MORALITY WHEN THE MIND IS OPAQUE: INTENT VS. OUTCOME ACROSS THE LIFESPAN IN YASAWA, FIJI by Rita Anne McNamara B.A., Washington University in St. Louis, 2009 M.A., The University of British Columbia, 2012 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Psychology) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) August 2016 © Rita Anne McNamara 2016 ii Abstract The ability to infer the presence and contents of other minds is one of the most powerful cognitive tools humans use to navigate our social worlds. Culture is an essential part of these social worlds. But how do mind and culture influence each other? Does culture merely shape the social situations that people navigate in the course of daily life, or does culture fundamentally alter the way that we perceive each other as we move through these social worlds? This dissertation examines how culture shapes mind through the specific example of people living in Yasawa, Fiji. Yasawan culture includes social norms that prohibit discussing others’ actions in terms of mental states – part of a wider phenomenon known as Opacity of Mind, documented in small-scale Indigenous societies and especially prevalent around the Pacific. This culturally-transmitted approach to thinking about minds offers an interesting contrast to the North American focus on minds and internal dispositions as the source of all behaviour. Across five studies, the research presented in this dissertation documents cross-cultural differences in how adults think about beliefs, thoughts, emotions, and social situations. This research also examines how underlying differences in everyday thinking about minds can be applied to social situations, resulting in different emphases on intent or outcome in moral judgments. These differences in intent vs. outcome focus are further shown to be more influenced by culture later in life; children in both cultures show similar degrees of intent focus while North American adults show greater intent focus and Yasawan adults show lower intent focus. This suggests that mental state inference and intentionality reasoning may be a part of core human cognition that is modulated by cultural influences – both increasing and decreasing mentalizing focus – into adulthood. More importantly, this work demonstrates the need to take cultural differences documented outside of urban laboratory research as a serious part of the research process. Cultural differences in adult iii psychological processing should not be considered as variation around an ideal prototype (conveniently documented in Western samples), but as reactions to specific socio-ecological pressures and historical influences that shape individuals into enculturated beings. iv Preface The studies reported here are collaborative projects, and I am the primary author of each. My primary contributions are study designs, fieldwork conducted in Yasawa, analyses, and authorship. The introductory review in chapter 1 and the conclusions in chapter 5 connect the current research to the wider body of work on these topics. They also are intended to integrate disparate threads of research and theory on individual and cultural-level influences on thinking about minds within social cognitive processing. I am the sole author of both of these chapters. These chapters present review of existing data and commentary on how the present research fits within this existing body of work; therefore these chapters do not require ethical review. Chapters 2, 3, and 4 are all manuscripts currently in preparation. All of the research presented in these chapters was approved by the University of British Columbia Behavioural Research Ethics Board (BREB certificate # H12-01044). For chapters 2 and 3, the author order is McNamara, R.A., Willard, A.K., Norenzayan, A., and Henrich, J. Chapter 2 studies 1 and 2 and for chapter 3 study 1 were designed as a collaboration between Willard and myself. Willard conducted field research with Indo-Fijian participants. I designed the experiment in chapter 3 study 2. I conducted the data analysis and wrote the manuscripts. For chapter 4, the author order is McNamara, R.A., Hamlin, J.K., and Henrich, J. General study design is based on previous work by Hamlin. I adapted study design to Yasawan contexts, helped construct stimuli along with research assistants in Hamlin’s lab, collected child and adult data in Yasawa and North America, analyzed the data, and wrote the manuscript. v Table of Contents Abstract .......................................................................................................................................... ii Preface ........................................................................................................................................... iv Table of Contents .......................................................................................................................... v List of Tables ................................................................................................................................ xi List of Figures ........................................................................................................................... xviii Acknowledgements .................................................................................................................... xxi Chapter 1. Introduction: Mentalizing as a Core Foundation of Culture ............................. 1 1.1 Theory, Evolution, and Development of Mentalizing ................................................... 2 1.1.1 Evolutionary History: Mentalizing in Non-Human Primates .............................................. 5 1.1.2 Mentalizing Development and Disorders ........................................................................... 6 18.104.22.168 General Patterns in Timing and Sequence of Mentalizing Development .................... 7 22.214.171.124 Individual and Cultural Influences on Mentalizing Development ............................. 10 126.96.36.199 Mentalizing Deficits in Psychological Disorders ....................................................... 11 1.2 Why Think About Minds? Mentalizing in Competition and Cooperation .................. 12 1.2.1 Define Behavioural Expectations Through Social Structure ............................................. 13 1.2.2 Mentalizing to Coordinate Action ..................................................................................... 14 1.2.3 Empathic Processing and Decisions About Cooperation or Competition ......................... 15 1.2.4 Thinking About Intent in Competition .............................................................................. 17 1.3 Mentalizing and Culture .............................................................................................. 18 1.3.1 Cultural Influences on Mentalizing Processes ................................................................... 19 1.3.2 Mentalizing and Overimitation: Cumulative Cultural Learning and Norms ..................... 20 1.3.3 Mentalizing and Religion ................................................................................................... 23 1.3.4 Mentalizing and Moralizing ............................................................................................... 26 1.4 Overview of Chapters .................................................................................................. 29 vi Chapter 2. Thinking About Thoughts When the Mind is Unknowable: Cross-Cultural Differences in Mental State Reasoning ..................................................................................... 32 2.1 Introduction .................................................................................................................. 32 2.1.1 Theory of Mind and Empathy in WEIRD Cultures ........................................................... 34 2.1.2 Assessing Thoughts about Minds: False Beliefs and Empathy Quotients ......................... 35 2.1.3 Mentalizing Across Cultures: Ethno-Psychologies and Social Selves .............................. 38 2.1.4 Opacity of Mind ................................................................................................................. 41 2.2 Overview of Studies ..................................................................................................... 44 2.2.1 Why Fiji? Indigenous Yasawans and Indo-Fijians ............................................................ 45 188.8.131.52 Yasawa Island ............................................................................................................. 46 184.108.40.206 Lovu Village, Viti Levu ............................................................................................. 47 2.3 Study 1: False Belief When Others’ Minds are Private ............................................... 48 2.3.1 Method ............................................................................................................................... 48 220.127.116.11 Participants ................................................................................................................. 49 18.104.22.168 Materials ..................................................................................................................... 49 2.3.2 Results ................................................................................................................................ 51 2.3.3 Discussion .......................................................................................................................... 54 2.4 Study 2: Measuring Empathy Across Cultures ............................................................ 54 2.4.1 Method ............................................................................................................................... 55 22.214.171.124 Participants ................................................................................................................. 55 126.96.36.199 Materials ..................................................................................................................... 56 188.8.131.52 Procedure .................................................................................................................... 57 2.4.2 Results ................................................................................................................................ 57 184.108.40.206 Cross-Cultural Measurement Model of EQ Answers ................................................. 59 220.127.116.11 Group-Level EQ Differences ..................................................................................... 67 vii 2.4.3 Linking False Belief and Empathy .................................................................................... 69 2.4.4 Discussion .......................................................................................................................... 71 2.5 General Discussion ...................................................................................................... 73 2.5.1 Implications and Future Directions .................................................................................... 74 Chapter 3. Weighing Outcome vs. Intent in Moral Matters: Priming Thoughts Makes Opaque Minds more Transparent ............................................................................................. 76 3.1 Introduction .................................................................................................................. 76 3.1.1 Moralizing as Applied Mental State Reasoning ................................................................ 78 3.1.2 Moralizing and Mentalizing in the Field: Research Sites in Fiji ....................................... 81 18.104.22.168 Field Sites: Yasawa and Lovu Village, Fiji ................................................................ 82 3.1.3 Overview of Studies ........................................................................................................... 86 3.2 Study 1: Cross-cultural Differences in Moral Judgment Intent/ Outcome Focus ........ 87 3.2.1 Method ............................................................................................................................... 87 22.214.171.124 Participants ................................................................................................................. 88 126.96.36.199 Materials ..................................................................................................................... 89 188.8.131.52 Procedure .................................................................................................................... 91 3.2.2 Results ................................................................................................................................ 94 184.108.40.206 Badness and Punishment Across Intent Conditions ................................................... 95 220.127.116.11 Comparing Intent vs. Outcome .................................................................................. 97 18.104.22.168 Predicting Intent Perception in Yasawa ................................................................... 100 3.2.3 Discussion ........................................................................................................................ 100 3.3 Study 2: Priming Intent .............................................................................................. 102 3.3.1 Method ............................................................................................................................. 102 22.214.171.124 Participants ............................................................................................................... 103 126.96.36.199 Materials ................................................................................................................... 103 viii 188.8.131.52 Procedure .................................................................................................................. 106 3.3.2 Results .............................................................................................................................. 107 184.108.40.206 Priming Effects on Badness and Punishment Across Intent Conditions .................. 108 220.127.116.11 Priming Effects on Intent vs. Outcome .................................................................... 111 3.3.3 Discussion ........................................................................................................................ 114 3.4 General Discussion .................................................................................................... 115 3.5 Conclusion ................................................................................................................. 119 Chapter 4. Learning to See (or Unsee) Mind: Culture Modulates Intent vs. Outcome Focus Across Development ....................................................................................................... 120 4.1 Introduction ................................................................................................................ 120 4.1.1 Why Intent? Mental State Reasoning and Moral Decision Making ................................ 123 4.1.2 Development of Intent vs. Outcome Focus in Moral Reasoning ..................................... 125 4.1.3 Cross-Cultural Differences in Mental State Reasoning: Opacity of Mind ...................... 126 4.1.4 Yasawa, Fiji ..................................................................................................................... 127 4.1.5 Assumptions and Predictions ........................................................................................... 128 4.2 Method ....................................................................................................................... 130 4.2.1 Participants ....................................................................................................................... 130 18.104.22.168 Age Groups ............................................................................................................... 130 4.2.2 Procedure ......................................................................................................................... 131 22.214.171.124 Yasawa ..................................................................................................................... 132 126.96.36.199 North America .......................................................................................................... 133 4.2.3 Stimuli .............................................................................................................................. 134 188.8.131.52 Puppet Shows ........................................................................................................... 134 184.108.40.206 Test Conditions ......................................................................................................... 138 220.127.116.11 Control Conditions ................................................................................................... 140 ix 4.3 Results ........................................................................................................................ 147 4.3.1 Analytical Approach ........................................................................................................ 147 4.3.2 Combined Analysis: Intent vs. Outcome Focus ............................................................... 147 18.104.22.168 Combined Test Conditions: When do Participants Choose Positive Intent? ........... 147 22.214.171.124 Combined Test Conditions: When do Participants Choose Positive Outcome? ...... 152 4.3.3 Targeted Analyses of Specific Intent/ Outcome Pairings ................................................ 156 126.96.36.199 Same Outcome, Different Intent .............................................................................. 157 188.8.131.52 Same Intent, Different Outcome .............................................................................. 162 184.108.40.206 Failed Attempts: Preference for Positive Intent Despite Negative Outcome? ......... 166 4.3.4 Control Conditions: Neutral Intentions and Targets of Action ........................................ 169 4.4 Discussion .................................................................................................................. 171 4.5 Conclusion ................................................................................................................. 176 Chapter 5. Conclusion: How Culture Interacts with Cognition to Shape Mind .............. 177 5.1 Psychological Universals, Core Cognition, and Enculturated Social Beings ............ 182 5.2 Where Do We Go From Here? .................................................................................. 185 5.3 What Does Learning About Yasawa Teach Us About Minds? ................................. 193 References .................................................................................................................................. 197 Appendix A Ch. 2 Cross-cultural Mental State Reasoning Supplement ............................. 229 A.1 ANOVA EQ Factor Means Predicting Curse of Knowledge ................................ 229 A.2 CFA Covariance Matrices ..................................................................................... 230 A.3 Releasing Intercepts Constraints for Partial Scalar Invariance ............................. 232 A.4 EQ Factor Means by Culture and Sex ................................................................... 232 x Appendix B Ch. 3 Weighing Outcome Vs. Intent Across Cultures Supplement ................ 234 B.1 Study 1 Full Multilevel Regression Tables ........................................................... 234 B.2 Individual-level Predictors of Intent Ratings in Yasawa ....................................... 238 B.3 Study 2 Full Multilevel Regression Tables ........................................................... 239 xi List of Tables Table 2.1 Least Squares Mean ratings of the likelihood Juan will look in each box across samples and knowledge conditions. 95% CI calculated using profile likelihood estimation in brackets. ✔✔ : box that should get highest rating if participants correctly understand false beliefs. ✔ : box that should be rated higher ratings if participants are partially passing false belief and getting distracted by the old location. ✖ : box that should be rated higher if participants are distracted by new information. .................................................................... 52 Table 2.2 EQ-short 3-factor EFA with oblimin rotation on data across all cultures shows factors for understanding others’ internal states, interest in caring for others, and ability to navigate social situations. Items are retained in factors if their loadings absolute values are ≥ |0.27|. We also report additional cross-loadings with item 1 in Factor 2 and items 21 and 1 in Factor 2, as suggested in subsequent CFA analysis. ............................................................. 61 Table 2.3 Significant variance and mean differences across samples show pooling is not appropriate. ........................................................................................................................... 64 Table 2.4 We use the loading structure found in our EFA on all data to fit CFA models 1 and 2 to Mturk sample data. Chi-square exact fit reported from DWLS and Model 2 adds a residual correlation cross loadings between item 1 on Factor 2, items 1 and 21 on Factor 3, and a residual correlation between items 3 and 12. We compare models using Satorra & Bentler (2010) chi-square adjustments. Our configural model on all four samples provides an approximate but not exact fit; we find evidence for configural and metric invariance, but must relax constraints on intercepts of items 1, 21, and 22 to find partial scalar invariance. This is sufficient evidence to compare samples on factor means. We find that the error variance on the residuals does not support strict invariance. ................................................ 66 xii Table 2.5 Simple slopes of EQ factors predicting box ratings for Box x EQ factors interactions. Internal mental states items predict false belief as expected only when ignoring overall sample differences. Within Yasawa, internal mental states factor predicts lower ratings of correct false belief box. Across samples, the EQ social situations factor predicts lower ratings of the wrong boxes. ................................................................................................... 70 Table 3.1 Intent/ Outcome Matrix for Intent conditions. Endorsements of stronger punishments against failed attempts indicate intent focus; stronger punishments of accidental violations indicate outcome focus. ........................................................................................................ 89 Table 3.2 Number of repeated observations and unique participants in study of baseline cross-cultural differences in intent/ outcome focus. While all participants were asked to judge up to 4 vignettes each, not all participants completed all 4 vignettes. Due to the continued relationships inherent to working with the Yasawan communities in this sample, Yasawan participants were the only group with the chance to participate in multiple years. 28 of the 54 Yasawan participants who answered failed attempts/ no violation conditions in 2013 also answered intent/ accident conditions in 2012, so that the total number of unique Yasawan participants is 151. ................................................................................................................ 