056. The effect of cognitive load on anthropomorphism was small and nonsignificant, but the mean differences were in the predicted direction, F(1, 90) = 3.193, p = .077, ω2 = .019, ηp2 = .034. Specifically, those in the high-load condition scored somewhat higher on anthropomorphism (M = 2.27, SD = 1.03) than those in the low-load condition (M = 1.92, SD = 0.89), which is consistent with predictions that suppressing the rational System 2 through higher cognitive load resulted in somewhat greater anthropomorphism. To test Hypothesis 2, regarding the effects of high cognitive load (i.e., suppression of System 2) on guilt and in-game aggression, two analyses were conducted. First, a t test was conducted. The effect of cognitive load on guilt (H2a) was not significant, t(92) = 1.632, p = .106, Cohen’s 2

3

7

Table 1. Zero-order correlations between variables 1.

2.

3.

4.

1. Aggression 2. Care Salience

.087

3. Guilt

.150

.366**

4. Anthropomorphism

.237**

.151

.417**

5. Video Game Experience

.472***

.031

.133

.207

Note. n = 94. ** = p < .05. *** = p < .001.

d = .334, with guilt in the high-load condition (M = 2.00, SD = 1.03) not significantly different from guilt in the low-load condition (M = 1.70, SD = 0.74). To test the effect of high cognitive load on in-game aggression (H2b), an ANCOVA was conducted. The effect of the covariate game experience was significant, F(1, 90) = 27.968, p < .001, ω2 = .017, ηp2 = .237, and positively related to bullets fired, B = 56.34, SE = 10.65, ηp2 = .256. In addition, there was a small effect of cognitive load on in-game aggression, F(1, 90) = 3.766, p = .055, ω2 = .014, ηp2 = .040. The means were in the predicted direction, specifically those in the high-load condition showed lower in-game aggression (M = 206.02, SD = 175.10) that those in the low-load condition (M = 278.24, SD = 227.26). Hypothesis 3 predicted a direct relationship between care salience and two variables: guilt and in-game aggression. The results of the correlations indicated that care salience was significantly correlated with guilt (r = .366, p < .001) but not with in-game aggression (r = .087, p = .406). Thus, H3 was partially supported. To test Hypothesis 4, that harm/care salience moderates the effect of condition on aggression and guilt, the PROCESS macro (Hayes, 2013; Model 1, 10,000 bootstrap samples, moderator mean centered for analysis) was used to test the moderating effect of care salience on aggression and guilt. Care salience did not moderate the effect of condition on in-game aggression (R2 chg = .000, p = .979). Thus, there is no evidence that care salience moderates the relationship between cognitive load and aggression. In terms of guilt, there was a small moderation effect of condition by care salience, condition b = .22, t(90) = 1.27, p = .207; care  condition b = .44, t(90) = 1.744, p = .084, but this was driven by participants 1 SD above the mean on care salience, b = .525, p = .0356, 95% CI = .036–1.02. Therefore, H5 was partially supported. A final exploratory analysis was conducted to determine if the effects of cognitive load on guilt and in-game aggression are due to a mediating variable such as anthropomorphism.3 Given that this is consistent with the logic

Analyses without the covariate are as follows. H1: The effect of cognitive load on anthropomorphism was F(1, 94) = 3.029, p = .085, ωp2 = .021, ηp2 = .031. H2b: The effect of cognitive load on in-game aggression was F(1, 94) = 1.739, p = .190, ωp2 = .007, ηp2 = .018. We would like to thank a reviewer for suggesting this analysis during the review process.

Ó 2017 Hogrefe Publishing

Journal of Media Psychology (2019), 31(1), 2–11


8

presented in the paper, we tested for this effect in a post hoc analysis. To examine the notion that anthropomorphism may mediate the relationship between cognitive load, guilt and in-game aggression, the PROCESS macro (Hayes, 2013; Model 4, 10,000 bootstrap samples) was used to test the mediating effect of anthropomorphism on aggression and guilt. Anthropomorphism was a negative predictor of aggression, b = 46.34, t = 2.13, p = .035, 95% CI = 89.42– 3.254, but there was no evidence for an indirect effect of cognitive load through anthropomorphism, b = 16.01, 95% CI = 49.58– 1.26. The results were similar for guilt, such that anthropomorphism positively predicted guilt, b = .37, t = 4.14, p < .001, 95% CI = .195–.554, with no evidence of an indirect effect of cognitive load through anthropomorphism, b = .13, 95% CI = .002–.326. Therefore, it appears anthropomorphism is strongly driving both guilt and aggression responses; however, this may be only slightly influenced via cognitive load.

Discussion Summary of Findings Our results indicate that, as hypothesized, under conditions of cognitive load, individuals tend to experience somewhat greater anthropomorphism. Furthermore, under higher cognitive load, players had lower in-game aggression as indicated by lower bullet count. Both of these findings are consistent with predictions. This may be due to the role of anthropomorphism on guilt and in-game aggression. The more anthropomorphism players felt toward in-game characters, the less in-game aggression players displayed, and the more guilt players experienced. While we found some support for the role of cognitive load on anthropomorphism, it is clear there is a strong effect of the perception of human essence on in-game play and postgame attitudes. In addition, findings indicated that care salience was not positively related to in-game aggression but was directly related to guilt, thus greater emphasis on the moral foundation of care resulted in greater guilt. Furthermore, these did not interact with the cognitive load manipulation. Thus, both cognitive load and the individual difference variable of care salience were related to the variables of interest: in-game aggression and guilt. Finally, we would like to note that player skill was a significant covariate in all analyses, perhaps because the longer that one exists in a game world, the less accustomed one is to seeing virtual characters as real people and instead as a means to an end. This logic is consistent both with our argument, and with

Journal of Media Psychology (2019), 31(1), 2–11

M. Krcmar & A. Eden, Cognitive Load in Video Games

the basic logic put forth by Hartmann (2011, 2012) in his discussion of media realism.

Theoretical and Practical Implications What conclusions can be drawn from the present findings? Consider, first, that higher cognitive load induced lower in-game aggression. This finding gives initial empirical support for Hartmann’s (2011, 2012) idea that video game play does in fact utilize two separate systems: the more intuitive System 1, in which players respond quickly based on gut intuitions, and the more rational and slower System 2, which encourages effective and strategic game play. If we combine theory relating to System 1 and System 2 processing with MFT, which suggests that moral reasoning is indeed an automatic process, the suppression of System 2 should result in less in-game aggression via the unsuppressed salience of care to players. Conversely, low load, or not suppressing System 2, should encourage players to play more typically, i.e., attempting to use the game rules and play the game. In the case of many violent video games, reliance on System 2 should result in more in-game aggression, especially if this is a functional choice. Again, the data are consistent with this rationale. In addition, the results for anthropomorphism are quite telling. By suppressing System 2 game play, players are prevented from discounting the perceptual cues of System 1. Thus, suppression of System 2 would result in more intuitive judgments. In the case of video game play, intuitive judgments, those based on our immediate and gut reactions, should cause us to perceive characters that look and behave human as human. To some extent, that is what happened. This perception, then, influenced the in-game aggression displayed and the experienced guilt after play. In sum, the intuitive system acts automatically, at some gut or base level. Cognitive load impaired System 2 thus leading to a greater automatic sense that the characters were human. Overall then, because the human perceptual system, which recognizes and categorizes others as human or nonhuman, is evolutionarily older (Moskowitz, 2005), humanappearing characters are typically judged to be human initially. Similarly, Haidt and Joseph (2008), in their conceptualization of moral foundations, argue that moral judgments are also innate, developed through the process of evolution and evolutionarily advantageous. Protection of others (i.e., the care foundation), for example, is one primary human moral response. From an evolutionary perspective it is useful to one’s own survival to protect and care for others. Thus, System 2, which might naturally engage to remind us that a video game is just a game (Hartmann, Toz, & Brandon, 2010), when suppressed

Ó 2017 Hogrefe Publishing


M. Krcmar & A. Eden, Cognitive Load in Video Games

9

would emphasize instead that harming others was wrong, that we should not shoot at others, and indeed that looking human (e.g., video game characters) means that someone is human. Given that the perception of human essence was a significant determinant of both guilt and in-game aggression, we can see that even in a game setting the initial reaction to things that look human is to treat them as such. The present findings are consistent with broader empirical research and theorizing about emotional processing and reactions to video games. Rationalization, as we might call it colloquially, served to increase enjoyment according to the research. In order to play a violent video game, it is essential to engage System 2. Behaviors such as accruing points and selecting weapons rely on rational processing. Rational processing allows for winning the game and enjoying violent behavior (Hartmann & Vorderer, 2010). Conversely, suppression of System 2 makes it more difficult to engage in immoral behavior. Although we did not measure enjoyment, it seems possible that a cognitive load manipulation might also serve to decrease enjoyment. After all, System 2 is needed to play a game and presumably enjoy it. Enhancement of the intuitive System 1 might serve to dampen enjoyment of aggression in many players. Furthermore, the significant relationship between harm/care salience and guilt suggests that although players may be influenced by cognitive processes enacted through System 1 and System 2, individual differences in moral salience can influence game play responses as well.

Studies 3 and 4), which may naturally increase cognitive load during play. In any case, each of these should help in our understanding of how players process games. Second, we utilized only one game. Because we wanted to investigate the effect of cognitive load on in-game aggression, we chose to begin our exploration with only one game. We reasoned that with each game, in-game aggression could be operationalized differently, making it difficult to derive a reliable and valid dependent variable. However, using only one game offers the possibility that effects were game specific, a potential threat sometimes referred to as message effects. Thus, future research should attempt to replicate these findings using other games. Finally, we may have had a floor effect on the guilt variable, as the mean for the composite measure was low. However, it was in line with previous research using this scale (Hartmann & Vorderer, 2010). In conclusion, it appears that video game play is in fact processed with two distinct cognitive systems and future theoretical work should consider the importance of these two distinct forms of processing during video game play. Furthermore, future work should continue to examine video game play as it relates to moral reasoning both in and outside the game world and in relation to System 1 and System 2 processing. Evidence here, and in other related studies, continues to support the notion that video game play can be a morally complex experience and one central to understanding how humans process social information in virtual worlds.

Limitations and Future Research

Acknowledgments

Although the findings from this study are promising, the study itself was not without limitations. We would like to stress that our data are merely suggestive of the relationships presented, and clearly future research must investigate these links further. The main limitations are as follows: First, the cognitive load induction is a relatively weak manipulation, which may have led to the smaller effects seen in this study. We externally manipulated cognitive load, but game controls all require cognitive processing allocation, which may vary with participant skill. Thus, it would be inaccurate to claim that those in the low-load condition were in fact playing under low load. Instead, playing the game itself presented players with some load; the test group merely experienced additional load. In addition, there are other ways to manipulate cognitive load. For example, the secondary task paradigm, where players are asked to attend to a task in addition to the primary activity, may offer a stronger manipulation. Alternately, the game controller may be altered to increase complexity during play (Przybylski, Deci, Rigby, & Ryan, 2014,

Ó 2017 Hogrefe Publishing

The authors would like to thank Tilo Hartmann for extensive input into an earlier version of this paper that was presented at the 100th annual conference of the National Communication Association, Chicago, IL.

References Bowman, N. D. (2016). Video gaming as co-production. In R. Lind (Ed.), Producing 2.0: The intersection of audiences and production in a digital world (Vol. 2, pp. 107–123). New York, NY: Peter Lang Publishing. Boyan, A., Grizzard, M., & Bowman, N. (2015). A massively moral game? Mass Effect as a case study to understand the influence of players’ moral intuitions on adherence to hero or antihero play styles. Journal of Gaming & Virtual Worlds, 7(1), 41–57. doi: 10.1386/jgvw.7.1.41_1 Bilandzic, H., & Busselle, R. (2008). Transportation and transportability in the cultivation of genre-consistent attitudes and estimates. Journal of Communication, 53, 508–529. doi: 10.1027/1864-1105/a000036 Caporael, L. R. (1986). Anthropomorphism and mechanomorphism: Two faces of the human machine. Computers in Human Behavior, 2(3), 215–234.

Journal of Media Psychology (2019), 31(1), 2–11


10

Deutsch, R., & Strack, F. (2006). Duality-models in social psychology: From opposing processes to interacting Systems. Psychological Inquiry, 17, 116–172. doi: 10.1207/s15327965pli1703-2 Duffy, S., & Smith, J. (2012). Cognitive load in the multi-player prisoner’s dilemma game: Are there brains in games? Available at SSRN 1841523. Eastin, M. S. (2006). Video game violence and the female game player: Self-and opponent gender effects on presence and aggressive thoughts. Human Communication Research, 32(3), 351–372. doi: 10.1111/j.1468-2958.2006.00279.x Eden, A., Tamborini, R., Grizzard, M., Lewis, R., Weber, R., & Prabhu, S. (2014). Repeated exposure to narrative entertainment and the salience of moral intuitions. Journal of Communication, 64(3), 501–520. doi: 10.1111/jcom.12098 Epstein, S., & Pacini, R. (1999). Some basic issues regarding dualprocess theories from the perspective of Cognitive-experiential Self-theory. In S. Chaiken & Y. Trope (Eds.), Dual process theories in social psychology (pp. 462–482). New York: Guilford Publishers. Erceg-Hurn, D. M., & Mirosevich, V. M. (2008). Modern robust statistical methods: An easy way to maximize the accuracy and power of your research. American Psychologist, 63(7), 591. Evans, J. St. B. T., & Stanovich, K. E. (2013). Dual-process theories of higher cognition: Advancing the debate. Perspectives on Psychological Science, 8(223–241), 263–271. Farrar, K., Krcmar, M., & McGloin, R. (2013). The perception of human appearance in videogames: Towards an understanding of the effects of player perceptions of game features. Mass Communication and Society, 16(3), 299–324. Gilbert, D. T., & Hixon, J. G. (1991). The trouble of thinking: Activation and application of stereotypic beliefs. Journal of Personality and Social Psychology, 60(4), 509–517. doi: 10.1037/0022-3514.60.4.509 Graham, J., Nosek, B. A., Haidt, J., Iyer, R., Koleva, S., & Ditto, P. H. (2011). Mapping the moral domain. Journal of Personality and Social Psychology, 101(2), 366–385. doi: 10.1037/a0021847 Green, C. S., & Bavelier, D. (2006). The cognitive neuroscience of video games. In P. Messaris & L. Humphreys (Eds.), Digital media: Transformations in human communication (pp. 211–224). New York, NY: Peter Lang. Grizzard, M., Tamborini, R., Lewis, R. J., Wang, L., & Prabhu, S. (2014). Being bad in a video game can make us morally sensitive. Cyberpsychology, Behavior, and Social Networking, 17(8), 499–504. doi: 10.1089/cyber.2013.0658 Haidt, J. (2001). The emotional dog and its rational tail: A social intuitionist approach to moral judgment. Psychological Review, 108, 814–834. doi: 10.1.1.124.9206 Haidt, J., & Graham, J. (2007). When morality opposes justice: Conservatives have moral intuitions that Liberals may not recognize. Social Justice Research, 20(1), 98–116. doi: 10.1007/ s11211-007-0034-z Haidt, J., & Joseph, C. (2008). The moral mind: How five sets of innate intuitions guide the development of many culture-specific virtues, and perhaps even modules. In P. Carruthers, S. Laurence, & S. Stich (Eds.), The innate mind (Vol. 3, pp. 367–392). NewYork, NY: Oxford University Press. Hartmann, T. (2011). Players’ experiential and rational processing of virtual violence. In S. Malliet & K. Poels (Eds.), Vice city virtue. Moral issues in digital game play (pp. 135–150). Leuven: Acco. Hartmann, T. (2012). Moral disengagement during exposure to media violence. In R. Tamborini (Ed.), Media and the moral mind. New York, NY: Routledge. Hartmann, T., Toz, E., & Brandon, M. (2010). Just a game? Unjustified virtual violence produces guilt in empathic players. Media Psychology, 13(4), 339–363. doi: 10.1080/15213269. 2010.524912

Journal of Media Psychology (2019), 31(1), 2–11

M. Krcmar & A. Eden, Cognitive Load in Video Games

Hartmann, T., & Vorderer, P. (2010). It’s okay to shoot a character: Moral disengagement in violent video games. Journal of Communication, 60(1), 94–119. doi: 10.1111/j.14602466. 2009.01459.x Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach. Guilford Press. Hoffner, C., & Levine, T. (2005). Enjoyment of mediated fright and violence: A meta-analysis. Media Psychology, 7, 207–237. Izard, C. E. (1993). The Differential Emotions Scale: DES IV–a method of measuring the subjective experience of discrete emotions. Newark, DE: University of Delaware. Joekel, S., Bowman, N., & Dogruel, L. (2012). Gut or Game? The influence of Moral Intuitions on Decisions in Video Games. Media Psychology, 15(4), 460–485. Klimmt, C., Schmid, H., Nosper, A., Hartmann, T., & Vorderer, P. (2006). How players manage moral concerns to make video game violence enjoyable. Communications, 31(3), 309–328. doi: 10.1515/COMMUN.2006.020 Krcmar, M., & Cingel, D. P. (2016). Moral Foundations Theory and decision making. In videogame play: Using real-life morality. Journal of Broadcasting and Electronic Media, 60, 87–103. Moskowitz, G. B. (2005). Social cognition: Understanding self and others. Hatfield, UK: Guilford Press. Nowak, K. L., Hamilton, M. A., & Hammond, C. C. (2009). The effect of image features on judgments of homophily, credibility, and intention to use as avatars in future interactions. Media Psychology, 12, 50–76. Nowak, K. L., & Rauh, C. (2006). The Influence of the Avatar on Online Perceptions of Anthropomorphism, Androgyny, Credibility, Homophily, and Attraction. Journal of Computer Mediated Communication, 11(1), 153–178. doi: 10.1111/j.10836101.2006.tb00308.x Petitte, O. (2013). Mass effect 3 by the numbers: 4 percent of players like shooting doctors in the face. PC Gamer. Retrieved from http:// www.pcgamer.com/2013/03/25/mass-effect-3-infographic Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of persuasion. Advances in Experimental Social Psychology, 129, 124–125. doi: 10.1016/s0065-2601(08)60214-2 Przybylski, A., Deci, E., Rigby, C. S., & Ryan, R. (2014). Competence-impeding electronic games and players’ aggressive feelings, thoughts, and behaviors. Journal of Personality and Social Psychology, 106(3), 441–457. doi: 10.1037/ a0034820 Przybylski, A. K., Weinstein, N., Ryan, R. M., & Rigby, C. S. (2009). Having to versus wanting to play: background and consequences of harmonious versus obsessive engagement in video games. Cyberpsychology and Behavior, 12(5), 485–492. doi: 10.1089/cpb.2009.0083 Schubert, T. (2009). A new conception of spatial presence: Once again, with feeling. Communication Theory, 19(2), 161–187. Slater, M. (2009). Place illusion and plausibility can lead to realistic behaviour in immersive virtual environments. Philosophical Transactions of the Royal Society London, 364, 3549–3557. Slater, M., Antley, A., Davison, A., Swapp, D., & Guger, C. (2006). A virtual reprise of the Stanley Milgram obedience experiments. PLoS ONE, 1(1), e39. Sloman, S. S. (2002). Two systems of reasoning. In T. Gilovich, D. Griffin, & D. Kahneman (Eds.), Heuristics and biases: The psychology of intuitive judgment (pp. 370–396). New York, NY: Cambridge University Press. Smith, E. R., & DeCoster, J. (2000). Dual process models in social and cognitive psychology: Conceptual integration and links to underlying memory Systems. Personality and Social Psychology Review, 4, 108–131. doi: 10.1207/S15327957PSPR0402_01

Ó 2017 Hogrefe Publishing


M. Krcmar & A. Eden, Cognitive Load in Video Games

Sheppes, G., & Meiran, N. (2007). Better late than never? On the dynamics of online regulation of sadness using distraction and cognitive reappraisal. Pers Soc Psychol Bull, 33(11), 1518–1532. Stanovich, K. (1999). Who is rational? Studies of individual differences in reasoning. Mahwah, NJ: Lawrence Erlbaum Associates. Stanovich, K. E., & West, R. F. (2000). Advancing the rationality debate. Behavioral and Brain Sciences, 23, 701–726. Tamborini, R., Weber, R., Eden, A., Bowman, N. D., & Grizzard, M. (2010). Repeated exposure to daytime soap opera and shifts in moral judgment toward social convention. Journal of Broadcasting & Electronic Media, 54(4), 621–640. doi: 10.1080/ 08838151.2010.519806 Weaver, A. J., & Lewis, N. (2012). Mirrored morality: An exploration of moral choice in video games. Cyberpsychology, Behavior, and Social Networking, 15, 1–5. Whitty, M. T., Young, G., & Goodings, L. (2011). What I won’t do in pixels: Examining the limits of taboo violation in MMORPGs. Computers in Human Behavior, 27(1), 268–275. doi: 10.1016/ jchb.2007.07.002 Zillmann, D. (2006). Dramaturgy for emotions from fictional narration. In J. Bryant & P. Vorderer (Eds.), Psychology of entertainment (pp. 215–238). Mahwah, NJ: Lawrence Erlbaum Associates. Received October 29, 2015 Revision received August 7, 2016 Accepted August 12, 2016 Published online April 28, 2017

Ó 2017 Hogrefe Publishing

11

Marina Krcmar Department of Communication Wake Forest University Winston-Salem, NC, 27106 USA krcmar@wfu.edu

Marina Krcmar (PhD, University of Wisconsin-Madison) is a professor at Wake Forest University. Her research focuses on the how children and adolescents use and are affected by media. Recent work has focused on video games as well as media targeting very young children. She has published 2 books, as well as many chapters and journal articles on these topics.

Allison L. Eden (PhD, Michigan State University) is an assistant professor of communication science at VU University Amsterdam, The Netherlands. She identifies and tests the functional role of enjoyment in attention to and selection of media content and the effects of media entertainment on moral attitudes and norms, well-being, and behavior.

Journal of Media Psychology (2019), 31(1), 2–11


The positive aspects of digital game play and its impact Topics covered include • Children’s reasons and motivations for getting involved in game play and how they benefit from such play • Exergames, designed to encourage physical activity • Games played by young people with learning difficulties • The importance of educational games and how they can benefit children’s knowledge and learning • A game application designed to help middle school students learn algebra

Mark Blades / Fran C. Blumberg / Caroline Oates (Editors)

Children and Digital Games (Series: Zeitschrift für Psychologie – Vol. 221/2) 2013, iv + 56 pp., large format US $49.00 / € 34.95 ISBN 978-0-88937-446-1 Children’s and adolescents’ exposure to and use of digital media has steadily increased in recent years, in both entertainment and education contexts. The growth of the internet and the development of interactive media such as computer games have opened up new issues that are as yet only partly understood. Furthermore, existing research has often been about the negative impact of media on young people. However, an activity that engages and motivates, such as game play, is one that can have a positive

www.hogrefe.com

impact, e.g., by stimulating knowledge and skills acquisition. This volume investigates the positive aspects of digital game play and its impact on children’s and adolescents’ learning, development, and physical activity. It contains contributions on the design, the use, and the benefits of digital games and provides a general overview of the appeal and educational ramifications of digital game play for youths.


Use movies to learn about positive psychology “This is the most important book about movies of our times.” Frank Farley, PhD, L. H. Carnell Professor, Temple University, Philadelphia, Former President of the American Psychological Association (APA)

Ryan M. Niemiec / Danny Wedding

Positive Psychology at the Movies

Using Films to Build Character Strengths and Well-Being 2nd edition 2014, xvi + 486 pp. US $59.00 / € 41.95 ISBN 978-0-88937-443-0 Also available as eBook Positive psychology is regarded as one of the most important developments in the field of psychology over the past century. This inspiring book uses movies as a medium for learning about the latest research and concepts, such as mindfulness, resilience, meaning, positive relationships, achievement, well-being, as well as the 24 character strengths laid out by the VIA Institute of Character. Films offer myriad examples of character strengths and other positive psychology concepts and are uniquely suited to learning about them and inspiring new ways of thinking. This book systematically discusses each of the 24 character strengths, balancing film discussion, related psychological research, and practical applications. Each chapter outlines Key Concepts, Relevant Research, an Exemplar from a key movie, Overuse/Underuse, Key

www.hogrefe.com

Enablers and Inhibitors, Practical Applications, International Cinema, and a Summary. Watching the films recommended in this book will help the reader to practice the skill of strengths-spotting in themselves and others, inspiring self-improvement. Practical resources include a suggested syllabus for a complete positive psychology course based on movies, a list of suitable movies for children, adolescents, and families as well as a list of questions for classroom and therapy discussions. Positive Psychology at the Movies is conceived for educators, students, practitioners, and researchers, but anyone who loves movies and wants to change their lives for the better will find it inspiring and relevant. Read this book to learn more about positive psychology – and watch these films to become a stronger person!


Original Article

They Are All Armed and Dangerous! Biased Language Use in Crime News With Ingroup and Outgroup Perpetrators Jeroen Vaes,1 Marcella Latrofa,2 Caterina Suitner,2 and Luciano Arcuri2 1

Department of Psychology and Cognitive Sciences, University of Trento, Rovereto, Italy

2

Department of Developmental Psychology and Socialization, University of Padova, Padova, Italy Abstract: The present research aims to verify the presence of linguistic biases in crime news reports (Study 1) and their role (Study 2) in activating a crime stereotype toward racial/ethnic minorities. In a first content analysis study, the natural occurrence of a set of linguistic biases was analyzed in Italian news articles that described comparable crimes committed by an in- or an outgroup aggressor. Results indicated that when the crime was committed by an outgroup (vs. ingroup) member, more aggravating and less attenuating adjectives were used. Moreover, the nationality of the perpetrator was not only mentioned more frequently, it also appeared in most cases as a noun. In Study 2, participants read a fictitious news article that either described an in- or outgroup criminal act with neutral or biased language. Their implicit associations between in- and outgroup members and weapons (vs. tools) were measured immediately afterward in the weapon paradigm. Results confirmed that a biased (vs. neutral) language use increased participantsâ&#x20AC;&#x2122; crime-related associations with the outgroup in general only when an outgroup criminal was staged. The role of media portrayals in determining the cognitive representations of racial/ethnic minorities is discussed. Keywords: media effects, crime news, linguistic biases, racial/ethnic minorities, stereotyping and prejudice

The media provide a window on the world that potentially molds our beliefs and expectations, especially about the lesser-known aspects of the environment we live in. Racial/ ethnic minorities are relatively unknown and scores of studies have provided empirical evidence on the negative impact of the media on stereotype formation and endorsement toward a broad range of social minorities (see Mutz & Goldman, 2010, for a review). The most common subject of analysis has been media portrayals that link crime to racial/ ethnic minorities. Little attention, however, has been paid to the linguistic aspects that make media portrayals of minority groups activate and maintain a social stereotype. Here we present two studies: The first documents the natural occurrence of biased language in Italian newspaper articles reporting crimes perpetrated by racial/ethnic minorities versus Italian ingroup members; the second study aims to show that such a biased language plays an important role in the activation of a crime stereotype only when it is used to describe members of the outgroup.