94 xiii Table 3.3 Unstandardized regression coefficients with 95% CI calculated by profile likelihood estimation in brackets. Model 1: Interactions between intent and outcome for good/ bad ratings and reward/ punish ratings across cultures. Larger values of the difference in interactions between good/ bad vs. reward/ punish indicate that the effect of outcome depends more on intent in good/ bad ratings than reward/ punish ratings. Model 2: differences between Accidents (positive intent with negative outcome) vs. Failed Attempts (negative intent with positive outcome) indicate relative importance of intent or outcome in judgments. Lower values indicate negative intent (failed attempt) is considered worse and punished more strongly than negative outcome. ................................................................... 98 Table 3.4 Unstandardized regression coefficients with 95% CI calculated by profile likelihood estimation in brackets. Model 1: Interactions between intent and outcome for good/ bad ratings and reward/ punish ratings across cultures. Model 2: differences between Accidents (positive intent with negative outcome) vs. Failed Attempts (negative intent with positive outcome) indicate relative importance of intent or outcome in judgments. Priming effects were strongest for intent in Yasawa. ................................................................................... 112 Table 4.1 Age distribution across age groups ............................................................................. 131 Table 4.2 Overview of Puppet Show stimuli for Yasawan participants in 2013. ....................... 144 Table 4.3 Overview of puppet show stimuli for Yasawan participants in 2014. All scenarios feature protagonist puppet attempting to open a box. ......................................................... 145 Table 4.4 Overview of puppet show stimuli for North American participants. All scenarios feature protagonist puppet attempting to open a box. ......................................................... 146 xiv Table 4.5 Culture x Age Group sample sizes for individual observations of puppet choices and unique participants who made repeated choices in shows that allowed puppets to be distinguished based on intent. The sample size of the subset of individual puppet choices that included choosing the puppet associated with the positive outcome are reported in parentheses next to the repeated choice observation sample sizes. The total number of unique participants is reported underneath the choice observations samples sizes. ........... 148 Table 4.6 Odds ratios with 95% CI calculated using cluster robust adjusted standard errors. Average odds of choosing positive intent across cultures and age groups. ........................ 150 Table 4.7 Culture x Age Group report of sample sizes for individual observations of puppet choices and unique participants who made repeated choices in shows that made it possible to distinguish puppets based on outcome. The sample size of the subset of individual puppet choices that included choosing the puppet associated with the positive intent are reported in parentheses next to the repeated choice observation sample sizes. The total number of unique participants is reported underneath the choice observations samples sizes. ........... 153 Table 4.8 Odds ratios with 95% CI calculated using cluster robust adjusted standard errors. Average odds of choosing positive outcome across cultures and age groups. ................... 154 Table 4.9 Raw counts of participants who chose positive intent when outcomes are held constant and total sample size from each age group across cultures. ............................................... 158 Table 4.10 Odds ratios with 95% CI calculated using profile likelihood and unadjusted standard errors. Average odds of choosing positive intent across cultures and age groups. ............. 159 xv Table 4.11 Raw counts of participants who chose positive outcome when intentions are the same. This is the only condition that showed a potentially significant difference between the two sub-samples of North American adults, so this table reports their raw counts separately. Total participants who chose positive outcome across both conditions are reported in the Total column. Uni. = university sample; Com. = community sample. ............................... 163 Table 4.12 Odds ratios with 95% CI calculated using profile likelihood and unadjusted standard errors. Average odds of choosing positive outcome across cultures and age groups. Uni. = university sample; Com. = community sample. ................................................................. 164 Table 4.13 Culture x Age Group report of sample sizes for puppet choices and unique participants in shows that present puppets producing outcomes that mismatched their intentions. The numbers of participants who chose positive intent are reported in parentheses next to the repeated choice observation sample sizes; total number of unique participants underneath the choice sample sizes. Average odds of choosing positive intent for each culture’s age groups reported with 95% CI calculated using cluster robust standard errors. .................................................................................................................................. 167 Table 4.14 Raw counts of participants who chose coordinated actor, target of help, or target of coordinated action out of total observations and unique participants. ................................ 169 xvi Table 4.15 Average Odds of participants from age groups of both cultures choosing the coordinated actor in the coordinated action/ stomper (neutral intent) show, target of help in the target of helping or hindering show, or target of coordinated action in target of neutral intent show. The neutral intent/ negative outcome and target of helping or hindering shows had repeated measurements of a few participants, so 95% CI (presented in brackets) were calculated using cluster robust standard errors; 95% CI shown in brackets for the target of neutral intent show were calculated using profile likelihood. ............................................ 171 Table A.1 Type III ANOVA Table of type 3 with Satterthwaite df approximation for curse of knowledge box ratings by condition and sample, controlling for EQ factors scores and sex. 229 Table A.2 Covariance Matrix and mean values of EQ items for North American adult Mturk sample (n=207) used to fit CFA models 1 & 2 ................................................................... 230 Table A.3 Covariance Matrix and mean values of EQ items for North American university student sample (n= 205) ...................................................................................................... 230 Table A.4 Covariance Matrix and mean values of EQ items for Indo-Fijian sample (n=89) ..... 231 Table A.5 Covariance Matrix and mean values of EQ items for Yasawan sample (n=60) ........ 231 Table A.6 Change in CFA model fit when releasing multiple group equality constraints in multiple group invariance analysis. Releasing constraints on EQ items 2, 21, and 22 allows for partial scalar invariance. ................................................................................................ 232 Table B.1 Full multilevel regression study 1 separate DVs good/ bad & reward/ punish. Profile Likelihood CI. ..................................................................................................................... 234 xvii Table A B.2 Study 1 stacked DVs Good/ Bad & Reward/ Punish, simple fixed effects and random effects only. Interaction terms in Table B.3 & B.4. Profile Likelihood CI. .......... 235 Table B.3 Study 1 stacked Good/ Bad & Reward/ Punish Model 1 Interaction terms. Profile Likelihood CI. ..................................................................................................................... 236 Table B.4 Study 1 stacked Good/ Bad & Reward/ Punish Model 2 Interaction terms. Profile Likelihood CI. ..................................................................................................................... 237 Table B.5 Individual predictors of ratings of study 1 accident/ on purpose question among Yasawan participants. Older participants rate violations as more intentional and participants living in the village longer rate violations as more accidental (but only marginally). ....... 238 Table B.6 Full multilevel regression study 2 separate DVs good/ bad & reward/ punish. Profile Likelihood CI ...................................................................................................................... 239 Table B.7 Study 2 stacked DVs Good/ Bad & Reward/ Punish, simple fixed effects and random effects only. Interaction terms in Table B.8 & B.9. Profile Likelihood CI. ....................... 240 Table B.8 Study 2 stacked Good/ Bad & Reward/ Punish Model 1 Interaction terms. Profile Likelihood CI. ..................................................................................................................... 241 Table B.9 Study 2 stacked Good/ Bad & Reward/ Punish Model 2 Interaction terms. Profile Likelihood CI. ..................................................................................................................... 242 xviii List of Figures Figure 2.1. Map of Fijian Archipelago showing locations of Yasawa Island and Lovu Village. . 46 Figure 2.2. Diagram of Curse of Knowledge task. A) Character 1, Juan, places an item in the blue box. B) Character 2, Miguel, moves the item to a new box, then re-arranges the 4 boxes. .................................................................................................................................... 50 Figure 2.3 Least Squares Mean ratings of the likelihood Juan will look in each box across samples and knowledge conditions. 95% CI Error bars. Higher blue bars indicate students rate Juan as more likely to look in that box than Yasawans. ✔✔ : box that the ball was originally placed in. If participants correctly understand false beliefs, they should rate this box as the highest likelihood of being looked in. ✔ : Plausible Box; the ball was not in this box when Juan left, but when he returns it is now in the location in the room where the first box with the ball was. As the original location but not the original container, participants should rate this higher if they are partially passing false belief by keeping track of the old location in the room. ✖ : the Implausible Box; the ball was not in this box when Juan left, and it is in a different location in the room than the original box. Participants should rate this box that as more likely to be looked in if they are thinking less about Juan’s false beliefs and are getting more distracted by new information. ................................................ 53 Figure 2.4 Group differences with SE error bars showing scores on all 22 EQ items A) without a sample x sex interaction and B) with the interaction. Though there is no significant overall difference in sex effect across samples, allowing for this difference in sex effects highlights the difference in size of the sex differences across samples. Error bars show standard errors................................................................................................................................................ 68 xix Figure 2.5 Mean scores on EQ factors accounting for sex differences across samples. A) Yasawans score the lowest on factor 1 (ability to interpret others’ internal mental states) while B) there are no group differences in factor 2 (connecting with others emotions) and C) Indo-Fijians score lowest on factor 3 (navigating social situations). Error bars show standard errors. ...................................................................................................................... 69 Figure 3.1 Map of Fijian Archipelago showing locations of Yasawa Island and Lovu Village. .. 83 Figure 3.2 Marginal mean ratings of A) how good or bad and B) how worthy of reward or punishment the norm violation was. Error bars indicate standard errors. ............................ 96 Figure 3.3 Marginal mean ratings of A/B) how good or bad and C/D) how worthy of reward or punishment the norm violation was by cultural group and prime. Error bars indicate standard errors. For both DVs, the prime had the largest effect in Yasawans’ ratings of failed attempts. .................................................................................................................... 110 Figure 4.1 1892 copy of Ernst Haeckel's embryo drawings. The deep similarity observed among very young embryos across diverse species has long been taken as evidence for shared ancestry. .............................................................................................................................. 121 Figure 4.2 2013 Yasawa Puppet shows: A) Ball Show, B) Box Show, C) Failed Attempt ....... 136 Figure 4.3 2014 Yasawa Puppet shows: A) Red vs. Yellow, B) Orange vs. Green, C) Teal vs. Purple, D) Blue vs. Green ................................................................................................... 137 Figure 4.4 Log Odds of choosing positive intent across successful attempts, failed attempts, and same outcome/ different intent shows. Results in log odds to preserve linear scale across age and cultural groups. Error bars show robust standard errors corrected for clustered data around multiple observations of individual participants. .................................................... 149 xx Figure 4.5 Log Odds of choosing positive outcome across successful attempts, failed attempts, and same intent/ different outcome shows. Results presented in log odds to preserve linear scale across age group and cultural comparisons. Error bars show cluster robust standard errors corrected for multiple observations. ......................................................................... 154 Figure 4.6 Average log odds of choosing positive intent when outcomes are the same for positive and negative outcomes across age ranges and cultures. Error bars show standard errors. For the positive outcome condition, adolescents in both cultural groups selected the positive intent every time. This produced undefined estimates for this age group in this condition that are therefore not graphed here. .................................................................................... 159 Figure 4.7 Average log odds of choosing positive outcome when intentions are the same for positive and negative intentions across age ranges and cultures. Error bars show standard errors. North American university and community adults were marginally significantly different, so we separate them here. .................................................................................... 164 Figure 4.8 Average log odds of choosing positive intent between failed helpers and failed hinderers. Error bars show cluster robust standard errors. .................................................. 168 Figure A.1 Mean scores on EQ factors for men and women across samples. A) Yasawans score the lowest on factor 1 (ability to interpret others’ internal mental states) while B) there are no group differences in factor 2 (connecting with others emotions) and C) Indo-Fijians score lowest on factor 3 (navigating social situations). Error bars show standard errors. .. 232 xxi Acknowledgements I am deeply indebted to the people of Yasawa, Fiji, for welcoming me into their community, as well as warmly and enthusiastically participating in all of my studies. I am especially grateful for the ongoing research support from Paula Tekei, whose tireless efforts as project manager and infinite enthusiasm for this research should be enough to grant him an honorary doctorate in his own right. I am further honoured to have had the chance to work with other Fijian research assistants across my seasons of fieldwork, with particular gratitude to Melaia Tikoitoga, Joji Savou, Siteri Kalouniuca, and Malakai Waqairadovu. I also wish to express my sincerest thanks to Sereana Naepi for her advice about how to respectfully approach research with Indigenous groups in general and Fijian communities in particular. I am especially grateful to Sereana for her patience in dealing with my ignorance to help me be a better researcher and her continued advice on how to build toward a fully collaborative and respectful research program that incorporates Indigenous and institutional perspectives. I wish to thank my research supervisors Joseph Henrich and Ara Norenzayan for the support and direction throughout this research process, and especially to Joe for giving me the space and encouragement to grow into a competent field researcher. I thank J. Kiley Hamlin for her continued support in bringing developmental perspectives into my research and in providing a guiding light on navigating the processes of building an academic career. I am most humbly grateful to Benjamin G. Purzycki for his generosity with his time and attention in reading papers, listening to presentations, bouncing ideas, and generally helping me grow into a well-rounded researcher. I am similarly grateful for the research mentorship and general fieldwork advice so kindly, freely, and generously given to me by Michelle Kline, Mathew Gervais, and Annie Wertz. I extend my humblest gratitude to Jeremy Biesanz and Victoria Savalei for their xxii instruction in statistical methods that make the current analyses possible. I also extend my gratitude to John Shaver for his continued, thorough, and sound advice toward deepening my ethnographic grasp of Fiji. I also thank Pascal Boyer and Karen Wynn for their mentorship through my undergraduate to postgraduate transition and for taking the chances on me that made it possible for me to set off on this deeply interdisciplinary path. And beyond the academics, I extend my gratitude to my family and friends. Though my mother Julie McNamara, father Mark McNamara, and sister Sara McNamara may not be familiar with the theories I babble on about or the statistical analyses I’m trying to get my mind around that day, their unwavering support is the only reason I’ve been able to get to the point of even writing this document. I also thank my aunt Anne McNamara for being the trailblazing woman in science that helped spark my love of research and dogged determination to break into the academy from our rural Illinois roots. I am further indebted to Sarah Topps for her continued presence as a rock of friendship that helped me weather countless storms in the course of my learning and research here at UBC. I also thank my fellow residents living at UBC’s Green College for their continued presence as a sympathetic ear as we all muddle through the stresses, trials, tribulations, and triumphs of learning how to be academics. Thank you also to the research assistants and fellow students in the psychology department who further helped me in growing from my state of unaware ignorance when I first set foot on UBC campus to my state of more knowledgeable ignorance today. Your kindness, dedication, and commiseration have been a vital part of my journey through this degree. 1 Chapter 1. Introduction: Mentalizing as a Core Foundation of Culture Our mentalizing abilities, or the abilities to infer the presence and contents of other minds, are among the key defining characteristics that make us human. These abilities likely evolved as a part of a broad suite of specialized cognitive mechanisms that facilitate learning from each other, competing or cooperating with each other, and generally living with each other in extended, complex social groups (Adolphis, 2009; Byrne & Whiten, 1989; Dunbar, 2009). The broad functions that these mentalizing processes carry out for us are often categorized under two sub-types of mentalizing processes: theory of mind and empathy. Theory of mind processing typically refers to inferences about others’ knowledge states, goals, and especially true or false beliefs that may differ from one’s own (Brüne & Brüne-Cohrs, 2006; Chris D Frith & Frith, 2006; Leslie, 2001; Premack & Woodruff, 1978). Empathy, on the other hand, typically refers to the emotional components that we use to orient toward shared emotional and affective experience. Empathic processing can be particularly important for encoding and decoding non-verbal communicative displays. These displays of emotional states can convey particular social meanings (e.g.: anger as cue to stop a behaviour, shaming to prevent someone from stepping out of their appropriate place in a social group, or sadness to elicit help and show submission after making a social faux pas; Ali, Amorim, & Chamorro-Premuzic, 2009; R. J. R. Blair, 2002; Cote & Hideg, 2011; Montgomery, Stoesz, & McCrimmon, 2013; Salovey & Mayer, 1990). In the following review, I seek to further elaborate how individual and group-level influences interact to shape these mentalizing processes. I pay special attention to ways that cultural variation may modulate mentalizing processes throughout. I first review dominant theoretic perspectives on mentalizing, then how these processes emerge across human evolution and development. I next turn to a more general discussion of why humans should metalize at all, 2 and specifically discuss what benefits mentalizing may confer in various social challenges inherent to living in groups. These social challenges may be adaptive pressures that helped promote mentalizing as the solution we are so prone to employing today. Finally, I focus on the interdependence between mentalizing and cultural forms like morality, religion, and cumulative cultural learning to map out ways that mentalizing enables and perpetuates cultural forms over time. 1.1 Theory, Evolution, and Development of Mentalizing Several dominant perspectives have emerged to explain how mentalizing processes are acquired in development, embodied in neural systems, and selected throughout evolutionary history. Theory theory posits that people develop a series of tacit theories, akin to scientific theories, to explain how minds work (Gopnik & Wellman, 1992). These theories remain tacit because people do not effortfully build or evoke them in the way of a true scientific theory. Because they are built on experience instead of explicit scientific experimentation, these tacit theories of mind are sometimes referred to as folk or naïve theories. To use these tacit theories, people intuitively and reflexively evoke them as needed to explain behaviour through the lens of intentionality and other psychological states (Mull & Evans, 2010; Wellman & Miller, 2006; Wellman, Cross, & Watson, 2001). Theory theory is often contrasted to simulation theory, which posits that people simulate in their minds how they would react in a given situation and use this simulation to impute the other person’s mental states (Goldman, 1989; R. M. Gordon, 1986). Importantly, simulation theory bypasses the need for people to develop an elaborate conceptual framework for minds (as is necessary in theory theory). Instead, people can merely evoke the neural mechanisms that would simulate the same mental states in themselves, reflect on what these states are, and use this simulated mental state information to infer what the other person 3 must be thinking or feeling. Mentalizing as simulation is thought to principally rely upon mirror neuron systems which engage to create the same neural activation patterns in observer and observed (Gallese & Goldman, 1998). A third key perspective posits that mentalizing, and theory of mind processing in particular, is built upon a series of evolved cognitive modules (Baron-Cohen, 1995a; Leslie, Friedman, & German, 2004). These modules are purpose-built cognitive mechanisms that take in particular kinds of inputs and analyze them toward particular outputs (H. C. Barrett & Kurzban, 2006; Frankenhuis & Ploeger, 2007; Tooby & Cosmides, 1992). For example, intentionality detection1 may form one of the most basic modules underlying mentalizing processing (Baron-Cohen, 1995a). Agent-oriented cognition differs from reasoning about physical or biological properties of entities in that the ends