Media and Stereotyping Most of the research that studied the role of the media in the formation of stereotypes has analyzed the content Journal of Media Psychology (2019), 31(1), 12â&#x20AC;&#x201C;23 DOI: 10.1027/1864-1105/a000216

of televised and written portrayals of racial/ethnic minority groups. Television in the United States, for example, has a documented history of underrepresenting, caricaturing, and marginalizing racial/ethnic minorities (Bogle, 2001; Greenberg & Baptista-Fernandez, 1980; Ramirez Berg, 1990) in a large variety of programs ranging from comedy (Mastro & Greenberg, 2000), sit-coms (Weisbuch, Pauker, & Ambady, 2009), reality shows (Dubrofsky & Hardy, 2008; Hasinoff, 2008; Oliver, 1994; Shugart, 2006), to news programs (Dixon & Linz, 2000; Entman, 1992, 1994; Entman & Rojecki, 2000; Mastro & Robinson, 2000). In the latter type of programs, the recurrent reported finding is that African Americans and Latino Americans are consistently overrepresented in criminal and law-breaker roles, while they are systematically underrepresented as victims and law-defenders. Also in Europe, similar biased media portrayals have been reported. In two local newspapers in Italy, for example, immigrants were overrepresented as offenders and underrepresented as victims by comparison with objective crime statistics (Di Nicola & Caneppele, 2004). That such biased representations lead to biased perceptions has been the center of a number of cross-sectional studies demonstrating that increments in media exposure Ă&#x201C; 2017 Hogrefe Publishing


J. Vaes et al., Linguistic Biases and Stereotype Activation

reinforce negative attitudes toward racial/ethnic minority groups (Dixon, 2008a; Mastro, Lapinski, Kopacz, & Behm-Morawitz, 2009; Ramasubramanian, 2010; Ramasubramanian & Oliver, 2007; Vaes, Latrofa, Vieno, & Pastore, 2015). Similarly, media exposure moderated the emotional concern experienced after reading a crime story, with heavy-news viewers perceiving more emotional discomfort about a Black (vs. White) perpetrator (Dixon & Maddox, 2005). Media priming studies have further corroborated this finding potentially providing evidence for a causal relation between biased media representations and stereotype activation. Ford (1997), for example, found that a comedy skit portraying African Americans stereotypically (poor, uneducated, and prone to violence) versus neutrally, activated stereotypical beliefs about African Americans and increased guilt attributions to a crime suspect identified as African American (see also Johnson, Adams, Hall, & Ashburn, 1997; Peffley, Shields, & Williams, 1996; Power, Murphy, & Coover, 1996). Despite this evidence, priming people with crime news that contains racial/ethnic protagonists does not always activate stereotypes. Indeed, several studies have individuated variables that moderate the effects of the media on stereotype activation. Mastro et al. (2009), for example, showed that gender mattered, showing that only women gave less favorable judgments when confronted with a Black male crime suspect. Some studies only found changes in the perception of some minorities (such as Latino immigrants or Asian Indian minorities), but not for others, such as European immigrants or African American minorities (Brader, Valentino, & Suhay, 2008, Ramasubramanian & Oliver, 2007). Also, biased media reports impact especially those people who have only little direct contact with the portrayed minority group (Mastro, Behm-Morawitz, & Ortiz, 2007) and lead to divergent outcomes as a function of the involvement of the reader (Igartua, 2013). While these studies focused on viewer or target characteristics, the impact of specific features of media portrayals have been understudied. Tan, Fujioka, and Tan (2000) posited that the valence of portrayals of racial and ethnic minorities on television (i.e., positive or negative) predicted stereotypic responses associated with these groups. Extending these initial findings, Mastro and Kopacz (2006) found that the prototypicality of the representation is more incisive than its valence. Specifically, they proposed that the more the representation of a racial/ethnic minority deviates from the White prototype or norm, the more the endorsement of negative stereotypes will increase. The present research focuses on another characteristic of media portrayals testing the occurrence and the consequences of linguistic biases in reporting crime news comparing

Ó 2017 Hogrefe Publishing

13

instances where the perpetrator is an in- or outgroup member.

The Role of Linguistic Biases Newscasts and newspapers are supposed to provide factual information on what happens in the world, but when covering crime news involving racial/ethnic minorities, content and style might change. Differences in the use of linguistic and style elements when describing crimes committed by an in- or outgroup member are understudied, even though they might distort the way people perceive and judge the crime itself, the criminal, and his/her social group at large. The literature offers a plethora of linguistic biases. Arguably, one of the most studied is language abstraction. Following the linguistic category model (LCM; Semin & Fiedler, 1991), the same event or behavior can be described on different levels of abstractness ranging from descriptive action verbs (e.g., kick), which are most concrete, via interpretive action verbs (e.g., attack), state verbs (e.g., hate) to the most abstract category, adjectives (e.g., aggressive). Building on this framework, the linguistic intergroup bias (LIB; see Maass, 1999, for a review) has shown that positive behaviors conducted by ingroup members and negative behaviors by outgroup members are described in a more abstract fashion than negative ingroup and positive outgroup behaviors. Geschke, Sassenberg, Ruhrmann, and Sommer (2010) demonstrated that news reports using an abstract compared with a concrete language to describe outgroup negative behavior led to the expression of higher levels of prejudice against outgroup minorities. Similarly, content analyses of the ideas and reflections of participants who read news stories of a crime committed by immigrants (vs. their economic contribution) reflected a more abstract and negative affective language (Fernández, Igartua, Moral, Palacios, Acosta, & Muñoz, 2012). Carnaghi and colleagues (2008) added nouns as an even more abstract level to the LCM. Labeling someone as an aggressor rather than aggressive contains categorical information that goes well beyond the person’s qualities; it reveals the individual’s essence. Therefore, one can expect that references to a person’s provenance (group membership) are omnipresent when describing negative actions of racial/ethnic minority members, while they are hardly specified when the actions of an ingroup member are described. Extending this reasoning, one can expect the overrepresentation of immigrants in crime news to be even more accentuated when social membership is expressed with nouns rather than adjectives. Such references put a person on a par with his/her group membership, and when used in the context of crime news, reinforce the idea that

Journal of Media Psychology (2019), 31(1), 12–23


14

J. Vaes et al., Linguistic Biases and Stereotype Activation

Table 1. Examples of linguistic biases that were coded in Study 1 Linguistic biases Abstract language

References to nationality Verb tense Descriptive adjectives

Examples

Descriptive verb

The aggressor “throws away the knife, goes home, washes”

Interpretative verb

The aggressor “tries to remove all traces”

State verb

The aggressor “did not accept the Western style of life of the daughter”

Adjective

The aggressor was “a master-father, as you say, but also very ignorant”

Adjective

“An Italian thirty-nine-year-old was put in jail . . .”

Noun

“Chiara, raped by a Tunisian . . .”

Active

“He hit and kicked her, dragged her behind the bushes and raped her”

Passive

“Yesterday morning the victim was cleaning the offices of a call center when she got assaulted”

Aggravating

“A Tunisian irregular immigrant sexually abused a 15-year-old”

Attenuating

“An uncensored employee that [was] pressured by the military at the end collapses and confesses”

Note. The text as it appears in the articles is indicated in italics, while the linguistic biases are bold-faced.

belonging to a racial/ethnic minority says something about a person’s criminal nature creating or maintaining a crime stereotype toward these minorities (Graf, Bilewicz, Finell, & Geschke, 2012). A linguistic bias that has received the most attention in the realm of sexual violence regards the use of an active or passive voice. Previous research demonstrated that passive voices (e.g., “The girl was raped by an older man” vs. “The older man raped the girl”) put the aggressor in the background emphasizing the role of the victim. Such language use has been shown to increase the perceived responsibility of the victim while decreasing the responsibility of the assailant (Bohner, 2001). Following the research on rape cases, such language use might decrease the perceived responsibility of ingroup perpetrators, an attenuation that may not be used in the case of outgroup criminals. More related to the content, and less to the syntax of the message, is the use of aggravating and/or attenuating adjectives in describing criminal acts. Such adjectives qualify the description of the acts, the aggressor, or even the victim, making the crime seem worse (aggravating) or less severe (attenuating). For example, describing the act as fierce (vs. unintentional), the aggressor as a previous offender (vs. uncensored), or the victim as defenseless (vs. armed) are all aggravating (vs. attenuating) the crime and the way the criminal will be perceived. Even though these adjectives often describe what really happened, they can be easily omitted or included to change the gravity of the reported crime. We expect that aggravating adjectives will appear more frequently, while attenuating adjectives will be omitted more easily when crimes are committed by a racial/ethnic minority rather than by ingroup members. In sum, journalists might create a range of systematic linguistic biases when describing crimes committed by racial/ethnic minorities compared with ingroup members, Journal of Media Psychology (2019), 31(1), 12–23

and in this way help to create and maintain a distorted and stereotyped image of minority group members (see Table 1 for examples of all the different types of linguistic biases that were coded).

The Present Research Two studies were conducted to investigate the presence of linguistic biases in news reports (Study 1) and their role (Study 2) in activating a crime stereotype toward racial/ethnic minorities. Study 1 compared the natural occurrence of a set of linguistic biases in the description of comparable criminal acts that appeared in Italian newspapers in which only the nationality of the aggressor differed, keeping all other characteristics constant as much as possible. Study 1 was therefore a content analysis study that tested the following hypothesis: Hypothesis 1 (H1): Counting the number of linguistic biases that naturally occur in the description of comparable crimes that were committed by in- or outgroup members, we expect to find: more references to the nationality (H1a) especially in the form of nouns compared with adjectives (H1b); a more abstract language (H1c); less use of a passive language to describe the crime (H1d); and more aggravating and less attenuating adjectives (H1e) when crimes were committed by outgroup compared with ingroup members. Study 2 consisted in an experimental study in which we created four fictitious crime news articles systematically manipulating both the aggressor’s nationality (ingroup vs. outgroup) and some linguistic elements that described the criminal act in a biased vs. a neutral way. A fifth control condition was created in which participants read a neutral and unrelated news report. Immediately afterward, Ó 2017 Hogrefe Publishing


J. Vaes et al., Linguistic Biases and Stereotype Activation

participants stereotype activation was measured and they were asked to indicate what punishment they would give to the aggressor in the crime new story. In Study 2 the following hypotheses were tested: Hypothesis 2 (H2): Reading a criminal act conducted by an outgroup member would increase the activation of a crime stereotype of the outgroup as a whole, especially when the crime is described using linguistic biases. When a neutral, more factual language is used to describe the same criminal act, the activation of the crime stereotype is expected to be comparable to the one in the control condition where no mention is made of the outgroup or any criminal acts. Instead, reading a criminal act conducted by an ingroup member is not expected to activate a crime stereotype and the use of linguistic biases is not expected to change this effect.

Hypothesis 3 (H3): Reading a crime news story that uses biased language would lead to a harsher punishment compared with when the same crime is described in a neutral manner only when the aggressor is an outgroup member.

Study 1 Method Article Selection The newspaper articles included in the content analysis were selected on the basis of an event search. Using the search engine provided by Google news (http://news. google.it), comparable criminal acts were selected in which only the nationality of the aggressor differed, keeping all other characteristics constant as much as possible. In addition, the nationality of the victim was registered and if possible both intergroup (outgroup aggressor/ingroup victim, ingroup aggressor/outgroup victim) and intragroup (ingroup aggressor/ingroup victim, outgroup aggressor/outgroup victim) crime stories were gathered. As a consequence, only crime stories in which the aggressor and the victim and their nationality were certain could be included. The typology (i.e., similar number of victims and aggressors, comparable context, use of the same weapon if any) and the cruelty of the crimes needed to be comparable, as well as the time of the year in which the crimes were committed. This search focused on criminal events that happened in the period between the beginning of 2008 and the end of 2009. According to these criteria, six types of criminal events were retained Ó 2017 Hogrefe Publishing

15

(see Table 2 for an overview and Electronic Supplementary Materials, ESM 1 for a full description). After this selection procedure, 16 criminal events were retained and 73 articles that reported on these events were found. Care was taken that these articles were from some local and some national newspapers that covered the complete political spectrum. Given that neither the type (local vs. national) of newspapers nor their political orientation influenced the dependent variables, they will no longer be discussed. Coding All articles were copied in a word processing file and given to four independent coders. The coders were trained to recognize and correctly classify all the indices. They were divided into two dyads that each received half of the total amount of articles. As such, each article was read and analyzed by two coders independently. Intercoder reliability was verified calculating Krippendorff’s alpha (Kα) for each pair of coders’ independent judgments and was generally high or satisfactory (see next paragraph). Afterward, the coders met in person and resolved the few inconsistencies that were present in their separate files. Analyses were conducted on this final file that included the coding on which both coders agreed. For the current study, we will focus on the following indices: – The total number of times the nationality of the aggressor was mentioned in terms of a noun (Kα1 = .83, 95% CI = .68–.96; Kα2 = .83, 95% CI = .97–1.00) or an adjective (Kα1 = .98, 95% CI = .96–1.00; Kα2 = .91, 95% CI = .83–.97), partialized for the total number of words used in each article. – The number of aggravating (Kα1 = .80, 95% CI = .62–.94; Kα2 = .90, 95% CI = .77–1.00) and attenuating adjectives (Kα1 = .57, 95% CI = .08–.89; Kα2 = .79, 95% CI = .00–1.00) mentioned in the article, partialized for the total number of words used in each article. – The language abstraction in the article was measured following the LCM (Coenen, Hedebouw, & Semin, 2006). This index is calculated based on the sum of descriptive action verbs (Kα1 = .88, 95% CI = .81–.93; Kα2 = .93, 95% CI = .84–.98), the sum of interpretive action verbs multiplied by 2 (Kα1 = .91, 95% CI = .83–.96; Kα2 = .85, 95% CI = .61–.97), the sum of state verbs multiplied by 3 (Kα1 = .81, 95% CI = .54–.96; Kα2 = .65, 95% CI = .34–.91), the sum of adjectives multiplied by 4 (Kα1 = .83, 95% CI = .68–.95; Kα2 = .83, 95% CI = .73–.92), and the sum of the nouns multiplied by 5 (Kα1 = .99, 95% CI = .97–1.00; Kα2 = .98, 95% CI = .97–.99). The result was divided by the total number of words used. Journal of Media Psychology (2019), 31(1), 12–23


42 –

Note. # = Total number of selected articles within a given category and event. Loc. = Number of newspaper articles with a local circulation. Nat. = Number of newspaper articles with a national circulation. L = Number of articles coming from a left-wing newspaper. R = Number of articles coming from a right-wing newspaper. N = Number of articles coming from a newspaper with a neutral political affiliation. Total = The total number of articles in a given category. % of Total = The percentage of the total number of articles in a given category.

4

29

6

42

4

29 64

9 5

36 –

14 6

27 23

5 11

50 54

12 10

46 –

22 4

31 15

2 7

54 61

8 5

39 –

13 6

25 33

8 10 10 24

% of Total

14

1 2 1 2 4

2

42

1 2 1 2 2 4

1 1 3 3 2 5 1 1 3 2 5

3

2

Father kills daughter Husband kills wife Total

2 5

58

3 0 2 2 4 0 0 3 2 1 3 1 1 3 3 2 5 1 2 2

1

Rape

3

1

2 2

2 3

4 1

2 5

5

1 1 3 3 2 5 1 1 3 3 2 5 1

2 1

2 2

5 Car incident

2 5 Gang rape

3

2

R L Nat Loc # Baby gang

Event

3

Loc # N

3 1

R L

1 2

Nat Loc

3 5

# N

2 0

R L

1 2

Nat Loc #

1

Circulation Circulation

N

3

Political aff.

Outgroup victim Ingroup victim

Political aff. Political aff.

Outgroup victim

Circulation

2

R Nat

L

Political aff. Circulation

Ingroup victim Ingroup aggressor Outgroup aggressor

Table 2. Summary of selected articles in the archival study (Study 1) 1

1

J. Vaes et al., Linguistic Biases and Stereotype Activation

N

16

Results and Discussion Given that several articles described the same event, this was treated as a random factor in the linear mixed models performed with the lmer function of the R package lme4 (Bates, Maechler, Bolker, & Walker, 2014; R Core Team, 2013).1 In order to test H1a and H1b that verifies whether national membership was mentioned quantitatively and qualitatively differently according to the aggressor’s nationality, the number of references to national membership was used as a dependent variable in a linear mixed model, with aggressor and victim nationality as independent factors, type of reference (adjective vs. noun) was a repeated factor, and type of event was included as a random factor. The data better fitted the tested model (AIC = 164.10, BIC = 196.92) compared with the null model (AIC = 176.96, BIC = 188.90), w2 = 26.86, df = 7, p = .00035. As can be seen in Table 3, there is a main effect of the aggressor’s nationality, which is further qualified by the type of reference. The number of references to the nationality of immigrant criminals is more frequent in the noun (M = 0.69, SD = 0.66) than adjective form (M = 0.31, SD = 0.43). The references to the nationality of Italian criminals are generally lower (close to zero), with no difference between the noun (M = 0.05, SD = 0.15) and adjective (M = 0.06, SD = 0.13) form. The interaction effects of the tested model are visually represented in Figure 1. Testing H1e, we looked at whether Italian and immigrant criminals are described with different types of adjectives. A mixed linear model with aggressor and victim nationality as independent factors, type of adjective (attenuating vs. aggravating) as repeated factor, and type of event as a random factor had a better fit (AIC = 1,451.9, BIC = 1,419.1) compared with the null model (AIC = 1441.6, BIC = 1429.7), w2 = 24.30, df = 7, p = .001. As can be seen in Table 4, there is a main effect of the aggressor’s nationality, which is further qualified by the type of adjective. Immigrant criminals are more frequently described with aggravating (M = 0.002, SD = 0.0009) than attenuating adjectives (M = 0.0001, SD = 0.0005), whereas Italian criminals are more frequently described with attenuating (M = 0.006, SD = 0.001) than aggravating adjectives (M = 0.0004, SD = 0.0004). The interaction effects of the tested model are visually represented in Figure 2. We also tested the role of linguistic abstraction (H1c) and the use of passive (vs. active) sentences (H1d), but none of

Given that a variable number of articles described the same event, we used a mixed model approach (Pinheiro & Bates, 2006) that treats each observation as a unit of analysis. Such an approach allows one to verify the contribution of the predictors (i.e., fixed effects), while controlling for the random source of variability (i.e., random effects) owing to the fact that data are grouped together (i.e., they refer to the same event). Moreover, they are robust to nested designs that are not balanced in terms of the number of observations (as in the present data set). Therefore, the event was treated as a random factor in the linear mixed models performed with the lmer function of the R package lme4 (Bates, Maechler, Bolker & Walker, 2014; R Core Team, 2013).

Journal of Media Psychology (2019), 31(1), 12–23

Ó 2017 Hogrefe Publishing


J. Vaes et al., Linguistic Biases and Stereotype Activation

17

Table 3. Parameter estimates of the model: references  aggressor  victim  reference type + (1|event) + (1|article) Fixed effects

Estimate

SE

df

t

p > |t|

(Intercept)

.044

.12

18.86

.38

.71

Aggressor nationality

.404

.19

19.56

2.11

.048

Victim nationality

.086

.19

19.56

.45

.66

Reference type

.03

.11

125.25

.29

.77

Aggressor: victim

.28

.27

19.43

1.03

.31

Victim: noun reference

.36

.19

125.25

1.92

.05

Aggressor: reference

.05

.19

125.25

.27

.79

Aggressor: victim : reference

.04

.26

125.25

.18

.85

Figure 1. The effect of the aggressor’s membership on the number and type of reference to group membership (error bars represent the standard error of the mean).

these linguistic features were related to the group membership of the criminals. Together, these findings partially confirm our first hypothesis. As we expected, the nationality of immigrant criminals is more often specified (H1a), and it is provided as a noun (rather than an adjective; H1b) when the aggressor is an immigrant compared with an ingroup member. Furthermore, immigrant criminals are more likely further characterized by aggravating than attenuating adjectives compared with Italian criminals (H1e).

Study 2 Going beyond documenting the presence of linguistic biases in the printed media, Study 2 was designed to test the role of linguistic biases in crime news in determining a crime stereotype associated with immigrant outgroup members. Therefore, both the aggressor’s group Ó 2017 Hogrefe Publishing

membership (Italian vs. African) and the linguistic style of the crime news were manipulated. In this way, we wanted to investigate whether it is the mention of the outgroup per se that creates people’s racial bias, or if the way journalists report these crimes through the use of linguistic biases is key in determining stereotype activation. The present study used the weapon paradigm (Payne, 2001, 2006) to implicitly measure people’s activation of a crime stereotype. The weapon paradigm measures the extent to which the appearance of a Black (vs. White) face facilitates the recognition of a gun (vs. a tool). The common reported finding in the United States is that guns are detected faster in the presence of a Black face, indicating the existence of a crime stereotype toward African Americans (Payne, 2001), a finding that has been called the weapon bias. In addition, participants’ explicit attitudes toward African immigrants were measured in all conditions and before they read one of the crime stories. In this way Journal of Media Psychology (2019), 31(1), 12–23


18

J. Vaes et al., Linguistic Biases and Stereotype Activation

Table 4. Parameter estimates of the model: references  aggressor  victim  adjective type + (1|event) + (1|article) Fixed effects

Estimate

SE

df

t

p > |t|

(Intercept)

3.67e 04

3.88e 04

2.85e+01

.95

.35

Aggressor

1.84e 03

6.43e 04

2.97e+01

2.87

.007

Victim

1.140e 04

6.43e 04

2.97e+01

.177

Adjective type

5.53e 06

4.56e 04

1.19e+02

.01

.99

Aggressor: victim

8.15e 04

9.07e 04

2.94e+01

0.899

.376

.860

Victim: Adjective type

4.954e 04

7.592e 04

1.191e+02

0.652

.515

Aggressor: Adjective type

2.093e 03

7.592e 04

1.191e+02

2.757

.007

Aggressor: victim : Adjective type

2.056e 04

1.070e 03

1.191e+02

.19

.858

Figure 2. The effect of the aggressor’s membership on the type of descriptive adjectives (error bars represent the standard error of the mean).

it was possible to ascertain that participants’ prejudice and contact with the outgroup before the manipulation was similar in all conditions. At the same time, it became possible to explore the moderating effects of people’s explicit prejudice on the extent to which they were influenced by the crime story, given that some studies have found that people’s initial prejudice potentiates media priming effects (Dixon, 2006; Dixon, 2008b; Peffley, et al., 1996).

Method Participants In all, 91 White Italian participants took part in the study (40 male and 49 female; two did not report their gender). To reach a larger range of participants besides university students, participants were approached in a local community center. Volunteers were led to a room that was set up appositely, in which participants responded individually to all the measures. The participants’ age ranged from 18 to 65 years (M = 26.42, SD = 9.47). Journal of Media Psychology (2019), 31(1), 12–23

Procedure All participants first responded to an adapted version of the Subtle and Blatant Prejudice Scale (Pettigrew & Meertens, 1995). Nine affirmations were indicative of participants’ blatant prejudice (e.g., “People of color come from less able races and this explains why they are not as well off as most Italians”), while seven items were presented to measure participants’ subtle prejudice (e.g., “How different or similar do you think people of color living here are to other Italians like yourself in the values they teach their children?”). Participants were asked to indicate their agreement with each statement (1 = completely disagree to 7 = completely agree). Subsequently, participants rated a set of positive (i.e., hospitable, athletic, warm, musical, and sociable) and negative (i.e., aggressive, criminal, dishonest, hostile, and violent) stereotypical traits indicating the extent to which each described African immigrants in general (1 = not at all to 7 = completely). Afterward, two feeling thermometers were added where participants indicated – separately for the personal and the societal level – how Ó 2017 Hogrefe Publishing


J. Vaes et al., Linguistic Biases and Stereotype Activation

favorable they or Italians in general felt toward African immigrants on a vertical scale labeled at 9-degree intervals from 0° (very cold, or unfavorable) to 38° (very warm, or very favorable). Finally, participants responded to an adapted Social Contact scale (Brown, 1995) indicating on a 4-point Likert scale (1 = none, 4 = five or more) how many African immigrants lived in their neighborhood, could be found among their friends, among their colleagues, and they had (had) as a fiancé. The quality of contact was measured with a single item (i.e., “Overall, how would you rate the quality of the contacts you have had with African immigrants?”) and was answered on a bipolar scale ranging from 1 (= negative) to 7 (= positive). Manipulation After filling out the explicit questionnaires, participants were asked to read one of five news stories. In the control condition the news story talked about the beneficial effects of acupuncture in reducing physical pain. In the experimental conditions, participants read a news story about an episode of mugging. The nationality of the aggressor (African immigrant vs. Italian) and the presence (vs. absence) of linguistic biases were varied between subjects in the four experimental conditions. The crime victim was Italian for immigrant criminals and an immigrant for Italian criminals, providing an intergroup scenario. The linguistic biases were taken from the results of Study 1 and from the literature on linguistic biases in intergroup relations. Specifically, the nationality of the aggressor was always identified with a noun (e.g., the Nigerian, the Italian). Aggravating adjectives (e.g., illegal, armed) were used to describe the aggressor and his acts, while attenuating adjectives were used to describe the victim (e.g., very young, defenseless). The crime itself was described Using the active form and an animal metaphor describing the criminal was also included (see Goff, Eberhardt, Williams, & Jackson, 2008). Differently, the language used in the unbiased condition was factually the same as in the other article, but the language remained more neutral (see Electronic Supplementary Materials, ESM 2 for examples of the full articles in English. The Italian originals can be obtained from the first author on request). Weapon Paradigm After the manipulation, participants performed a priming task that was identical to that used by Payne (2001; Study 1). Participants were instructed to classify, as fast and as accurately as possible, a target picture as either a gun or a tool. Each target was preceded by the picture of a Black or White face (priming stimulus). Participants were instructed to ignore the primes and focus on the target pictures. The prime remained on the screen for 200 ms and was immediately replaced by the target picture lasting Ó 2017 Hogrefe Publishing

19

200 ms and replaced by a white screen until participants responded. After each response there was a 500-ms intertrial interval. First, participants completed 10 practice trials, after which they responded to 128 test trials. Four different stimuli were used for each prime (White or Black faces) or target (tool or gun) category. The prime–target pairs were presented in a random order. Crime Punishment Only in the experimental conditions, participants were asked to evaluate the crime and indicate what punishment they would give to the aggressor on a scale ranging from 1 to 7 (1 = no punishment, 2 = less than 1 year of imprisonment, 3 = between 1 and 5 years of imprisonment, 4 = between 5 and 10 years of imprisonment, 5 = between 10 and 20 years of imprisonment, 6 = life imprisonment, and 7 = death sentence). At the end of this judgment task, participants were fully debriefed and thanked for their participation.

Results Explicit Measures The explicit attitudes measures showed a satisfactory reliability (α = .82, M = 2.52, SD = 0.98 for blatant prejudice; α = .76, M = 3.69, SD = 0.81 for subtle prejudice; α = .77, M = 4.98, SD = 1.01 for the attribution of positive stereotypes; α = .88, M = 4.55, SD = 1.22 for the attribution of negative stereotypes; α = .64, M = 1.17, SD = 0.72 for the quantity of contact; and M = 5.21, SD = 1.38 for the quality of contact) allowing us to calculate mean scores for each participant. Each index was analyzed in a one-way ANOVA comparing the five conditions, but no significant differences between the various conditions emerged (all p values > .51). Weapon Race Bias In order to calculate the weapon race bias, outliers were removed using the same cut-off criteria as in Payne (2001), so that reaction times (RT) faster than 100 ms and slower than 1000 ms, as well as incorrect responses, were dropped. The analyses were conducted on a natural log-transformation of the original RTs. A weapon race index (WRI) was computed subtracting the mean of the stereotype congruent trails (Black–gun, White–tool) from the mean of stereotype incongruent trials (White–gun, Black– tool) and. As such, higher values indicated a stronger bias in identifying guns compared with tools after the presentation of a Black versus a White face prime. As expected, a one-way ANOVA revealed a main effect of the experimental condition on the WRI, F(4, 86) = 2.60, p = .04, ηp2 = .11. To better understand the meaning of this effect, two independent contrasts were calculated comparing the influence of linguistic biases in the case of an immigrant or an Italian aggressor. For each type of target a contrast code was given Journal of Media Psychology (2019), 31(1), 12–23


20

to the different language conditions (2 = biased condition, 1 = unbiased condition, and 1 = control condition). Confirming H2, contrast comparisons indicate that the WRI was stronger in the immigrant condition that used linguistic biases (M = 0.10, SD = 0.07) compared with both the immigrant unbiased (M = 0.03, SD = 0.06) and the control condition (M = 0.05, SD = 0.08), t(51) = 2.83, p = .007, d = .79. Importantly, the immigrant unbiased and the control condition did not differ from one another. In the case of an Italian aggressor, no significant differences between the various conditions emerged (see Table 5). Crime Punishment The level of punishment participants chose for the aggressor was analyzed comparing the role of linguistic biases separately for each type of target (African immigrant vs. Italian). In the immigrant conditions, participants expressed a marginally higher level of punishment for the African perpetrator in the biased (M = 3.50, SD = 1.15) compared with the unbiased condition (M = 2.78, SD = 1.11), t(34) = 1.91, p = .06, d = .66. By contrast, there was no difference in the level of punishment for the Italian perpetrator between the biased (M = 3.47, SD = 0.80) and unbiased conditions (M = 3.42, SD = 0.84), t(34) = .18, ns. No moderating role of the participants’ explicit judgments on the WRI and their judgment of the crime was found.

Discussion Study 2 demonstrated that the content and stylistic characteristics of crime news about an outgroup aggressor can change our perception of the whole outgroup category associating them more easily with a crime stereotype. Comparing their performance with the control condition, those participants who read a crime news article that portrayed a black African criminal using biased language increased their tendency to associate guns with African immigrants in general. When the article had the same factual content, but used a more neutral language or when the aggressor was an ingroup member, no differences with the control condition were observed. As such, these findings confirm H2 and indicate that not the simple mention, but the way members of racial/ethnic minorities are depicted in the media is detrimental to the development and maintenance of crimerelated stereotypes that typically target such social groups. In this study, participants were also required to assign a punishment to the aggressor. Confirming H3 only in the case of an outgroup criminal, biased language use tended to increase participants’ punishment, a judgment that could truly influence the fate of crime suspects belonging to a racial/ethnic minority. Unlike Peffley and colleagues (1996), none of the reported findings were moderated by participants’ initial explicit Journal of Media Psychology (2019), 31(1), 12–23

J. Vaes et al., Linguistic Biases and Stereotype Activation

Table 5. Mean weapon race bias as a function of condition (Study 2) Condition

M

SD

Immigrant Biased

.10

a

.07

Immigrant Unbiased

.03b

.06

Control

.05b

.08

Condition

M

SD

Italian Biased

.07

a

.07

Italian Unbiased

.05a

.06

Control

.05a

.08

Note. Numbers with a different superscript differ from each other within the same column, p < .05.

attitudes toward African immigrants. Both individuals with positive or negative explicit attitudes or outgroup contact were equally influenced by the changes in the crime reports suggesting that biased language use might be equally important in creating as well as in maintaining people’s stereotypic associations with members of racial/ethnic minorities. It is important to note, however, that we cannot rule out the possibility that the measurement of participants’ explicit attitudes at the beginning of the study influenced the reading process of the crime news articles. These explicit measures might have made participants more aware of the existence of certain stereotypes and emphasized our stereotype activation findings. Still, given that these attitudes were measured in all conditions, this procedure cannot account for the relative differences in stereotype activation between conditions. Another limitation of the current study lies in the fact that we manipulated the nationality of both the aggressor and the victim. This procedure was followed in order to create an intergroup situation in all conditions. Still, future research could add the within-group situations (ingroup aggressor and victim/outgroup aggressor and victim) to ascertain whether biased language use triggers the activation of the crime stereotype regardless of the nationality of the victim (both ingroup and outgroup). All in all, increasing the awareness among newsmakers about their language use is one of the most concrete applicative implications of the current study. Journalists and editors decide how certain news facts are told to the general public often using a language that emphasizes and potentially sensationalizes certain details. While this might be done to attract the readers’ attention and boost sales, the present study showed that such seemingly small linguistic choices increase people’s implicit stereotypes only when the aggressor is a member of a racial/ethnic minority.

General Discussion The mass media provide people with a unique window on reality. Especially when focusing on racial/ethnic minorities, a group with which the general public often has less familiarity, media input can become pervasive and the only source of information. The way in which the media filter and shape reality then becomes an important variable in molding and even distorting people’s beliefs and perceptions about the Ó 2017 Hogrefe Publishing


J. Vaes et al., Linguistic Biases and Stereotype Activation

world they live in. When focusing on people’s beliefs about racial/ethnic minorities, the most common studied subjects have been crime and criminality, and media portrayals that link crime to racial/ethnic outgroups (Mutz & Goldman, 2010). This emphasis is warranted given the consistent overrepresentation of racial/ethnic minorities in law-offending roles, not only compared with the White majority, but also in relation to actual crime statistics (Di Nicola & Caneppele, 2004; Dixon & Linz, 2000). Sometimes it seems to be the only context in which they appear in the media. The association “Carta di Roma” (see Morcellini, Binotto, Bruno, & Lai, 2009), for example, that monitors the Italian media for biased reporting of racial/ethnic minorities, has found that 76.2% of the times immigrants are mentioned in the Italian media, they appear in crime news. But is it indeed enough to simply mention racial/ethnic minority members in the context of a crime story to activate people’s crime-related stereotypes? Early priming studies in social psychology have demonstrated that the presentation of a group member (i.e., name or image) is a sufficient condition to activate the associated group stereotypes in perceivers’ minds (e.g., Bargh, 1999; Devine, 1989), therefore suggesting an affirmative answer to this question. However, recent developments in this field have underlined the flexibility of such effects emphasizing the importance of personal, contextual, and motivational variables (e.g., Blair, 2002; Dasgupta, 2013) in the activation of people’s prejudice and stereotypes. In line with these more recent findings, the present research effort documented and manipulated linguistic biases in crime news that were expected to moderate the activation of a crime stereotype toward members of racial/ethnic minorities. In Study 1, a set of linguistic biases were analyzed in crime news articles that described comparable crimes committed by an ingroup or an outgroup member. Results indicated that when the crime was committed by an outgroup member, the nationality of the perpetrator was not only mentioned more frequently, it appeared in most cases as a noun. This means that the mass media more likely refer to a single outgroup aggressor using a categorical label. In line with Carnaghi and colleagues’ (2008) work, such references essentialize the person’s criminal nature and link it to his or her group membership making generalizations more likely. In addition, even though the crimes that were committed by ingroup and outgroup members were similar in terms of gravity and the conditions in which they occurred, the latter ones were described using more aggravating and less attenuating adjectives. Even though language abstraction has been found in a large range of spontaneously written texts and has been used to favor the ingroup over the outgroup (see Maass, 1999, for a review), the crimes described in the current study did not use a significantly more abstract language Ó 2017 Hogrefe Publishing

21

when the crime was committed by an outgroup rather than an ingroup member, at least, at the level of the verbs that were used to describe the crimes. When having the choice to refer to the nationality of the outgroup criminal in terms of an adjective or a noun, however, clearly the most abstract noun was used suggesting that language abstraction occurred in the current crime news articles, but not at all possible levels. In addition, no evidence was found for the more consistent use of the active voice when describing the criminal act of an outgroup compared with an ingroup member. This hypothesis was derived from work on sexual harassment and rape in which the passive voice is regularly used increasing the co-responsibility of the victim. The current results seem to suggest that such a mechanism cannot be extended to crime news articles in general. In Study 2, four different experimental conditions were created that varied the group membership of the perpetrator and the victim together with the use of a neutral or biased language. Results suggested that the simple mention of an outgroup target in the realm of a crime story was not a sufficient condition to activate crime-related stereotypes. Only when biased language was used to describe an outgroup perpetrator’s acts was an increase in stereotype activation observed. Besides a theoretical advance, these findings also have practical implications. Newsmakers in general and journalists in particular make choices on what events will be covered in the media and how they are reported. The current research especially speaks to the stylistic/linguistic choices that are made when portraying racial/ethnic minority members in crime reports. While newsmakers might be under pressure to color and sensationalize their news reports, the present findings plead for a more factual-based and neutral language when reporting crimes in the media, especially when the crime suspect is a member of a racial/ethnic minority. Electronic Supplementary Materials The electronic supplementary material is available with the online version of the article at http://dx.doi.org/10.1027/ 1864-1105/a000216 ESM 1. Table (PDF) Detailed description of the articles that were analyzed in Study 1. ESM 2. Text (PDF) Examples of the articles that we used in Study 2.

References Bargh, J. A. (1999). The cognitive monster: The case against the controllability of automatic stereotype effects. In C. Shelly & Y. Trope (Eds.), Dual-process theories in social psychology (pp. 361–382). New York, NY: Guilford Press.

Journal of Media Psychology (2019), 31(1), 12–23


22

Bates, D., Maechler, M., Bolker, B., & Walker, S. (2014). lme4: Linear mixed-effects models using Eigen and S4. R package version 1.17. Retrieved from http://CRAN.R-project.org/package=lme4 Blair, I. V. (2002). The malleability of automatic stereotypes and prejudice. Personality and Social Psychology Review, 6, 242–261. doi: 10.1207/S15327957PSPR0603_8 Bogle, D. (2001). Toms, coons, mulattoes, mammies, and bucks: An interpretive history of Blacks in American films. New York, NY: Continuum. Bohner, G. (2001). Writing about rape: Use of the passive voice and other distancing text features as an expression of perceived responsibility of the victim. British Journal of Social Psychology, 40, 515–529. doi: 10.1348/014466601164957 Brader, T., Valentino, N. A., & Suhay, E. (2008). What triggers public opposition to immigration? Anxiety, group cues, and immigration threat. American Journal of Political Science, 52, 959–978. doi: 10.1111/j.1540-5907.2008.00353.x Brown, R. (1995). Psicologia Sociale del Pregiudizio [Social psychology of prejudice]. Bologna, Italy: Il Mulino. Carnaghi, A., Maass, A., Gresta, S., Bianchi, M., Cadinu, M., & Arcuri, L. (2008). Nomina sunt omina On the inductive potential of nouns and adjectives in person perception. Journal of Personality and Social Psychology, 94, 839–859. doi: 10.1037/ 0022-3514.94.5.839 Coenen, L. H. M., Hedebouw, L., & Semin, G. R. (2006). The linguistic category model (LCM) manual. Amsterdam, The Netherlands: Free University Amsterdam. Dasgupta, N. (2013). Implicit attitudes and beliefs adapt to situations: A decade of research on the malleability of implicit prejudice, stereotypes, and the self-concept. Advances in Experimental Social Psychology, 47, 233–279. doi: 10.1016/ B978-0-12-407236-7.00005-X Devine, P. G. (1989). Stereotypes and prejudice: Their automatic and controlled components. Journal of Personality and Social Psychology, 56, 5–18. doi: 10.1037/0022-3514.56.1.5 Di Nicola, A, & Caneppele, S (2004). Media e Criminalità: La Rappresentazione della Criminalità nei Giornali della Provincia di Padova [Media and criminality: Crime representations in the local journals of the province of Padova]. Retrieved from http://transcrime.cs.unitn.it/tc/37.php Dixon, T. L. (2006). Schemas as average conceptions: Skin tone, television news exposure, and culpability judgments. Journalism and Mass Communication Quarterly, 83, 131–149. doi: 10.1177/107769900608300109 Dixon, T. L. (2008a). Network news and racial beliefs: Exploring the connection between national television news exposure and stereotypical perceptions of African Americans. Journal of Communication, 58, 321–337. doi: 10.1111/j.1460-2466.2008.00387.x Dixon, T. L. (2008b). Who is the victim here?: The psychological effects of overrepresenting White victims and Black perpetrators on television news. Journalism: Theory, Practice, & Criticism, 9, 582–605. doi: 10.1177/1464884908094160 Dixon, T. L., & Linz, D. (2000). Overrepresentation and underrepresentation of African Americans and Latinos as lawbreakers on television news. Journal of Communication, 50, 131–154. doi: 10.1111/j.1460-2466.2000.tb02845.x Dixon, T. L., & Maddox, K. B. (2005). Skin tone, crime news, and social reality judgments: Priming the stereotype of the dark and dangerous Black criminal. Journal of Applied Social Psychology, 35, 1555–1570. doi: 10.1111/j.1559-1816.2005.tb02184.x Dubrofsky, R. E., & Hardy, A. (2008). Performing race in Flavor of Love and The Bachelor. Critical Studies in Media Communication, 25, 373–392. doi: 10.1080/15295030802327774 Entman, R. M. (1992). Blacks in the news: Television, modern racism and cultural change. Journalism Quarterly, 69, 341–361. doi: 10.1177/107769909206900209

Journal of Media Psychology (2019), 31(1), 12–23

J. Vaes et al., Linguistic Biases and Stereotype Activation

Entman, R. M. (1994). Representation and reality in the portrayal of Blacks on network television news. Journalism Quarterly, 71, 509–520. doi: 10.1177/107769909407100303 Entman, R. M., & Rojecki, A. (2000). The Black image in the White mind: Media and race in America. Chicago, IL: University of Chicago Press. Fazio, R.H., & Olson, M. A. (2003). Implicit measures in social cognition research: Their meaning and use. Annual Review of Psychology, 54, 297–327. doi: 10.1146/annurev.psych.54. 101601.145225 Fernández, I., Igartua, J. J., Moral, F., Palacios, E., Acosta, T., & Muñoz, D. (2012). Language use depending on news frame and immigrant origin. International Journal of Psychology, 48, 772–784. doi: 10.1080/00207594.2012.723803 Ford, T. E. (1997). Effects of stereotypical television portrayals of African-Americans on person perception. Social Psychology Quarterly, 60, 266–275. http://www.jstor.org/stable/2787086 Geschke, D., Sassenberg, K., Ruhrmann, G., & Sommer, D. (2010). Effects of linguistic abstractness in the mass media: How newspaper articles shape readers’ attitudes toward migrants. Journal of Media Psychology: Theories, Methods, and Applications, 22, 99–104. doi: 10.1027/1864-1105/a000014 Goff, P. A., Eberhardt, J. L, Williams, M. J., & Jackson, M. C. (2008). Not yet human: Implicit knowledge, historical dehumanisation, and contemporary consequences. Journal of Personality and Social Psychology, 94, 292–306. doi: 10.1037/0022-3514.94.2.292 Graf, S., Bilewicz, M., Finell, E., & Geschke, D. (2012). Nouns cut slices: Effects of linguistic forms on intergroup bias. Journal of Language and Social Psychology, 32, 62–83. doi: 10.1177/ 0261927X12463209 Greenberg, B. S., & Baptista-Fernandez, P. (1980). Hispanic-Americans: The new minority on television. In B. S. Greenberg (Ed.), Life on television: Content analysis of US TV drama (pp. 3–13). Norwood, NJ: Ablex. Hasinoff, A. A. (2008). Fashioning race for the free market on America’s Next Top Model. Critical Studies in Media Communication, 25, 324–343. doi: 10.1080/15295030802192012 Igartua, J. J. (2013). Attitudinal impact and cognitive channeling of immigration stereotypes through the news. Revista Latina de Comunicación Social, 68, 599–621. doi: 10.4185/RLCS-2013-992en Johnson, J., Adams, M., Hall, W., & Ashburn, L. (1997). Race, media, and violence: Differential racial effects of exposure to violent news stories. Basic and Applied Social Psychology, 19, 81–90. doi: 10.1207/s15324834basp1901_6 Maass, A. (1999). Linguistic intergroup bias: Stereotype perpetuation through language. Advances in Experimental Social Psychology, 31, 79–121. doi: 10.1016/S0065-2601(08)60272-5 Mastro, D. E., Behm-Morawitz, E., & Ortiz, M. (2007). The cultivation of social perceptions of Latinos: A mental models approach. Media Psychology, 9, 347–365. doi: 10.1080/15213260701286106 Mastro, D. E., & Greenberg, B. S. (2000). The portrayal of racial minorities on prime time television. Journal of Broadcasting and Electronic Media, 44, 690–703. doi: 10.1207/ s15506878jobem4404_10 Mastro, D., & Kopacz, M. (2006). Media representations of race, prototypicality, and policy reasoning: An application of selfcategorization theory. Journal of Broadcasting & Electronic Media, 50, 305–322. doi: 10.1207/s15506878jobem5002_8 Mastro, D. E., Lapinski, M. K., Kopacz, M. A., & Behm-Morawitz, E. (2009). The influence of exposure to depictions of race and crime in TV news on viewer’s social judgments. Journal of Broadcasting and Electronic Media, 53, 615–635. doi: 10.1080/ 08838150903310534 Mastro, D. E., & Robinson, A. L. (2000). Cops and crooks: Images of minorities on primetime television. Journal of Criminal Justice, 28, 385–396. doi: 10.1016/S0047-2352(00)00053-2

Ó 2017 Hogrefe Publishing


J. Vaes et al., Linguistic Biases and Stereotype Activation

Morcellini, M., Binotto, M., Bruno, M., & Lai, V. (2009). Ricerca Nazionale su Immigrazione e Asilo nei Media Italiani [National research on immigration and asylum in the Italian media] University of Rome, La Sapienza. Retrieved from http://www.unhcr. it/news/dir/91/view/733/ricerca-nazionale-su-immigrazione-e-asilonei-media-italiani-73300.html Mutz, D. C., & Goldman, S. K. (2010). Mass media. In J. F. Dovidio, M. Hewstone, P. Glick, & V. M. Esses (Eds.), The Sage handbook of prejudice, stereotyping, and discrimination (pp. 241–258). Thousand Oaks, CA: Sage. Oliver, M. B. (1994). Portrayals of crime, race, and aggression in ‘‘reality-based’’ police shows: A content analysis. Journal of Broadcasting and Electronic Media, 38(2), 179–192. doi: 10.1080/08838159409364255 Payne, B. K. (2001). Prejudice and perception: The role of automatic and controlled processes in misperceiving a weapon. Journal of Personality and Social Psychology, 81, 181–192. doi: 10.1037/0022-3514.81.2.181 Payne, B. K. (2006). Weapon bias: Split-second decisions and unintended stereotyping. Current Directions in Psychological Science, 15, 287–291. doi: 10.1111/j.1467-8721.2006. 00454.x Peffley, M., Shields, T., & Williams, B. (1996). The intersection of race and crime in television news stories: An experimental study. Political Communication, 13, 309–327. doi: 10.1080/ 10584609.1996.9963120 Pettigrew, T. F., & Meertens, R. W. (1995). Subtle and blatant prejudice in Western Europe. European Journal of Social Psychology, 25, 57–75. doi: 10.1002/ejsp.2420250106 Pinheiro, J., & Bates, D. (2006). Mixed-effects models in S and S-PLUS. New York: Springer Science & Business Media. Power, J. G., Murphy, S. T., & Coover, G. (1996). Priming prejudice: How stereotypes and counter-stereotypes influence attribution of responsibility and credibility among ingroups and outgroups. Human Communication Research, 23, 36–58. doi: 10.1111/ j.1468-2958.1996.tb00386.x R Core Team. (2013). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Retrieved from http://www.R-project.org/ Ramasubramanian, S. (2010). Television viewing, racial attitudes, and policy preferences: Exploring the role of social identity and intergroup emotions in influencing support for affirmative action. Communication Monographs, 77, 102–120. doi: 10.1080/03637750903514300 Ramasubramanian, S., & Oliver, M. B. (2007). Activating and suppressing hostile and benevolent racism: Evidence for comparative media stereotyping. Media Psychology, 9, 623–646. doi: 10.1080/15213260701283244 Ramirez Berg, C. (1990). Stereotyping in films in general and of the Hispanic in particular. Howard Journal of Communication, 2, 286–300. doi: 10.1080/10646179009359721 Shugart, H. (2006). Ruling class: Disciplining class, race, and ethnicity in television reality court shows. Howard Journal of Communication, 17, 79–100. doi: 10.1080/10646170600656839 Semin, G. R., & Fiedler, K. (1991). The linguistic category model, its bases, applications and range. European Review of Social Psychology, 2, 1–30. doi: 10.1080/14792779143000006 Tan, A., Fujioka, Y., & Tan, G. (2000). Television use, stereotypes of African Americans and opinions on affirmative action: An effective model of policy reasoning. Communication Monographs, 67, 362–371. doi: 10.1080/03637750009376517 Vaes, J., Latrofa, M., Vieno, A., & Pastore, M. (2015). Exposure to politicized media and prejudice against immigrants in Italy: Identifying its impact and psychological mediators. Psicologia Sociale, 2, 141–160.

Ó 2017 Hogrefe Publishing

23

Vergani, L. (2009, October 10). È vero che il 72% della stampa è di sinistra? Facciamo quattro conti [Is it true that the 72% of the press is left-wing oriented? Taking stock]. Retrieved from http://www.verosimile.it/post/-vero-che-il-72-della-stampa–disinistra-facciamo-quattro-conti/catid/24 Weisbuch, M., Pauker, K., & Ambady, N. (2009). The subtle transmission of race bias via televised nonverbal behavior. Science, 326, 1711–1714. doi: 10.1126/science.1178358 Received May 4, 2015 Revision received July 26, 2016 Accepted August 17, 2016 Published online April 28, 2017 Jeroen Vaes University of Trento Department of Psychology and Cognitive Sciences Corso Bettini 31 38068 Rovereto Italy jeroen.vaes@unitn.it

Jeroen Vaes, PhD is an associate professor of social psychology at the University of Trento, Italy. His current research interests focus primarily on processes of dehumanization and sexual objectification, social perceptions in intergroup relations and media effects.

Marcella Latrofa has obtained her PhD in Cognitive-social psychology at the University of Padova, Italy. Her research interests focus on intergroup relations, discrimination processes, stereotyping and sexual objectification. Currently, she is a professional psychotherapist.

Caterina Suitner, PhD, is currently an assistant professor of social psychology at the University of Padova, Italy. Her research focuses on the relation between language and social cognition and on the role of writing direction on the representation of social targets. She is also interested in cross-cultural psychology and stereotyping.

Luciano Arcuri is a retired professor of social psychology at the University of Padova, Italy. His main research interest is in the area of social cognition, with particular interest in the use of techniques involving implicit measures.