of an action sequence are referenced to justify the rationale behind the preceding actions. Instead of looking at each step as the cause of the next (for example: the ball rolled forward because of inertia, the box blocked the ball’s path so the ball bounced back toward its original location) the end of the action sequence is the reason for the prior actions (the ball wanted to get to the other side of the box, so it jumped over the box to get there). These basic agency detection modules also underlie the tendency to consider minds and bodies as separate entities under naïve mind-body dualism (Bloom, 2004). Additional, very basic cognitive modules that regulate eye gaze detection combine with agency/ intent detection to build into shared (Baron-Cohen, 1995b) or joint attention (Carpenter, Nagell, Tomasello, Butterworth, & Moore, 1998; Tomasello & Farrar, 1986) processes that allow for collaborative 1 Intentionality Detection has elsewhere referred to as agency detection (J. L. Barrett & Johnson, 2003; Iizuka & Di Paolo, 2007; Kumar & Srinivasan, 2012; theory of bodies: Leslie, 1994), and the closely related perception of purpose, functions, or goals under teleological thinking (Csibra & Gergely, 1998; 2011b; Gergely & Csibra, 2003; Kelemen & Rosset, 2009) 4 action and learning. A further theory of mind module covers the ability to understand that agents can act with beliefs or knowledge states that are based upon false perceptions of the world (Baron-Cohen, 1995a). Under this modular perspective on mentalizing abilities, the tendency to see social actions through the lens of internal mental states derives from the joint action of simpler cognitive mechanisms that evolved for specific aspects of the mentalizing process. Surely there is some truth to all three theoretical orientations; the neural activation patterns seen when mentalizing processes are measured in brain scans, plus the consistent pattern of development across many cultural backgrounds, do hint at possible neural and basic cognitive architectural features that underlie mentalizing processes (Apperly, 2008; Carruthers, 2008; Saxe, 2005). Tacit theories of minds as are evident early in development likely form part of core cognition similar to folk biology and folk physics (Carey & Spelke, 1994; Spelke & Kinzler, 2007). The more basic, encapsulated, and modular cognitive computational aspects of mentalizing may combine with lived experience to update the core theories with learned statistical regularities in one’s social environment that provide for more flexible mentalizing than would be afforded by strictly modular cognition (Carey, 2011; Carey & Spelke, 1994; Gopnik & Wellman, 2012; Machery, 2008; Perreault, Moya, & Boyd, 2012). While effortful reasoning about others’ goals and intentions might be driven more by one’s learned theories about how minds work, the emotional contagion aspects of empathic processing may be especially reliant upon mirror neurons and simulation, making the observer literally feel the same feelings as the observed (Baird, Scheffer, & Wilson, 2011; Rockwell, 2008). However, little of the above research and theorizing has fully considered the additional layering of culturally-transmitted norms that might be giving further structure to folk psychologies of various cultural groups (Luhrmann, Padmavati, Tharoor, & Osei, 2015). These 5 norms provide a group-level structure to social interactions that further stabilize behaviour into predictable patterns. These reliable behavioural patterns can then allow individuals within those groups to develop appropriate schemas and scripts that lead people to infer roughly similar folk psychologies within the same cultural group. If we restrict our research to just one cultural group, we run the risk of mistakenly concluding that the folk psychology of that culture is the only possible folk psychology and therefore a basic design feature of the system. 1.1.1 Evolutionary History: Mentalizing in Non-Human Primates Much of the work on how human mentalizing capacities evolved comes from comparative studies focusing especially on our closest living non-human relatives: chimpanzees (Pan troglodytes) and bonobos (Pan paniscus; both species are hereafter referred to as chimps). Chimps do clearly show some aspects of mentalizing abilities seen in humans, though they do not appear to have the full suite of mentalizing abilities we have (Emery & Clayton, 2009; Povinelli & Giambrone, 2001; Tomasello, Call, & Hare, 2003). In particular, chimps seem to understand that others have different perspectives and goals than their own, but they do not seem to interpret the possibility of others having wrong information or false beliefs (Call & Tomasello, 2008). Chimps appear to use mentalizing processing more toward competitive than collaborative or cooperative ends. For example, chimps can tell when another does not see a desired item (i.e. a food reward) and can use deception to keep a competitor from finding the desired food item (Hare, Call, & Tomasello, 2006; Hare, Call, Agnetta, & Tomasello, 2000). Young chimps raised around humans respond to human cues soliciting help (Warneken, 2006), but they do not appear to respond to the desire states of other chimps to choose actions that will be mutually beneficial (Brosnan, Henrich, Mareno, Lambeth, & Schapiro, 2009). Importantly, chimps can and do learn socially by watching and emulating each others’ actions to develop cultural differences across 6 chimp troops (Shipton & Nielsen, 2015; Whiten, Horner, & de Waal, 2005; Whiten, McGuigan, Marshall-Pescini, & Hopper, 2009). Importantly, though, this social learning through copying among chimps appears more as direct replication of action sequences without reference to the model’s goal states. Because chimps appear to prefer to reach their own desired end of a given action sequence (emulate) rather than copy all elements of the sequence a model shows based upon some concept of what the model’s goals are (imitation and over-imitation), chimps have less of a tendency to keep new socially learned tricks that can be built upon by further innovation in future generations (Legare & Nielsen, 2015). Joint attention is one particularly important element central to the collaborative learning we see in humans that develops early in human infancy but is not clearly present among chimps; chimps will orient their gaze toward what others are attending to, but they do not seem to refer back to the model to jointly gain new information about the object they both are attending to (Carpenter et al., 1998; Miklósi & Soproni, 2005; Okamoto-Barth & Tomonaga, 2006). This collaborative social intelligence appears to be the cognitive specialization that sets human mentalizing abilities apart from other apes (Andrews, 2001; Brüne & Brüne-Cohrs, 2006; E. Herrmann, Call, Hernàndez-Lloreda, Hare, & Tomasello, 2007; Povinelli & Preuss, 1995). 1.1.2 Mentalizing Development and Disorders The complexity of human mentalizing processes becomes most evident when we consider the ways these processes emerge across development and how they may go awry in various disorders. The majority of this research comes from Western samples, though some scant evidence from more collectivistic cultures points to ways that social environments further shape mentalizing. 7 220.127.116.11 General Patterns in Timing and Sequence of Mentalizing Development Infants as young as two days old show very basic agency detection by distinguishing biological from non-biological movement (Simion, Regolin, & Bulf, 2008). At around the same time, days-old infants also show basic elements of empathy and emotional contagion by crying at the sound of other infants’ crying (Sagi & Hoffman, 1976). This shared distress, expressed as crying and facial displays of anger and sadness at the sound of another infant crying in pain, continues on well into the first year of life (Geangu, Benga, Stahl, & Striano, 2010). Days-old infants also reliably copy adults’ facial expressions and gestures (Meltzoff & Moore, 1977; 1983). Infants as young as 3-months-old begin to make social judgments about helpful vs. unhelpful characters (Hamlin & Wynn, 2011; Hamlin, Wynn, & Bloom, 2007). By around eight months, infants selectively prefer characters who were prosocial to other prosocial characters and antisocial to other antisocial characters (Hamlin, Wynn, & Bloom, 2011) and begin to privilege intent over outcome (Hamlin, 2013). By nine months, infants begin to interpret agentic actions in terms of teleology or goals, though they may not be making conceptual references to agent’s internal beliefs or desires yet (Csibra & Gergely, 1998). By 12 months, infants hit a major milestone when they begin to reliably and flexibly orient their attention toward what other agents are attending to (even when agency cues are minimal, see: S. C. Johnson, 2003). These joint attention episodes gradually progress from being aware of a shared object of attention, to following another’s gaze, to actively directing others’ attention, often for the purposes of explicitly learning about the object of shared attention (Carpenter et al., 1998; Tomasello & Farrar, 1986). By 15 months, infants begin to pass non-verbal false belief tests (Onishi & Baillargeon, 2005) and begin to integrate mental states like desires and intentions into their understanding of how people’s emotions and goals are causally 8 related to one another (Saxe, Carey, & Kanwisher, 2004). Though infants show affective empathy in the form of shared distress very early in life, the ability to translate empathy into attempts to help only starts to emerge after many infants’ first birthday (Roth-Hanania, Davidov, & Zahn-Waxler, 2011). Though the earliest attempts to help are strongest when the distressed person is the infant’s mother, infants do start to help even unfamiliar adults based upon emotional cues by their second birthday (Svetlova, Nichols, & Brown, 2010; Zahn-Waxler, Radke-Yarrow, Wagner, & Chapman, 1992). By around 1.5 to 2 years, infants and toddlers begin to distinguish real vs. imagined scenarios to engage in pretend play (Leslie, 1987). Two-year-olds also show signs of understanding actions in terms of desires (i.e. Sally wants the ball), and the social cognitive precursors to understanding beliefs may present in infancy (Onishi & Baillargeon, 2005; Surian, Caldi, & Sperber, 2007), but the full cognitive suite of cognitive faculties needed to explicitly distinguish one’s own belief from another’s false belief about the world appear not to be fully present until around children’s fourth birthday (Wellman & Woolley, 1990). The basic cognitive mechanisms that begin to emerge in early infancy and suggest modularity may provide a different system of mentalizing processing than the later-developing system that handles explicit reasoning about beliefs (Saxe, Carey, & Kanwisher, 2004). Part of the developmental delay in linking up these two systems may stem from lower executive control abilities in younger children, such that they can infer a difference in what they know and what a target knows, but they cannot suppress their own knowledge to give the correct answer (hence evidence for 15-month-olds passing visual but not verbal false belief tasks; see: Apperly & Butterfill, 2009; Heyes, 2014; Leslie, German, & Polizzi, 2005; Sabbagh, Xu, Carlson, Moses, & Lee, 2006) 9 This acquisition of belief concepts is a big milestone; toddlers also begin to try to cover up transgressions with basic deception at around 2-years-old, but their ability to maintain the deception does not fully emerge until they can also begin to grasp belief at around 4-years-old (A. D. Evans & Lee, 2013; Talwar & Lee, 2002). By around age 5 or 6, children’s comprehension of false beliefs expands to include the idea that others can have beliefs about other peoples’ beliefs (Wimmer & Perner, 1983). As children become better at these second-order ‘beliefs about beliefs’ tasks, they also become better at explicitly integrating intent into verbal moral judgments and better at verbal deception (Fu, Xiao, Killen, & Lee, 2014; Talwar, Gordon, & Lee, 2007). By around 6-8-years-old, children begin to grasp the disconnect between what is said and what is meant in figurative speech like irony and metaphor (Filippova & Astington, 2008; Lucariello & Mindolovich, 1995; Rundblad & Annaz, 2010). By around ages 9-11, children start to understand when someone makes a social mis-step by saying or doing something socially unacceptable even when the actor does not grasp their mistake (Baron-Cohen, O'Riordan, Stone, Jones, & Plaisted, 1999). This faux pas comprehension milestone is particularly sophisticated in that it involves understanding the mental states of the socially inappropriate actor and the potentially offended recipient of the social faux pas. Moving in to later childhood and adulthood, the ability to further understand and interpret additional layers of meaning and belief gradually emerge, though additional levels beyond first-order beliefs (i.e. understanding that others can have beliefs different to one’s own) and second-order beliefs (i.e. understanding that others can have beliefs about other people’s beliefs) do not appear to be as fast and effortless (Devaine, Hollard, & Daunizeau, 2014; Kinderman, Dunbar, & Bentall, 1998). The non-automaticity of these higher order belief tasks is further elaborated in 10 experiments with adults that show these kinds of social cognitions take extra effort (Back & Apperly, 2010; Van Overwalle & Vandekerckhove, 2013). 18.104.22.168 Individual and Cultural Influences on Mentalizing Development Evidence for modularity and core cognitive features in mentalizing processing suggest that many of these capacities should develop in roughly the same pattern and at the same time regardless of cultural context. Though cross-cultural studies of very young infants and children are relatively rare, there is broad consistency in the timing and sequence of especially false belief understanding across cultures (Callaghan et al., 2005). However, some family and culture-level influences have been shown to influence the timing and pattern of certain mentalizing processes. Children whose parents use more mental state words and who have older siblings tend to develop mentalizing abilities slightly earlier (Carpendale & Lewis, 2004). Language acquisition also parallels and facilitates theory of mind development (Bretherton & Beeghly, 1982; Ruffman, Slade, Rowlandson, Rumsey, & Garnham, 2003). As demonstrated by the example of deaf children and different generations of signers across the evolution of Nicaraguan sign language, the introduction of more nuanced mental state terms lead to significant increases in false belief task performance, even within the same people (Meristo et al., 2007; Pyers & Senghas, 2009). However, there are important differences in some cultural groups, especially small-scale societies around the Pacific and in Central America that normatively suppress explanations of behaviour based upon unseen motivations (a.k.a. Opacity of Mind norms: Duranti, 2015; Robbins & Rumsey, 2008; A. Rumsey & Robbins, 2008) that may lead to slight delays in passing false belief tasks (H. C. Barrett, Broesch, Scott, He, Baillargeon, Di Wu, et al., 2013a; A. Mayer, 2013; Wassmann, Trauble, & Funke, 2013). More generally, greater focus on communal rather than individualistic values may also present slight variations in which aspects of 11 mentalizing processes emerge first (for example, greater communal focus predicts later ages of majority passing false belief: Liu, Wellman, Tardif, & Sabbagh, 2008; communal focus on agreement and avoiding disputes may lead to later understanding that opinions can be highly diverse: Shahaeian, Peterson, Slaughter, & Wellman, 2011) and where they are located in neural systems (Kobayashi Frank & Temple, 2009). More community-oriented cultural groups also tend to emphasize agreement; this focus on getting along and avoiding conflict tends to also coincide with more restrictions on expressing emotions – especially negative emotions – which may also reduce recognition of these displays when they do occur (Fernandez, Carrera, Sanchez, Paez, & Candia, 2000; Matsumoto et al., 2008; Shao, Doucet, & Caruso, 2015). This cultural modulation of especially negative emotional displays also shows up early in childhood (Borke, 1973). Though sympathy with negative emotions may lead to development of prosocial behaviour in Western cultures (Trommsdorff, Friedlmeier, & Mayer, 2007), greater emphasis on relational obligations may provide an alternate route to prosociality in more collectivistic groups (Kärtner, Keller, & Chaudhary, 2010). 22.214.171.124 Mentalizing Deficits in Psychological Disorders Some of our greatest insights into how mentalizing processes work come from cases of psychological disorders where mentalizing goes awry (Brüne & Brüne-Cohrs, 2006). Autism spectrum disorders present marked deficits in the ability to infer others’ beliefs and difficulty with social situations (Baron-Cohen, Leslie, & Frith, 1985; Baron-Cohen, Wheelwright, Skinner, Martin, & Clubley, 2001; U. Frith, Morton, & Leslie, 1991; Wakabayashi, Baron-Cohen, Uchiyama, et al., 2006a). People with autism spectrum disorders may have trouble recognizing emotions, but are still empathically moved by and concerned about others’ emotions (S. B. Cohen, Wheelwright, Hill, Raste, & Plumb, 2001; A. P. Jones, Happé, Gilbert, Burnett, & 12 Viding, 2010). Psychopathy, on the other hand, is often marked by perspective taking abilities that are similar to populations without diagnosed psychological dysfunction (A. P. Jones et al., 2010; Mullins-Nelson, Salekin, & Leistico, 2006). Instead, psychopathy tends to involve marked deficits in being moved by the emotional experiences of others (Ali & Chamorro-Premuzic, 2010; Lishner, Swim, Hong, & Vitacco, 2011; Malterer, Glass, & Newman, 2008). Both autism and psychopathy have associations with amygdala abnormalities (R. Blair, 2008), though psychopathy also has links to specific impairments in the mirror neuron system (Fecteau, Pascual-Leone, & Théoret, 2008). Schizophrenia is another disorder often associated with abnormalities in mentalizing (Corcoran, Mercer, & Frith, 1995; Sprong, Schothorst, Vos, & Hox, 2007), but often as a case of over-mentalizing rather than under-mentalizing as seen in autism and psychopathy (K. Gray, Jenkins, Heberlein, & Wegner, 2011). Schizophrenia is associated with disorganized thinking about both first and second order belief tasks (Mazza, De Risio, Surian, Roncone, & Casacchia, 2001) and in ability to mirror other’s behaviours, but not the tendency to care about others’ emotions (Haker & Rössler, 2009). The mentalizing abnormalities in schizophrenia appear to also be related to irregularities in the mirror neuron system (Mehta et al., 2014) and issues with attention regulation in executive control that are evident even in relatives who do not show clinical signs of the disease (Snitz, Macdonald, & Carter, 2006). 1.2 Why Think About Minds? Mentalizing in Competition and Cooperation As outlined above, the past four decades of research into mentalizing processes paint a broad picture of both typical and atypical patterns in mentalizing development. However, understanding how these cognitive mechanisms emerge tells us precious little about why we might have them in the first place. Mentalizing is a strong tool in our cognitive toolkit that we can deploy to solve the basic cooperative and competitive dilemmas inherent to social life 13 (Baron-Cohen, 1995a). It would be reasonable to hypothesize that mentalizing may have evolved expressly for this purpose. Though mentalizing has some clear advantages in these domains, how often does mentalizing occur in basic cooperative and competitive situations? What other sources might individuals reference in making these social decisions? In this section, I review evidence for mentalizing processes involved in primate social life, mentalizing in coordinated action, and end with the role of empathic and intentionality/ belief processing in cooperation and competition. 1.2.1 Define Behavioural Expectations Through Social Structure As mentioned above, some elements of human social cognitive abilities are also apparent in closely related primate species. These similarities in social cognition are likely due to a common tendency for primates to band together as social groups. However, the cooperation benefits we get from being so gregarious – increased access to mates, help raising young, help finding food, help in defending against predators – are also the competitive costs to living in groups – competition for mates, resources, and potential to attract extra attention from predators. Across primate species, dominance hierarchies are a common and effective solution to dealing with cooperation and competition dilemmas inherent to social life (Cummins, 1996). These dominance hierarchies provide clear, shared expectations about how individuals should behave toward each other based upon rank (Seyfarth, Cheney, & Bergman, 2005). Importantly, however, referencing hierarchical rank to solve social decision-making issues need not entail inference about others’ mental states (beliefs, desires, goals); it merely requires knowledge about their relative rank and the set of privileges and obligations that come with that rank. So, while referencing the social system around an individual can resolve a wide array of ambiguities 14 around how, when, and whom to cooperate or compete with, presence of dominance relationships may actually undercut the need to use mentalizing as much as humans do. 1.2.2 Mentalizing to Coordinate Action Beyond behavioural expectations within established, hierarchical social networks, the ability to coordinate behaviour is, at a more general level, the key prerequisite to both cooperation and competition. More than just repeating another’s actions, behavioural coordination requires the ability to predict the other actor’s behaviour and modify one’s own behaviour to compliment theirs. This may require some degree of mentalizing to anticipate behaviour, though again these behavioural predictions might be based solely upon knowledge of the situational constraints (like place within a hierarchy) rather than inference about internal states like desires or beliefs. Given that situational constraints on social behaviour may be an evolutionarily older solution to coordinated action problems, it is reasonable to expect that humans will preferentially refer to these situational constraints first, and only engage in additional internal mental state inferences when the situational information is insufficient. Observations that more complex mentalizing processes around beliefs are also more effortful (Back & Apperly, 2010; Keysar, Lin, & Barr, 2003) and that many adults do not engage extra mentalizing processing unless specifically motivated to do so (Epley, Waytz, & Akalis, 2008b; Waytz et al., 2010) suggest that mentalizing might not be the first strategy used in initiating a coordinated action sequence. However, neurocognitive evidence does point to certain mentalizing processes that are activated during coordinated action. At its most basic level, the ability to detect contingent motion (a core feature of agency detection: S. C. Johnson, 2003) is a necessary skill needed to evade predators or capture prey. Having some inkling about what the other agent wants (a desire inference) can further refine 15 decisions about how to respond. As discussed in section 126.96.36.199, human infants are able to interpret and respond to cues that signal the target’s desires long before they can reliably respond to questions about beliefs. If thinking about a person’s desires is enough to direct contingent helping behaviours, then children who can represent desires but cannot yet represent beliefs should have little trouble in completing helpful actions. The literature indicates that this is indeed the case; toddlers do help adults based upon displays of thwarted desires (S. I. Hammond, 2014; Warneken, 2006). Joint attention, shared understandings about what is expected during the task at hand, and even shared neural activations of motor pathways associated with mirror neuron systems all help bring individuals into correctly coordinated action sequences (Newman-Norlund, van Schie, van Zuijlen, & Bekkering, 2007; Sebanz, Bekkering, & Knoblich, 2006). Activities that require participants to perform action sequences in synchronous time with each other also boost interpersonal affiliation and heighten cooperation by increasing perspective taking, boosting mutual motor area activation, and decreasing psychological distance to make it easier to infer others’ goals (Baimel, Severson, & Baron, 2015; Wiltermuth & Heath, 2008). Engaging in synchronous action also helps direct participants to specifically focus on the mental states of their synchronization partners (Baimel, 2015). 