Journal of Media Psychology (2019), 31(1), 12–23


Original Article

Why User Comments Affect the Perceived Quality of Journalistic Content The Role of Judgment Processes Patrick Weber, Fabian Prochazka, and Wolfgang Schweiger Department of Communication, University of Hohenheim, Stuttgart, Germany Abstract: User comments on news websites are frequently uncivil and are not supported by reasoned argumentation. These characteristics can have negative effects on the perceived quality of the commented-on journalistic content, yet to date, it remains unclear how such effects occur. We propose three mechanisms that assume that the effect of user comments depends on how deliberately and elaborately the quality of the commented-on news item is judged. We conducted an experiment (N = 633) in which we varied the level of civility and reasoning in the comments accompanying a news article and the brand of the news website on which it was presented. The results showed that a lack of reasoning in the comments decreased the perceived quality of the news item irrespective of brand awareness, but only with high elaboration during judgment. Incivility in the comments decreased the perceived quality of the journalistic content, but only with low elaboration, and only with an unknown news brand. We discuss different psychological mechanisms that can explain this pattern of effects. Keywords: user comments, user-generated content, online news, judgment, civility

The implementation of comment features on news websites gave rise to high hopes. Scholars have highlighted their potential for extending the democratic public sphere of deliberative discussion and opinion formation (e.g., Weber, 2014). Journalists and the news industry anticipated that user-generated content such as comments would help to increase website traffic, build brand loyalty, and help them to remain competitive in the news market (Vujnovic, 2011). The reality, however, has been largely disappointing and has increasingly caused concern among scholars, journalists, and the public. The main reason is that the quality of the discussion in the comment sections is frequently low. A substantial proportion of the comments are unnecessarily disrespectful and impolite, that is, user comments are often uncivil (e.g., Coe, Kenski, & Rains, 2014), and comments are often lacking in reasoned argumentation (Ruiz et al., 2011). Facing this reality, journalists fear that they pay dearly for the increased website loyalty of active commenters. One central concern is that the low quality of user comments might affect the perceived quality of their work and damage their brands (e.g., Meltzer, 2014). Recent research (Houston, Hansen, & Nisbett, 2011; Sikorski & Haenelt, 2016; Wallsten & Tarsi, 2015), especially research on the impact of low-quality comments, Journal of Media Psychology (2019), 31(1), 24â&#x20AC;&#x201C;34 DOI: 10.1027/1864-1105/a000217

shows that these concerns are justified. Ash, Peeling, and Hettinga (2010) found that readers liked journalistic articles less when they were accompanied by banal usergenerated comments. Kim and Sun (2006) found that low-quality comments negatively influenced the perceived quality of a news article (cited from Lee, 2012). Prochazka, Weber, and Schweiger (2016) showed that both incivility and a lack of reasoning in user comments can have negative effects on the perceived quality of the journalistic article. The present study extends this line of research by attempting to explain how the quality of user-generated comments affects the perceived quality of the commented-on journalistic content. We argue that the characteristics of user comments, especially incivility and a lack of reasoning, are cues that can affect perceptions of quality. To advance the theory on the effects of user-generated comments, we propose three mechanisms that potentially explain these effects and derive boundary conditions for them. An experimental study is conducted to test how these boundary conditions influence the effects of incivility and lack of reasoning. The findings contribute to our understanding of how user comments affect the audienceâ&#x20AC;&#x2122;s quality assessments. Ă&#x201C; 2017 Hogrefe Publishing


P. Weber et al., Why User Comments Affect Perceived News Quality

A Theory of How Audiences Judge the Quality of Journalistic Content When asked about their perceptions of the quality of journalistic content, recipients largely refer to the characteristics of news that correspond with professional journalistic values (cf. van der Wurff & Schoenbach, 2014). Such values are rooted in normative conceptions about the role of journalism in society and what standards it should comply with to fulfill this function. Urban and Schweiger (2014) reviewed the literature on such normative quality indicators and described six core values: diversity, the inclusion of a variety of points of views in the news; impartiality, the avoidance of bias; relevance, a focus on information of current importance to society; accuracy, the truthfulness of information; comprehensibility, the presentation of news in a digestible fashion; and compliance with ethical standards such as tolerance and respect. But recipients’ judgments about the quality of news do not focus exclusively on such criteria. Research shows that beyond receiving information and orientation about current affairs, the audience seeks other gratifications, especially entertainment and relaxation (e.g., van der Wurff & Schoenbach, 2014). Therefore, affective responses such as being entertained or feeling interested are also important facets of users’ judgments of the quality of a journalistic message (Meijer, 2013). Although news consumers appreciate journalistic core values, they are not very good at detecting differences between news items that vary in their adherence to journalistic core values (Urban & Schweiger, 2014). We propose two explanations for why it might be difficult for users to evaluate the quality of journalistic content on the Internet, at least when judging whether the article meets core quality standards. First, on the basis of the distinction between online and memory-based judgments (Hastie & Park, 1986), we assume that quality judgments are generally not formed online, but are based on memory. A judgment that is made online is formed at the time the relevant information is taken in, that is, inferences regarding quality are made while processing the news item and they are stored in memory together with the article’s content. Such online judgments are most likely when the individual knows that the respective judgment will be needed, when the individual is motivated to form a judgment, and when the individual has sufficient resources to encode, understand, store, and simultaneously evaluate relevant information (Hastie & Park, 1986; Hastie & Pennington, 1989). It is unlikely that these conditions are met when reading everyday news. Most of the time, readers’ primary goal is to inform themselves or to find a distraction, not to evaluate journalistic quality. Furthermore, we assume that quality Ó 2017 Hogrefe Publishing

25

judgments are usually not of high personal relevance to readers. Therefore, most quality judgments are likely to be memory based, which means that judgment-relevant inferences are not made and stored when the stimulus is perceived, but judgment-relevant information about the object is retrieved from memory when a judgment is called for. This implies that judgments about journalistic content will depend on what judgment-relevant information the individual can recall about the previously read news item. Considering the characteristics of information processing and news reading on the Internet, it seems unlikely that recipients are able to retrieve all of the information from memory that might be relevant for quality judgments. Reading online tends to involve scanning and cursory reading (Weinreich, Obendorf, Herder, & Mayer, 2008). Reading from a screen also tends to be shallower than off-screen reading, and it is more difficult to sustain high reading performance over time (Dillon, 1992). This can impair reading comprehension (Wästlund, Reinikka, Norlander, & Archer, 2005). Moreover, online-specific features of news presentation such as hypertext linking tend to increase the cognitive demands of news consumption (relative to traditional news formats; e.g., Eveland & Dunwoody, 2002). This not only depletes the cognitive resources necessary for making online judgments, but can also, in turn, impair the recall of information from the news. The likely outcome of these characteristics of news processing online is that users simply do not encode and/or store all of the information necessary for evaluating the quality of a news item in terms of journalistic core values (e.g., whether the article featured different viewpoints), and hence this information cannot be retrieved when attempting to form a memory-based judgment. The second reason it might be difficult for users to evaluate the quality of journalistic content is that recipients’ interpretations of the core values may not be as specific as normative theory assumes them to be. Normative theories of journalistic quality propose observable indicators for the various quality criteria; for example, providing a balanced representation of the different viewpoints on an issue is indicative of impartiality. If recipients have only a vague conception of the core values, their conceptual knowledge about journalistic quality will not prepare them well to attend to the respective features of the news item, such as whether it offers different viewpoints and represents them in a balanced fashion. Consequently, selectively stored information from the news item may not include this judgment-relevant information. When later asked to evaluate the quality of the news item, the recipients might be unable to retrieve the relevant information from memory, which would hinder them from accurately assessing whether it complied with the respective journalistic standard. Journal of Media Psychology (2019), 31(1), 24–34


26

Information processing theories such as the elaboration likelihood model (Petty & Cacioppo, 1986) suggest that in situations where recipients have low motivation and/or an inability to make accurate judgments, they are highly likely to use other information than just the content of the message to judge the quality of the journalistic content. Information that is not processed as part of the substantial and judgment-relevant content of a message is commonly referred to as peripheral cues. In line with the assumption that peripheral cues affect quality judgments, research shows that the attributes of the sender of a news article affect how recipients judge its quality. Fichter and Jonas (2008) manipulated the name of the news organization and the layout of an article in an experimental study, and found that when an article was attributed to a news brand that was considered to be reputable and deliver high-quality journalism, readers rated the article as higher in quality than when the same article was attributed to a news brand that was considered more sensational and sloppy. The results of this study showed that the name of the news brand serves as a peripheral cue that activates the brand image, and that readers rely on the brand image rather than the content of the article to judge the article’s quality. We assume that beyond the brand name, the characteristics of user comments can also serve as peripheral cues in judging the quality of a news item. In the following section, we speculate about the mechanisms by which incivility and a lack of reasoning in user comments might affect quality judgments and derive testable predictions for the boundary conditions on the effects of user comments.

How and When the Quality of Comments Affects News Quality Judgments Thus far, we have argued that users’ quality judgments comprise judgments of a news item’s professional journalistic values and hedonic qualities, that such judgments are likely to be based on information that can be recalled when the judgment is made, and that users have a low motivation and limited capability to make such judgments in Internet news use. Therefore, we suggested, users resort to peripheral cues such as incivility and a lack of reasoning in the comments that accompany a news item to come to a judgment. To understand exactly how these peripheral cues might affect quality judgments, we complement the theory with assumptions about the relative salience of news vs. comments characteristics during news use and their relative accessibility during judgment. When a reader is exposed to an article that is accompanied by user comments, features such as name-calling, vulgar language, short sentences

Journal of Media Psychology (2019), 31(1), 24–34

P. Weber et al., Why User Comments Affect Perceived News Quality

(i.e., indicators of incivility, cf. Coe et al., 2014), and probably also a lack of substantiation for stated opinions (i.e., a lack of reasoning) are more salient than the core qualities of the journalistic content such as diversity or relevance. Such salient features are likely to be encoded and stored in memory with less conscious effort (Lang, 2009) than that required for the complex information represented in the journalistic text. Consequently, this information is presumably more accessible when retrieving information from memory to judge the news item’s quality (cf. Koriat & Ma’ayan, 2005, for evidence that suggests a positive association between ease of encoding and accessibility in memory). On the basis of this, we see three mechanisms by which information from the comments might affect memory-based judgments of a news item’s quality. First, if the content of the comments is more accessible and comes to mind more easily when forming a memorybased judgment, it could simply be misremembered as stemming from the professionally produced news item rather than from the user-generated comments. Or it is remembered without source information so that the judge is unaware of the information’s origin. In any case, if it is used for inferring the quality of the professional journalistic content, the impact of comment characteristics is a result of misattribution. Second, highly salient characteristics such as incivility could serve as heuristic cues. Heuristics can be defined as learned if–then associations (Chen & Chaiken, 1999) in which the if part is a heuristic cue, and the then part is an attribute of an object that is to be judged, for example, its quality. Judgment heuristics are based on knowledge and beliefs; hence, if a feature of user comments acts as a valid heuristic cue in judging a news item’s quality, then there must be a belief that this feature of users’ discourse is predictive of a journalistic article’s quality. Such a belief could be based on users’ experience that only journalistic content that is somehow “problematic” (e.g., biased) causes “problematic” (i.e., uncivil) comments. Indeed, Ksiazek, Peer, and Zivic (2015) found that biased news items were accompanied by more uncivil user comments than were nonbiased news items. As a result of observing such an association during news use, users might come to believe that the quality of comments and the quality of the journalistic content that triggered these comments are associated, and hence they might use the salient features of the comments to judge a news item’s quality heuristically (heuristic judgment). Third, obtrusive features of the comments, such as incivility, could induce a negative affect during reception that might lead to affect-congruent, that is, negatively biased, subsequent judgments (affect induction; see Kühne, 2012).

Ó 2017 Hogrefe Publishing


P. Weber et al., Why User Comments Affect Perceived News Quality

Misattribution and heuristic judgment are both based on memory of information from user comments that come to mind when a judgment of the news item’s quality is called for. However, when the impact of comment features on the judgment is mediated by misattribution, this is purely a result of false memory. Mediation through heuristic judgment, on the other hand, requires a belief or background knowledge that makes the comment feature a heuristic cue, that is, which makes it a predictor of the to-be-judged criterion (see Krueger, 2012). Such a belief and the reliance on a decision rule are not necessary for misattribution to occur. Furthermore, both misattribution and heuristic judgment differ from affect induction in that the latter does not depend on memory of comment features during judgment. If the effects are mediated by affect, the process is that an affective state is induced during reception, and subsequent judgments are primarily based on this affect, not on recalled information. Through which of these mechanisms the characteristics of user comments affect quality judgments is likely to depend on how much the recipient elaborates when making the judgment. We assume that it requires more cognitive effort to recognize and remember whether there was a reasoned argument than whether there was incivility (e.g., rude language, name-calling) because the former contains more complex information. As a lack of substantiation for stated opinions is less salient than incivility, recipients need to search more intensely to retrieve that information from memory. When they do so, the lack of reasoning may be misremembered as being a quality of the journalistic content rather than the comments. However, this effect is only likely to occur when recipients take the time to actually think about the quality of the news and try to retrieve relevant information from memory while judging the article’s quality, that is, when they elaborate. We therefore expect that the degree of elaboration when making a judgment moderates the effect of a lack of substantiation in user comments on assessments of a news item’s quality, and predict a negative effect of lack of substantiation only for judgments that involve a high level of elaboration (Hypothesis 1). The effect of incivility is more likely to be mediated by the other two proposed mechanisms. Emotional reactions are more likely to be elicited by incivility than by a lack of substantiation. Furthermore, the stimulus characteristics of incivility are vivid and easily perceivable, which are key qualities of heuristic cues. The elaboration likelihood model (Petty & Cacioppo, 1986) predicts that heuristic and emotion-based judgments are most likely to occur when the recipient does not engage in extensive issue-relevant thinking and memory searching to form the judgment, that is, judgments that involve a low level of elaboration. Consequently, we expect that the degree of elaboration Ó 2017 Hogrefe Publishing

27

when making a judgment moderates the effect of incivility in user comments, and predict a negative effect of incivility in comments on the perceived quality of the commented journalistic content only for judgments that involve a low level of elaboration (Hypothesis 2).

Multiple Cues Our theoretical discussion rests on the conceptualization of incivility and a lack of reasoning in user comments as peripheral cues in news quality judgments, and our aim is to explore how these features affect judgments in conjunction with other known peripheral cues, especially the media brand. Therefore, we pose: Research Question 1 (RQ1): Do the moderated effects that incivility and lack of reasoning in comments have on recipients’ assessments of the news item’s quality differ between known and unknown news brands?

Method Design, Procedure, and Sample We conducted a randomized 3  2  2-factorial online experiment in which participants read a news article on a website and then completed a questionnaire. The between-subjects factors were the brand of the news website (two known and reputable news brands vs. an unknown fictitious brand), civility in the comments that accompanied the article (uncivil vs. civil), and reasoning in the comments (unsubstantiated vs. substantiated). Invitations to complete the survey were distributed by the participants of a university course on online communication using e-mails and posts to their social networks, and the link to the questionnaire was also advertised by a local online newspaper. Screenshots of the stimuli were integrated into the online questionnaire, which started with questions regarding the participant’s demographic characteristics. The participants were then randomly assigned to one of the 12 experimental conditions. They were instructed to read the subsequent article as they would ordinarily read news online, and were then exposed to the stimulus. They could stay on the stimulus page as long as they wanted. Once they had finished reading, the participants responded to a series of questions about the quality of the news article. The sample consisted of 633 respondents. The majority of the respondents were female (61.9%), and the average age was 31.4 years (SD = 13.7); 37.9% held a college or Journal of Media Psychology (2019), 31(1), 24–34


28

university degree, and for 45.7% the highest level of education was graduation from high school.

Stimulus Materials A screenshot of a news website that contained a short journalistic article accompanied by user comments was embedded in the questionnaire as the stimulus. The article discussed the relative risks of the legalization of marijuana consumption, was balanced in terms of discussing the pros and cons of the issue, and remained constant across all conditions. The news brand was manipulated by embedding the article into the website designs of two real German news sites, Focus Online and Spiegel Online, and the website of a fictitious and thus unknown news brand (Fakt Aktuell). We deliberately decided to use two known brands to test the generalizability of the conditional effects of the comments across known news brands. The two chosen news brands are prestigious news media in Germany, and were among the most visited German news sites at the time this study was conducted (Krei, 2014). Both brands have a reputation for producing highly professional journalism (Kaltenhäuser, 2005). However, the brands differ in other dimensions of their brand images: Spiegel is considered to be more left-wing oriented and critical, with an emphasis on investigative reporting, whereas Focus is considered to be more conservative (Kaltenhäuser, 2005). Despite these differences, both news sources represent typical online outlets of traditional journalistic news magazines that adhere to the hard-news paradigm, and that allow for only limited audience participation. Below the article, there was a comments section that contained four reader comments in each condition. The first comment was in favor, the second and third comments against, and the fourth comment in favor of the legalization of marijuana consumption. In addition to the order of the comments, everything except reasoning and incivility in the comments was kept constant across the experimental conditions. Following Govier’s (2010) definition of argument, the level of reasoning in the four comments was manipulated by varying the amount of substantiation for the position that was advocated in the comments. Substantiated user comments stated a conclusion (e.g., “Marijuana should be illegal”) and also gave reasons for this conclusion (e.g., “Because it can act as a gateway drug”). Unsubstantiated user comments expressed opinions without giving reasons for the stated conclusion (e.g., “Marijuana should be legalized, that is clear”). Following the conceptualization of incivility by Coe et al. (2014), we varied the language that was used in the

Journal of Media Psychology (2019), 31(1), 24–34

P. Weber et al., Why User Comments Affect Perceived News Quality

comments to manipulate civility. The comments in the uncivil conditions featured informal and emotional language such as slander, insults, and discrimination against social groups; for example, calling other commenters “dumb idiots” who “don’t get the point about marijuana.” The comments in the civil conditions were reasonable and polite, and included phrasing such as, “If I may, my opinion on the topic is. . .,” or politely addressed other commenters with, “Please think about. . .” To test whether these manipulations of the individual comments affected the overall perceptions of the comments section, we conducted a pretest of our stimulus materials in which the participants (N = 100, 56.0% female, Mage = 34.29 years [SD = 14.6], 45.5% graduated from high school) were instructed to evaluate the comments. The stimulus materials and the experimental design in this pretest were the same as in the main study, except that there was no experimental factor for the media brand. Five items assessed perceived incivility (e.g., “The comments were impolite”) on 5-point Likert-type scales (1 = do not agree at all to 5 = agree completely). Scores were averaged to create an index of perceived incivility (Cronbach’s α = .94, M = 3.43, SD = 1.39). Perceived lack of reasoning was measured with five items (e.g., “The comments only voiced opinions but no reasons were given for these opinions”) using the same scales, and the scores were averaged (Cronbach’s α = .88, M = 3.18, SD = 1.11). A twofactor analysis of variance (ANOVA) showed that the manipulations resulted in corresponding changes in the perception of the comments sections. Perceived incivility was influenced by the civility manipulation, F(1, 95) = 256.15, p < .01; Mcivil = 2.11, Muncivil = 4.49, but not the reasoning manipulation, and not by their interaction (all p > .05). Perceived lack of reasoning was solely affected by the reasoning manipulation, F(1, 95) = 53.27, p < .01; Msubstantiaed = 2.49, Munsubstantiated = 3.78; all other p > .05. These results indicate that the manipulations resulted in corresponding perceptions of the comments sections, and confirm that civility and reasoning were manipulated independently. We conducted another stimulus test after the main study was completed using the same experimental design as in the pretest to examine the possibility that the manipulations of the comments affected the perceived realism of the stimuli. Participants (N = 90, 57.8% female, Mage = 36.4 years, SD = 14.4, 51.1% graduated from high school) were asked to rate the realism of the comments on the same 5-point scales using four items (e.g., “Comments like these are typical examples of user comments,” Cronbach’s α = .82). The scores were averaged to create an index of perceived realism. A grand mean of 3.80 (SD = 0.88) shows that, on average, the stimuli were perceived as realistic.

Ó 2017 Hogrefe Publishing


P. Weber et al., Why User Comments Affect Perceived News Quality

Furthermore, a two-factor ANOVA revealed that the perceived realism was unaffected by the manipulation of civility, F (1, 86) = .17, p = .68, by the manipulation of reasoning, F (1, 86) = .77, p = .38, and by their interaction, F(1, 86) = 1.99, p = .16. Thus, a lack of perceived realism is unlikely to confound any effects of the manipulations. The two stimulus tests were not included in the main study for two reasons. For the sake of external validity we did not want to draw the participants’ attention specifically to the comments in the main study – we wanted them to read the webpage as they would usually read news online. We also wanted to avoid any effects of the measurement of the perception of the comment on any other measurements (e.g., the quality of the news item) and vice versa. Nevertheless, we wanted to ensure that in principle the overall perception of the comments below the article corresponded with the stimulus manipulation while, at the same time, the comments’ perceived realism was invariant across manipulations. Although this was shown to be the case by the stimulus tests, these tests were conducted with separate samples and we cannot say for sure that the results generalize to the sample of the main study. But since all samples share core demographic properties (participants were dominantly female, over 30 years of age on average, and highly educated), and presumably also key psychographic properties, we are confident that, also in the main study, the participants’ perception of the comments corresponded with the manipulations, and that no problems of internal validity arise from the stimulus materials due to variations in perceived realism.

Measures Dependent Variable The participants’ assessments of the quality of the journalistic article in the main study were measured with 14 items (Urban & Schweiger, 2014). They were asked to indicate their level of agreement with each item using a 5-point Likert-type scale (1 = do not agree at all to 5 = agree completely). The items reflect journalistic values that are central to recipients’ conceptions of quality: diversity (“The article considers different opinions”), impartiality (“The article is objective”), relevance (“The article reports relevant facts and arguments”), accuracy (e.g., “The article does not contain errors or contradictions”), comprehensibility (“The article is comprehensible”), and compliance with ethical standards (e.g., “The article is adequate in tone”). Further items measured judgments of hedonic qualities such as the level of interest, entertainment, and liking. Responses to the items were averaged to create an index of the perceived quality of the news item (Cronbach’s α = .88, M = 3.11, SD = 0.65). Ó 2017 Hogrefe Publishing

29

Moderator The time that the respondents spent making their quality judgments was recorded to obtain a measure of the judgment process, that is, the degree of elaboration. Response latencies are frequently used as unobtrusive, cost-efficient measures of the degree of elaboration when making judgments in computer-assisted questionnaires (c.f., Mayerl, 2013). This operationalization is based on the assumption that answering survey questions in a deliberative fashion is more time consuming than answering questions in a spontaneous mode. When respondents engage in issue-relevant thinking, and retrieve relevant memory content while making a judgment, they take longer to provide their answers than when they rely on shortcuts to make their judgment. On average, it took respondents 74.27 s to provide their answers to the items that measured their quality assessments (SD = 76.41 s). We conducted an ANOVA to check whether the response latencies differed between the experimental groups. None of the experimental factors affected the speed with which respondents judged the article’s quality: news brand, F(2, 621) = 1.67, p = .19, civility, F(1, 621) = .23, p = .63, and reasoning in the comments, F(1, 621) = 1.49, p = .22. Furthermore, none of the interactions between the experimental factors was significant (all p  .18). Thus, the effects of the treatment in the subsequent analyses are not confounded with the effects of the response latencies. Covariate We measured the stimulus exposure time as a proxy for how attentively the stimulus was processed. Because attention and reading time can affect how much and which information is encoded and stored during stimulus exposure, and which information is consequently available when making subsequent judgments, we controlled for exposure time (M = 95.68 s, SD = 52.80) in our analyses. ANOVA results showed that the stimulus exposure time did not differ significantly between the experimental groups (all p  .16).

Results A regression analysis was conducted to test the predictions and the research question. This method can easily incorporate a continuous moderator variable (response latency), and provides a convenient way to determine which experimental groups differ in complex experimental designs. As hypothesis testing should be limited to actual hypotheses, the model specification was limited to the hypothesized effects. The hypotheses imply one two-way interaction between lack of reasoning and elaboration (measured by

Journal of Media Psychology (2019), 31(1), 24–34


30

response latency), and one two-way interaction between incivility and response latency. The research question implies three-way interactions of these two variables with the media brand. Therefore, in addition to these interactions, the model included all of the lower-order variables that define these interactions. Lack of reasoning was entered as a dummy-coded variable (0 = substantiated comments, 1 = unsubstantiated comments), as was incivility (0 = civil comments, 1 = uncivil comments). Response latency was mean centered (so that zero indicated the average response latency) to simplify the interpretation of the interactions involving response latencies. The two known news brands, Spiegel and Focus, were included as dummy variables and the unknown brand served as the reference category. The model included a constant and controlled for exposure time, respondents’ gender (1 = female, 0 = male), and age. Table 1 displays the estimated coefficients of the model1. Consistent with H1, the significant interaction between lack of reasoning and response latency indicates that the degree of elaboration during judgment moderated the effect of a lack of reasoning on ratings of the news article’s quality. In the moderated multiple regression, this effect must be interpreted in relation to its lower-order constituents, for which conditional regression weights were estimated (see, e.g., Hayes, 2013, for details on moderated multiple regression). Therefore, bLack of reasoning quantifies the effect of lack of reasoning at the average response latency for the unknown news brand. The negative coefficient shows that there was a nonsignificant negative effect of a lack of reasoning in the unknown brand at the average response latency. The negative interaction between lack of reasoning and response latency indicates that this negative effect increased significantly as the response latency increased, quantified as an increase of .004 for each 1-s increase in the response latency above the average. The insignificant three-way interactions that include the media brand indicate that the conditional negative effect of lack of reasoning did not significantly change as a function of the news brand. Overall, these findings are consistent with H1 as they show that a lack of reasoning in the comments had an increasingly negative effect on the perceived quality of the commented-on news item the more the participants elaborated while making their judgment, regardless of the news brand. To fully test the prediction in H1 it is necessary to know at which response latencies the lack of reasoning in the 1

P. Weber et al., Why User Comments Affect Perceived News Quality

Table 1. Multiple regression analysis predicting the rated quality of the news item from incivility in the comments, lack of reasoning in the comments, news brand, and response latency Rated quality of the news item Variable

b

SE

Constant

2.930**

.123

Incivility

.109

.089

Lack of reasoning

.069

.089 .109

Spiegel

.105

Focus

.078

.111

Response latency

.000

.000

Incivility  Response latency

.003*

.001

Lack of reasoning  Response latency

.004*

.002

Spiegel  Response latency

.001

.001

Focus  Response latency

.000

.003

Spiegel  Incivility

.183

.127

Focus  Incivility

.053

.127

Spiegel  Lack of reasoning

.027

.127

Focus  Lack of reasoning

.139

.126

Spiegel  Incivility  Response latency

.009**

.003

Focus  Incivility  Response latency

.006

.004

Spiegel  Lack of reasoning  Response latency

.000

.003

Focus  Lack of reasoning  Response latency

.003

.004

Gender

.119*

.054

Age

.001

.002

Exposure time

.002**

.001

adj. R2

.036

F

2.180**

Note. N = 633. *p < .05. **p < .01.

comments had a significant negative effect on the ratings of the article’s quality, as the interaction coefficient only shows that the change in the effect of lack of reasoning was significant. Instead of arbitrarily picking response latency values at which to estimate the effect of reasoning on quality assessments, we used the Johnson–Neyman (JN) technique (Johnson & Neyman, 1936) to probe the focal interaction (cf. Hayes, 2013, pp. 238, ff. for a review of the technique’s application). The JN technique computes the values of the moderator (response latency) at which the independent variable (lack of reasoning in the comments) had a significant effect on the dependent variable (quality ratings). These values were computed based on the regression model using the JN technique as implemented in the PROCESS macro proposed

Note that this model did not include an interaction between incivility and lack of reasoning and no higher-order interactions with the news brand and response latency that involved this interaction. The corresponding effects were not hypothesized and we wanted to avoid excessive multiple testing to reduce the problem of Type I error inflation. However, to explore whether this modeling affected our results, we specified a full analysis of covariance model that controlled for all possible interactions. This model showed that there was no significant interaction between incivility and reasoning, it yielded the same results with respect to our hypotheses as the more parsimonious model, and therefore we only report the results of the latter.

Journal of Media Psychology (2019), 31(1), 24–34

Ó 2017 Hogrefe Publishing


P. Weber et al., Why User Comments Affect Perceived News Quality

by Hayes (2013). A single solution was generated, revealing that the effect of lack of reasoning in the comments on perceived article quality was significant at response latencies of 40.12 s above the average and higher. The effect’s point estimates are shown in Figure 1. At 40.12 s above the average response latency, the effect of lack of reasoning in the comments becomes significantly negative (note that the 95% confidence intervals of the effect’s estimates do not include zero only for response latencies above 40.12 s) and is estimated as b = –.23 (SE = .12); that is, at this point, the lack of reasoning in the comments significantly decreased the perceived article quality by about 0.23 scale points compared with the conditions in which the opinions were substantiated. As predicted, the negative effect occurred only at higher response latencies: Only a small proportion of the sample (about 10%) elaborated for 1.91 min (40.12 s above the average) or longer when making their quality judgments. There was no effect of a lack of reasoning at lower latencies; that is, the lack of reasoning did not affect the perceived quality of the article when respondents elaborated less and judged the article’s quality more quickly. Turning to the effects of incivility in the comments, a significant interaction between incivility and response latency (see Table 1) indicates that the degree of elaboration during judgment moderated the effect of incivility on ratings of the commented article’s quality. The positive interaction between incivility and response latency indicates that the negative effect of incivility that was estimated at the mean response latency in the unknown brand (bIncivility in Table 1) was significantly attenuated as response latencies increased above the average. This implies that incivility had a negative effect when recipients judged the article’s quality very quickly (i.e., at below average response latencies), but the effect disappeared when recipients elaborated more during judgment. However, this conditional effect was qualified by a significant three-way interaction between incivility, news brand, and response latency (see Table 1). With reference to RQ1, this implies that the degree of elaboration moderated the effect of incivility in the comments on assessments of the article’s quality, but the moderation effect differed for the different news brands. We therefore probed the interaction of incivility in the comments and response latency separately for each news brand using the JN technique. The analysis of the unknown news brand revealed that the effect of incivility on perceived article quality was significant at response latencies of 18.25 s below the average and lower (see Figure 2). At 18.25 s below the average response latency, the effect of incivility was significantly negative (b = –.19, SE = .10). The effect was also negative at lower response latencies, but did not differ

Ó 2017 Hogrefe Publishing

31

Figure 1. The conditional effect of a lack of reasoning in user comments on perceived article quality as a function of the response latency when judging the article’s quality.

significantly from zero at higher response latencies. Consistent with H2, the negative effect occurred only at low response latencies: Approximately 39% of the respondents in the unknown news brand condition elaborated for 56 s (18.25 s below the sample average) or less when making their quality judgments. The analysis of the two known brands revealed that there were no effects of incivility. Uncivil comments in the Focus Online condition did not affect the article’s perceived quality (b = –.16, SE = .09, p = .07) and there was no conditional effect of incivility (bincivility  latency = –.002, SE = .003, p = .48). The same was true for Spiegel Online (bincivility = .07, SE = .10, p = .51; bincivility  latency = –.005, SE = .004, p = .15). Taken together, the results partially support H2: Incivility in the comments decreased the perceived quality of the journalistic content when the participants only elaborated a little, but only with the unknown news brand.

Journal of Media Psychology (2019), 31(1), 24–34


32

Figure 2. The conditional effect of incivility in user comments on perceived article quality as a function of the response latency when judging the article’s quality in the unknown news brand.

Discussion The experimental results were consistent with our hypotheses. A lack of reasoning in user comments decreased the perceived quality of the commented-on journalistic content, but only when recipients elaborated during judgment. We explain this effect in terms of misattribution: The lack of substantiation for the positions advocated in the comments was misremembered as being a quality of the journalistic article itself. Interestingly, this conditional effect did not depend on the news brand. This finding has two implications. First, having a reputation for being a highly professional journalistic news medium does not protect against the effect. If comments do not feature a reasoned argument, they will negatively affect the perceived quality of the news article, at least when readers really think about its quality by trying to recall relevant information on which to base their judgment. Second, it is unlikely that heuristic judgment or affect infusion explain the effect because such judgment processes are expected only with low elaboration, Journal of Media Psychology (2019), 31(1), 24–34

P. Weber et al., Why User Comments Affect Perceived News Quality

whereas we observed the effect with high elaboration. Thus, a lack of reasoning in comments probably does not serve as a heuristic cue. And even though it might induce affect, this does not influence readers’ quality judgments. Incivility in the comments decreased the perceived quality of the commented-on journalistic content, but only with low elaboration. However, this effect occurred only for the unknown news website, which was not associated with a professional and reputable news brand and thus provided no information on which the judgment could be based. This suggests that the effect of incivility arose from a heuristic judgment process, and that incivility is an inferior heuristic cue to which recipients resort when no other informative heuristic cues (e.g., the brand name) are available. If incivility was a dominant cue or the effect was mediated by induced affect, incivility in the comments would have been much more likely to affect the perceived quality, irrespective of the news brand. In addition, if the effect was not the result of heuristic judgment, it should not have occurred only with low elaboration. Another implication of the result is that having a reputation for being highly professional and delivering high-quality journalism protects against the negative effect of incivility in user comments that occurs when readers make a quick judgment on the quality of the journalistic content. The size of the estimated coefficients in our analysis suggests that effects are rather small. However, an adequate interpretation must consider that they are estimated from data that were gathered in relatively natural settings (compared with a laboratory where other potential influences can be eliminated), that they represent the effect of just one exposure to low-quality comments, and that the dependent variable is probably affected by a variety of other factors and might be especially hard to influence when the actual quality of an article is constant. Thus, it is only realistic that user comments are a minor factor in quality perceptions, and it is more surprising that the comments’ quality does have a causal effect. It seems possible that such small effects, when they accumulate over time, lead to more substantial and generalized effects of user-generated comments, like, for example, reduced trust in the news in general. We proposed three mechanisms that explain the effects of incivility and lack of reasoning in user comments, derived the boundary conditions for the effects, and empirically tested the predictions that were deduced from this theory. The data were consistent with the predictions and so we inferred that the proposed processes were at work. However, we did not measure these processes directly. Therefore, future research could try to measure, for example, what information is recalled when judging the quality of a news item to test our proposed mechanisms more directly. Ó 2017 Hogrefe Publishing


P. Weber et al., Why User Comments Affect Perceived News Quality

This study relied on the measurement of response latency to assess the degree of elaboration. However, we cannot be sure that longer response latencies indicated that the respondents tried to retrieve more relevant information from memory while making their judgments. Although our results are consistent with this assumption, future research should test the hypotheses using another operationalization of the degree of elaboration. Finally, this study relied on a news article that featured a single issue as the stimulus, and it was conducted using a sample of respondents that was not representative of the whole population. Another design feature of our experiment that might limit the generalizability of the results is that we used a special configuration of (un-)civil and (un-)reasoned comments. This raises the question of whether our results are valid for other configurations as well, for example, when there are more or less than four comments, when there are other mixtures of high- and low-quality comments, or when there are comments that vary regarding other qualities than civility and reasoning. Whether our results can be generalized to a wider population, to a wider range of news topics, and are observable with diverse configurations of comments remain open questions. We are more confident regarding generalizations across different news outlets. By including two known news websites in the experimental design that represent typical online counterparts of traditional news magazines that allow for limited audience participation, we are confident that our results would generalize to other websites of this category. However, the sources that we used are typical sources for hard news. Thus, it remains unclear whether the effects that we found can also be observed with tabloid sources that are more inclined to report soft news, and where readers might place a different value in other readers’ comments. Furthermore, it is questionable whether the results would generalize to news websites that allow for more, and rely more on, active participation of lay communicators than just commenting (like, e.g., social news sites). We recommend that these boundary conditions of our study be addressed in future research.

References Ash, E., Peeling, A., & Hettinga, K. (2010, August). Does content matter? The effects of type and number of user-generated comments on news stories. Paper presented at the Association for Education in Journalism & Mass Communication National Conference, Denver, CO. Chen, S., & Chaiken, S. (1999). The heuristic-systematic model in its broader context. In S. Chaiken & Y. Trope (Eds.), Dualprocess theories in social psychology (pp. 73–96). New York, NY: Guilford Press. Coe, K., Kenski, K., & Rains, S. A. (2014). Online and uncivil? Patterns and determinants of incivility in newspaper website

Ó 2017 Hogrefe Publishing

33

comments. Journal of Communication, 64(4), 658–679. doi: 10.1111/jcom.12104 Dillon, A. (1992). Reading from paper versus screens: A critical review of the empirical literature. Ergonomics, 35(10), 1297–1326. doi: 10.1080/00140139208967394 Eveland, W. P., & Dunwoody, S. (2002). An investigation of elaboration and selective scanning as mediators of learning from the Web versus Print. Journal of Broadcasting & Electronic Media, 46(1), 34–53. Fichter, C., & Jonas, K. (2008). Image effects of newspapers. Journal of Psychology, 216(4), 226–234. doi: 10.1027/00443409.216.4.226 Govier, T. (2010). A practical study of argument. Belmont, CA: Cengage Learning. Hastie, R., & Park, B. (1986). The relationship between memory and judgment depends on whether the judgment task is memory-based or on-line. Psychological Review, 93(3), 258–268. doi: 10.1037/0033-295X.93.3.258 Hastie, R., & Pennington, N. (1989). Notes on the distinction between memory-based versus on-line judgments. In J. N. Bassili (Ed.), On-line cognition in person perception (pp. 1–17). Hillsdale, NJ: Erlbaum. Hayes, A. F. (2013). Introduction to mediation moderation, and conditional process analysis. New York, NY: Guilford Press. Houston, J. B., Hansen, G. J., & Nisbett, G. S. (2011). Influence of user comments on perceptions of media bias and thirdperson effect in online news. Electronic News, 5(2), 79–92. doi: 10.1177/1931243111407618 Johnson, P. O., & Neyman, J. (1936). Tests of certain linear hypotheses and their application to some educational problems. Statistical Research Memoirs, 1, 57–93. Kaltenhäuser, B. (2005). Abstimmung am Kiosk [Vote at the kiosk]. Wiesbaden, Germany: DUV. Kim, E.-M., & Sun, Y.-H. (2006). The effect of replies in Internet news on the audience. Korean Journal of Journalism & Communication Studies, 50, 33–64. Koriat, A., & Ma’ayan, H. (2005). The effects of encoding fluency and retrieval fluency on judgments of learning. Journal of Memory and Language, 52(4), 478–492. doi: 10.1016/j.jml. 2005.01.001 Krei, A. (2014). Online-IVW im Juli [Online-IVW in July]. Retrieved from http://www.dwdl.de/zahlenzentrale/47078/wm_hilft_newssites_ starkes_debuet_fuer_heftigco_/ Krueger, J. I. (2012). The (Ir)rationality project in social psychology. In J. I. Krueger (Ed.), Frontiers of social psychology. Social judgment and decision making (pp. 59–75). New York, NY: Psychology Press. Ksiazek, T. B., Peer, L., & Zivic, A. (2015). Discussing the news. Digital Journalism, 3(6), 850–870. doi: 10.1080/21670811. 2014.972079 Kühne, R. (2012). Media-induced affects and opinion formation. Living Reviews in Democracy, 3, 1–20. Retrieved from https:// www.lrd.ethz.ch/index.php/lrd/index Lang, A. (2009). The limited capacity model of motivated mediated message processing. In R. L. Nabi & M. B. Oliver (Eds.), The SAGE handbook of media processes and effects (pp. 193–204). Los Angeles, CA: Sage. Mayerl, J. (2013). Response latency measurement in surveys. Survey Methods: Insights From the Field. Retrieved from http://surveyinsights.org/?p=1063 Meijer, I. C. (2013). Valuable journalism. journalism, 14(6), 754–770. doi: 10.1177/1464884912455899 Meltzer, K. (2014). Journalistic concern about uncivil political talk in digital news media. The International Journal of Press/ Politics, 20(1), 85–107. doi: 10.1177/1940161214558748

Journal of Media Psychology (2019), 31(1), 24–34


34

Petty, R. E., & Cacioppo, J. T. (1986). Communication and persuasion. New York, NY: Springer. Prochazka, F., Weber, P., & Schweiger, W. (2016). Effects of civility and reasoning in user comments on perceived journalistic quality. Journalism Studies. Advance online publication. doi: 10.1080/1461670X.2016.1161497 Ruiz, C., Domingo, D., Mico, J. L., Diaz-Noci, J., Masip, P., & Meso, K. (2011). Public shere 2.0? The International Journal of Press/Politics, 16(4), 463–487. doi: 10.1177/1940161211415849 Sikorski, C. von., & Haenelt, M. (2016). Scandal 2.0. Journalism & Mass Communication Quarterly. Advance online publication. doi: 10.1177/1077699016628822 Urban, J., & Schweiger, W. (2014). News quality from the recipients’ perspective. Journalism Studies, 15(6), 821–840. doi: 10.1080/1461670X.2013.856670 van der Wurff, R., & Schoenbach, K. (2014). Civic and citizen demands of news media and journalists. Journalism & Mass Communication Quarterly, 91(3), 433–451. doi: 10.1177/ 1077699014538974 Vujnovic, M. (2011). Participatory journalism in the marketplace. Economic motivations behind the practices. In J. B. Singer, A. Hermida, D. Domingo, A. Heinonen, S. Paulussen, T. Quandt, . . ., M. Vujnovic (Eds.), Participatory journalism. Guarding Open Gates at online newspapers (pp. 139–154). West Sussex, UK: Wiley-Blackwell. Wallsten, K., & Tarsi, M. (2015). Persuasion from below? Journalism Practice. Advance online publication. doi: 10.1080/ 17512786.2015.1102607 Wästlund, E., Reinikka, H., Norlander, T., & Archer, T. (2005). Effects of VDT and paper presentation on consumption and production of information. Computers in Human Behavior, 21(2), 377–394. doi: 10.1016/j.chb.2004.02.007 Weber, P. (2014). Discussions in the comments section: Factors influencing participation and interactivity in online newspapers’ reader comments. New Media & Society, 16(6), 941–957. doi: 10.1177/1461444813495165 Weinreich, H., Obendorf, H., Herder, E., & Mayer, M. (2008). Not quite the average. ACM Transactions on the Web, 2(1), 1–31. doi: 10.1145/1326561.1326566 Received January 4, 2016 Revision received August 30, 2016 Accepted September 5, 2016 Published online April 28, 2017

Journal of Media Psychology (2019), 31(1), 24–34

P. Weber et al., Why User Comments Affect Perceived News Quality

Patrick Weber University of Hohenheim 540G 70593 Stuttgart Germany p.weber@uni-hohenheim.de

Patrick Weber (PhD, University of Zurich) is a postdoctoral researcher at the University of Hohenheim, Germany. His research focuses on media effects and persuasion, especially in public communication in online media.

Fabian Prochazka (MA, University of Salzburg) is a research associate and PhD candidate at the University of Hohenheim. His research interest is the changing relationship of journalism and its audiences. He focuses on trust in news media, perceptions of journalistic quality, and the role of social media in public discourse.

Wolfgang Schweiger (PhD, LMU Munich) is a professor of communication at the University of Hohenheim, Germany. His research interests cover a broad range of issues from corporate communication and evaluation, media use and effects research, online research, risk communication, and empirical methods.

Ó 2017 Hogrefe Publishing


How to provide culturally sensitive care for clients with PTSD and related disorders “The field of cultural clinical psychology takes an important stride forward with this carefully edited volume on the cultural shaping of posttraumatic stress disorder.” Andrew G. Ryder, PhD, Associate Professor of Psychology, Concordia University, Montreal, QC, Canada

Andreas Maercker / Eva Heim / Laurence J. Kirmayer (Editors)

Cultural Clinical Psychology and PTSD 2019, x + 236 pp. US $62.00 / € 49.95 ISBN 978-0-88937-497-3 Also available as eBook This book, written and edited by leading experts from around the world, looks critically at how culture impacts on the way posttraumatic stress disorder (PTSD) and related disorders are diagnosed and treated. There have been important advances in clinical treatment and research on PTSD, partly as a result of researchers and clinicians increasingly taking into account how “culture matters.” For mental health professionals who strive to respond to the needs of people from diverse cultures who have experienced traumatic events, this book is invaluable. It presents recent research and and practical approaches on key topics, including:

www.hogrefe.com

• How culture shapes mental health and recovery • How to integrate culture and context into PTSD theory • How trauma-related distress is experienced and expressed in different cultures, reflecting local values, idioms, and metaphors • How to integrate cultural dimensions into psychological interventions Providing new theoretical insights as well as practical advice, it will be of interest to clinical psychologists, psychiatrists, and other health professionals, as well as researchers and students engaged with mental health issues, both globally and locally.


Aviation Psychology and Applied Human Factors

nline free o issue le samp

Official Organ of the European Association for Aviation Psychology (EAAP) and the Australian Aviation Psychology Association (AAvPA) Editor-in-Chief Harald Kolrep University of Applied Sciences for Media, Communication, and Management, Berlin Associate Editors Cristina Albuquerque, Lisbon, Portugal André Droog, Groningen, The Netherlands Hinnerk Eißfeldt, Hamburg, Germany

ISSN-Print 2192-0923 ISSN-Online 2192-0931 ISSN-L 2192-0923 2 issues per annum (= 1 volume)

Subscription rates (2019) Individuals US $143.00 / € 102.00 (print & online) Institutions From US $283.00 / € 219.00 (print only; pricing for online access can be found in the journals catalog at hgf.io/journals2019) Postage / Handling US $8.00 / € 6.00

About the Journal Aviation Psychology and Applied Human Factors publishes innovative, original, high-quality applied research covering all aspects of the aerospace domain. In order to make the journal accessible to both practitioners and scientific researchers, the contents are broadly divided into original scientific research articles and papers for practitioners. The fully peer-reviewed Original Articles cover a variety of methodological approaches, ranging from experimental surveys to ethnographic and observational research, from those psychological and human factors disciplines relevant to the field, including social psychology, cognitive psychology, and ergonomics. High-quality critical review articles and meta-analyses cover particulars topic of current scientific interest. Shorter studies are published as Research Notes. APAHF in Practice consists of less technically written, but still fully peer-

www.hogrefe.com

Monica Martinussen, Tromsø, Norway Michaela Schwarz, Vienna, Austria Matthew J. W. Thomas, Goodwood, Australia

reviewed articles covering a wide range of topics, such as comments on incidents and accidents, innovative applications of aviation psychology, and reviews of best practices in industry. Finally, the journal’s News and Announcements section features past and upcoming events around the world, association news, interviews, and similar. Manuscript Submissions All manuscripts must be sent to the journal’s email address, journal@eaap.net. Electronic Full Text The full text of the journal – current and past issues (from 2011 onward) – is available online at econtent.hogrefe.com/loi/apf (included in subscription price). A free sample issue is also available there. Abstracting Services The journal is abstracted  /  indexed in PsycINFO, PSYNDEX, and Academic Index.


Original Article

Effects of Human vs. ComputerControlled Characters and Social Identity Cues on Enjoyment Mediation Effects of Presence, Similarity, and Group Identification Jorge Peña,1 Jannath Ghaznavi,1 Nicholas Brody,2 Rui Prada,3 Carlos Martinho,3 Pedro A. Santos,3 Hugo Damas,3 and Joana Dimas3 1

Department of Communication, University of California, Davis, CA, USA University of Puget Sounds, Tacoma, WA, USA

2 3

INESC-ID and Instituto Superior Técnico, Universidad de Lisboa, Portugal Abstract: This study explored how group identification, avatar similarity identification, and social presence mediated the effect of character type (avatars or agents) and social identity cues (presence or absence of avatars wearing participants’ school colors) on game enjoyment. Playing with teammate avatars increased enjoyment indirectly by enhancing group identification. In addition, the presence of social identity cues increased enjoyment indirectly by augmenting identification with one’s avatar. Unexpectedly, playing in multiplayer mode in the presence of social identity cues decreased enjoyment, whereas playing in multiplayer mode in the absence of social identity cues increased enjoyment. Social presence was not a reliable mediator. The findings supported media enjoyment and social identity theories, and highlighted how virtual character type and identification processes influence enjoyment. Keywords: enjoyment, avatars and agents, identification, social identity, social presence

When playing video games, people’s subjective experience depends on both technical and social factors. For example, playing against human-controlled opponents increases enjoyment compared with playing against computercontrolled characters (Weibel, Wissmath, Habegger, Steiner, & Groner, 2008). Players also show increased physiological arousal when virtual characters are introduced as human-controlled avatars instead of computercontrolled agents (Lim & Reeves, 2009). Players also prefer human instead of computer-controlled partners (McGee, Merritt, & Ong, 2013). Participants played a capture game with an artificial teammate and with a “presumed” human teammate (i.e., an artificial teammate that they believed was a human teammate). Participants chose the “presumed human” over the artificial teammate for a follow-up game even though the teammates were the same (McGee et al., 2013). In addition, when playing single and multiplayer video games, players take sides and form groups to cooperate with teammates and compete against other groups. Ó 2017 Hogrefe Publishing

In Halo’s online multiplayer mode, players are assigned to teams of avatars clad in red or blue armor, and ingroup and outgroup social identity is frequently color coded in video games to ease with player recognition. In this context, social identity and self-categorization theories (SIT/ SCT/SIDE) predict that salient social identity cues (e.g., playing video games with avatars or agents wearing participants’ school colors) may increase positive ingroup biases and negative outgroup biases (Tajfel & Turner, 1986; Turner, Hogg, Oakes, Reicher, & Wetherell, 1987; Spears & Postmes, 2015). If so, the presence of shared social identity cues may increase video game enjoyment based on the activation of positive ingroup biases compared with playing video games in the absence of shared social identity cues. Virtual characters wearing participants’ school colors may also increase similarity-based identification (Cialdini, Borden, Thorne, Walker, Freeman, & Sloan, 1976; Van Looy, Courtois, De Vocht, & De Marez, 2012), which may in turn augment game enjoyment. Journal of Media Psychology (2019), 31(1), 35–47 DOI: 10.1027/1864-1105/a000218


36

This study attempts to integrate the effects of playing with human or computer-controlled virtual characters and social identity cues on players’ game enjoyment into a single study. Enjoyment is defined as a pleasant response to media use that has physiological, cognitive, and affective components (Vorderer, Klimmt, & Ritterfeld, 2004). Enjoyment is a product of playing with human or computer partners, storyline and characters, the type of game controller (e.g., joystick or motion-based controller), cooperative or competitive game mode, along with games being able to satisfy basic player needs such as competence, autonomy, and relatedness (Schmierbach, Xu, Oeldorf-Hirsch, & Dardis, 2012; Tamborini, Bowman, Eden, Grizzard, & Organ, 2010). In addition, we examine whether playing with human or computer-controlled virtual characters and social identity cues may not only directly augment enjoyment but may also do so indirectly by influencing (a) social presence, (b) identification with one’s virtual character, and (c) identification with a gaming group or clan. Social presence is a key experience afforded by video games (Lee, 2004; Lombard & Ditton, 1997), and is defined as the degree of awareness, attention, mutual comprehension, and affective and behavioral interdependence in interactions with a real or virtual entity (Harms & Biocca, 2004). Believing that a virtual partner is human-controlled and not a computer-controlled character increases social presence (Guadagno, Blascovich, Bailenson, & McCall, 2007). In addition, playing against humans increases spatial presence (i.e., feelings of being in the game world) compared with playing against the computer (Ravaja, Saari, Turpeinen, Laarni, Salminen, & Kivikangas, 2006). This study seeks to contribute to this literature by studying how social presence mediates the effects of playing with human or computer-controlled characters and social identity cues on game enjoyment. Though these effects have not been examined, experiencing more social presence when employing a virtual reality system is linked to increased enjoyment (Heeter, 1995). In addition, increased social presence is linked to higher enjoyment of e-commerce websites (Cyr, Hassanein, Head, & Ivanov, 2007). Identification with an online group refers to attraction to a recreational or task-oriented ingroup (Van Looy et al., 2012; Wang, Walther, & Hancock, 2009). In particular, identification with an online gaming group is linked to team satisfaction and intentions to perform well in future matches (Eveleth & Eveleth, 2010). In the present study, players may identify more with their ingroup than with an outgroup, especially if the game ingroup is composed of human-controlled avatars instead of computer-controlled agents, and also if virtual characters display the participants’ school colors. Identification with a virtual character refers to players integrating a game character’s properties into their self-concept (Klimmt, Hefner, & Vorderer, 2009), and such identification with virtual characters Journal of Media Psychology (2019), 31(1), 35–47

J. Peña et al., Virtual Characters and Social Identity Effects

should increase game enjoyment (Van Looy et al., 2012). Similarity identification is a key dimension of player identification processes as people are more likely to identify with virtual characters that they perceive as more similar to themselves (Van Looy et al., 2012). Thus, playing with human-controlled characters and interacting with avatars wearing participants’ school colors should augment identification with one’s avatar, which may in turn trigger increased enjoyment. Below we articulate theoretical models predicting that virtual character type and social identity cues will affect game enjoyment, along with evidence articulating how this process should be mediated by social presence, group identification, and similarity identification.