1.2.3 Empathic Processing and Decisions About Cooperation or Competition Looking beyond coordinated actions, emotional experiences and empathic processes have also been implicated in competitive and cooperative action. Cognitively, shared emotional experiences, especially relating to empathetic negative affect arousal, may cue greater effort toward understanding another person’s behaviour. Conversely, shared experience of extreme emotions or highly arousing stimuli might cue individuals to direct their attention to the same elements of the environment (Nummenmaa et al., 2012). Empathic concern for the emotional 16 states (especially in alleviating the negative emotional states) of others has long been thought to be an important precursor to altruism and cooperation. The relationship between taking on another person’s perspective, appreciating their emotional experiences, and increased cooperation/ altruistic helping gets stronger from middle childhood into adulthood (Eisenberg & Miller, 1987; Marcus, Telleen, & Roke, 1979; Underwood & Moore, 1982). However, the extent to which empathy aids in motivating actual costly helping is less clear. The elements of empathy that cue attention to others’ needs and motivate action might be relevant, but a sense of self-other overlap (rather than emotions per se) may be more important for promoting meaningful help that requires even small sacrifices in time, energy, or resources (Cialdini, Brown, Lewis, Luce, & Neuberg, 1997; Neuberg et al., 1997). Another element of cooperative action that might benefit from empathic processing comes when cooperation falls short for unintended reasons. If the aggrieved party feels an offender’s negative affect around guilt or shame after they accidentally failed to cooperate, then this shared affect may help maintain a cooperative relationship by ameliorating feelings of resentment (Rumble, Van Lange, & Parks, 2010). Despite the oft-emphasized cooperative benefits gained from empathic processing, awareness of and ability to engage with others’ emotions can also be used toward competitive ends. As discussed in section 188.8.131.52, psychopathy is a personality profile marked by reduced empathic processing. More specifically, people who score high on trait psychopathy are able to understand and engage with others emotions, but the emotional experiences others have do not create emotional contagion effects that would modulate behaviour toward less self-maximization or antisocial action. A very similar empathic profile occurs in Machiavellian personalities, the key difference being that Machiavellians tend to also have better impulse control than is common among people with psychopathic personality traits (Paulhus & Williams, 2002). In contrast to 17 autisms spectrum populations, Machiavellians tend to be very socially savvy but less concerned about whether their actions cause negative emotions in others. Machiavellians can use their high social skills and emotional intelligence to manipulate others (Ali et al., 2009; Cote, DeCelles, McCarthy, Van Kleef, & Hideg, 2011; Kilduff, Chiaburu, & Menges, 2010). 1.2.4 Thinking About Intent in Competition In addition to empathic processing, inferences about a competitor’s goals, intentions, and knowledge states are also key components to the competitive strategies people use. Neurocognitive evidence suggests that brain areas engaged in reasoning about others’ intentions are also engaged in competitive interactions (Gallagher, Jack, Roepstorff, & Frith, 2002). Importantly, these intentionality-focused processing areas are most readily engaged when the stakes are high enough to care about how much reward can be gained out of the competitive interaction (Halko, Hlushchuk, Hari, & Schürmann, 2009). This competition-driven intentionality focus can also activate joint attention and shared representation of action sequences typical of coordinated behaviour within largely cooperative interactions discussed above (Ruys & Aarts, 2010). The additional incentive structure in competitive contexts seems to also help children with high-functioning autism reason about belief above and beyond their performance on tasks that are more neutral than competitive (Chang & Cheung, 2016; C. C. Peterson, Slaughter, Peterson, & Premack, 2013). The need to detect deception as a mismatch between a potential interaction partner’s statements and intentions may be a fundamental adaptive pressure that pushed the evolution of mentalizing processes (Andrews, 2001). Along these lines, Machiavellians appear to favour highly competitive interactions because they do readily infer intent but lack empathic concern for their competition partners (Paal & Bereczkei, 2007). 18 1.3 Mentalizing and Culture While mentalizing abilities have the above-mentioned benefits for social interactions between and among individual actors, these individual-level benefits do not adequately explain the full extent of human mentalizing. In particular, the individual-level benefits fall short of explaining why humans in particular appear to be unique in our ability to think not just about what others want, but also about what they intend, what they believe, that their beliefs may be different from our own, and that these beliefs may be true or false. For example, chimps also regularly compete with each other both within and across groups, but they do not appear to have the same sophistication and elaboration in their mentalizing cognitive toolkit. I argue that the additional complexity in human mentalizing processes show their real power in their ability to sustain long-term social groupings through culture. This culture, or the body of ideas, knowledge, values, norms, and beliefs that is shared and transmitted within particular social groups, is the defining feature of human sociality. Mentalizing is at the heart of the ways that we learn the contents of culture as lessons accumulated across generations through cumulative cultural learning. Mentalizing is also a key ingredient behind our ability to form and sustain groups though norm psychology (the suite of cognitive mechanisms that support learning, transmitting, and maintaining norms; see: Chudek & Henrich, 2011; Sripada & Stich, 2006). In this section, I review evidence for the interplay between mentalizing abilities and cultural learning. This cultural learning is important both for accumulated practices around artefact manufacture/ food production and for becoming appropriately socialized into local group norms. I further explore how specific subsets of cultural norms and beliefs ensconced within religion and morality rely upon mentalizing processes to sustain long-term cooperation in groups of increasing size and relational distance. 19 1.3.1 Cultural Influences on Mentalizing Processes As mentioned in section 184.108.40.206, there are some cross-cultural differences in specifics of timing and sequence in various mentalizing abilities, though the overall trend is quite stable. In this section, I turn to specific culturally transmitted practices and social structures that may produce different mentalizing orientations and different folk theories of mind in different cultures (Lillard, 1998). Perhaps the most important window into how differences in how people think about minds across cultures may emerge comes from differences in the ways that adults think about what children can and cannot be taught at various stages of development. In cultures where community values are more important than individual values, the ideal child is more often industrious and obedient rather than inquisitive – more along the lines of the old adage, ‘children should be seen and not heard’ (Lancy & Grove, 2010). Teaching in traditional societies is often more about scaffolding experiences after children are deemed fit to start learning; very few instances in the ethnographic record explicit, Western-style pedagogy (Kline, Boyd, & Henrich, 2013; Lancy & Grove, 2010). Often, teaching does not start until around age 5, when kids are believed to have developed the reasoning skills and emotional control necessary to begin learning vital skills to succeed in society. The common age when people in many traditional societies believe children have become reasonable enough to begin teaching them also coincides with the developmental milestone when children begin to understand and interpret behaviour in terms of beliefs instead of just desires. In contrast to these traditional learning scenarios that tend to focus on informal teaching through structured observation, formal education in classroom settings rely far more upon explicit knowledge transfer through direct pedagogy. Different cultures rely on formal and informal education to different degrees, and each mode of education has distinct effects on 20 children’s cognitive development (Falgout & Levin, 1992; Scribner & Cole, 1973). Broadly speaking, societies that are more focused on individual values and less concerned about managing uncertainty also tend to emphasize more independent thought and innovation. On the other hand, cultural groups that are more community-oriented and that face greater threat from resource, pathogen, or environmental challenges tend to emphasize conformity and dissuade students from learning by asking questions (Hofstede, 1986; Joy & Kolb, 2009; Lancy & Grove, 2010). These different approaches to socializing children based on emphasis of individual action or reference to existing social norm systems likely has a measurable impact on the ways that mentalizing processes are used across the lifespan. 1.3.2 Mentalizing and Overimitation: Cumulative Cultural Learning and Norms Social learning through imitation – copying that references the model’s goals – starts us off onto a cultural learning path almost as soon as we enter the world; as such, human infants and toddlers are effective, flexible, and judicious imitators. However, infants and toddlers may also be less inclined to repeat a model’s actions if these actions do not have a clear link to the task’s end goal or if there is a clear external reason (like a physical constraint) for the additional actions (Gergely & Király, 2002; McGuigan & Whiten, 2009). Tendencies to faithfully replicate actions apparently irrelevant to the end purpose get stronger as abilities to infer beliefs increase throughout middle childhood and into adulthood (McGuigan & Makinson, 2011). The observation that autistic children tend not to overimitate further suggests that reasoning about models’ intentions are especially important in overimitation (Marsh, Pearson, Ropar, & Hamilton, 2013). This tendency to overimitate is essential to our ability to accumulate and socially transmit the lessons of our ancestors that lies at the heart of cumulative cultural evolution (Shipton & Nielsen, 2015; Danchin et al., 2011; Mesoudi, 2011). Overimitation is 21 particularly relevant to two key ingredients to human culture: 1) acquiring complex behaviours like food processing and artefact manufacture and 2) acquiring norms – both social and non-social. A common cue that triggers overimitation is causal opacity. An action sequence is causally opaque when the learners interpret the actions to be intentionally directed toward an end goal, but the relationship between that end goal and each step is not obvious. Both chimps and human children will replicate extra actions in a model’s action sequence if they do not immediately infer how each action is causally related to the end product (Horner & Whiten, 2005). However, only humans seem to adopt the belief that these extra actions are done for a purpose and that this purpose should be replicated in their own behaviour. This expectation to be taught develops along with more general theory of mind processing and may be a specific adaptation behind human cultural learning (Csibra & Gergely, 2011a; Wellman & Lagattuta, 2004). This learning stance is most obvious after around 5 years of age, when children begin thinking more in terms of beliefs rather than more basic desire states (McGuigan, Whiten, Flynn, & Horner, 2007). The learning stance leading to extra imitation may be related to more basic teleological reasoning about objects as ‘being for something,’ or having a function, and these functional characteristics as intrinsic, long-term properties (Csibra & Gergely, 2006; German & Barrett, 2005; German & Johnson, 2002; Kelemen, 1999; Kelemen & Rosset, 2009). In situations where the causal relationship between actions and final products are not obvious, overimitation may be a route into learning about causality. Children may spontaneously encode all of the actions a model performs as somehow causally meaningful to the end product and construct meaning out of these actions to further facilitate learning (Lyons, Damrosch, Lin, Macris, & Keil, 2011; Lyons, Young, & Keil, 2007). This kind of causal learning may be 22 particularly important for culturally accumulated learning leading to increasingly complex toolkits (Boyd, Richerson, & Henrich, 2011; Richerson & Boyd, 2006; Tennie, Call, & Tomasello, 2009) and more extensive food processing techniques that allowed for more thorough expansion into new environments (J. Henrich, 2015; J. Henrich & Henrich, 2010). An important caveat to this causal learning story behind overimitation, however, is that the causal stories people might report once pressed to explain how things work may be totally irrelevant to the true underlying physical relationships (as is illustrated in the case of Fijian food taboos in J. Henrich & Henrich, 2010). If this is the case, then causal learning from overimitation may have more to do with learning norms, or social rules governing behaviour, than learning causality in the scientific sense (Ben Kenward, Karlsson, & Persson, 2011). When models communicate a strictly instrumental purpose to their action sequence, children are more likely to deviate from the model’s sequence and innovate their own solutions to emulate the end goal (Legare, Wen, Herrmann, & Whitehouse, 2015). The flexibility in this conventional vs. instrumental imitation strategy may be especially helpful in transmitting necessary causal processes while allowing for new innovations to accumulate (Legare & Nielsen, 2015). On the other hand, the conventional stance is a very basic component of ritual, which is itself a powerful means of transmitting norms to children and perpetuation norms within communities (Graham & Haidt, 2010; Rossano, 2012). By the time children are able to understand pretence at around age 2-3, they readily understand norms as shared ideas about how things should be done, learn new norms quickly, and selectively enforce these normative rules within the specific contexts they apply to (Rakoczy & Schmidt, 2013; Wyman, Rakoczy, & Tomasello, 2009). Children can understand the extent to which extra steps in an overimitation process are not directly causally important. When children teach an action sequence they learned through overimitation, they do 23 so in a way that primarily conveys normative information (Kenward, 2012; Keupp, Behne, & Rakoczy, 2013). Of course, normativity and artefact thinking need not be mutually exclusive products of overimitation. When first learning about a new artefact, children (especially children 5-years-old and older) tend to treat their use as more defined by convention (Casler, Terziyan, & Greene, 2009; Schillaci & Kelemen, 2014). In many traditional and small-scale societies, artefact manufacturing processes are indeed matters of normative conventions that are not subject to individual innovation; children are given very limited freedom to deviate from the prescribed methods of producing artefacts, which further perpetuates very specific and consistent traditional forms (Lancy & Grove, 2010). Having discussed how mentalizing provides the foundation to socially learning norms, we next turn to two norm systems that further rely upon mentalizing processes to support and maintain social structures and long term cooperation: religion and morality. 1.3.3 Mentalizing and Religion Mentalizing processes have been theorized to be the evolutionary root of religious experiences; our ability to infer the minds of others may be extended to infer the presence and contents of unseen supernatural agents’ minds (J. L. Barrett & Keil, 1996; Bering, 2006; Boyer, 2001; Guthrie, 1995). While these theories that focus on mentalizing as the root of religion posit that religious experience may be the result of excessive perception of other minds (i.e. hyperactive agency detection: Bering, 2006) or seeing human-like traits in non-human objects within anthropomorphism (Guthrie, 1995), other research suggests that humans are not the promiscuous mentalizers these theories suggest (Epley et al., 2008b; Waytz et al., 2010). Instead, we may be more likely to resort to a mentalistic explanation for phenomena only when we are already feeling socially isolated (Epley, Akalis, Waytz, & Cacioppo, 2008a) or when other 24 causal frameworks fail to adequately explain an objects’ behaviour (Waytz et al., 2010). Further, it does not appear that adults avoid excessive mentalizing by learning to suppress seeing human in everything; children too appear to favour mentalistic explanations only for human rather than object explanations (Wellman & Lagattuta, 2004). Therefore, the fact that humans around the world and throughout history have long applied mental state reasoning capacities to infer the presence and contents of unseen, supernatural agent minds is not likely just because our mind-reading capacities are always on high alert, but more likely due to specific situational influences. Though the above theories about mentalizing and religion are at least partially supported by evidence that some forms of supernatural belief might be directly related to mentalizing (see for example Willard & Norenzayan, 2013; Wlodarski & Pearce, 2016), the vast majority of religious experience seems to be more dependent upon the situational constraints that are transmitted through culture (W. M. Gervais & Henrich, 2010). Within culturally-transmitted religious traditions, individual differences in mentalizing ability does also relate to some differences in religious belief (W. M. Gervais, 2013b; Norenzayan, Gervais, & Trzesniewski, 2012). However, this relationship between mentalizing and religion may be more important for belief than ritual practice, and may therefore be more important for some traditions than others (e.g. Protestant Christianity places far more emphasis on correct belief and correct thought than other traditions, which more often focus on ritual and religious practice regardless of belief, see: A. B. Cohen & Hill, 2007; Graham & Haidt, 2010; Laurin & Plaks, 2014). However, even if a religious tradition is more focused on ritual than belief, that need not necessarily indicate that mental state reasoning is not evoked in these religious ritual activities. As discussed above, certain aspects of ritual displays like synchrony may engage mentalizing processes to facilitate coordinated action and leads to better cooperation (Baimel et al., 2015; 25 Wiltermuth & Heath, 2008). Importantly, the cooperative aspects of synchronized action may make it easier for ritual performers to tune into each other’s minds and take on each other’s perspectives to promote social cohesion. Adopting a ritualized stance in performing action sequences may also engage the conventional learning stance that promotes overimitation, thereby facilitating norm transmission and maintenance. Both of these benefits of ritualized behaviour relate to individuals honing in on each other’s intentional states and lining up actors’ intentions to produce reliably consistent patterns of coordinated actions. Another benefit of ritual displays is that these can help broadcast the trustworthiness of ritual practitioners to observers (Purzycki & Arakchaa, 2013). These ritual displays then can communicate intentions to be part of a group and further support long-term group bonding and cohesion (Soler, 2012; Sosis, 2000; Whitehouse, 2002; Xygalatas et al., 2013). Looking beyond ritual displays, certain aspects of religious belief further aid in sustained and extended prosocial behaviour and rely upon believers making specific inferences about supernatural agents’ minds. Using teleological reasoning to infer purpose behind random events may also push people to look for reasons why certain fortunes or (especially) misfortunes befall specific others. These teleological beliefs often take shape in the form of witchcraft, karma, or similar beliefs seeking explanations for why particular individuals might have particular outcomes, often as a result of failing to fulfill religious or other social obligations (Homans, 1941; Norenzayan et al., 2015; Purzycki & McNamara, 2015). Mind-body dualism also promotes beliefs like souls persisting after the physical death of the body, existence of disembodied minds like ghosts and other supernatural agents, and may combine with teleological thinking to analyze why supernatural agents might levy punishments or rewards that are associated with afterlife beliefs (Atkinson & Bourrat, 2011; Bering & Bjorklund, 2004; Shariff & Rhemtulla, 2012; 26 Willard & Norenzayan, 2013). Beliefs that supernatural agents care about human actions (inferences about what these supernatural agents are potentially attending to), that they can know what people are doing (inferences about supernatural agents’ knowledge and beliefs about people), and that they might punish people for doing things deemed inappropriate (predicting supernatural agent behaviour) has been shown to support cooperation with increasingly large groups of unrelated individuals (D. D. P. Johnson & Bering, 2006; D. D. P. Johnson & Krüger, 2004; Purzycki et al., 2016; Roes & Raymond, 2003; Schloss & Murray, 2011); though the punishment part may be more important in some traditions than the moralizing part of these beliefs (Watts, Greenhill, & Lieberman, 2015). Importantly, the specific contents of belief rely upon mentalizing processes both for initial belief acquisition, additional norm learning around what supernatural agents are typically thought to be interested in, and for making judgments about what supernatural agents might know and do in response to believers’ actions. These functions of religious belief and practice have little adaptive benefit at the individual level. However, at the cultural level, they help bind groups into cooperative units and support increasingly distant cooperative networks that help support humans in our hypersocial cultural niche. 