The Effects of Playing With Human or Computer-Controlled Virtual Characters Previous studies agree on the prediction that playing with human instead of computer-controlled game partners affects key gaming experiences. For example, playing with humans in multiplayer games is more enjoyable because it is less predictable and more suspenseful in comparison with playing single player games (Klimmt & Hartmann, 2008). Playing with human instead of artificial game partners is also linked to differences in interactivity. Playing with human-controlled avatars may allow more communicating, competing, and strategizing compared with playing with computer-controlled agents (Weibel et al., 2008). There is a perceptual component involved in this process as players led to believe that they are playing with humancontrolled avatars instead of computer-controlled agents show increased enjoyment (Weibel et al., 2008), social presence (Guadagno et al., 2007), and arousal (Lim & Reeves, 2009) in otherwise identical game interactions. Players’ mere perception that they are interacting with human instead of artificial characters may affect game experiences even if in reality they have been playing with artificial characters all along. One additional approach is to experimentally manipulate game mode, an objective technological affordance, in order to examine the effects of playing with human or computer-controlled characters (Ravaja et al., 2006). In this context, participants show increased spatial presence or feelings of being in the game world when playing with human instead of computer-controlled characters (Ravaja et al., 2006). Along these lines, Johnson, Wyeth, Clark, and Watling (2015) examine the effects of playing with human or computer-controlled characters by asking participants to play a first-person shooter game in single or multiplayer mode. Playing with human-controlled characters increases relatedness or sense of connection with others and flow or absorption in the game activity (Johnson et al., 2015). Playing a bowling video game with a human partner Ó 2017 Hogrefe Publishing


J. Peña et al., Virtual Characters and Social Identity Effects

also increases feelings of relatedness compared with playing in single player mode, and relatedness is linked to game enjoyment (Tamborini et al., 2010). Overall, studies examining the effects of playing video games with human or computer characters manipulate either subjective factors (e.g., the belief that one is playing with avatars or agents, Guadagno et al., 2007; Weibel et al., 2008) or objective factors (e.g., game mode as a means to enable single player or multiplayer experiences, Johnson et al., 2015; Ravaja et al., 2006) in order to affect enjoyment, social presence, and relatedness. Consistent with this research, we manipulate game mode to enable interactions with human-controlled avatars or computer-controlled agents. In general, people experience higher enjoyment, social presence, and relatedness when playing with human instead of computer-controlled characters (Johnson et al., 2015; Ravaja et al., 2006). Though the link between playing with human or computer-controlled characters and identification with one’s gaming group has not been examined, when considering the relatedness findings above, it is likely that identification with one’s game group is facilitated when playing with human instead of computer-controlled characters, and this in turn should increase enjoyment. Note that single and multiplayer game modes may both increase identification with one’s avatar as people may identify with their character’s storyline, appearance, morals, etc. (Van Looy et al., 2012; Vorderer et al., 2004). However, given the present experiment featuring short-term play, minimal story, and no character development, along with evidence on the preference for playing with human instead of computer-controlled partners, we expect higher similarity identification with one’s character after playing with human avatars instead of computer-controlled agents. Thus: Hypothesis (H1): Participants playing with humancontrolled avatars will experience (a) increased enjoyment compared with those playing with computer-controlled agents, and this effect will be mediated by (b) social presence, (c) similarity identification, and (d) group identification.

Social Identity Effects According to SIT, people’s sense of self contains personal and social identity components. Personal identity includes unique individual attributes (e.g., personal taste, style, behaviors), whereas social identity refers to membership in social groups coupled with how self-esteem is linked to group membership (e.g., being a student at a prestigious school, Tajfel & Turner, 1986). Building on SIT, SCT expects that categorizing oneself into a social group leads to perceiving self and partners through the lens of the group’s prototype instead of personal traits (i.e., depersonalization Ó 2017 Hogrefe Publishing

37

effects, Turner et al., 1987). To trigger depersonalization effects, social identity cues must be perceptually salient and fit the person (Turner et al., 1987). Building on SIT and SCT, the social identity model of deindividuation effects (SIDE) predicts that anonymous online interactions exacerbate social identification effects (Lea & Spears, 1992; Spears & Lea, 1994; Spears & Postmes, 2015). SIDE forecasts that anonymous online partners with a shared social identity (e.g., avatars wearing participant’s school colors) should behave more positively to each other compared with anonymous online partners with no shared social identity (Lea & Spears, 1992). In online contexts, social identity cues should weigh heavily when people are anonymous and lack information about each other. SIT/SCT/SIDE studies examine their predictions in minimal group experimental paradigms in which unacquainted participants are sorted into groups and develop positive ingroup biases and negative outgroup biases on the basis of trivial criteria such as preference for photographs or liking of abstract painters (Tajfel & Turner, 1986). Minimal group effects have been observed in online interactions in which communicators remain visually anonymous to each other, as seemingly trivial information including usernames (Tanis & Postmes, 2007) and work instructions emphasizing the individual or the group (Lea & Spears, 1992) operate as cues to social identity in virtual teams. For example, anonymous online communicators that were categorized into a virtual group based on shared zodiac signs report increased group identification compared with participants who did not have the same zodiac sign (Wang et al., 2009; Yilmaz & Peña, 2014). Based on the activation of a positive ingroup bias, enjoyment should increase when playing in the presence of ingroup social identity cues compared with groups in which avatars and agents share no social identity. Shared social identity cues may also increase social presence as the activation of a positive ingroup bias may lead participants to pay more attention and feel more connected to ingroup characters wearing their school colors, and such increase in social presence may mediate the direct effects of social identity cues on enjoyment. Presence of ingroup social identity cues may also increase group and similarity identification because participants will self-categorize into gaming teams displaying ingroup cues such as participants’ school colors in real life (Turner et al., 1987). Thus: Hypothesis 2 (H2): Participants playing in groups with virtual characters dressed in their school colors will experience increased (a) enjoyment compared with those in groups with avatars dressed in different colors, and this effect will be mediated by (b) social presence, (c) group identification, and (d) similarity identification. Journal of Media Psychology (2019), 31(1), 35–47


38

Interaction Effects As noted above, playing with human-controlled avatars instead of computer-controlled agents should intensify enjoyment, social presence, similarity-based identification with an avatar, and group identification. Additionally, shared social identity cues should also augment players’ enjoyment, social presence, similarity identification, and group identification as these cues should trigger positive ingroup biases and self-categorization processes. Considering the above, it is possible that these main effects also result in a statistical interaction effect. Because there is no single theoretical model that encapsulates the synergistic effects of the manipulations, we put forward a research question asking: Research Question 1 (RQ1): How does playing with human-controlled avatars instead of computercontrolled agents in the presence or absence of avatar social identity cues influence (a) enjoyment, (b) social presence, (c) similarity identification, and (d) group identification?

Method Participants Undergraduate students at a large public university in the U.S. (N = 216) participated in the experiment. 166 participants were female. The participants were 21.5 years old in average (SD = 3.60). Most of the participants were Asian (41.4%), followed by Caucasian (35.3%), Hispanic (12.1%), African-American (4.2%), and other (7%) ethnic origin. There were 109 participants in the social identity cues present condition and 106 participants in the social identity cues absent condition. In addition, 90 participants played with computer-controlled agents and 125 participants played with human-controlled avatars. Computer and video game use was measured in a 1-7 Likert-type scale. The participants used computers (M = 6.92, SD = 0.45) more frequently than video games (M = 2.44, SD = 1.83), t(215) = 35.83, p < .001.

Materials and Procedure The video game used in this study was the result of a multidisciplinary collaboration between computer scientists and communication researchers. The video game puts players in the role of plane crash survivors in an island with an active volcano. Players spent in-game days (i.e., game turns) collecting resources to leave the island before the volcano’s eruption. The participants could either collect Journal of Media Psychology (2019), 31(1), 35–47

J. Peña et al., Virtual Characters and Social Identity Effects

resources that benefitted the group (i.e., wood to build a raft to get out of the island) or individuals (i.e., gold to keep and thus individually “win” the game). While collecting resources, participants would notice that the volcano grew angrier and that there was another group of stranded survivors (i.e., a computer-controlled outgroup, Figure 1). Before coming to the laboratory, participants filled out an online survey measuring traits and demographic information. The survey took 10–15 minutes to complete. A week later, they played one single or multiplayer video game match in the laboratory. Upon arriving to the lab, the participants were randomly assigned to play the game in either the social identity cues present or absent conditions (Figure 1). The game depicted two three-person groups onscreen (i.e., ingroup and outgroup). In the social identity cues present condition, all ingroup human-controlled avatars and computer-controlled agents wore the participant’s school color and outgroup agents wore a rival school’s color. In the social identity cues absent condition, ingroup and outgroup avatars and agents wore different colors, and no color was repeated. The game session and exit survey took about 45 minutes. The laboratory had visually anonymous isolated cubicles and was arranged as a small LAN party with four gaming desktop computers with large screens. Experimental sessions included other visually anonymous participants playing the game in adjacent cubicles with dividers. Participants played the game using noise-cancelling headphones and were not allowed to talk. The experimental video game did not allow exchanging in-game messages. Single player game sessions also included other participants in the same room playing the game in single player mode. Single and multiplayer sessions had a maximum of three participants, and the study did not pit groups of participants simultaneously playing against each other. Outgroup agents were controlled by the computer, and they collected supplies from all resource sites to ensure that participants were aware of their presence in the virtual island. After playing the game, the participants completed an exit survey with the dependent variables and an awareness check.

Independent Variables Human vs. Computer-Controlled Characters The participants played the experimental video game in single player mode in which the other two members of the ingroup were computer-controlled or in multiplayer mode in which human participants controlled the other two ingroup characters (Figure 1). This procedure is consistent with game mode manipulations as a means to enable interactions with human-controlled avatars or computercontrolled agents (Johnson et al., 2015; Ravaja et al., 2006). In both modes the computer operated the outgroup Ó 2017 Hogrefe Publishing


J. Peña et al., Virtual Characters and Social Identity Effects

39

Figure 1. Ingroup and outgroup camps in the social identity cues present condition (A) and in the social identity cues absent condition (B).

(A)

(B)

characters played and the game looked the same to allow for proper experimental control. The participants were not informed about whether their teammates in the game were controlled by other human players or by the artificial intelligence of the game. Social Identity Cues In the social identity cues present condition, ingroup avatars wore the participant’s school color whereas outgroup avatars wore the rival school’s color. In the social identity cues absent condition, ingroup and outgroup avatars wore different colors, and no two avatars wore the same color (Figure 1). The avatar dress color manipulations were congruent with minimal group conditions outlined above.

Dependent Variables We used validated scales to measure the key factors representing subjective game experiences. The intercorrelations between the outcome variables appear in Table 1. Enjoyment This factor was measured after playing the experimental video game with Lin, Gregor, and Ewing’s (2008) 1 to 7 Likert-type scale. The wording of the items was slightly modified to refer to playing a video game instead of visiting websites. The scale had 12 items (e.g., “While playing the game, I was deeply engrossed”, “Playing the game was deeply worthwhile”, α = .91; M = 4.14, SD = 1.00). Social Presence This factor was assessed with Harms and Biocca’s (2004) social presence measure. The co-presence and behavioral Ó 2017 Hogrefe Publishing

interdependence subscales (12 items) were selected as they best represented the study’s goals. Co-presence is the degree to which an observer feels accompanied and in which self and partners are aware of each other (e.g., “My presence was obvious to my team members”, “I caught my team members’ attention”). Behavioral interdependence refers to the extent to which user’s behavior affects and is affected by partners’ behavior (e.g., “My team member’s behavior was closely tied to my behavior,” “I reciprocated my team member’s actions”). The scale was reliable (α = .89; M = 3.13, SD = 0.98). Similarity Identification This factor was measured with Van Looy and associates’ (2012) subscale. The subscale had 6 items such as “My character resembles me” and “I identify with my character” and it was reliable (α = .95; M = 3.03, SD = 1.48). Group Identification This factor was measured with Wang and associates’ (2009) 10-item scale and it included items such as “I see myself as a member of this group” and “I felt loyal toward the group.” This scale was reliable (α = .91; M = 4.05, SD = 1.20).

Awareness Check After completing the procedures and scales described above, participants were asked if they knew the aim of the experiment and whether there was a theme or connection between the different sections in the study. The awareness check questions were adapted from Bargh Journal of Media Psychology (2019), 31(1), 35–47


40

J. Peña et al., Virtual Characters and Social Identity Effects

Table 1. Intercorrelation matrix among outcome variables Social presence Social presence

Pearson Correlation

1

Sig. (2-tailed) Enjoyment

Similarity identification

Group identification

Enjoyment

Similarity identification

Group identification

.29**

.34**

.46**

.000

.000

.000

N

216

216

216

216

Pearson correlation

.29**

1

.32**

.52**

Sig. (2-tailed)

.000

.000

.000

N

216

216

216

216

Pearson correlation

.34**

.32**

1

.37**

Sig. (2-tailed)

.000

.000

.000

N

216

216

216

216

Pearson correlation

.46**

.52**

.37**

1

Sig. (2-tailed)

.000

.000

.000

N

216

216

216

216

**Correlation is significant at the .01 level (two-tailed).

and Chartrand’s (2000) funneled debriefing technique, and it included seven questions (e.g., “What did you think the purpose of this experiment was?” “Did you think that any of the experimental tasks were related in any way? In what way were they related?” “Did you notice any particular pattern or theme in the game?”). To clarify whether participants were aware that they played with human or computer-controlled characters, two of the authors independently read and coded all of the participants’ responses. Participants’ responses were coded in regards to whether participants showed awareness of having played with other humans or computer-controlled characters (0 = Absence of human player or computercontrolled character references; 1 = Presence of human player or computer-controlled character references). Intercoder reliability was calculated based on a randomly chosen 20% of the total number of awareness checks, and the coding was acceptable (kappa = .84). There were no significant differences in regards to making references about playing with other humans or computer-controlled characters between the single (6 presence of human or computer player references; 85 absence of human or computer player references) and multiplayer mode (7 presence of human or computer player references; 118 absence of human or computer player references, w2 (1) = 1.55). There were also no differences in regards to references about playing with human or computer-controlled characters between the social identity cues present (5 presence of human or computer player references; 104 absence of human or computer player references) and the social identity cues absent conditions (8 presence of human or computer player references; 99 absence of human or computer player references, w2 (1) = .80). Participants thus did not refer to playing with avatars or agents to be the main goal of the study. Journal of Media Psychology (2019), 31(1), 35–47

Figure 2. Visual representation of mediation paths among variables.

Results The Effects of Playing With Human or Computer-Controlled Characters on Enjoyment We explored the data’s independence using the methods recommended by Kenny, Kashy, and Cook (2006) because the participants were run in the presence of others. The results showed that all scores were independent (ZEnjoyment = .05, p = ns; Zsocial presence = .06, p = ns; Zsimilarity identification = .04, p = ns; Zgroup identification = .07, p = ns). To address Hypothesis 1, model 4 mediation analyses were performed with PROCESS (Hayes, 2013). Enjoyment was the focal outcome variable based on the hypotheses above. The predictor variables were the character type and social identity cues manipulations. Group identification, similarity identification, and social presence were run as parallel mediators (see Figure 2). This resulted in a model testing for direct and indirect human vs. computer-controlled character effects and another model testing for direct and indirect social identity effects. Descriptive statistics on the dependent variable and mediators appear in Table 2. Ó 2017 Hogrefe Publishing


J. Peña et al., Virtual Characters and Social Identity Effects

41

Table 2. Means and (standard deviations) for the experimental conditions Character type

Social identity cues

Human-controlled

Computer-controlled

Present

Absent

Enjoyment

4.27 (0.98)

3.96 (0.98)

4.11 (1.01)

4.17 (1.00)

Social presence

3.25 (0.61)

2.95 (0.68)

3.22 (0.63)

3.02 (0.67)

Group identification

4.30 (1.14)

3.71 (1.19)

4.12 (1.19)

3.99 (1.21)

Similarity identification

3.17 (1.45)

2.84 (1.50)

3.21 (1.53)

2.86 (1.41)

As seen in Table 3, playing with human-controlled avatars increased enjoyment indirectly by augmenting group identification in comparison with playing with computer-controlled agents. Zero (0) did not fall within the bootstrapped confidence intervals, thus indicating that the playing with avatars – group identification – enjoyment indirect effect was significant (Hayes, 2013). This finding supported H1c. Note that the total effect of playing with avatars on enjoyment was significant and thus, at first glance, this ostensibly supported H1a. However, the direct effect was insignificant. This suggests that the more accurate interpretation is H1c instead of H1a, as significant total effects but insignificant direct effects imply that the relationship between virtual character type and enjoyment would not be significant without the group identification mediator. This further reinforced the importance of the playing with avatars – group identification – enjoyment relationship. In addition, playing with avatars increased social presence compared with playing with agents but social presence was not linked to enjoyment (Table 3). Also, playing with avatars had near-significant positive effects on similarity identification, and similarity identification was linked to increased enjoyment (Table 3). Nevertheless, zero (0) fell within the bootstrapped confidence intervals, thus indicating that the playing with avatars – similarity identification – enjoyment indirect effect was insignificant (Hayes, 2013). Overall, the results supported H1c. The remaining hypotheses were disconfirmed. In addition, relative to the shared social identity cues absent condition, the social identity cues present condition (i.e., playing with avatars wearing participants’ school colors) increased enjoyment indirectly by increasing similarity identification (Table 4). Though the presence of social identity cues had only near-significant positive effects on similarity identification while similarity identification was positively linked to enjoyment, zero (0) did not fall within the bootstrapped confidence intervals, thus indicating that the presence of shared social identity cues – similarity identification – enjoyment indirect effect was statistically significant (Hayes, 2013). Note that the direct effect of social identity cues on enjoyment was insignificant, thus further reinforcing the importance of the social identity Ó 2017 Hogrefe Publishing

cues – similarity identification – enjoyment indirect effect. Group identification was linked to enjoyment but this factor did not mediate the link between social identity cues and enjoyment. In addition, the presence of social identity cues comparatively increased social presence, but social presence itself was not associated with enjoyment (Table 4). The findings confirmed H2d. The remaining hypotheses were not supported.

Interaction Effects of Human or Computer-Controlled Characters and Social Identity Cues In response to the study’s research question, interaction effects between the character type and social identity cues conditions qualified the enjoyment effects reported above, F(1, 216) = 8.13, p = .005, partial η2 = .04. Participants who played with human-controlled avatars in the social identity cues absent condition enjoyed the game the most. When playing with human-controlled avatars, participants in the social identity cues absent condition enjoyed the game more (M = 4.48, SD = 0.91) than those in the social identity cues present condition (M = 4.09, SD = 1.05), t(123) = 2.21, p < .05, d = .40. Participants who played with human-controlled avatars in the social identity cues absent condition also experienced more enjoyment than those who played with computer-controlled agents in the social identity cues absent condition (M = 3.77, SD = 0.98), t(105) = 3.82, p < .001, d = .74. Additionally, among participants playing with computer-controlled agents, those in the social identity cues present condition reported more enjoyment (M = 4.15, SD = 0.95) compared with those in the social identity cues absent condition (M = 3.77, SD = 0.98), though this difference was only near-significant, t(89) = 1.87, p = .07, d = .39. There were no enjoyment differences among participants playing with human or computer-controlled characters in the social identity cues present condition, t(107) = .34, p = ns, d = .07. There were no interaction between the character type and social identity cues conditions on social presence F < 1, similarity identification F(1, 216) = 1.55, p = ns, and group identification, F(1, 216) = 1.16, p = ns. Journal of Media Psychology (2019), 31(1), 35–47


42

J. Peña et al., Virtual Characters and Social Identity Effects

Table 3. Mediation effects of social presence, group identification, and similarity identification on the link between character type and enjoyment (N = 216)

(a) Social presence model summary

R

R-sq

.22

.05

F 11.35

Model

Coeff

SE

Constant

2.66

.15

18.20

t

Character type

.30

.09

3.37

(a) Group identification model summary

R

R-sq

F

.24

.06

13.59

Model

Coeff

SE

Constant

3.12

.27

11.77

t

Character type

.59

.16

3.69

(a) Similarity identification model summary

R

R-sq

F

.11

.01

2.60

Model

Coeff

SE

t

Constant

df1 1.00 p

LLCI

ULCI

2.3683

2.9437

.1234

.4712

.0009** df1 1.00

LLCI

ULCI 3.6436

.2751

.9073

.0003** df1 1.00

1.8543

3.1790

.0728

.7281

.0000 .11

(b) Enjoyment model summary

R

R-sq

.54

.30 SE

t 6.63

p .11 ULCI

7.49

.31

214.00 LLCI

1.61

2.09

df2

p

.34

Coeff

p .0003**

2.5979

.20

Constant

214.00

p

.33

Model

df2

.0000

2.52

22.19

p .0009**

.0000

Character type

F

df2 214.00

df1 4.00 p .0000**

df2 211.00

p .0000**

LLCI

ULCI

1.4649

2.7045

Social presence

.05

.10

.48

.63

.1526

.2509

Group identification

.38

.06

6.65

.0000**

.2660

.4904

Similarity identification

.10

.04

2.26

.03*

.0125

.1818

(c’) Character type

.05

.12

.38

.70

.1935

.2862

Total effects model on enjoyment model summary

R

R-sq

F

.16

.02

5.35

1.00

df2 214.00

p .02*

Model

Coeff

SE

p

LLCI

ULCI

Constant

3.64

.23

16.09

.0000

3.1941

4.0857

.32

.14

2.31

.02*

.0469

.5859

(c) Total effect of X on Y

Effect

SE

t

p

LLCI

ULCI

.32

.14

2.31

.02*

.0469

.5859

(c’) Direct effect of X on Y

Effect

SE

t

p

LLCI

ULCI

.05

.12

.38

.1935

.2862

Character type

t

df1

Total, direct, and indirect effects

Indirect effect of X on Y

.70

Effect

Boot SE

BootLLCI

Total

.27

.08

.12

BootULCI .45

Social presence

.02

.03

.0434

.0836

Group identification

.22

.07

.1087

.3774*

Similarity identification

.03

.02

.0014

.1020

Notes. Character type (Computer-controlled agent = 1, Human-controlled Avatar = 2). *Correlation is significant at the .05 level (two-tailed). **Correlation is significant at the .01 level (two-tailed).

Discussion This study set out to examine how people’s experience with video games was influenced by technical and social factors. A key technical factor is the ability to play games with human or computer-controlled characters as people can play the same video game in collaboration with humancontrolled avatars or with artificial agents, “bots,” etc. Journal of Media Psychology (2019), 31(1), 35–47

In the present study we manipulated game mode in order to enable playing with human or computer-controlled characters (Johnson et al., 2015; Ravaja et al., 2006). Playing with human-controlled avatars indirectly increased enjoyment by augmenting group identification in comparison with playing with computer-controlled agents. This finding is congruent with how human-controlled characters increased feelings of relatedness, and implies that playing Ó 2017 Hogrefe Publishing


J. Peña et al., Virtual Characters and Social Identity Effects

43

Table 4. Mediation effects of social presence, group identification, and similarity identification on the link between social identity cues and enjoyment (N = 216)

(a) Social presence model summary

R .15

R-sq .02

df1 1.00

Model

Coeff

Constant

2.83

.14

20.19

.0000

Social identity cues

.20

.09

2.22

.03*

(a) Group identification model summary

R .06

SE

F 4.91

R-sq .003

Coeff 3.86

.26

14.91

Social identity cues

.13

.16

.81

(a) Similarity identification model summary

R

Model

Coeff

Constant

.01 SE

df1

.66

Constant

R-sq

p

F

Model

.12

SE

t

1.00

t

p .0000 .42

F

df1

3.07

1.00

t

p

2.51

.32

7.89

.0000

Social identity cues

.35

.20

1.75

.08

(b) Enjoyment model summary

R .55

Model

Coeff

Constant

2.31

R-sq .30 SE .32

F

df1

22.75

4.00

t

p

7.18

.0000

df2 214.00 LLCI

p .03* ULCI

2.5542

3.1069

.0216

.3702

df2

p

214.00

.42

LLCI

ULCI

3.3459

4.3653

.1895

.4534

df2

p

214.00

.08

LLCI

ULCI

1.8791

3.1318

.0438

.7463

df2 211.00 LLCI 1.6718

p .000* ULCI 2.9380

Social presence

.07

.10

.69

.49

.1304

.2712

Group identification

.38

.06

6.77

.000**

.2687

.4895

Similarity identification

.10

.04

2.37

.02*

.0169

.1862

(c’) Social identity cues

.15

.12

1.31

.19

.3835

.0769

Total effects model on enjoyment model summary

R .03

R-sq .0007

df1

.16

Model

Coeff

Constant

4.22

.22

19.48

.05

.14

.39

Social identity cues

SE

F

1.00

t

p .0000 .69

df2

p

214.00

.69

LLCI

ULCI

3.7932

4.6474

.3232

.2156

Total, direct, and indirect effects (c) Total effect of X on Y

Effect

(c’) Direct effect of X on Y

Effect

.05 .15

SE .14 SE .12

t

p

.39

.69

t

p

1.31

.19

LLCI .3232 LLCI .3835

ULCI .2156 ULCI .0769

Indirect effect of X on Y Effect

Boot SE

BootLLCI

BootULCI

Total

.10

.08

.0537

.2616

Social presence

.01

.02

.0236

.0722

Group identification

.05

.06

.0727

.1674

Similarity identification

.04

.03

.0009

.1117*

Note. Social identity cues (Absence = 1, Presence = 2). *Correlation is significant at the .05 level (two-tailed). **Correlation is significant at the .01 level (two-tailed).

with others enhanced membership and loyalty towards one’s gaming group, and this in turn augmented game enjoyment. This is also congruent with how MMORPG players showed increased favoritism and trust for ingroup gaming groups or guilds (Guegan, Moliner, & Buisine, 2015; Ratan, Chung, Shen, Williams, & Poole, 2010). This finding resonates with Sassenberg’s (2002) distinction between common bond groups (formed by attachment Ó 2017 Hogrefe Publishing

between group members) and common identity groups (formed by attachment to the group as a whole), and highlights the importance of attachment to the online gaming group as a whole as opposed to interpersonal attachment to other players. Considering that it is unlikely that participants formed attachment to players in short-term anonymous gaming groups that could not exchange messages, one contribution of this study is showing how playing Journal of Media Psychology (2019), 31(1), 35–47


44

with avatars instead of agents increased identification with a common identity gaming group (Sassenberg, 2002), and how this in turn increased game enjoyment. This link offers an exciting theoretical mechanism for the effects of playing with avatars or computer agents on enjoyment, and thus it deserves future study. For example, note that the indirect effect of group identification was more reliable than the direct effect of playing with avatars or computer agents on enjoyment, and was also stronger than the indirect effects of social presence and similarity identification. In addition, playing a video game in the presence of social identity cues (i.e., avatars and agents clothed in participants’ school colors) increased similarity identification with one’s avatar, and similarity identification was associated with increased enjoyment. This indirect effect is congruent with research stressing how similarity identification with virtual characters is fundamentally related to game enjoyment (Klimmt & Hartmann, 2008; Van Looy et al., 2012). The presence of social identity cues – similarity identification – enjoyment indirect effect is also congruent with SIT/SCT predictions that salient social identity cues that fit the perceiver can lead to depersonalization effects. The findings are also consistent with the SIDE prediction that cues to a shared social identity (e.g., salience of participants’ school colors) in anonymous online interactions increases ingroup attachment (Spears & Postmes, 2015). This finding expanded the scope of social identity effects from impression formation, social influence and power in online discussions, the formation of norms in virtual groups, and collaboration in virtual teams (for a review, see Spears & Postmes, 2015) to how social identity cues increase game enjoyment by enhancing similarity identification with one’s avatar. It is also worth noting that social presence was a less reliable mediator of the link between playing with avatars or agents and social identity cues on enjoyment in comparison with the indirect effects of group identification and similarity identification. One possibility is that the manipulations were too weak to influence social presence to begin with. This was not the case as playing with human instead of computer-controlled characters increased social presence. Playing in the presence instead of absence of social identity cues also augmented social presence (see Tables 3 and 4). However, social presence was simply not associated with enjoyment in both of these conditions. Another possibility is that social presence and game enjoyment are unrelated mechanisms. Increased social presence may not be a requisite for game enjoyment as players may potentially find a game pleasing without needing to pay attention and coordinate with other players. A final possibility is that social presence has been defined and measured differently across studies (Lee, 2004; Lombard & Ditton, 1997). Journal of Media Psychology (2019), 31(1), 35–47

J. Peña et al., Virtual Characters and Social Identity Effects

While previous studies have focused on social presence as immersion and transportation into the game world (e.g., Heeter, 1995), this study focused on social presence as the degree of awareness, attention, and interdependence in mediated interactions (Harms & Biocca, 2004). Future studies should determine exactly which facet of social presence is influenced by playing with avatars, agents, and social identity cues, and also outline precisely which social presence factor is related to game enjoyment. In addition, participants playing with human-controlled avatars in the absence of social identity cues reported higher enjoyment compared with those playing with human-controlled avatars in the presence of social identity cues. Assuming that playing with humans was more unpredictable and harder to coordinate, it is possible that the presence of avatars wearing their school colors made the participants pay more attention to and expect more from human teammates, thus decreasing enjoyment. This interpretation is congruent with the “black sheep effect,” in which deviant ingroup members are more negatively evaluated (Marques & Paez, 1994). In other words, playing with avatars in the presence of social identity cues likely heightened positive ingroup biases that were not fully met because human actions (e.g., collecting resources, finishing a game turn) were less coordinated. In one example of this, participants experienced greater competence when playing with artificial agents instead of other human participants because agents played more effectively (Johnson et al., 2015). This prediction should be tested in future studies, as it implies that shared social identity may not always increase game enjoyment, especially when ingroup members’ performance disappoints.

Limitations This study has several limitations. The participants were not committed gamers and had low video game experience. For instance, women made up the majority of the sample. This may have affected the results as, in general, women dislike games that lack social interactions or feature too much violence or competition (Hartmann & Klimmt, 2006). Additionally, players usually know whether they are playing a single or multiplayer video game. Thus, the generalizability of the findings is not fully guaranteed. The present findings may not generalize to single and multiplayer commercial video games in which players fully know that the characters they encounter are human or artificial. The findings may apply to Turing test-based games in which the real or artificial nature of game partners is kept hidden for people to guess. Another limitation was that the experimental video game allows for increased experimental control but had lower production values compared with commercial video games. Ó 2017 Hogrefe Publishing


J. Peña et al., Virtual Characters and Social Identity Effects

For example, participants could not communicate using text messages or voice chat, and therefore our findings may not generalize to more sophisticated video games. Participants were not explicitly informed whether they were playing with human or computer-controlled characters and this may also appear as a limitation. In some studies participants were made unequivocally aware of playing with avatars or agents (even if this was not truly the case), whereas in other studies the game mode itself was manipulated. While we found empirical effects of character type (avatar or agent), the participants themselves did not mention game mode or character type to be a main feature of the study. Considering this, why did the participants react differently to avatars and agents if this factor was not made explicit and participants did not mention such factor in the awareness check? One possibility is that participants inferred that they were playing with human or computercontrolled characters based on game behavior and, at the same time, this factor itself was not perceived to be a main part of the study. Players may encounter more coordination issues when playing with human instead of computercontrolled characters (Johnson et al., 2015), and these behaviors may have clued in the participants to the real or artificial nature of their teammates. Participants may have noticed that they were playing with avatars or agents based on game behavior but it is still possible that this factor did not raise enough suspicions for participants to comment on this regard. If participants noted that they were playing with avatars or agents but did not believe this to be a main factor in the study, then by definition they were unaware of the influence of this experimental manipulation (even if it did affect their perceptions or behaviors). Perception and behavior can be affected with no need of awareness of influence (Bargh & Chartrand, 2000). A final limitation was the effect size of the findings. The indirect effect of similarity identification on the influence of social identity cues on enjoyment was small, whereas the indirect effect of group identification was medium in size (Tables 3 and 4). This is congruent with meta-analyses showing that experimental social identity manipulations range from medium to small sizes, and that experimental studies yield smaller effects than non-causal social identity studies (van Zomeren, Postmes, & Spears, 2008). Overall, the mediation results should be interpreted with caution because of their effect sizes.

Conclusion Sociotechnical factors can critically affect user experience when playing video games. In particular, game mode and bestowing avatars with social identity cues affected players’ game enjoyment. Group and similarity identification were Ó 2017 Hogrefe Publishing

45

key mediators of the link between virtual character type and social identity cues on enjoyment. Social presence proved to be a less reliable mediator. This sheds light on the psychological mechanisms involved in playing video games, and shines a spotlight on the crucial role of identification with gaming groups and virtual characters in predicting enjoyment. These findings illustrate that we have the capacity to socially create more enjoyable gaming experiences by varying game modes and highlighting or concealing social identity cues. Acknowledgments This work was supported by national funds provided through Fundação para a Ciência e a Tecnologia (FCT) (UID/CEC/50021/2013) and by the INVITE project (UTA-Est/MAI/0008/2009) funded by FCT under a UT-Austin/Portugal cooperation agreement.

References Bargh, J. A., & Chartrand, T. L. (2000). The mind in the middle: A practical guide to priming and automaticity research. In H. T. Reis & C. M. Judd (Eds.), Handbook of research methods in social and personality psychology (pp. 253–285). New York: Cambridge University Press. Cialdini, R. B., Borden, R. J., Thorne, A., Walker, M. R., Freeman, S., & Sloan, L. R. (1976). Basking in reflected glory: Three (football) field studies. Journal of Personality and Social Psychology, 34, 366–375. doi: 10.1037/0022-3514.34.3.366 Cyr, D., Hassanein, K., Head, M., & Ivanov, A. (2007). The role of social presence in establishing loyalty in e-Service environments. Interacting with Computers, 19, 43–56. doi: 10.1016/ j.intcom.2006.07.010 Eveleth, D. M., & Eveleth, A. B. (2010). Team identification, team performance and leader-member exchange relationships in virtual groups: Findings from massive multi-player online role play games. International Journal of Virtual Communities and Social Networking, 2, 52–66. doi: 10.4018/jvcsn. 2010010104 Guadagno, R. E., Blascovich, J., Bailenson, J. N., & McCall, C. (2007). Virtual humans and persuasion: The effects of agency and behavioral realism. Media Psychology, 10(1), 1–22. doi: 10.1080/15213260701300865 Guegan, J., Moliner, P., & Buisine, S. (2015). Why are online games so self-involving: A social identity analysis of massively multiplayer online role-playing games. European Journal of Social Psychology, 45, 349–355. doi: 10.1002/ejsp.2103 Harms, C., & Biocca, F. (2004). Internal consistency and reliability of the networked minds social presence measure. In M. Alcaniz & B. Rey (Eds.), Seventh annual international workshop: Presence. Valencia: Universidad Politécnica de Valencia. Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach. New York: The Guilford Press. Heeter, C. (1995). Communication research on consumer VR. In F. Biocca & M. R. Levy (Eds.), Communication in the age of virtual reality (pp. 191–218). Hillsdale, NJ: Lawrence Erlbaum Associates. Journal of Media Psychology (2019), 31(1), 35–47


46

Hartmann, T., & Klimmt, C. (2006). Gender and computer games: Exploring females’ dislikes. Journal of Computer-Mediated Communication, 11, 910–931. doi: 10.1111/j.1083-6101.2006.00301.x Johnson, D., Wyeth, P., Clark, M., & Watling, C. (2015, April). Cooperative game play with avatars and agents: Differences in brain activity and the experience of play. In CHI ‘15 Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems (pp. 3721–3730). New York. Klimmt, C., & Hartmann, T. (2008). Mediated interpersonal communication in multiplayer videogames: Implications for entertainment and relationship management. In E. A. Konijn, M. Tanis, S. Utz, & S. B. Barnes (Eds.), Mediated interpersonal communication (pp. 309–330). London: Routledge. Kenny, D. A., Kashy, D. A., & Cook, W. L. (2006). Dyadic data analysis. New York, NY: Guilford. Klimmt, C., Hefner, D., & Vorderer, P. (2009). The video game experience as “true” identification: A theory of enjoyable alterations of players’ self-perception. Communication Theory, 19, 351–373. doi: 10.1111/j.1468-2885.2009.01347.x Lea, M., & Spears, R. (1992). Paralanguage and social perception in computer-mediated communication. Journal of Organizational Computing and Electronic Commerce, 2, 321–341. doi: 10.1080/10919399209540190 Lee, K. M. (2004). Presence, explicated. Communication Theory, 14, 27–50. doi: 10.1111/j.1468-2885.2004.tb00302.x Lin, A., Gregor, S., & Ewing, M. (2008). Developing a scale to measure the enjoyment of Web experiences. Journal of Interactive Marketing, 22, 40–57. doi: 10.1002/dir.20120 Lim, S., & Reeves, B. (2009). Computer agents versus avatars: Responses to interactive game characters controlled by a computer or other player. International Journal of HumanComputer Studies, 68, 57–68. doi: 10.1016/j.ijhcs.2009.09.008 Lombard, M., & Ditton, T. (1997). At the heart of it all: The concept of presence. Journal of Computer-Mediated Communication, 3. [Advance online publication]. doi: 10.1111/j.1083–6101.1997. tb00072.x Marques, J. M., & Paez, D. (1994). The “black sheep effect”: Social categorization, rejection of ingroup deviates, and perception of group variability. In W. Stroebe & M. Hewstone (Eds.), European review of social psychology (Vol. 5, pp. 37–68). Chichester, UK: Wiley. doi: 10.1080/14792779543000011 McGee, K., Merritt, T., & Ong, C. (2013). Understanding differences in enjoyment: Playing games with human or AI team-mates. In Y. G. Ji (Ed.), Advances in affective and pleasurable design (pp. 446–451). Boca Raton, FL: CRC Press. Ratan, R. A., Chung, J. E., Shen, C., Williams, D., & Poole, M. S. (2010). Schmoozing and smiting: Trust, social institutions, and communication patterns in an MMOG. Journal of ComputerMediated Communication, 16, 93–114. doi: 10.1111/j.10836101.2010.01534.x Ravaja, N., Saari, T., Turpeinen, M., Laarni, J., Salminen, M., & Kivikangas, M. (2006). Spatial presence and emotions during video game playing: Does it matter with whom you play? Presence: Teleoperators and Virtual Environments, 15, 381–392. doi: 10.1162/pres.15.4.381 Sassenberg, K. (2002). Common bond and common identity groups on the Internet: Attachment and normative behavior in on-topic and off-topic chats. Group Dynamics: Theory, Research, and Practice, 6, 27–37. doi: 10.1037/1089-2699.6.1.27 Schmierbach, M., Xu, Q., Oeldorf-Hirsch, A., & Dardis, F. E. (2012). Electronic friend or virtual foe: Exploring the role of competitive and cooperative multiplayer video game modes in fostering enjoyment. Media Psychology, 15, 356–371. doi: 10.1080/ 15213269.2012.702603

Journal of Media Psychology (2019), 31(1), 35–47

J. Peña et al., Virtual Characters and Social Identity Effects

Spears, R., & Lea, M. (1994). Panacea or panopticon? The hidden power in computer-mediated communication. Communication Research, 21, 427–459. doi: 10.1177/009365094021004001 Spears, R., & Postmes, T. (2015). Group identity, social influence, and collective action online: Extensions and applications of the SIDE model. In S. S. Sundar (Ed.), The handbook of psychology and communication technology (pp. 23–46). West Sussex, UK: Wiley-Blackwell. Tajfel, H., & Turner, J. C. (1986). The social identity theory of intergroup behavior. In S. Worchel & W. Austin (Eds.), The psychology of intergroup relations (2nd ed., pp. 7–24). Chicago, IL: Nelson-Hall. Tamborini, R., Bowman, N. D., Eden, A., Grizzard, M., & Organ, A. (2010). Defining media enjoyment as the satisfaction of intrinsic needs. Journal of Communication, 60, 758–777. doi: 10.1111/j.1460-2466.2010.01513.x Tanis, M., & Postmes, T. (2007). Two faces of anonymity: Paradoxical effects of cues to identity in CMC. Computers in Human Behavior, 23, 955–970. doi: 10.1016/j.chb. 2005.08.004 Turner, J. C., Hogg, M. A., Oakes, P. J., Reicher, S. D., & Wetherell, M. S. (1987). Rediscovering the social group: A self-categorization theory. New York: Basil Blackwell. Van Looy, J., Courtois, C., De Vocht, M., & De Marez, L. (2012). Player identification in online games: Validation of a scale for measuring identification in MMOGs. Media Psychology, 15, 197–221. doi: 10.1080/15213269.2012.674917 van Zomeren, M., Postmes, T., & Spears, R. (2008). Toward an integrative social identity model of collective action: Quantitative research synthesis of three socio-psychological perspectives. Psychological Bulletin, 134, 504–535. doi: 10.1037/0033-2909.134.4.504 Vorderer, P., Klimmt, C., & Ritterfeld, U. (2004). Enjoyment: At the heart of media entertainment. Communication Theory, 14, 388–408. doi: 10.1111/j.1468-2885.2004.tb00321.x Wang, Z., Walther, J. B., & Hancock, J. T. (2009). Social identification and interpersonal communication in computermediated communication: What you do versus who you are in virtual groups. Human Communication Research, 35, 59–85. doi: 10.1111/j.1468-2958.2008.01338.x Weibel, D., Wissmath, B., Habegger, S., Steiner, Y., & Groner, R. (2008). Playing online games against computer- vs. humancontrolled opponents: Effects on presence, flow, and enjoyment. Computers in Human Behavior, 24(5), 2274–2291. doi: 10.1016/j.chb.2007.11.002 Yilmaz, G., & Peña, J. (2014). The influence of social categories and interpersonal behaviors on future intentions and attitudes to form subgroups in virtual teams. Communication Research, 41, 333–352. doi: 10.1177/ 0093650212443696 Received October 14, 2015 Revision received August 26, 2016 Accepted September 18, 2016 Published online Apri 28, 2017 Jorge Peña Department of Communication University of California 367 Kerr Hall Davis, CA, 95616 USA jpena@ucdavis.edu

Ó 2017 Hogrefe Publishing


J. Peña et al., Virtual Characters and Social Identity Effects

Jorge Peña (PhD, Cornell University) is an Associate Professor at the Department of Communication at University of California, Davis. His research focuses on cognitive, affective, and behavioral processes involved in online collaboration and play.

Jannath Ghaznavi (PhD, University of California, Davis) is an Instructional Designer at California State University, Northridge. Her research focuses on media uses and effects, particularly the influence of media and technology on society and perceptions of marginalized groups.

Nicholas Brody (PhD, University of Texas at Austin) is an Assistant Professor at the Department of Communication Studies at the University of Puget Sound. His research interests include bystander behavior during cyberbullying incidents and interpersonal communication and language use in mediated contexts.

Rui Prada (PhD, Instituto Superior Técnico) is an Associate Professor of Computer Science at Instituto Superior Técnico, Universidade de Lisboa (IST) and a Senior Researcher at the Intelligent Agents and Synthetic Characters Group (GAIPS), INESC-ID Lisbon. He conducts research on social intelligent agents, human-agent interaction, computer games and applied gaming, and user centered and experience design.

Ó 2017 Hogrefe Publishing

47

Carlos Martinho (PhD, Instituto Superior Técnico) is an Assistant Professor of Computer Science and Engineering at Instituto Superior Técnico, Universidade de Lisboa (IST) and a Senior Researcher at the Intelligent Agents and Synthetic Characters Group (GAIPS), INESC-ID Lisbon. He studies believable synthetic characters, tools to support creativity, and artificial intelligence and affective computing to enhance user experience with computer and video games.

Pedro A. Santos (PhD, Technische Universität Chemnitz) is an Assistant Professor of Mathematics at Instituto Superior Técnico, Universidade de Lisboa (IST) and a Senior Researcher at the Intelligent Agents and Synthetic Characters Group (GAIPS), INESC-ID Lisbon. His research interests include functional analysis, serious games, artificial intelligence, and complex systems. He has also published computer and board games.

Joana Dimas (PhD, Instituto Superior Técnico) does research at the intersection between social psychology, information systems, and computer engineering. She is interested in improving computer game characters and enhancing player experience by combining cognitive science and game design.

Hugo Damas (MSc, Instituto Superior Técnico) is a computer scientist working to advance technological and scientific understanding of artificial intelligence. He is currently focusing on how to develop and evaluate artificial intelligence, and is concentrating on a career as game developer and researcher.

Journal of Media Psychology (2019), 31(1), 35–47


Research Report

Investigating the Negation of Media Stereotypes Ability and Motivation as Moderators Florian Arendt Department of Communication Science and Media Research, University of Munich (LMU), Germany Abstract: We investigated the negation of media stereotypes. Negation refers to an internal attempt to negate stereotypic content (“No! This is not true!”). The process of negation is important because a critical assessment of stereotypic content can be beneficial for stereotype and prejudice reduction. This fact is a crucial reason why readers’ disagreement regarding simplified stereotypic depictions is of central interest for mass communication research and media literacy campaigns. Importantly, factors that can increase negation are of special interest. Although the ability and motivation to process stereotypic content can be theoretically identified as potential influencing factors, mediastereotype research has not yet tested the influence of these factors on negation. In Experiment 1 (N = 347), we manipulated the motivation to negate by presenting awareness material. We informed some of the participants that the news media often do not represent the world as it is, but sometimes do so in a stereotypic way. Analyses revealed that participants who received the awareness material before reading negated to a higher extent. In Experiment 2 (N = 223), we investigated the impact of ability by manipulating the time participants had to negate stereotypic content. The ability to negate was assumed to be higher the more time the participants had to process stereotypic information. As hypothesized, negation was higher when there was more time available. Interestingly, the increase in effect size was dampened the more time was available, which indicated a curvilinear relationship. Implications for media-literacy campaigns are discussed. Keywords: media stereotypes, negation, ability, motivation, media literacy, awareness

Research has shown that the mass media tend to present simplified pictures of specific social groups (Harris, 2009, pp. 64–106). Exposure to such simplified mediated information can influence perceptions of and behavior toward stereotyped groups (Roskos-Ewoldsen, RoskosEwoldsen, & Carpentier, 2009). This is why readers’ disagreement regarding simplified stereotypic depictions – often disparaging when applied to out-groups – should be of scholarly concern. One important concept in this regard is negation. It refers to an internal attempt to negate stereotypic information (e.g., “No! This is not true!”; see Strack & Deutsch, 2004). Although research has gathered much knowledge about the content and effects of media stereotypes (Mastro, 2009), the lack of empirical evidence regarding the negation of media stereotypes is surprising. As critical assessments of stereotypic content can be beneficial for stereotype and prejudice reduction (Arendt, 2013), targeted scholarly attention to the factors that stimulate such assessments is important. We investigated the influence of ability- and motivation-related factors on the negation of stereotypic content.

Journal of Media Psychology (2019), 31(1), 48–54 DOI: 10.1027/1864-1105/a000198

Negation of Media Stereotypes Negation describes an internal attempt to negate encoded information (Peters & Gawronski, 2011). This cognitive process occurs during exposure and is, therefore, an online validity judgment (see Hastie & Park, 1986). Previous research has shown that the negation of stereotypic information can reduce effects on stereotyping (Gawronski, Deutsch, Mbirkou, Seibt, & Strack, 2008; Gregg, Seibt, & Banaji, 2006; Kawakami, Dovidio, Moll, Hermsen, & Russin, 2000; Peters & Gawronski, 2011). Importantly, recent mass-communication research has also shown that negation can be beneficial. For example, Arendt (2013) tested a mediation model where news-stereotype exposure was modeled to influence stereotype-related outcomes through its effect on negation. It has been revealed that negation during exposure reduced the news effect. This has been called the “dampening effect of negation.” Arendt, Marquart, and Matthes (2015) replicated this effect in the realm of political advertising. As negation reduces the effect of media stereotype exposure, negation training could be

Ó 2017 Hogrefe Publishing


F. Arendt, Negation of Media Stereotypes

part of media literacy campaigns (see Ramasubramanian, 2007).

Stimulating Negation Although negation could be an easy-to-use stereotypereduction strategy, it is important to note that negation requires a certain amount of processing time, intention, and cognitive capacity (Strack & Deutsch, 2004). Only if there is enough time and the intention to assess the content critically will recipients negate media stereotypes. Unfortunately, ability (e.g., time constraints) and motivation (e.g., no intention) can be low (see Strack & Deutsch, 2004). Individuals can choose between multiple cognitive strategies based on the current situation (see e.g., Olson & Fazio, 2009). According to Fiske and Taylor (1991), humans are “motivated tacticians”: Humans strategically allocate cognitive resources and thus use a processing style emphasizing either speed and ease or accuracy and rational logic. The processing choice depends on the ability and the motivation to process the information. Negation involves a more thoughtful, deliberate form of processing (Strack & Deutsch, 2004). Unfortunately, the process of negation seems to be restricted in the mass communication context. For example, negation may decrease enjoyment. Accordingly, we assumed that individuals typically only have a low motivation to start negation.

Experiment 1 Negation requires a particular level of intention (Strack & Deutsch, 2004). We hypothesized that the negation of media stereotypes is stronger in individuals with high levels of motivation (Hypothesis 1). Experiment 1 investigated whether awareness material can stimulate negation. We tested whether simply telling individuals that they should negate may be enough to reduce a detrimental effect.

Method We utilized a 2 (media-stereotype exposure: yes or no)  3 (awareness-material presentation: before exposure, after exposure, none) factorial experimental design. All individuals read eight crime articles merged with filler texts. We used the “criminal foreigner” news stereotype (Arendt, 2010). Offenders were labeled as foreigners or not. The decision to use the criminal foreigner stereotype builds on previous research showing that crime news tends to associate specific social (minority) groups with negative Ó 2017 Hogrefe Publishing

49

behavior. In Austria, where both experiments were conducted, the social category foreigners has been shown to be overrepresented as criminals in a leading news outlet (Arendt, 2010). Participants A total of 347 students who were enrolled in an introductory lecture on research methods participated. Of these participants, 79% were women. The participants ranged in age from 17 to 49 years (M = 20.62, SD = 3.53). The majority indicated their nationality as Austrian (70%), followed by German (20%), or other (10%). Media-Stereotype Exposure The crime texts were constructed based on original articles from a tabloid-style newspaper. In the treatment condition (n = 173), offenders were labeled as foreigners (mediastereotype exposure). The foreign nationality of the offender was mentioned once in the headline and one more time in the body of each text. Individuals in the control condition (n = 174) read the same crime articles, but the foreign nationalities were not mentioned. Awareness-Material Exposure We informed participants that the news media often do not represent the world as it is, but sometimes do so in a stereotypic way. They read awareness material informing them that there were scientific studies that had investigated how some news media outlets tended to overrepresent specific social groups such as foreigners as criminals, even when official statistics did not support this. We closed this awareness information by appealing to the participants to consider this information when reading the news. We used three conditions. In one condition, participants received this awareness material before reading (n = 115). A second group received this awareness material after reading (n = 124). This condition was utilized to increase the confidence in the interpretation of results (see later). A third group received the awareness treatment at the end of the study as a part of the debriefing (n = 108). This third condition acted as a control, because the awareness material could not have influenced reading or the negation measure. Negation We used a measure already employed by previous research (Arendt, 2013). Participants were asked on a 7-point scale ranging from 1 (= I totally disagree) to 7 (= I totally agree) about their agreement regarding four statements (e.g., “During the reading of the texts, I repeatedly thought ‘No! This description is not true’”; “I repeatedly negated the crime texts, because they were stereotypic and prejudiced”; “When a criminal was depicted, I thought ‘No! This description is too stereotypic and prejudiced’”; Journal of Media Psychology (2019), 31(1), 48–54


50

F. Arendt, Negation of Media Stereotypes

“I repeatedly thought ‘No! This information is not true – these texts do not represent reality as it is’”; M = 4.13, SD = 1.46, α = .86). We utilized factor analysis to confirm unidimensionality. The outcome was one factor, which explained 70.57% of the variance. Statistical analyses were performed with the factor scores of this variable. Procedure Participants were individually welcomed by a male experimenter. First, they got an overall introduction to the “reading study.” Next, they sat down in front of a computer in individual research cubicles. The articles and the awareness material were presented on the screen. After the reading part was over, participants filled out a computer-administered questionnaire. Finally, they were thanked and debriefed.