1.3.4 Mentalizing and Moralizing Like religion, morality is often associated with supernatural belief, helps maintain long-term cooperation in groups, and also fundamentally relies upon mentalizing processes. This dissertation focuses on morality as the subset of social rules between conventions and formal legal codes. Moral norms differ from conventions in that morals tend to relate more to core group values, present more of a threat to the social group if they are violated, and tend to evoke stronger emotional and social responses than conventional norms (for example, killing someone 27 without a socially acceptable reason is far worse than wearing your sleeping clothes to work). Importantly, responses to moral norm violations nearly always involve some assessment of the perpetrator’s intent – their goals and beliefs while committing various acts – to gauge society’s response to the violation (Berg-Cross, 1975; Cushman, 2008; K. Gray, Young, & Waytz, 2012; Knobe, 2004). As such, moral reasoning can more generally be thought of as a specific instance of applied mental state reasoning. Moral judgments involve some calculation of perpetrator intent even among small-scale traditional cultures that place less emphasis on mental states as explanations for behaviours (though intent focus is highly variable across the cultures sampled, see: H. C. Barrett et al., 2016). However, not all violations are created equal. Moral norms fall into various categories that are fairly consistent across cultures (Graham et al., 2011; Rozin, Lowery, Imada, & Haidt, 1999; Shweder, Much, Mahapatra, & Park, 1997). Some categories like harm are far more attuned to intent, while other categories like purity are more focused on outcome and the fact that any violation occurred (L. Young & Saxe, 2011; L. Young & Tsoi, 2013). That someone meant to poison their rival makes the act far worse than an accidental dose of a chemical mistaken for something harmless, but any amount of dirt in one’s food is too much dirt – whether or not the cook meant to put it in the dish. The cognitive systems responsible for responding to moral norm violations also appear to ride upon emotional systems related to empathic processing. While some moral judgments may be driven by effortful, reasoned consideration (J. D. Greene & Haidt, 2002), much of the immediate response to moral norm violations appears to be driven by emotions that are later rationalized to explain the observer’s response (Haidt, 2001). In line with the supposition that moral reasoning fundamentally relies upon mental state reasoning, populations with varying degrees of dysfunction in mentalizing abilities show distinct patterns of moral judgments that 28 differ from populations without marked neurocognitive impairments. In general, people with lower levels of empathic concern – either as a sub-clinical individual difference (Gleichgerrcht & Young, 2013), as diagnosed high-functioning autism-spectrum traits (Gleichgerrcht et al., 2013), or as psychopathic traits (Tassy, Deruelle, Mancini, Leistedt, & Wicker, 2013) – tend to report being more willing to choose utilitarian options that minimize damage to a larger number of people by causing devastating harm to just a single person. People with lesions in the ventromedial prefrontal cortex, brain areas specifically associated with theory of mind processing, tend to pay less attention to intent in their moral judgments. Participants with lesions in these theory of mind areas tend to preference outcome by judging malicious intent in failed attempts less harshly when not associated with a bad outcome and bad outcomes even if they were accidental. People without any brain lesions and people with brain lesions in parts of the brain not associated with mentalizing, on the other hand, still focus more on intent for making their moral judgments (L. Young, Bechara, Tranel, Damasio, Hauser, & Damasio, 2010a). Children diagnosed with autism-spectrum disorders do distinguish between moral and conventional violations in ways that suggest some basic orientation toward empathic concern about others’ distress (Moran et al., 2011). People with high-functioning autism spectrum diagnoses do seem to be sensitive to the affective components of moral vs. conventional norm violations, though they focus less on intent and violations’ effects on others’ welfare (Zalla, Barlassina, Buon, & Leboyer, 2011). Similarly, criminal offenders who exhibited high levels of psychopathic traits were more likely to say that accidents – bad outcomes without malicious intent – were significantly less forbidden than comparison groups without measured psychopathic traits, perhaps due to less connection to the negative emotional experience of the accidental victim (L. Young, Koenigs, Kruepke, & Newman, 2012). Therefore, moral reasoning 29 appears to be profoundly intertwined with mental state reasoning that is applied both toward maintaining the integrity of social groups and is transmitted through cultural norms. 1.4 Overview of Chapters In this dissertation, I examine how specific cultural norms, as exemplified by the Opacity of Mind norms in Yasawan culture, can produce cross-cultural differences in how people think about thoughts. Indigenous Fijians living in small communities on Yasawa Island, Fiji, are of particular focal interest for this research, because Opacity of Mind norms present in Indigenous Fijian culture specifically restrict thinking about others’ inner mental states. Across five studies, I examined cross-cultural patterns of mentalizing about beliefs, thoughts, emotions, and social situations. I also examined how adults and children apply mental state reasoning to make moral judgments. First, chapter 2 examines how adults living in Indigenous and non-Indigenous communities in Fiji and urban communities in North America respond to different psychological measures that access theory of mind processing and emotional/ empathic processing. Because only Yasawans have Opacity of Mind norms proscribing thought about thoughts, only Yasawans should show lower focus on mentalizing processes. Specifically, the Yasawan participants should only indicate reduced emphasis on internal mental states accessed through theory of mind processing, while emotional/ empathic processing and social savvy should be unaffected. In chapter 3, potential underlying differences in baseline thinking about internal mental states are further elaborated in the context of moral reasoning. As mentioned above, moral reasoning and judgments about moral norm violations can be seen as an applied theory of mind task – even if the cultural value around answering questions about mind might prohibit talking about others’ mental states, making judgments about moral norm violations based upon perpetrator intent requires a background calculation of perpetrator’s mental state. Thus, moral 30 reasoning may serve as a proxy to evoke mental state reasoning even if direct measures may be affected by participants’ reluctance to self-report mental state focus. This can then get around the potential issue that any cultural differences seen in chapter 2 may not be due to genuine difference in cognitive activation, but rather due to reluctance to self-report mental state focus if mental state focus is deemed socially inappropriate. Chapter 3 study 1 examines baseline differences in intent vs. outcome focus in judgments about moral norm violations among Yasawan, non-Indigenous Indo-Fijian, and North American adults. Again, because Yasawans are the only cultural group with Opacity of Mind norms, they should be the only group to show markedly higher focus on outcome. However, if Yasawans are also attending to intent, then this can be taken as evidence that cognitive mechanisms for mental state reasoning are indeed being activated and used to make these judgments. Chapter 3 study 2 tests whether underlying differences in chronic activation of mental state reasoning may be at the root of potential differences in intent vs. outcome focus. Yasawan adults and North American university students were explicitly reminded to think about thoughts or actions before making judgments about moral norm violations. Reminders to think about thoughts should boost intent focus, while reminders to think about actions should boost outcome focus (a similar paradigm showed this pattern among North American and Indian adults, see: Laurin & Plaks, 2014). If Opacity of Mind norms in Yasawa are reducing focus on mental states, then explicitly reminding Yasawans to think about thoughts should boost focus on intent to levels similar to other cultural groups that may be more actively focused on minds without such reminders. If Yasawan adults do indeed shift to judgments that are more clearly focused on intent when reminded to think about thoughts, then this is strong evidence that any underlying differences at baseline are likely not the result of theory of mind mechanisms functioning differently in these two cultural contexts. 31 Rather, differences in mental state reasoning seen across these cultural groups are likely due to different levels of mentalizing mechanisms’ activation during everyday life. Chapter 4 examines the developmental trajectory of mental state reasoning within the context of intent vs. outcome focus in moral judgments. For this study, Yasawan and North America children and adults were asked to select characters from simplified moral norm vignettes that present contrasts between positive and negative intentions and outcomes. By presenting these contrasts as forced choices, this study gives some insight into the hierarchy of preferences people of either culture may show at different points across the lifespan. If cross-cultural differences in intent vs. outcome focus seen among adults are the result of group-level differences that are largely independent from cultural learning, then the children of either cultural group should look more like the adults of their culture than like each other. Conversely, if adult cross-cultural differences only emerge with the appropriate cultural learning, and if theory of mind processing and intentionality reasoning are indeed part of human core cognition, then the children should be most similar and the adults of either culture should diverge. Further, because these simplified moral norm vignettes are non-verbal, they may also help bridge potential problems with translation in the adult story vignettes used in chapter 3 studies 1 and 2. Taken together, these five studies give a glimpse of how cultural context interacts with cognition across development to shape individual minds into encultrated beings. 32 Chapter 2. Thinking About Thoughts When the Mind is Unknowable: Cross-Cultural Differences in Mental State Reasoning 2.1 Introduction The ability to interpret other’s thoughts, goals, and desires from their behaviour is a key skill humans rely on to navigate the complicated social worlds we live in. People living in Western societies often take this as a matter of course, some with such confidence that they state people can be understood and read like a book (Lillard, 1998). The Fundamental Attribution Error, a reasoning bias that over-emphasizes internal dispositions rather than situation constraints to explain others’ behaviours, is a related phenomenon that can be seen as stemming from this confidence that behaviour is a reliable indicator of underlying mental states, (E. E. Jones & Harris, 1967). Though this is a pervasive way of parsing the social world in the West, people from other cultures have been shown to be less inclined to focus strictly on the mind as the driver of behaviour (Krull et al., 1999). Ethnographies and developmental studies from small scale societies show greater variation in focus on mind and later ages of full theory of mind reasoning in especially more collectivistic societies (H. C. Barrett et al., 2016; H. C. Barrett, Broesch, Scott, He, Baillargeon, Di Wu, et al., 2013a; Duranti, 2015; A. Mayer, 2013). These conflicting patterns of theory of mind reasoning present a puzzle: if the ability to anticipate other’s actions is so vital to human social life, is it possible that this ability is not culturally universal? This chapter addresses this issue by comparing cultural groups with vastly different normative approaches to understanding others. We use behavioural and self-report measures of how people in these cultures approach the problem of other minds to assess how inferences about others’ thoughts, feelings, and beliefs vary as a result of the cultural environments people navigate. 33 We first briefly review how psychologists, philosophers, and primatologists have discussed how theory of mind processing is used to infer unseen mental states and how empathy is used to understand others’ emotions. Though these may represent distinct aspects of mentalizing abilities, theory of mind and empathy have often been conflated in samples from Western, Educated, Industrialized, Rich, Democratic (WEIRD) societies that are overrepresented in the current psychological record (J. Henrich, Heine, & Norenzayan, 2010). We next discuss how cultural beliefs about minds – hereafter referred to as cultural models of mind – differ in Western vs. non-Western cultural groups. Importantly, the Western emphasis on individualism makes the mind the locus of the self, the self the all-important arbiter of behaviour, and thus behaviour indicative of internal mental states. In more collectivistic non-Western groups, the relationship often forms the most important unit of behaviour. This focus on relationships and social standing rather than personal motives and desires may similarly reduce the mentalistic focus in collectivistic groups by making situations more salient than internal states and dispositions (Choi & Nisbett, 1998). In two studies, we investigate how theory of mind and empathy are affected by differences in cultural norms defining how knowable (or unknowable) minds are. In the first study, we use a continuous measure of adult reasoning about other a character’s false beliefs about the world. In study 2, we use a self-report questionnaire that assesses internal mental processing, interest in emotions, and facility with social situations across three groups: one WEIRD, one with no specific norms stating about minds as unknowable, and one with norms emphasizing that the mind is opaque and unknowable. We find that the participants from the cultural group with norms stipulating that minds are unknowable self-report significantly lower fluency in interpreting others’ internal mental states, but not emotions or social situations. We 34 find that lower self-reported focus on understanding internal mental states found in study 2 further explains group-level differences in ability to use knowledge of a character’s false belief about the world to predict their behaviour found in study 1. 2.1.1 Theory of Mind and Empathy in WEIRD Cultures In the decades since Premack and Woodruff (1978) coined the term “Theory of Mind” (ToM), as the ability to “[impute] mental states” (p. 515), we still remain surprisingly unclear about just what this ability really entails. ToM is often conflated with empathy, a similar concept relating to how we infer other’s minds, first introduced into English as a “process of humanizing objects, of reading or feeling ourselves into them” (Titchener, 1924, p. 417; Titchener, 1909; 1924). Premack and Woodruff immediately acknowledge the deep conceptual similarity between ToM and empathy; they attempt to distinguish the two by predicting that empathy will be restricted to purpose and motivation, but will lack any conception of the other party’s knowledge state. Despite near-exponential growth in research on the topic, ToM research covers so many sub-concepts and sub-processes as to make the conflation between ToM and empathy the least of our conceptual troubles (Schaafsma, Pfaff, Spunt, & Adolphs, 2015). Much of the insight into human ToM has been through one of two lenses: dysfunction or development (Schaafsma et al., 2015). Of the many techniques used to assess ToM functioning, variants of the False Belief Task (Baron-Cohen et al., 1985; H. C. Barrett, Broesch, Scott, He, Baillargeon, Di Wu, et al., 2013a; Birch & Bloom, 2007; Callaghan et al., 2005; Dennett, 1978; A. Mayer & Trauble, 2012) and Empathy Quotient (EQ: Baron-Cohen & Wheelwright, 2004; Berthoz, Wessa, Kedia, Wicker, & Grèzes, 2008; Freeth, Sheppard, Ramachandran, & Milne, 2013; Norenzayan et al., 2012) have been particularly fruitful in research including participants from many different cultural groups and of varying degrees of cognitive impairment. This 35 chapter focuses on variants of these two approaches used to measure theory of mind among Western samples to look at how humans approach the problem of other minds from the perspective of systematic cultural differences. 2.1.2 Assessing Thoughts about Minds: False Beliefs and Empathy Quotients In response to Premack and Woodruff’s seminal paper, Dennett (1978) proposed an experimental paradigm for investigating ToM: the false belief task. Dennett uses the example of a classic Punch and Judy puppet show gag, where children squeal with delight at the sight of Punch’s labours under the mistaken belief that Judy is still in the box that he is attempting to toss off of a cliff. In reality, the children have already seen that Judy has escaped from the box. Dennett’s proposed experiment similarly compels the participant to act in a way that pits their knowledge about the true state of the world against their inferences about what a character with incorrect knowledge about the world. If a participant is indeed tracking the inner mental states of the character, then they should expect the character to act in accordance with their mistaken beliefs (and so participants should expect Punch to persist in heaving the box off of the cliff, believing that Judy is still inside). However, if the participant is not distinguishing the character’s knowledge and their own updated knowledge of the real world, then they should expect the character to act on the true state of the world without any clear idea of how the character may have updated their incorrect knowledge. Shortly after Dennett proposed this study, Wimmer and Perner (1983) made this test a reality through short stories. Children answered questions about these stories and showed false belief comprehension that got progressively better from 3 to 9 years of age. Soon thereafter, Baron-Cohen, Leslie, and Frith (1985) adapted Wimmer and Perner’s task into an even simpler, vignette-based binary ToM measure (the Sally-Anne task). 36 Baron-Cohen and colleagues make the connection that the cognitive dysfunction underlying autism-spectrum disorders may be caused by limitations in ToM reasoning. In the following years, theory of mind tasks increased in complexity to measure ability in first and second-order mind recognition. First-order mind recognition tasks (like the Sally-Anne Task) require the participant to track the knowledge state of one character, while second-order mind recognition tasks require participants to track multiple characters’ beliefs about each other. Data gathered from first and second-order mind recognition tasks have made great strides in illuminating early ToM development. However, neither were able to distinguish subtleties in adult ToM functioning – even among adults diagnosed with confirmed ToM deficits in high-functioning Autism and Asperger’s Syndrome (Bowler, 1992; Happé, 1994). In order to develop a more nuanced test applicable to adult ToM, Baron-Cohen and colleagues leveraged some cross-culturally stable patterns in ability to connect facial expressions to ascriptions of others’ internal mental states to develop the Mind in the Eyes task (Baron-Cohen, 1996; Baron-Cohen, Jolliffe, Mortimore, & Robertson, 1997; S. B. Cohen et al., 2001). Despite the Mind in the Eyes task’s ability to distinguish even very high-functioning autism spectrum cases from sub-clinical and otherwise normal populations, it could not point to individual differences in specific aspects of ToM functioning. In subsequent attempts to develop relatively quick, self-administered tests of where individuals might fall on the Autism spectrum (e.g. the Autism Spectrum Quotient or AQ: Baron-Cohen et al., 2001), ToM again becomes difficult to disentangle from empathy. Given the clear evidence that autism results from problems with understanding others’ minds, Baron-Cohen and colleagues went on to develop the Empathy Quotient (EQ) as a question-based, individual difference measure that purports to capture both the cognitive (ToM) and affective/ emotional components of mental state reasoning (Baron-Cohen & Wheelwright, 2004). 37 Meanwhile, a parallel set of theories began to emerge around another EQ: the Emotional Quotient, or Emotional Intelligence (EI: Bar-On, 1996; Gardner, 1983; J. D. Mayer, Salovey, & Caruso, 2004; Salovey & Mayer, 1990). Instead of focusing on deficits and cognitive dysfunction, emotional intelligence attempts to explain why some people seem to be especially good at understanding emotions (both others’ and their own) and at using this emotional understanding toward successful social interaction. Despite the clear conceptual similarity, EI’s original formulations included many domains that extend well beyond the empathic concern that is often bound up with ToM (R. J. R. Blair, 2002). Perhaps because of EI’s focus on ability and EQ’s focus on disability, the EI literature became especially focused on assessing how high EI might facilitate job performance in sectors like business and medicine (Faye et al., 2011; Ilangovan, Scroggins, & Rozell, 2007; Singh, 2010; Yadav & Iqbal, 2009). Though the initial theorizing around EI remained distinct from the ToM literature, subsequent research has included populations known to be low in ToM (participants on the autism spectrum) and high in ToM but low in EI (individuals with psychopathic traits). This research with individuals on the autism spectrum or with psychopathic traits further suggests that, while the emotion recognition and application aspects of EI are indeed associated with ToM, they remain distinguishable (R. J. R. Blair, 2002; Ferguson & Austin, 2010; Montgomery et al., 2013). Of the studies that explicitly compare EI with ToM, the vast majority rely on tasks that focus on emotional recognition in facial cues (such as the Mind in the Eyes or similar tasks) rather than the self-report measure of ToM and empathy captured by the EQ (Ali et al., 2009; Ferguson & Austin, 2010; Montgomery et al., 2013); studies with children also may incorporate first and second-order ToM tasks (A. P. Jones et al., 2010). 