Results and Discussion Hypothesis 1 predicted that the awareness material moderated the effect of media-stereotype exposure. Analysis of variance (ANOVA) was used to test this hypothesis. We found a main effect for media-stereotype exposure, F(1, 341) = 56.07, p < .001, η2 = .14, and for awareness-material exposure, F(2, 341) = 6.19, p = .002, η2 = .04. More importantly, we hypothesized a two-way interaction. As hypothesized, there was a significant interaction, F(2, 341) = 3.22, p = .041, η2 = .02. Figure 1 depicts this interaction effect. As expected, analyses of simple effects revealed that those who received the awareness material before exposure showed the strongest news-stereotype exposure effect, t(113) = 7.09, p < . 001, d = 1.33. Although on a lower level, the reading of stereotypic articles also influenced negation in those receiving the awareness material after reading, t(122) = 3.45, p = .001, d = 0.62, as well as in those who did not receive the awareness treatment, t(106) = 2.87, p = .005, d = 0.55. It was not surprising that (journalism and mass communication) students in our sample negated even in the absence of an awareness-material treatment before reading. This is in line with previous research (Arendt, 2013). The most important finding is that those who received the awareness material before exposure showed the strongest effect. This supports Hypothesis 1.1 Explorative analyses found a significant difference between participants who received the awareness material 1

Figure 1. Experiment 1: Influence of a motivation-related factor (awareness material) on negation. The awareness material was presented before reading, after reading, or after the measurement of negation (i.e., “none”).

before and after reading in those who did not read the stereotypic articles, t(118) = 3.47, p = .001, d = 0.65. The presentation immediately before answering the negation questions may have introduced a social desirability bias: To present themselves as critical thinkers, participants might have chosen higher response options on the negation measure. This effect, however, was additive in nature (i.e., the media-stereotype exposure effect was parallel to the control condition). This fact is important when interpreting the results of the first experiment: We found a significant difference between those who received the awareness material before and after reading in participants who did not read the stereotypic articles. We return to this finding in the general discussion section.

Experiment 2 As the negation of stereotypical content requires a certain amount of processing time (Strack & Deutsch, 2004), it was hypothesized that the negation of media stereotypes is stronger the more time individuals have available for negation (Hypothesis 2). Time constraints should limit negation.

Method We utilized a 2 (media-stereotype exposure: yes or no)  4 (available time for negation: 100, 500, 1,000, 4,000 ms)

As a supplement to the six experimental conditions used to test the hypothesis, we additionally included a seventh condition for a methods experiment. We tested whether there is a difference between a control group (A) using crime articles and a control group (B) using crimeunrelated articles. The question was whether the priming of the crime concept in control group A would be able to elicit a comparable (or different) effect to that elicited by control group B. This was not the case: There was no significant difference between both control conditions on negation, t(107) = 1.15, p = .254, d = 0.22. Although the predicted interaction effect was also consistent with the hypothesis (and even more pronounced) had we used control group B, F(2, 342) = 4.93, p = .008, η2 = .03, we used control condition A for the analysis of the 2  3 design as theorized a priori to data collection. Furthermore, we collected additional variables in the questionnaire that we do not report in the manuscript.

Journal of Media Psychology (2019), 31(1), 48–54

Ó 2017 Hogrefe Publishing


F. Arendt, Negation of Media Stereotypes

factorial experimental design. All individuals were exposed to newspaper headlines. As in Experiment 1, we used the criminal foreigner news stereotype. Offenders were labeled as foreigners or not. Furthermore, we manipulated the presentation time of those headlines, ranging from very short flashes to rather long displays. Participants A total of 223 students were enrolled in an introductory lecture on research methods. This was the same lecture as in Experiment 1, but from a different year. Thus, none of the individuals participating in Experiment 1 participated in Experiment 2. Of these individuals, 82.1% were women. The participants ranged in age from 17 to 50 years (M = 20.67, SD = 3.69). The majority indicated their nationality as Austrian (69.9%), followed by German (18.5%), or other (11.6%). Experimental Manipulation We did not rely on the presentation of whole stereotypic articles as in Experiment 1, because this impeded the ability to manipulate exposure time. Thus, we decided to use headlines instead. Participants were exposed to stereotypic headlines (n = 112) or matched nonstereotypical headlines (n = 111). Whereas participants in the treatment condition were exposed to news stereotypes (e.g., “Romanian arrested”), individuals in the control condition were not (e.g., “Offender arrested”). We utilized a method developed by Chartrand and Bargh (1996), adapting it for the present experiment. All headlines appeared at one of four locations on the computer screen, equidistant from the fixation point at angles of 4500 , 13500 , 22500 , and 31500 (i.e., one in each of the four quadrants). The location order was randomly varied. Participants’ cover task for this “attention and vigilance task” was to press one of two buttons to indicate on which side of the screen the headline appeared. After a practice block using four unrelated bogus headlines, the test block consisted of 20 headlines. All headlines were between two and seven words in length and relied on an easy language. The amount of time between headline presentations randomly varied from 2 to 7 s to enhance the cover story in the attention and vigilance task. Participants were told to focus on the central asterisk presented at the start of each trial. Next, the headlines were presented as short flashes or for longer presentation times (100, 500, 1,000, 4,000 ms). The very short 100-ms duration was chosen as a control (n = 55), because it has been shown that more time is required to move the eyes away from the initial fixation point toward the parafoveally presented stimulus (Chartrand & Bargh, 1996). It was not 2

51

possible for participants to read the headlines in this condition, even if they immediately looked toward the location of the flash. The 500-ms duration allowed for a quick scan of the headline (n = 52). This duration enabled the ability to negate, as participants were able to encode the stereotypic information. The 1,000-ms (n = 55) and 4,000-ms (n = 61) durations enabled the reading of the whole headline.2 Negation We used the same negation measure as employed in Experiment 1, but used the word headline instead of crime texts (M = 4.39, SD = 1.23, α = .85). Again, we utilized factor analysis to confirm unidimensionality and the outcome was one factor, which explained 69.43% of the variance. Statistical analyses were performed with the factor scores. Procedure The procedure was very similar to Experiment 1. After the attention and vigilance task was finished, participants filled out a computer-administered questionnaire including the negation measure.

Results and Discussion Hypothesis 2 predicted that the available time participants had moderated the effect of media-stereotype exposure. An ANOVA was calculated to test this hypothesis. As in Experiment 1, we found a main effect for media-stereotype exposure, F(1, 215) = 37.92, p < .001, η2 = .15, and available time, F(3, 215) = 3.48, p = .017, η2 = .05. More importantly, analyses revealed the hypothesized two-way interaction, F(3, 215) = 5.67, p = .001, η2 = .07. Figure 2 depicts this interaction effect. As in Experiment 1, we conducted analyses of simple effects. Exposure to stereotypic headlines did not influence those participants who got all of the headlines presented for just 100 ms, t(53) = 0.13, p = .900, d = 0.03. By contrast, the 500-ms headline presentation produced a significant effect, t(50) = 2.23, p = .030, d = 0.62. The size of the effect increased for the 1,000-ms duration, t(53) = 4.55, p < .001, d = 1.24. The strongest effect (on a descriptive level) was observed for those who were able to deeply read the headlines for 4,000 ms, t(59) = 4.93, p < .001, d = 1.25. This supports Hypothesis 2. This experiment showed that an ability-related factor can increase negation. Interestingly, increases in the low levels of available time produced stronger effects on negation than did additional increases in the higher levels. There was virtually no difference between the 1,000-ms and 4,000-ms conditions. This indicates that there was a curvilinear effect of available time on negation.

The experimental script can be obtained upon request.

Ó 2017 Hogrefe Publishing

Journal of Media Psychology (2019), 31(1), 48–54


52

Figure 2. Experiment 2: Influence of an ability-related factor (available time to negate) on negation.

General Discussion Although the social-cognition literature has revealed that negation requires a certain amount of time and level of intention (Strack & Deutsch, 2004), media-stereotype research has not yet tested the effect of these factors on negation. We present supporting experimental evidence by showing that negation is stronger in those who are aware of possible stereotypic content (Experiment 1) and in those who have more time available to process the stereotypic information (Experiment 2).

Implications for Campaigns Despite the findings’ contribution to media-stereotyping literature, these findings also have implications for medialiteracy campaigns. Negation could be an easy-to-use strategy to reduce effects of stereotypic content (see Ramasubramanian, 2007). The present paper presents an easy-to-use stereotype-reduction strategy (Experiment 1). Merely raising awareness that the news may include stereotypic depictions, paired with an appeal to consider this fact when exposed to the news (“awareness” appeal), increased negation. In addition, the second experiment showed that increasing exposure time increased negation. This information can also be used in campaigns, for example, by using a “take-your-time” appeal: Individuals can change their processing habits in specific exposure situations. It is a valuable strategy to increase processing time (i.e., how long one ruminates on stereotypic content). Nevertheless, this strategy is challenged by the fact that modern media environments foster fast processing of apparently unlimited information. Furthermore, new communication technologies elicit changes in media use habits (e.g., smartphone use with more distractions). Thus, tailored campaign messages are needed. Journal of Media Psychology (2019), 31(1), 48–54

F. Arendt, Negation of Media Stereotypes

Media-stereotype research helps to identify specific target situations. For instance, a campaign could try to persuade target audiences to critically assess stereotypic crime news content (i.e., target situation). Participants should try to recognize stereotypic depictions (e.g., awareness appeal) and take their time to negate the information (e.g., take-your-time appeal). Although recipient-based strategies can be used to overcome possible negative media effects, what journalists report is not without consequences as well. Media-based strategies to prejudice reduction are also worth noting (see Ramasubramanian, 2007). For example, research has revealed that the presentation of counter-stereotypic content can have beneficial consequences: The use of counter-stereotypes can reduce prejudice-related outcomes (e.g., a reduction in the activation of stereotypes and negative evaluations; see Gawronski et al., 2008).

Limitations As with every study, the present one has its limitations. First, we used a self-report scale for negation measurement. Such measures can be influenced by a social desirability bias. In Experiment 1, we found a significant difference between those who received the awareness material before and after reading in participants who did not read the stereotypic articles (see Figure 1). Individuals receiving the awareness material after reading might have overestimated the presence of stereotypical news content in the articles without any stereotypic depictions. The presentation of the awareness material after reading might have contributed to a more pronounced social desirability bias: Individuals might have decided to present themselves as critical thinkers although they did not negate during reading. It is important to note, however, that the size of the effect of media stereotype exposure – as predicted by the hypothesis – was the same in those who received the awareness material after exposure and in those who did not receive awareness material at all. Therefore, this finding does not influence the interpretation of the hypothesis test: Individuals who received the awareness material before exposure showed the strongest effect of stereotypic content on negation. Nevertheless, this unexpected finding can be a good starting point for future research. For example, future research can refine the negation measure by using real-time-response measurement procedures. This would increase the clarity in the interpretation of results (e.g., it will help to separate out the independent contribution of a social desirability bias and the impact of awareness material in Experiment 1). Negation – defined as an online validity judgment (Hastie & Park, 1986) – is likely to be caused by different processes. Importantly, the online Ó 2017 Hogrefe Publishing


F. Arendt, Negation of Media Stereotypes

process has been shown to reduce effects on stereotyping (see Negation of Media Stereotypes). It seems to be unlikely that an increase in the retrospective negation measure due to a social desirability bias positively contributes to stereotype reduction as online negation.3 Second, we used a student sample for economic reasons. Communication students are generally more critical toward media content. Although this is likely to have increased the baseline level of negation in absolute terms, we were interested in the relative differences between the conditions. However, future research should replicate the findings with a general population sample. Third, we cannot make any claims about the duration of effects. We only investigated short-term effects on negation in laboratory experiments. Future research can test whether these effects are stable and whether repeated exposure to awareness material (e.g., in an antiprejudice campaign) can chronically stimulate negation.

Conclusion Despite the limitations, the present study shows that the negation of media stereotypes can be influenced by motivation- and ability-related factors. Because exposure to media stereotypes can influence perceptions of and behavior toward stereotyped groups (Roskos-Ewoldsen et al., 2009), a targeted research effort is important, as it could contribute to media consumers’ critical reading of the daily news, and thus possibly to a reduction in some of the detrimental consequences of media stereotypes regarding stereotyping and prejudice. In fact, increased levels of critical reading including negation may even change the news content itself, as news organizations might change their reporting style owing to changed reader interests.

References Arendt, F. (2010). Cultivation effects of a newspaper on reality estimates and explicit and implicit attitudes. Journal of Media Psychology, 22, 147–159. doi: 10.1027/1864-1105/a000020 Arendt, F. (2013). Dose-dependent media priming effects of stereotypic newspaper articles on implicit and explicit stereotypes. Journal of Communication, 63, 830–851. doi: 10.1111/ jcom.12056

3

53

Arendt, F., Marquart, F., & Matthes, J. (2015). Effects of right-wing populist political advertising on implicit and explicit stereotypes. Journal of Media Psychology, 27, 178–189. doi: 10.1027/ 1864-1105/a000139 Chartrand, T., & Bargh, J. (1996). Automatic activation of impression formation and memorization goals: Nonconscious goal priming reproduces effects of explicit task instructions. Journal of Personality and Social Psychology, 71, 464–478. doi: 10.1037/0022-3514.71.3.464 Dixon, T. L. (2009). “He was a Black guy”: How news’ misrepresentation of crime creates fear of Blacks. In R. A. Lind (Ed.), Race/gender/media: Considering diversity across audiences, content, and producers (pp. 24–29). Boston, MA: Pearson Education. Fiske, S. T., & Taylor, S. E. (1991). Social cognition. New York, NY: McGraw-Hill. Gawronski, B, Deutsch, R., Mbirkou, S., Seibt, B., & Strack, F. (2008). When “just say no” is not enough: Affirmation versus negation training and the reduction of automatic stereotype activation. Journal of Experimental Social Psychology, 44, 370– 377. doi: 10.1016/j.jesp.2006.12.004 Gregg, A., Seibt, B., & Banaji, M. (2006). Easier done than undone: Asymmetry in the malleability of implicit preferences. Journal of Personality and Social Psychology, 90, 1–20. doi: 10.1037/ 0022-3514.90.1.1 Harris, R. J. (2009). A cognitive psychology of mass communication. New York, NY: Routledge. Hastie, R., & Park, B. (1986). The relationship between memory and judgment depends on whether the judgment task is memory-based or on-line. Psychological Review, 93, 258–268. doi: 10.1037/0033-295X.93.3.258 Kawakami, K., Dovidio, J., Moll, J., Hermsen, S., & Russin, A. (2000). Just say no (to stereotyping): Effects of training in the negation of stereotypic associations on stereotype activation. Journal of Personality and Social Psychology, 78, 871–888. doi: 10.1037/0022-3514.78.5.871 Mastro, D. (2009). Effects of racial and ethnic stereotyping. In J. Bryant & M. B. Oliver (Eds.), Media effects. Advances in theory and research (pp. 325–341). New York, NY: Taylor & Francis. Olson, M. A., & Fazio, R. H. (2009). Implicit and explicit measures of attitudes: The perspective of the MODE model. In R. E. Petty, R. H. Fazio, & P. Briñol (Eds.), Attitudes: Insights from the new implicit measures (pp. 19–63). New York, NY: Psychology Press. Peters, K., & Gawronski, B. (2011). Are we puppets on a string? Comparing the impact of contingency and validity on implicit and explicit evaluations. Personality and Social Psychological Bulletin, 37, 557–569. doi: 10.1177/0146167211400423 Ramasubramanian, S. (2007). Media-based strategies to reduce racial stereotypes activated by news stories. Journalism and Mass Communication Quarterly, 84(2), 249–264. doi: 10.1177/ 107769900708400204 Roskos-Ewoldsen, D. R., Roskos-Ewoldsen, B., & Carpentier, F. D. (2009). Media priming. An updated synthesis. In J. Bryant & M. B. Oliver (Eds.), Media effects. Advances in theory and research (pp. 74–93). New York, NY: Taylor and Francis. Strack, F., & Deutsch, R. (2004). Reflective and impulsive determinants of social behavior. Personality and Social Psychology Review, 8, 220–247. doi: 10.1207/s15327957pspr0803_1

Research on the so-called mismemory effect may be relevant in this context as well (e.g., Dixon, 2009). The mismemory effect was described as a phenomenon where an “incomplete” crime script generates strong expectations for “missing” details of the story. This can lead readers to “fill in” the “gaps” in the news content. These filled-in content elements are then memorized as a supplement to the “real” story details. The presentation of the awareness material after reading might have contributed to a more pronounced mismemory effect: The awareness material might have primed memory traces related to the missing details of the crime script, which in turn have increased the retrospective negation measure. Future research should elaborate on this.

Ó 2017 Hogrefe Publishing

Journal of Media Psychology (2019), 31(1), 48–54


54

Received April 14, 2015 Revision received April 26, 2016 Accepted May 17, 2016 Published online January 18, 2017

Florian Arendt Department of Communication Science and Media Research University of Munich (LMU) Oettingenstraße 67 80538 Munich Germany florian.arendt@univie.ac.at

Journal of Media Psychology (2019), 31(1), 48–54

F. Arendt, Negation of Media Stereotypes

Florian Arendt (PhD) is a postdoctoral researcher (“Akademischer Rat”) at the Department of Communication Science and Media Research, LMU Munich, Germany. His research focuses on media stereotyping, political communication, health communication, and research methods.

Ó 2017 Hogrefe Publishing


Instructions to Authors Journal of Media Psychology (JMP) is committed to publishing original, high-quality papers which cover the broad range of media psychological research. This peer-reviewed journal focuses on how human beings select, use, and experience various media as well as how media (use) can affect their cognitions, emotions, and behaviors. It is also open to research from neighboring disciplines as far as this work ties in with psychological concepts of the uses and effects of the media. In particular, it publishes multidisciplinary papers that reflect a broader theoretical and methodological spectrum and comparative work, e.g., cross-media, cross-gender, or crosscultural. As JMP is intended to foster Open Science Practices, authors are offered to publish their data and materials (i.e., stimuli and surveys) as Electronic Supplementary Material on the publisher’s website at http://econtent@hogrefe. com. In line with the Peer Reviewers’ Openness Initiative, authors may be asked by reviewers to share their data and materials at any stage of the reviewing process. Journal of Media Psychology publishes the following types of article: Original Articles, Theoretical Articles, Research Reports, Pre-Registered Reports.

Manuscript submission: All manuscripts should in the first instance be submitted electronically at http://www.editorialmanager.com/ jmp. Detailed instructions to authors are provided at https:// www.hogrefe.com/j/jmp Copyright Agreement: By submitting an article, the author confirms and guarantees on behalf of him-/herself and any coauthors that the manuscript has not been submitted or published elsewhere, and that he or she holds all copyright in and titles to the submitted contribution, including any figures, photographs, line drawings, plans, maps, sketches, and tables, and that the article and its contents do not infringe in any way on the rights of third parties. The author indemnifies and holds harmless the publisher from any third party claims. The author agrees, upon acceptance of the article for publication, to transfer to the publisher the exclusive right to reproduce and distribute the article and its contents, both physically and in nonphysical, electronic, or other form, in the journal to which it has been

Ó 2019 Hogrefe Publishing

submitted and in other independent publications, with no limitations on the number of copies or on the form or the extent of distribution. These rights are transferred for the duration of copyright as defined by international law. Furthermore, the author transfers to the publisher the following exclusive rights to the article and its contents: 1. The rights to produce advance copies, reprints, or offprints of the article, in full or in part, to undertake or allow translations into other languages, to distribute other forms or modified versions of the article, and to produce and distribute summaries or abstracts. 2. The rights to microfilm and microfiche editions or similar, to the use of the article and its contents in videotext, teletext, and similar systems, to recordings or reproduction using other media, digital or analog, including electronic, magnetic, and optical media, and in multimedia form, as well as for public broadcasting in radio, television, or other forms of broadcast. 3. The rights to store the article and its content in machinereadable or electronic form on all media (such as computer disks, compact disks, magnetic tape), to store the article and its contents in online databases belonging to the publisher or third parties for viewing or downloading by third parties, and to present or reproduce the article or its contents on visual display screens, monitors, and similar devices, either directly or via data transmission. 4. The rights to reproduce and distribute the article and its contents by all other means, including photomechanical and similar processes (such as photocopying or facsimile), and as part of so-called document delivery services. 5. The right to transfer any or all rights mentioned in this agreement, as well as rights retained by the relevant copyright clearing centers, including royalty rights to third parties. Online Rights for Journal Articles: Guidelines on authors’ rights to archive electronic versions of their manuscripts online are given in the ‘‘Guidelines on sharing and use of articles in Hogrefe journals’’ on the journal’s web page at https://www. hogrefe.com/j/jmp October 2016

Journal of Media Psychology (2019), 31(1)


Hogrefe OpenMind Open Access Publishing? It’s Your Choice! Your Road to Open Access Authors of papers accepted for publication in any Hogrefe journal can now choose to have their paper published as an open access article as part of the Hogrefe OpenMind program. This means that anyone, anywhere in the world will – without charge – be able to read, search, link, send, and use the article for noncommercial purposes, in accordance with the internationally recognized Creative Commons licensing standards.

The Choice Is Yours 1. Open Access Publication: The final “version of record” of the article is published online with full open access. It is freely available online to anyone in electronic form. (It will also be published in the print version of the journal.) 2. Traditional Publishing Model: Your article is published in the traditional manner, available worldwide to journal subscribers online and in print and to anyone by “pay per view.” Whichever you choose, your article will be peer-reviewed, professionally produced, and published both in print and in electronic versions of the journal. Every article will be given a DOI and registered with CrossRef.

www.hogrefe.com

How Does Hogrefe’s Open Access Program Work? After submission to the journal, your article will undergo exactly the same steps, no matter which publishing option you choose: peer-review, copy-editing, typesetting, data preparation, online reference linking, printing, hosting, and archiving. In the traditional publishing model, the publication process (including all the services that ensure the scientific and formal quality of your paper) is financed via subscriptions to the journal. Open access publication, by contrast, is financed by means of a one-time article fee (€ 2,500 or US $3,000) payable by you the author, or by your research institute or funding body. Once the article has been accepted for publication, it’s your choice – open access publication or the traditional model. We have an open mind!


The latest knowledge on how to tackle the complexities of hoarding disorder “If you wish to help those who suffer with the debilitating problem of hoarding, get this book and learn from these experienced scientist–practitioners.” Michael A. Tompkins, PhD, ABPP, Co-Director, San Francisco Bay Area Center for Cognitive Therapy; Assistant Clinical Professor, University of California at Berkeley

Gregory S. Chasson / Jedidiah Siev

Hoarding Disorder (Series: Advances in Psychotherapy – Evidence-Based Practice – Volume 40) 2019, viii + 76 pp. US $29.80 / € 24.95 ISBN 978-0-88937-407-2 Also available as eBook Hoarding disorder, classified as one of the obsessive-compulsive and related disorders in the DSM-5, presents particular challenges in therapeutic work, including treatment ambivalence and lack of insight of those affected. This evidence-based guide written by leading experts presents the latest knowledge on assessment and treatment of hoarding disorder. The reader gains a thorough grounding in the treatment of choice for hoarding – a specific form of CBT interweaved with psy-

www.hogrefe.com

choeducational, motivational, and harm-reduction approaches to enhance treatment outcome. Rich anecdotes and clinical pearls illuminate the science, and the book also includes information for special client groups, such as older individuals and those who hoard animals. Printable handouts help busy practitioners. This book is essential reading for clinical psychologists, psychiatrists, psychotherapists, and practitioners who work with older populations, as well as students.


Assessment and treatment of Internet addiction “This excellent book is a pleasure to read. At a time when clinicians are scrambling to learn what they can about the rapidly developing problem of Internet addiction, this book offers them an excellent place to start.” Hilarie Cash, PhD, Chief Clinical Officer and Co-Founder of reSTART Life, PLLC, Fall City, WA – the first residential treatment program for Internet addiction in the US

Daria J. Kuss / Halley M. Pontes

Internet Addiction (Series: Advances in Psychotherapy – Evidence-Based Practice – Volume 41) 2019, iv + 86 pp. US $29.80 / € 24.95 ISBN 978-0-88937-501-7 Also available as eBook This book examines how you can identify, assess, and treat Internet addiction in the most effective manner. Internet use has become an integral part of our daily lives, but at what point does it become problematic? What are the different kinds of Internet addiction? And how can professionals best help clients? This compact, evidence-based guide written by leading experts from the field helps disentangle the debates and controversies around Internet addiction, including social media addiction

www.hogrefe.com

and Internet gaming disorder, and outlines the current assessment and treatment methods. The book presents a 12–15 session treatment plan for Internet and gaming addiction using the method and setting with the best evidence: group CBT. Printable tools in the appendix help clinicians implement therapy. This accessible book is essential reading for clinical psychologists, psychiatrists, psychotherapists, counsellors, social workers, teachers, as well as students.

Profile for Hogrefe

Journal of Media Psychology  

Journal of Media Psychology