38 2.1.3 Mentalizing Across Cultures: Ethno-Psychologies and Social Selves In her review, Lillard (1998) summarizes the Western model of mind as the belief that the mind is the seat of all thoughts and emotions. For Westerners, the mind is the place where the independent, discrete self is located (Markus & Kitayama, 1991; Taylor, 1989). This Western model of mind also incorporates an intuitive mind/ body dualism with the mind, and by extension the self, being located within the physical medium of the brain (or the eyes, see: Starmans & Bloom, 2012). This also creates a strong separation between the physical body and the mind that inhabits it (Bloom, 2004). One consequence of this mind-body dualism is a distinction between rational thought and other kinds of more physical, embodied, experiential mental processes (this separation between agency and experience is further reflected in a recent two factor, primarily Western model of mind reported in H. M. Gray, Gray, & Wegner, 2007). Another consequence is the Western paradox of mind reading and the problem of solipsism – if one’s mind is fundamentally separate from all others, how is anyone to know that other minds do in fact exist? The Western model of mind counters this problem with a keen sense that other minds are knowable – that we can come to comprehend the contents of another minds – despite the physical separation between the observer and the observed. Part of the Western solution to this solipsistic quandary is the belief that thoughts and other internal mental states are the mediators between the mind and the physical world. Thus, if one can observe another’s behaviour, the Western model of mind posits that these behaviours are caused by the mind, and therefore directly indicative of that mind’s contents. The Western model of mind further recapitulates the centrality of the individual by interpreting others’ actions through the lens of intent (Kelemen, 1999; Kelemen & Rosset, 2009; Malle, 2006b; Malle & Knobe, 1997; Mull & Evans, 2010; Woodward, 2009). Individuals’ subjective perspectives may lead to very different 39 conclusions about the same underlying ‘objective’ reality. Similarly, a person’s responses to events depend on the individuals’ thoughts about that event – and therefore their observable actions are inferred to be more indicative of their underlying character than the situation. This common Western assumption that internal thoughts and characteristics are more causally important than situations leads to biases in reasoning like dispositionism and the Fundamental Attribution Error (Ross, 1977). The most striking difference that separates the Western model of mind from other non-Western conceptions is in the importance (or non-importance) of social context and relational ties. Social context rarely factors in to Western models of mind, but the embeddedness of one’s social self is inextricable from many non-Western models of mind. The relationship – the space in and between individuals - is the focal point, not the individual (Anae, 2010; Brison, 2001; T. Brown, 2006; Carsten, 1995; Groark, 2011; Heine, 2001; Kavapalu, 1995; Lillard, 1998; A. Rumsey, 2000; Taylor, 1989). The underlying cultural factors that promote this emphasis on social connections vs. independence (cultural tightness or looseness: Gelfand, Nishii, & Raver, 2006; Gelfand et al., 2011) relate in part to the current and historical physical and social instability these cultural groups faced (Bauer, Cassar, Chytilová, & Henrich, 2011; Faulkner, Schaller, Park, & Duncan, 2004; Fincher & Thornhill, 2011; Hruschka & Henrich, 2013; Hruschka et al., 2014; Putnam, 1993; Van de Vliert, 2011; Vandello & Cohen, 1999). Conversely, the greater institutional and legal infrastructure cultures have to buffer against these sources of existential threat, the more likely individuals are to branch out from known sources of support like family and operate as autonomous units (Hruschka et al., 2014; Kay, Shepherd, Blatz, Chua, & Galinsky, 2010; Norris & Inglehart, 2004). Social hierarchy (power distance: Hofstede, 1983), and the complimentary norms dictating how to interact with others in one’s 40 social group (specifically, how strictly one is expected to adhere to norms of politeness or other highly-structured sets of behavioural expectations, as are found in cultures of honour and face: Boiger, Güngör, Karasawa, & Mesquita, 2014; Ho, 1976; Leung & Cohen, 2011; Vandello, Cohen, & Ransom, 2008) may limit the extent to which individual actions can be seen as clearly indicative of actors’ true underlying mental states rather than necessary reactions to the situation that are irrelevant to the actor’s true intent. For example, cultures that value hierarchy and social order (high power distance, high value on maintaining Face through proper respect for duties and obligations based on one’s place in the hierarchy) tend to have much more circumscribed norms limiting expression of especially negative emotions like anger, contempt, and disgust (Matsumoto, Takeuchi, Andayani, Kouznetsova, & Krupp, 1998); less fluency in recognizing these emotional displays (Fernandez et al., 2000; Matsumoto, 1989); and expectations of emotional self-censorship (Matsumoto et al., 2008). Extreme emotion – especially anger – is sometimes even thought to be physically dangerous, particularly in traditional societies (Lillard, 1998; McNamara, 2014). Some of the first studies to address whether ToM was indeed a unitary phenomenon around the world focused on young children’s performance on first and second-order False Belief tasks originally developed in the West. Early evidence showed there was indeed a common trend of children becoming more likely to pass these ToM tasks as they got older, though the rate of change in the proportion that pass and exact age when a majority of children pass varies substantially (Wellman et al., 2001). Even within a broader collectivist cultural group, studies on large-scale societies with a common Chinese heritage showed mainland Chinese children passing much earlier than children from Hong Kong or Singapore (Liu et al., 2008) – a surprising finding that may partially be accounted for by mainland Chinese values that 41 are a bit less collectivistic than either Hong Kong or Singapore (S. Lau, 1992). Children from collectivist cultures also understand that opinions may be diverse later in development than children raised in more individualistic cultures (Shahaeian et al., 2011; Wellman & Liu, 2004). Broadening the research to include small-scale societies shows much wider variation, with hunter-gatherer Baka in the rainforests of Central West Africa passing at far younger ages than small-scale agriculturalists in Quechua-speaking Peru (Wellman et al., 2001; Wellman & Miller, 2006). Looking even further into small-scale societies that have been ethnographically documented to down-play the possibility of knowing other minds (see “Opacity of Mind” section 2.1.4 below), even these children do show a progressive improvement with age but pass on average at a much later age than children in other societies (H. C. Barrett, Broesch, Scott, He, Baillargeon, Di Wu, et al., 2013a; Callaghan et al., 2005; A. Mayer & Trauble, 2012). 2.1.4 Opacity of Mind While philosophers and psychologists hashed out the quantitative details of whether and how theory of mind operated in all phases of development and in various forms of psychopathology, anthropologists – especially those working in and around the Pacific – began to challenge the very idea that all people intuitively and reflexively interpret others’ behaviour in terms of their inner mental states (Duranti, 2015; Groark, 2011; Hollan & Throop, 2011; Lillard, 1998; A. Rumsey & Robbins, 2008; Throop & Hollan, 2008; for a review, see: A. Mayer, 2013). Many ethnographers working with small-scale societies, especially in Papua New Guinea (PNG), Samoa, Fiji, and even communities in the Arctic and among the Maya, reported a consistent pattern of responses claiming that it is impossible to truly know the contents of another mind – in other words, that the mind is within the opaque container of the body, and only the person with that mind can know its contents. These opacity claims might reflect real differences in how their 42 respondents approach the problem of others’ minds (Robbins, 2008; Robbins & Rumsey, 2008), or they may just be products of ethnographers not fully understanding the local perspective on how to properly conceptualize minds (Duranti, 2008; Groark, 2011). Most respondents within these societies readily endorse statements that others have minds with contents that differ from their own (a basic step toward a first-order ToM understanding), but reject the idea that they, as a different person, can actually know what those contents are (Hollan & Throop, 2011; Throop, 2008). More specifically, the opaque part of Opacity of Mind claims appears to based on the observation that norms of politeness and potential desire to deceive make a person’s exterior displays non-informative of good vs. bad intent, and therefore essentially useless in determining what lies beneath (however, this does not seem to stifle extensive conversation and supposition about what observable actions might actually tell about one’s true intent; see: Groark, 2008; Schieffelin, 2008; Stasch, 2008; Throop, 2008). Despite the qualitative evidence that people in Opacity of Mind cultures may have different beliefs about how informative external behaviours are and how this should inform their own behaviours, there is precious little systematic, quantitative data on how ToM in these cultures explicitly compares to people in other places. Studies on the presence and developmental trajectory of ToM in small-scale societies show mixed evidence. Most rely on first order false belief tasks. The general trend supports the notion that false belief understanding may be a cross-cultural universal, and that it usually develops within the first decade of life (Callaghan et al., 2005; A. Mayer & Trauble, 2012). However, looking-time data suggests that False Belief understanding may indeed show up much earlier, despite cultural and linguistic barriers (H. C. Barrett, Broesch, Scott, He, Baillargeon, Di Wu, et al., 2013a). 43 Systematic data among adults is even harder to come by; some of our work shows that Fijians (both Indigenous and descended from indentured workers shipped over from India, hereafter referred to as Indo-Fijians) think other minds have characteristics and capacities that significantly differ from the Western model of mind (Willard & McNamara, n.d.). These mental capacities data suggest that the two-factor agency (rational thought) / experience (embodied/ emotional thought) distinction common to the Western model of mind (H. M. Gray et al., 2007) does not map on to either Indo-Fijian or Indigenous Fijian conceptions of either human or supernatural minds (Willard & McNamara, n.d.). However, this mental capacities data cannot distinguish whether there is a group-level difference between Indo-Fijians (who do not adhere to Opacity of Mind norms) vs. Indigenous Fijians (who do hold some Opacity of Mind norms). Further, though this does suggest that culture matters in how people conceive of others’ minds, it that study did not directly measure ToM or empathic functioning. One other study that also includes indigenous Fijians in Yasawa, Fiji, plus people in PNG and the Amazonian rainforests, does explicitly compare first-order false belief performance among adults (H. C. Barrett & Broesch, 2008). They use a binary choice task and an adult-oriented Curse of Knowledge task (see Study 1 in section 2.3 below; Birch & Bloom, 2007) that gauges more nuanced false belief understanding. Interestingly, results from the binary false belief task show all small-scale societies report just a bit less expectation that a character will look in the correct false belief location than Westerners, except the group from PNG who are much less likely to focus on the expected false belief answer. Similarly, the more continuous false belief measure in their Curse of Knowledge task shows Yasawans and some Amazonian groups clearly identify the correct original container (Fijians showed patterns similar to Westerners but with smaller effect sizes). However, a sample from an African pastoralists group and a PNG sample showed no clear 44 discernment among the choices. This suggests that, at a bare minimum, the African pastoralist and PNG samples may not use false belief comprehension to predict a fictional character’s behaviour in this scenario. These results suggest people in these small-scale societies may indeed be less fluent in interpreting other’s internal mental states. However, these false belief measures still tell us little about how empathic processes might be operating. 2.2 Overview of Studies This chapter seeks to address the gap in the literature on how theory of mind and empathy are impacted by culture. Here, we will explicitly compare how participants from different cultures within and across countries approach other minds through understanding thoughts or emotions, and how these mentalizing tendencies inform their use of knowledge about others’ beliefs to predict behaviour. In study 1, we use an adult-appropriate false belief task, the Curse of Knowledge task (Birch & Bloom, 2007), to compare North American university students and Indigenous Fijians living in Yasawa, Fiji. Yasawans should rate characters as less likely to look in the correct false belief location than North Americans. In study 2, we further test mental state focus by using the short version of the EQ (Wakabayashi, Baron-Cohen, Wheelwright, et al., 2006b) to systematically compare North Americans, Indo-Fijians living on the Fijian mainland, and Yasawans. The Indigenous Fijian and Indo-Fijian comparisons are particularly illuminating to separate out the effects of cultural background and country, as both groups share the same country-level institutional structures but differ greatly in both heritage and current day-to-day cultural practice. Because our Yasawan sample has a background of Opacity of Mind beliefs, we predict that they will self-report less fluency in thinking about and interpreting others’ mental states compared with other cultural groups. However, this should not impact Yasawans’ interest in emotions or comfort with social situations. Finally, we link up university students’ and 45 Yasawans’ Curse of Knowledge and EQ results to determine whether any differences in false belief expectations are related to self-reported ability to understand others’ internal mental states. We predict that the internal mental states factor of the EQ, but not understanding emotions or social situations, should explain any group-level differences in Curse of Knowledge ratings. 2.2.1 Why Fiji? Indigenous Yasawans and Indo-Fijians One of the most important contributions of this chapter comes from our within-country, cross-cultural comparison of two different groups in a non-WEIRD society. As a small Pacific Island nation, Fiji has a number of attributes that make it a particularly interesting focal site for the present study. First among these is that Indigenous Fijian culture is among the Pacific Island cultures that have been documented to exhibit Opacity of Mind norms. Further, much of the official governance in Fiji is still largely in the hands of Indigenous Fijians, creating greater opportunity for these traditional norms to persist even in the face of modernization and globalization. In addition to traditional Indigenous Fijian beliefs about opaque minds, Fiji also has a large population of non-Indigenous people of Indian descent who were brought into Fiji as labourers on sugar cane farms in the days of the British Empire. The comparison with Indo-Fijians is particularly helpful in that they share the larger country-level institutional structures with Indigenous Fijians, but come from a distinct, non-Western cultural tradition. In this section, we detail the site where our Indigenous Fijian participants live on Yasawa Island and the site where our Indo-Fijian participants live on the main island of Viti Levu (see Figure 2.1). 46 Figure 2.1. Map of Fijian Archipelago showing locations of Yasawa Island and Lovu Village. 220.127.116.11 Yasawa Island As members of a culture with Opacity of Mind norms, our Yasawan participants are the focus of both studies 1 and 2. The people of Yasawa, Fiji, live in traditional villages of around 70-150 adults, where the primary means of subsistence is through fishing and horticulture. Villages are set up around the traditional Fijian hierarchy that draws upon kinship ties to culminate power in a hereditary chief. The kinship relationships among individuals are the primary normative framework for deciding whom to cooperate with and how to allocate resources for day-to-day tasks ranging from finding and preparing food to building houses (McNamara & Henrich, 2016; Nayacakalou, 1955; 1957). Traditional practices and normative obligations based on kinship are the primary means of organizing social interactions, creating a more relational than individual sense of self (Brison, 2001; A. Rumsey, 2000). Holding to tradition is also considered a primary characteristic of what it means to be Fijian and carries political as well as interpersonal connotations throughout Fiji (France, 1969; Jolly, 1992). Traditional practices and beliefs are themselves bolstered by a combination of Christian and Turtle IslandYasawaIslandTaveuni IslandMatagi IslandLaucala IslandQamea IslandVanuaLevuLabasaPacific HarbourBeqa IslandSuvaOvalauIslandGau IslandVatuleleIslandLautokaNadiNausoriSigatokaViti Levu50 kmVanua Balavu IslandNorthernLau GroupSouthernLau GroupSavusavuKadavu IslandThe Fiji IslandsCopyright 2001 Under Watercoloursvisit www.underwatercolours.com for purchase & usage information10 3020 40Waya IslandWayasewa IslandNarara IslandKuata IslandNaukacuvuIsland Nanuya Balavu IslandDrawaqa IslandNamaraNatawaYalobiWayalevuSomosomoMarouSosoGunuYageta IslandNaviti IslandMatayalevuNacula IslandTavewa IslandMatacawa IslandNanuya Levu IslandNanuya Lailai IslandVuakeSisiliNaculaYasawa IslandSawa-i-lau IslandNanuya-i-yata IslandTeciVawa IslandYawini IslandYasawairaraBukamaNabukeruNanuya-i-ra IslandYasawa IslandGroupCopyright 2001 Under Watercoloursvisit www.underwatercolours.com for usage informationPacific HarbourBeqa IslandSuvaOvalau IslandVatulele IslandLautokaNadiBa VatukoulaTavuaRakirakiNausoriSigatokaViti LevuCopyright 2001 Under Watercoloursvisit www.underwatercolours.com for usage informationLovu%Village%Western'Vi*'Levu' 47 traditional ancestor spirit beliefs. These traditional ancestor spirits, or Kalou-vu, are the deified progenitors of the clans and phratries at the heart of traditional Fijian social hierarchies. The Kalou-vu are believed to care about traditional norms by affecting the health and fortune of those who deviate from traditions, but can also be called upon for traditional medicine among those who lead a proper traditional lifestyle (Katz, 1999; Shaver, 2014). Importantly, the Kalou-vu are tied specifically to the kin groups who form their progeny, and are often also bound in space to the location where the clans originated. This creates a deep social and spiritual bond that combines the individual, the kin network, and the lands traditionally inhabited by those kin groups into the vanua (the land and the people of that land), an essential component of indigenous Fijian identity (Abramson, 2000; Jolly, 1992; Turner, 1988; Williksen-Bakker, 1990). Christianity has also become a major component of what it means to be Fijian, which exists in partially syncretic, interconnected tension with traditional beliefs and ties to the land (Newland, 2004; Ryle, 2010; Tomlinson, 2007). The major Christian denominations in Yasawa are Wesleyan Methodism and the Assemblies of God Pentecostalism, though several others exist around Fiji. 18.104.22.168 Lovu Village, Viti Levu We include data from non-Indigenous Fijians as a comparison group in study 2. Our Indo-Fijian sample was primarily recruited from Lovu village, located on the main island of Viti Levu near Lautoka; some additional data comes from Indo-Fijian participants living in the nearby towns of Nadi and Ba. Indo-Fijians are a diaspora population who came to Fiji as indentured labourers between 1879 and 1912 (Gillion, 1962). Today, Indo-Fijians subsist primarily as wage labourers or sugar cane farmers. Indo-Fijians are primarily Hindu or Muslims, with a minority of Sikhs and Christian converts. While the Hindu tradition includes many gods, 48 the present participants reported belief that all gods are aspects of one single God (Willard, 2016). Though Yasawans do have some connection to the Lautoka area, it is unlikely that members of either sample have interacted with a member of the other sample; hence these two samples can be considered different, but geographically related, cultural groups. Despite their differences in cultural background, both Indigenous Yasawans and Indo-Fijians provide a useful contrast to the highly individualistic models of mind and self common in North America (M. M. Gervais, 2013a; Kelly, 1988; Kline et al., 2013). Yasawans and Indo-Fijians are highly collective with strong family ties that are often relied upon in times of need. Both also have more hierarchically structured social roles than commonly seen among North Americans. Other work in India suggest that the hierarchical power relations there lead to different patterns of emotional intelligence, though little evidence has been gathered to explicitly show any deficits in theory of mind or belief in the mind as unknowable (Sharma, Deller, Biswal, & Mandal, 2009). These differences in social relationships and religious beliefs make it possible to separate out the effects of heritage culture and current country-level institutional influences in how individuals come to understand others’ minds. 2.3 Study 1: False Belief When Others’ Minds are Private 2.3.1 Method Data in this study were collected in conjunction with study 2, data reported in chapter 3, and two additional projects on cross-cultural differences in beliefs about supernatural agents’ 49 minds and how these beliefs affect moral concerns.2 Analyses in this chapter focus on participants’ own perception of other human minds through mentalizing. 22.214.171.124 Participants We recruited 60 Yasawans (34 women; 19-80 yrs., mean = 40.24; 5-15 yrs. formal education, mean = 9.22) to participate in the June-July 2012 field season. From September 2012 to January 2013, we collected data from a further 205 (172 women; 17-32 yrs., mean = 19.85; 12-17 yrs. formal education, mean = 12.92) university students in Vancouver, B.C. Yasawans’ participation was strictly voluntary, and students were remunerated with course credit. 126.96.36.199 Materials We further measure participants’ ability to infer other’s minds using Birch and Bloom’s (2007) Curse of Knowledge task, an adult-appropriate extension of the false-belief task. As with classic false belief tasks, the Curse of Knowledge task also focuses on a story about a character, Juan, who places an item in a blue box (see Figure 2.2 A), then a second character, Miguel, moves the item to a new location after Juan leaves. The Curse of Knowledge adds difficulty beyond classic false belief tasks by including multiple boxes and changing the boxes’ arrangement (see Figure 2.2 B). 2 See McNamara, Willard, and Norenzayan (n.d.) for beliefs about supernatural agent minds and their impact on moral reasoning. Cultural differences in perception of agentic vs. experiential capacities of supernatural and human minds are reported in Willard & McNamara (n.d.). Both of these manuscripts are available from McNamara by request. 50 Figure 2.2. Diagram of Curse of Knowledge task. A) Character 1, Juan, places an item in the blue box. B) Character 2, Miguel, moves the item to a new box, then re-arranges the 4 boxes. Participants rate the likelihood that Juan will look in each location on a 1-7 scale. The 1-7 likelihood rating gives a more subtle look at how certain locations might pull participants toward a wrong location than the typical binary choice in children’s false belief tasks; even if participants correctly identify that Juan is most likely to look in the original blue box, higher ratings of other locations can indicate that there is still some residual appeal to incorrectly guess that Juan will look there. Across three between-subjects conditions, participant get different information about the item’s new location: 1) In the ignorance condition, participants are only told that the item was moved, but not where the new location is. 2) In the knowledge-implausible condition, participants are told the item is now in the purple box. The purple box is an implausible place for Juan to look because it is both a different color and a different orientation than where Juan originally placed the item. 3) In the knowledge-plausible condition, participants are told the item is in the red box, which is a different color but the same orientation as the original box. A B 51 The green box is never mentioned in any condition, so it acts as a distractor. Participants should primarily expect Juan to look in the blue box (original container) and secondarily expect the red box (original orientation). However, the knowledge-plausible condition should boost attention toward the red box because it should take more effort to inhibit this additional location knowledge in order to accurately guess what Juan will do based on his false belief. 2.3.2 Results We first look to see whether the North American university students and Yasawans differ on their Curse of Knowledge ratings. We use multilevel modeling to account for the multiple ratings (one for each box) every participant reported. We fit models in R (R Development Core Team, 2008) using the lme4 (Bates, 2010), lmerTest (Kuznetsova, Brockhoff, & Christensen, 2014), and LMERConvenienceFunctions (Tremblay & Ransijn, 2015) packages. We further account for potential sex differences, with women as the reference group. Yasawans and students rated each box significantly differently across the three knowledge conditions (F(6, 717.88) = 5.08, p < 0.001).3 The least squares mean ratings for the four boxes by knowledge condition and sample are reported in Error! Reference source not found. and Figure 2.3. Even when we control for sex differences,4 Yasawans rate Juan’s likelihood of looking in the original container as significantly lower than students. Though Yasawans consistently rate Juan as significantly less likely to look in the original blue box than the students, Yasawans’ ratings are closest to the students’ ratings in the Ignorance condition, when they are not given any additional information about the ball’s new location. Further, giving participants extra 3 We use type III ANOVA with Satterthwaite approximation for the degrees of freedom to calculate this overall difference. See Appendix A for ANOVA table. 4 This also holds if we control for EQ scores, as described in study 2. 52 distracting information about the new location as the implausible purple box (both a new box and new location) makes the difference between students and Yasawans even greater than when participants are given no information about where Miguel moved the ball. Box Information Students Yasawa Difference Students - Yasawa Old Box, New Location (Blue Box) No Info ✔✔ 6.14 [5.79, 6.55] 4.62 [3.78, 5.31] -1.52*** [-2.35, -0.69] New Box, Old Location ✔✔ 6.14 [4.27, 5.03] 3.48 [2.3, 3.83] -2.66*** [-3.46, -1.85] New Box, New Location ✔✔ 6.52 [2.42, 3.18] 3.82 [2.37, 3.90] -2.70*** [-3.51, -1.89] New Box, Old Location (Red Box) No Info 4.64 [2.28, 3.04] 3.12 [3.1, 4.62] -1.52*** [-2.35, -0.69] New Box, Old Location ✔ 4.77 [6.09, 6.89] 5.26 [3.02, 4.50] 0.49 [-0.31, 1.29] New Box, New Location 3.62 [3.21, 4.01] 2.54 [1.73, 3.21] -1.07* [-1.89, -0.27] New Box, New Location (Purple Box) No Info 2.81 [2.05, 2.85] 3.12 [3.63, 5.11] 0.31 [-0.52, 1.14] New Box, Old Location 1.95 [1.92, 2.72] 2.54 [2.38, 3.86] 0.59 [-0.21, 1.39] New Box, New Location ✖ 2.43 [5.80, 6.58] 4.38 [2.68, 4.12] 1.94*** [1.13, 2.75] Distractor (Green Box) No Info 2.70 [4.42, 5.20] 3.74 [4.50, 5.93] 1.05* [0.22, 1.88] New Box, Old Location 2.09 [1.55, 2.33] 2.59 [1.84, 3.28] 0.50 [-0.30, 1.30] New Box, New Location 2.30 [1.62, 2.41] 2.99 [1.99, 3.42] 0.69 [-0.12, 1.50] Significance codes: ***<0.001, **<0.01, *<0.05, †<0.1 Table 2.1 Least Squares Mean ratings of the likelihood Juan will look in each box across samples and knowledge conditions. 95% CI calculated using profile likelihood estimation in brackets. ✔✔ : box that should get highest rating if participants correctly understand false beliefs. ✔ : box that should be rated higher ratings if participants are partially passing false belief and getting distracted by the old location. ✖ : box that should be rated higher if participants are distracted by new information. 53 Figure 2.3 Least Squares Mean ratings of the likelihood Juan will look in each box across samples and knowledge conditions. 95% CI Error bars. Higher blue bars indicate students rate Juan as more likely to look in that box than Yasawans. ✔✔ : box that the ball was originally placed in. If participants correctly understand false beliefs, they should rate this box as the highest likelihood of being looked in. ✔ : Plausible Box; the ball was not in this box when Juan left, but when he returns it is now in the location in the room where the first box with the ball was. As the original location but not the original container, participants should rate this higher if they are partially passing false belief by keeping track of the old location in the room. ✖ : the Implausible Box; the ball was not in this box when Juan left, and it is in a different location in the room than the original box. Participants should rate this box that as more likely to be looked in if they are thinking less about Juan’s false beliefs and are getting more distracted by new information. No Info ✔✔ New Box, Old Location ✔✔ New Box, New Location ✔✔ No Info New Box, Old Location ✔ New Box, New Location No Info New Box, Old Location New Box, New Location ✖ No Info New Box, Old Location New Box, New Location Old Box, New Location (Blue Box) New Box, Old Location (Red Box) New Box, New Location (Purple Box) Distractor (Green Box) 0 1 2 3 4 5 6 7 Likelihood Rating Students Yasawa 54 2.3.3 Discussion In study 1, we find evidence that Yasawans and North American university students respond to false belief information differently. Though university students are still swayed by new information that the object is in the old location but a new container – information that the character with the false belief could not know - Yasawans report lower expectations that Juan will look in the original container and are more likely to be swayed by new information about an object’s true location. Though this may indicate a genuine difference in false belief processing, it may also reflect some differences in the ways that students vs. Yasawans respond to the demands of the task. For example, because Yasawan culture is more strictly hierarchical, Yasawans may be deferring more to the experimenter out of respect for the experimenter. To more directly target what aspects of mind Yasawans may be focusing differently on minds as compared to other cultural groups, we next use a self-report measure that taps into multiple aspects of mental state reasoning. If Yasawan culture is genuinely supressing thoughts about thoughts, only theory of mind aspects of mental state reasoning should be affected; thoughts about emotions or social situations should not be markedly less endorsed in Yasawa than among other groups. 2.4 Study 2: Measuring Empathy Across Cultures We use the EQ short (Wakabayashi, Baron-Cohen, Wheelwright, et al., 2006b) to compare how populations with and without normative constraints on mentalizing approach understanding others’ actions as a function of their mental states like emotions and thoughts. Yasawans, as a cultural group with opacity of mind norms, are of primary interest. We compare Yasawans’ self-reported mentalizing to self-reports from Indo-Fijians and North Americans. Indo-Fijians are an interesting comparison sample because they share many of the same institutional structures and constraints as native Fijians, but with differing traditional social 55 hierarchies and norms. We predict that Indo-Fijians will differ from Yasawans by self-reporting more focus on internal mental states than Yasawans. This contrast can provide some evidence for how the specifics of cultural norms, even when the overarching national social instructions are the same, might produce marked difference in the ways people view the world. 2.4.1 Method As with study 1, data for study 2 in this chapter was collected in conjunction with data reported in chapter 3 and two additional projects on beliefs about supernatural agents’ minds and moral concerns in different cultural contexts and religious belief systems.5 The analyses in this chapter focus on how participants think about other human minds through mentalizing. 188.8.131.52 Participants Yasawans and Indo-Fijians in Lovu village near Nadi were asked to participate during the June-July 2012 field season. We recruited 60 Yasawans (34 women; 19-80 yrs., mean = 41.58; 5-15 yrs. formal education, mean = 9.12) and 89 Indo-Fijians (46 women; 17-65 yrs., mean = 33.91; 7-17 yrs. formal education, mean = 11.82). From September 2012 to January 2013, we recruited a further 205 (172 women; 17-32 yrs., mean = 19.85; 12-17 yrs. formal education, mean = 12.90) university students in Vancouver, B.C. and 205 (99 women; 18-72 yrs., mean = 33.88; 10-21 yrs. formal education, mean = 14.38) American adults were recruited on Amazon’s Mechanical Turk. Yasawans’ participation was strictly voluntary, Indo-Fijians completed measures along with a longer interview and were given $1 FJD for their time, students were 5 Participants’ beliefs about supernatural agents’ minds and moral judgments are reported in McNamara, Willard, and Norenzayan (n.d.). Data on how humans perceive the agentic vs. experiential capacities of supernatural and human minds are reported in Willard & McNamara (n.d.). Both of these manuscripts are available from McNamara by request. 56 remunerated with course credit, and Mturk workers were paid $1.50 for 30 minutes of answering survey questions. 184.108.40.206 Materials Because the vast majority of psychological research to date has focused on people from Western, Educated, Industrialized, Rich, Democratic (WEIRD) societies (J. Henrich et al., 2010), we use psychological tools designed to measure how people from these cultures understand minds. In this study, we focus on the short version of the empathy quotient (EQ, Baron-Cohen & Wheelwright, 2004; Wakabayashi, Baron-Cohen, Wheelwright, et al., 2006b). Baron-Cohen & Wheelwright (2004) first introduced the EQ as a 40-item scale used to measure how good people are at understanding certain aspects of other’s minds – specifically, knowledge of socially appropriate behaviour and ability to infer other’s emotions and feelings – that are often deficient among people with autism spectrum disorders. Items are worded to either provoke agreement or disagreement; each item is answered on a 4-point likert scale. When scoring the EQ items, disagreement items are reverse coded, and answers that equate to a strong empathizing response (strongly agree on agreement items, strongly disagree on disagreement items) are given two points; weaker empathizing answers are given 1 point, and all other responses are given zero points. In looking for ways to shorten the EQ, Wakabayashi and colleagues (2006) found evidence for a single factor underlying the EQ scale and identified 22 items (EQ-short scale) that loaded at 0.40 or higher in a principle components analysis. Though the EQ-short was originally found to have a single component in the original short scale development data, subsequent studies on the psychometric properties of the long version of the EQ suggest a 3-factor structure in both English-speaking (Allison, Baron-Cohen, Wheelwright, 57 Stone, & Muncer, 2011; Lawrence, Shaw, Baker, Baron-Cohen, & David, 2004; Muncer & Ling, 2006) and French-speaking samples (Berthoz et al., 2008). 220.127.116.11 Procedure Fluent, bi-lingual research assistants translated and back-translated the scales from the original English into Standard Fijian or Hindi on-site. The Indo-Fijian participants were mostly interviewed in English, but Hindi translations were available in case of confusion. Yasawan participants were not sufficiently familiar with English, but were fluent in Standard Fijian. Yasawan and Indo-Fijian participants were then interviewed in person by a research assistant (Indigenous Fijian in Yasawa, Indo-Fijian or Indigenous Fijian among Indo-Fijians) and indicated their answers to questions on printed sheet showing a 1-4 number line (the Yasawan sample further had smiley faces to help orient less numerate villagers to the correct portions of the scale). The Standard Fijian lexicon is much smaller than English, so translations, even by our research assistant translators, were difficult. The research assistant who interviewed our Yasawan sample kept an English version with him to help explain concepts in the items if participant were ever confused about a question. Due to time constraints, Yasawans received the EQ separately from any other questionnaire, but the Indo-Fijians, students, and Mturk samples all answered the EQ along with a battery of other questionnaires. All North American participants responded in private on computers. 2.4.2 Results To determine what the EQ-short can reveal about how our samples approach others’ minds, we first look at the scale’s internal consistency, via Cronbach’s !, to get a sense of how 58 well the scale appears to be measuring the same thing across our samples. Because our goal is to identify underlying causal structures that may be producing particular patterns of answers to the EQ items, we begin with an exploratory factor analytic (EFA) approach (Floyd & Widaman, 1995). We fit the EFA model to the full cross-cultural data set in R using the psych package (Revelle, 2011b). We first determine the number of factors that best predicts the patterns of answers and settle upon 3 factors as sufficient to describe any possible universal pattern in the data across samples. These factors include items for 1) perceptions of others’ internal states, 2) interest in connecting emotionally with others, and 3) comfort with social situations. We next attempt to confirm the stricter assumptions of the measurement model implied by our three-factor EFA results by constructing a confirmatory factor analytic (CFA) model and run it on the Mturk data in R using the lavaan (Rosseel, 2012) and semTools packages (Pornprasertmanit, Miller, Schoemann, & Rosseel, 2013). By applying the stricter assumptions of the CFA model using the structure implied by the EFA results, we get a more precise estimate of how the items are clustering together around possible latent commonalities among them. However, exact fit from CFA models based on less restrictive EFA models – even when re-run on the same data the EFA model was derived from – are rare (van Prooijen & Van der Kloot, 2001). We find a CFA model that fits the Mturk data with sufficient evidence for fit (note previous CFA approaches to EQ factors have trouble achieving exact fit, see: Allison et al., 2011; Lawrence et al., 2004; Muncer & Ling, 2006). Because our end-goal is to make claims about group-level differences, we then first test whether our three-factor model as derived from the full data also applies to the each sample’s data. We use group measurement invariance analysis to simultaneously check our model’s exact and approximate fit to each sample’s data 59 (Meredith, 1993; Milfont & Fischer, 2010). We find evidence for partial strong invariance, with the three-factor structure, loadings, and intercepts for all but 3 items showing partial measurement invariance. We take this as evidence that our measurement model can indeed indicate meaningful differences in EQ performance across our samples. We look across the factors to find that Yasawans do self-report less fluency in thinking about other’s internal mental states like thoughts, as we expect given their Opacity of Mind norms. Further, we find the Indo-Fijians show a greater discomfort with social situations that is distinct from Yasawans but similar to other work among mainland Indians. This further indicates the Opacity of Mind effects we see in the Yasawans may be due to the specific, non-shared norms they have separate from even other groups within the same country. 18.104.22.168 Cross-Cultural Measurement Model of EQ Answers Using the psych (Revelle, 2011a) package for R (R Development Core Team, 2008), we find some preliminary indication that the scale does not work as well outside of North America: estimates of reliability using Cronbach’s alpha showed good internal reliability for North American Mturk workers (Mturkers: α=0.91), students (α=0.86), and Indo-Fijian samples (α=0.75). However, the Yasawan sample showed only moderate reliability (α=0.67). Further, many of the EQ items appear to correlate negatively with other items for both the Indo-Fijian and Yasawan samples. 22.214.171.124.1 EQ Exploratory Factor Analysis With this indication that the EQ may not be equally reliable across groups, we next ran exploratory factor analyses. Parallel analysis (Horn, 1965) and minimum average partial (MAP) tests (Velicer, 1976) across all four samples suggest between two and four factors, so we proceed 60 to run an EFA with 3 factors. We extract our initial 3-factor model based on the on the full cross-cultural dataset. This gives us both the best sample size to fit our models to and the best likelihood of finding a model that reasonably satisfies the patterns in answers across all samples. We use oblimin rotation to allow for correlations among factors because all factors should be related via underlying mentalizing ability. We find the three factors account for 44% of the overall variance, with factor 1 accounting for 14%, factor 2 10%, and factor 3 09%. Items that fall into each factor are shown in Table 2.2. Items appear to cluster into factors relating to ability to interpret others’ internal states, desire to care for others, and ability to navigate social situations. Factor 1 is most strongly correlated with factor 2 (ρ = 0.46); factors 1 and 2 are both relatively less correlated with factor 3 (Factor 1 ρ = 0.23, Factor 2 ρ = 0.22). This suggests that ability to interpret internal states and caring about others’ emotions are highly relevant to each other, but have relatively less to do with social graces. We retain the oblimin solution due to the high correlations between factors 1 and 2. 61 EQ Item Factor Standardized Loadings Factor 1: Ability to predict and interpret others' internal states 16 I can sense if I am intruding, even if the other person doesn’t tell me 0.65 14 I can easily tell if someone else is interested or bored with what I am saying 0.65 20 I can tell if someone is masking their true emotion 0.6 6 I can pick up quickly if someone says one thing but means another 0.54 10 I am quick to spot when someone in a group is feeling awkward or uncomfortable 0.52 1 I can easily tell if someone else wants to enter a conversation 0.48 19 I can easily work out what another person might want to talk about 0.4 21 I am good at predicting what someone will do 0.39 9 I am good at predicting how someone will feel 0.33 18 I can tune into how someone else feels rapidly and intuitively 0.3 13 Other people tell me I am good at understanding how they are feeling and what they are thinking 0.27 Factor 2: Interest in caring for others and connecting to others' emotions 2 I really enjoy caring for other people 0.58 13 Other people tell me I am good at understanding how they are feeling and what they are thinking 0.5 18 I can tune into how someone else feels rapidly and intuitively 0.44 15 Friends usually talk to me about their problems as they say that I am very understanding 0.42 9 I am good at predicting how someone will feel 0.37 22 I tend to get emotionally involved with a friend's problems 0.35 12 I don't tend to find social situations confusing 0.32 8 I find it easy to put myself in somebody else's shoes 0.31 19 I can easily work out what another person might want to talk about 0.3 1 I can easily tell if someone else wants to enter a conversation 0.13 Factor 3: Ability to navigate social situations/ social graces 4 I often find it difficult to judge if something is rude or polite 0.63 17 Other people often say that I am insensitive, though I don’t always see why 0.57 7 It is hard for me to see why some things upset people so much 0.55 11 I can’t always see why someone should have felt offended by a remark 0.51 5 In a conversation, I tend to focus on my own thoughts rather than on what my listener might be thinking 0.44 3 I find it hard to know what to do in a social situation 0.3 21 I am good at predicting what someone will do -0.08 1 I can easily tell if someone else wants to enter a conversation 0.08 Table 2.2 EQ-short 3-factor EFA with oblimin rotation on data across all cultures shows factors for understanding others’ internal states, interest in caring for others, and ability to navigate social situations. Items are retained in factors if their loadings absolute values are ≥ |0.27|. We also report additional cross-loadings with item 1 in Factor 2 and items 21 and 1 in Factor 2, as suggested in subsequent CFA analysis. 62 126.96.36.199.2 CFA Measurement Model for EQ Data We next use the EFA output to inform a 3-factor confirmatory factor analysis (CFA) model that we run on the same Mturk data using lavaan (Rosseel, 2012; see Appendix A for covariance matrices). We first test the more stringent CFA fit of our initial 3-factor EFA model on the on the Mturk sample because their Cronbach’s α indicated the this sample showed the best internal consistency on EQ items and because they have the highest population-level similarities to the initial populations used to design the measure (Baron-Cohen & Wheelwright, 2004; Wakabayashi, Baron-Cohen, Uchiyama, et al., 2006a; Wakabayashi, Baron-Cohen, Wheelwright, et al., 2006b). Therefore, if we are going to find an underlying factor structure that best fits the understanding of other minds that the EQ is designed to capture, then the Mturk sample is our best bet. Due to small sample sizes within each cultural group (especially our two Fijian samples), we use weighted least squares means and variance adjusted (WLSMV) estimation fit in the lavaan package (Rosseel, 2012). WLSMV estimation is more stable under small sample sizes and simultaneously helps account for possible abnormalities in standard errors calculated on small sample sizes (Beauducel & Herzberg, 2006; Hox, Maas, & Brinkhuis, 2010; Rhemtulla, Brosseau-Liard, & Savalei, 2012). We use the chi-square test scaling methods suggested in Satorra and Bentler (2010) to account for these mean and variance-robust adjustments in the WLSMV estimates. Our first CFA model includes correlations among all factors and loadings of each item onto factors as outlined in Table 2.2. This model did not fit the data (χ2(201)=255.28, p=0.006), a common problem EFA result replications face – even on the exact same data (van Prooijen & Van der Kloot, 2001). Approximate fit indices suggest the model mis-fit to the data is higher than desired, but not extreme (SRMR = 0.06, RMSEA = 0.04 [0.02, 0.05], CFI = 0.94). Tests of 63 close fit and not-close fit find that the largest residuals (Root Mean Square Error of Approximation, RMSEA) are not larger than 0.05 (p = 1.00) and significantly smaller than 0.07 (p < 0.001) – both show that the model fit is not exact, but it is close (MacCallum, Browne, & Sugawara, 1996). We run Lagrange multiplier tests to determine if any modifications might significantly improve the model. We find that adding cross-loadings between Factor 3 and EQ items 1 and 21, a cross-loading between item 1 and Factor 2, and an additional correlated residual between items 3 and 12 (items that both include the phrase “social situations”) show the highest modification indices (Whittaker, 2012), so we added these parameters for our second model. Our additions to the second model increase the fit to the data so much that we fail to find evidence of misfit to the Mturk sample data in our WLSMV-adjusted chi-square test (χ2(197)= 215.11, p=0.18). The approximate fit indices show this fit is close (SRMR = 0.05, RMSEA = 0.02 [0.00, 0.04], CFI = 0.98). RMSEA tests of close fit and not-close fit further show the second model’s RMSEA is not significantly greater than 0.05 (p = 1.00), and is significantly smaller than 0.07 (p < 0.001). The second model’s additional parameters do present a significant improvement in the model fit (Δχ2(4) = 31.40, p > 0.001), so we use this model for our invariance analysis. 188.8.131.52.3 Multiple-Group Measurement Invariance Analysis For our first step in our invariance analysis, we confirm that data from each sample are sufficiently different that pooling the samples together is inappropriate (see Table 2.3). 64 χ2 (df) Variances Means Omnibus 1740.17 (759)*** 305.71 (66)*** North American vs. Fijian 989.83 (253)*** 127.60 (22)*** Yasawan vs. Indo-Fijian 436.00 (253)*** 111.04 (22)*** Mturk vs. Students 366.81 (253)*** 99.79 (22)*** Significance codes: ***<0.001, **<0.01, *<0.05, †<0.1 Table 2.3 Significant variance and mean differences across samples show pooling is not appropriate. Since we cannot reasonably pool the samples together based on similarity in variances or means, we next look for measurement invariance across our samples to determine if their responses to the EQ measurement model applies to each sample in equivalent ways (Hirschfeld & Brachel, 2014; Milfont & Fischer, 2010; Steenkamp & Baumgartner, 1998). Fit indices for various levels of invariance are shown in Table 2.4. We fail to find exact fit for configural invariance when robust means and variances adjustments are applied to the chi-square estimates, but the approximate fit indices and test of close and not-close fit suggest the mis-fit is not severe. We therefore carry on to test for metric invariance. We next check metric invariance by constraining all groups to have the same factor loadings. If all samples show the same loadings onto factors, then we have evidence that participants are responding to the items in approximately the same way across all samples, and, therefore, the items are reflective of the underlying constructs in equivalent ways. We find similar approximate fit, close fit test of RMSEA value as ≤ 0.05, (p = 1) and insufficient evidence for not-close fit (RMSEA ≥ 0.07, p < 0.001). Tests of differences in the unadjusted chi-square tests show that adding constraints of equal loadings does not significantly worsen the exact fit (Δχ2(90) = 96.72, p = 0.30). This test of exact-fit differences we have sufficient evidence for metric invariance across samples. 65 We next check for scalar invariance by constraining EQ item intercepts to be equal across samples. Establishing scalar invariance is important for confirming that individuals who score the same values on the latent constructs would also score the same on the observed items measured. If the samples showed evidence of the same item intercepts, then we can conclude that all samples’ average values on the underlying constructs are approximately equal. Establishing scalar invariance is the bare minimum necessary to continue on to compare populations on factor means (Steenkamp & Baumgartner, 1998). Our approximate fit indices are again roughly similar to our configural and metric invariance models, but the exact fit gets significantly worse between constraining the loadings to be the same and adding the item intercepts constraints (Δχ2(57) = 99.72, p < 0.001.). We therefore conclude that we have insufficient evidence for full measurement invariance, with configural and metric invariance holding but not scalar invariance. We test for partial scalar invariance by assessing univariate score tests for which intercepts should be freed to improve model fit (Bentler & Chou, 1992; Kaplan & Wenger, 1993). We find that releasing equality constraints on intercepts for items 2, 21, and 22 would most improve model fit (see Appendix A for details on changes to the model for each released constraint). After releasing these constraints, we find that the chi-square difference test between weak and partial scalar invariance models is no longer significant. From this, we conclude we have sufficient evidence for partial scalar invariance, which can allow us to compare samples on factor means. 66 EFA to CFA DWLS χ2 (df) MV Robust χ2 (df) SRMR RMSEA [.95 CI] CFI Mturk model 1 126.03 (201) 255.28 ** (201) 0.06 0.04 [0.02, 0.05] 0.94 Mturk model 2 94.63 (197) 215.11 (197) 0.05 0.02 [0.00, 0.04] 0.98 Δχ2 Model 1 - Model 2 21.40 *** (4) Group Invariance Configural 784.03 (788) 996.12 *** (788) 0.07 0.05 [0.04, 0.05] 0.91 Metric (Loadings) 1302.47 *** (878) 1195.27 *** (878) 0.09 0.05 [0.05, 0.06] 0.86 Δχ2 Configural - Metric 96.72 (90) Scalar (Intercepts) 1535.59 (935) 1346.59 *** (935) 0.11 0.06 [0.05, 0.07] 0.79 Δχ2 Metric - Scalar 99.44 *** (57) Partial Scalar (Intercepts) 1437.63 *** (926) 1290.35 *** (926) 0.10 0.06 [0.05, 0.06] 0.84 Δχ2 Metric - Partial Scalar 58.61 (48) Strict (Residuals) 1890.97 *** (992) 1536.79 *** (992) 0.11 0.07 [0.06, 0.07] 0.76 Δχ2 Partial Scalar - Strict 151.03 *** (66) Significance codes: ***<0.001, **<0.01, *<0.05, †<0.1 Table 2.4 We use the loading structure found in our EFA on all data to fit CFA models 1 and 2 to Mturk sample data. Chi-square exact fit reported from DWLS and Model 2 adds a residual correlation cross loadings between item 1 on Factor 2, items 1 and 21 on Factor 3, and a residual correlation between items 3 and 12. We compare models using Satorra & Bentler (2010) chi-square adjustments. Our configural model on all four samples provides an approximate but not exact fit; we find evidence for configural and metric invariance, but must relax constraints on intercepts of items 1, 21, and 22 to find partial scalar invariance. This is sufficient evidence to compare samples on factor means. We find that the error variance on the residuals does not support strict invariance. Given that we suspect Yasawans may be less habitually inclined to think about thoughts than other groups, the possibility that the samples do not have the same average underlying levels of reported fluency in each of these three mentalizing domains would not be surprising. 67 We continue on from our partial scalar invariance to check for partial strict invariance with constraints on residuals being equal across groups. A significant chi-square test indicates the strict model significantly worsens fit, so we conclude that only partial strong invariance is supported. 184.108.40.206 Group-Level EQ Differences Having established that our four samples show sufficient measurement invariance, we now look to see how groups differ on observed means of EQ item scores within each factor. We run four linear regressions (EQ total scores and total item scores for each factor), each taking the Mturk sample and women as the comparison groups. We also check for interactions between sample and sex. Across samples, women score higher on all EQ items combined (F(1, 516) = 4.52, p = 0.03; see for example (Baron-Cohen, 2002), but there are no significant overall differences across the samples (F(3, 516) = 1.73, p = 0.16), and the differences between women and men do not significantly differ across samples (F(3, 516) = 0.55, p = 0.65). We compare each sample against the Mturkers and find Yasawans score the lowest overall. Yasawans, as a group, have significantly lower total EQ scores than Mturkers (b = -2.26, CI.95 [-4.37, -0.14], p = 0.04), but this difference is no longer significant if we allow samples to differ in how much higher women score on the EQ (Yasawan women score an average of 1.91 points lower than Mturk women: CI.95 [-4.76, 0.95], p = 0.19; Yasawan men score an average of 2.58 points lower than Mturk men: CI.95 [-5.74, 0.58], p = 0.11). So if we take all of the samples and ignore the different groups they come from, it looks as though Yasawans report lower EQ mentalizing. However, if we allow for women and men to differ in unique ways across the samples, then we see that the biggest differences are between Yasawan women vs. other groups; Yasawan men cluster together with Canadian students and Indo-Fijians (see Figure 2.4). 68 Figure 2.4 Group differences with SE error bars showing scores on all 22 EQ items A) without a sample x sex interaction and B) with the interaction. Though there is no significant overall difference in sex effect across samples, allowing for this difference in sex effects highlights the difference in size of the sex differences across samples. Error bars show standard errors. We next turn to the three factors to see whether different portions of the EQ can reveal how each group tends to approach other minds. Because the interaction terms between sex and sample were not significant for any factor, we use an effect code for sex (women coded as +0.5, men coded as -0.5) to account for sex differences and report the average values across women and men in each sample. Figure 2.5 shows plots of these factor means across samples. Men and women as a whole do not significantly differ on their scores for the factor containing items relating to inferring others’ internal mental states, (F(1, 535) = 0.85, p = 0.36), but there is a significant difference across samples (F(3, 535) = 9.25, p < 0.001). This difference is driven by the Yasawans, who score an average of 2.26 points lower than Mturkers (CI.95 [-4.44 -1.82], p < 0.001). This difference in tendency to think about the inner workings of other’s minds is what we would expect in a culture that has norms proscribing inference about others’ inner mental states. The scores on the factor containing items about emotions show no sex or sample differences. The factor containing questions about understanding social situations shows a significant overall sex 16 18 20 22 24 Women Men Item Score Sum EQ Scale Total without Sex x Sample interaction Mturk Students Indo-Fijian Yasawa A16 18 20 22 24 Women Men Item Score Sum EQ Scale Total with Sex x Sample Interaction Mturk Students Indo-Fijian Yasawa B 69 difference (F(1, 533) = 8.61, p = 0.003) with women reporting more comfort in navigating social situations. Samples are also significantly different on the social situations factor (F(3, 533) = 15.38, p < 0.001). This difference driven by the Indo-Fijians, who score an average of 2.26 points lower than Mturkers (CI.95 [-2.89, -1.44], p < 0.001). This difference may be in part due to the highly competitive within-group dynamics seen in other studies of in-group favouritism using economic game measures (Willard, 2016). Figure 2.5 Mean scores on EQ factors accounting for sex differences across samples. A) Yasawans score the lowest on factor 1 (ability to interpret others’ internal mental states) while B) there are no group differences in factor 2 (connecting with others emotions) and C) Indo-Fijians score lowest on factor 3 (navigating social situations). Error bars show standard errors. 2.4.3 Linking False Belief and Empathy We next investigate how the EQ factor scores predict participants’ box ratings across samples and knowledge conditions. We allow likelihood ratings of each box to be differently affected by EQ factors by adding an interaction with box for each factor (Box x EQ factor 1 + Box x EQ factor 2 + Box x EQ factor 3). We use the Aiken and West (1991) simple slopes 7 9 11 13 Item Score Sum A EQ Mental States 8 9 10 11 12 Item Score Sum B EQ Emotions 4 5 6 7 8 Item Score Sum C EQ Social Situations 70 method to look at how each EQ factor predicts box ratings. By shifting the box we use as the reference group in our Box categorical variable, we can look at the effect the EQ factors have on each box’s ratings. Because false beliefs are internal, unseen mental states, we would expect the factor 1 internal mental states items to have the strongest relationship with curse of knowledge box ratings. Table 2.5 shows our simples slopes results. Boxes differ by Condition & Sample (Box x Knowledge Condition x Sample) Ignoring Sample differences (Box x Knowledge Condition) Boxes differ by Condition & Sample, Samples separated All participants (n=247) All participants (n=247) Yasawa (n=52) Students (n=195) Box EQ factor b (SE) b (SE) b (SE) b (SE) Old box, New location (blue) Internal mental -0.02 (0.03) 0.08 (0.03)* -0.19 (0.08)* 0.00 (0.04) Emotions 0.00 (0.05) -0.09 (0.05)† 0.02 (0.09) 0.00 (0.05) Social situations 0.03 (0.05) -0.02 (0.05) -0.01 (0.10) 0.04 (0.06) New box, Old location (red) Internal mental -0.01 (0.03) 0.01 (0.03) 0.03 (0.08) -0.02 (0.04) Emotions 0.06 (0.05) 0.04 (0.05) 0.12 (0.09) 0.04 (0.05) Social situations -0.06 (0.05) -0.06 (0.05) -0.18 (0.10)† -0.02 (0.06) New box, New location (purple) Internal mental 0.01 (0.03) -0.05 (0.03) 0.03 (0.08) 0.02 (0.04) Emotions 0.04 (0.05) 0.10 (0.05)* 0.06 (0.09) 0.03 (0.05) Social situations -0.11 (0.05)* -0.09 (0.05)† -0.08 (0.10) -0.12 (0.06)* Distractor (green) Internal mental 0.06 (0.03)† 0.00 (0.03) 0.04 (0.08) 0.04 (0.04) Emotions 0.01 (0.05) 0.06 (0.05) -0.12 (0.09) 0.05 (0.05) Social situations -0.14 (0.05)** -0.11 (0.05)* -0.14 (0.10) -0.14 (0.06)* Significance codes: ***<0.001, **<0.01, *<0.05, †<0.1 Table 2.5 Simple slopes of EQ factors predicting box ratings for Box x EQ factors interactions. Internal mental states items predict false belief as expected only when ignoring overall sample differences. Within Yasawa, internal mental states factor predicts lower ratings of correct false belief box. Across samples, the EQ social situations factor predicts lower ratings of the wrong boxes. Higher scores on the EQ internal mental states items only predict higher original box ratings when we ignore the overall differences between the samples; students’ overall higher self-reported tendency to think about mental states seems to explain this cultural difference. Within the student sample, EQ does not predict ratings of the correct original box, likely due to a ceiling effect on their ratings (6 or 7 out of 7, see Figure 2.3). However, Yasawans with higher 71 scores on the EQ internal mental state items rate that Juan is significantly less likely to look in the original blue box (p = 0.03). Higher scores on the EQ factor 2 items measuring interest in others’ emotions, predicts higher ratings of the wrong (Purple new box, new location) box when we ignore the sample differences. We see this effect is strongest in Yasawa, though it is not significant (p = 0.22). In general, higher social skills in EQ factor 3 also predict that participants in both samples will give lower ratings to the wrong boxes (any box other than the original blue box). This is especially true for the new box (red), new location (purple) and distractor (green) boxes, which may indicate that social skills can also help navigate away from distracting wrong information. Except in the plausible knowledge condition (when they are given distracting new information about where the ball moved to), the Yasawans rate the new box in the old location (red box) lower than students. Yasawans appear to be more influenced by location information but less oriented to the ball’s original location than the student sample. 2.4.4 Discussion What does study 2 tell us about how these cultures approach minds? The EQ-short is an imperfect tool, but multiple group measurement invariance analysis shows that it fits reasonably well to our diverse samples. We find three underlying factors to be driving EQ answers; our three-factor model corroborates previous work on this scale in the UK, where the scale was originally developed (Allison et al., 2011; Lawrence et al., 2004; Muncer & Ling, 2006). As an added bonus, we find these factors underlying the much shorter EQ-short (the full EQ is 40 questions with 20 distractors, the EQ-short is only 22 items). This shorter version is much easier to implement, especially in field settings. 72 We take pains to ensure that our measurement model is an accurate reflection of the data from all four samples. Though previous studies on the psychometric properties of the EQ do look for the latent causal structure behind the measure and identify similar factors, they do not proceed with group measurement invariance analysis to confirm that the measure applies equally well to samples with and without autism spectrum disorder diagnoses. Studies that seek to validate the EQ outside English-speaking populations rarely include explicit cross-cultural comparisons, but instead focus on one cultural group at a time (Berthoz et al., 2008; Ruta, Mazzone, Mazzone, Wheelwright, & Baron-Cohen, 2011; Wakabayashi, Baron-Cohen, Uchiyama, et al., 2006a). Therefore, we cannot be sure from study to study whether there are any group-level cultural differences. The few studies that do make explicit cross cultural comparisons on either the EQ or other measures that tap similar constructs like the autism quotient (Baron-Cohen et al., 2001) or emotional intelligence (Salovey & Mayer, 1990) often do not take the extra step to ensure the scale is measuring mentalizing constructs in the same way across groups (Freeth et al., 2013; Sharma et al., 2009). Without ensuring there is sufficient evidence that each group is independently approaching the measure in at least approximately the same way, we cannot be sure whether the group-level differences that might arise are due to legitimate differences in the way each group thinks about other minds, or whether they are just understanding the questions differently (Milfont & Fischer, 2010). Importantly, the factor measuring ability to understand other’s internal mental states shows Yasawans to be significantly lower on just this facet of ToM. We would expect this pattern given their cultural background of Opacity of Mind norms. Previous work on the autism quotient across cultures shows students from mainland India performing more poorly on the social skills sub-scale compared with students from the UK (Freeth et al., 2013), which 73 corroborates our finding that Indo-Fijians also feel less comfortable in social situations. This supports our preliminary predictions that Indo-Fijians may be more like Indians than Yasawans, even though they both live in the same country, with the same legal system and same institutional controls. This further speaks to the importance of non-shared norms within cultural groups for shaping the ways that people understand others’ minds. 2.5 General Discussion We show that cultural norms that proscribe inference about others internal thoughts, beliefs, and motivations leads to group-level difference in self-reported fluency in thinking about others’ thoughts. The different emphases on minds appears to drive lower expectations that a character will look in the correct false belief location on a first-order false belief task. Further, this also leads to lower self-reported ability to infer other’s internal mental states, but not their ability to understand social situations or tune in to other’s emotions. Importantly, we show that the restricted focus on mentalizing processes that we find among our Yasawan participants applies only to thinking about internal mental states, but not emotions or social situations. Populations with restricted mentalizing abilities, including those on the autism spectrum, tend to self-report less facility with thinking about all three facets of mentalizing. Therefore, we suggest that these cultural differences in mentalizing performance arise from greater focus on the social and interpersonal, rather than internal, aspects of mind. We further suggest that that this differential focus stems from cultural exposure to particular norms that reduce belief about how knowable other minds are, which may lead to differences in how people in different cultural groups think about minds. We suggest that cultures with norms that restrict beliefs about how much one can know about the contents of another’s mind do not make people with these norms less able to think about minds. Rather, cultural exposure to these norms 74 lead people in these cultures to focus on different aspects of others’ minds, resulting in less apparent facility in processing others’ internal mental states. 2.5.1 Implications and Future Directions Because the EQ items that focused on emotion showed approximately equal and high ratings across all groups, our results suggest that the mental processes involved in inferring other’s emotions may be more equally accessible in these cultural groups. At the same time, the tendency to infer knowledge states and thoughts may be more restricted to a functional universal (Norenzayan & Heine, 2005). The present research further suggests that different facets of theory of mind may interact with each other differently in different cultures to lead people to similar social behaviours. For example, though group-level differences in tendency to think about internal mental states explains much of the difference in false belief performance between North American students and Yasawans, we also find evidence that social skills may help orient people away from the wrong answers. These social skills may also be a proxy for comfort and fluency with understanding appropriate social behaviour, ability to tactfully navigate social situations, and greater ability to read social situations for cues to how to respond that do not require mental state inference. Therefore, greater competency in reading the situation, provided that situations are sufficiently constraining on behaviour, may help people get to correct mentalizing solutions without referencing mental states (Walker, Smith, & Vul, 2015). Further research on this possible mentalizing/ reading the situation trade-off is needed to explore this possibility. Though neither the Curse of Knowledge task nor the EQ were designed for non-Western cultural groups, they can corroborate more qualitative, ethnographic accounts of cross-cultural differences in how people try to understand other minds. Though we take pains to establish that participants in each cultural group were approaching the questions in the same way, the fact that 75 these tools were not designed with non-Western populations in mind may still present a hurdle to accurate measurement of precisely how people in these cultures think about other minds. Future studies that explicitly start from how non-Western groups approach minds, perhaps as a function of the social networks and relational self, may more accurately capture the nuances of how thinking about minds shapes human behaviour in diverse cultural contexts. The present research also cannot inform us about how these differences arise either in development or in the cultural history of different groups. Though we show evidence for differences among adults, these data cannot distinguish whether this cultural difference arises early in development or comes later in life as children are socialized into fully enculturated adults. It may be that mentalizing abilities are relatively consistent across cultures in early childhood, but are then modulated by cultural forces later in life. Or, these differences in tendency to think about minds may be present from the earliest social interactions children are exposed to. The present data also do not tell us about how the broader cultural trends in norms that might be shifting people’s perceptions of other minds come to be. One suggestion is more stringent constraints on behaviour that make situations more informative about one’s motivation than their true goals and desires. Even further than that, this does not tell us why these normative differences exist, or what they do in the cultural contexts where the have arisen and persist. Future research into how these norms perpetuate groups in various cultural contexts is needed to explore these possible developmental and social origins behind the observed cultural differences we see today. 76 Chapter 3. Weighing Outcome vs. Intent in Moral Matters: Priming Thoughts Makes Opaque Minds more Transparent 3.1 Introduction In the span of two and a half months, two boys’ lives – 18-year-old Michael Brown in Ferguson, Missouri, and 17-year-old Laquan McDonald in Chicago, Illinois – ended at the hands of law enforcement officials. Shortly thereafter, two court proceedings, separated by approximately 300 miles and a little more than 365 days, hinged upon what two groups of men and women decided was going on in those police officers’ heads. That the fatal bullets came from Darren Wilson’s and Jason Van Dyke’s guns was not in dispute. Instead, the key question for these jurors was of intent: did the officers have reasonable fear for their own lives to warrant using fatal force – or – did they willfully continue shooting with the intent of ending these boys’ lives? For people living in Western, Educated, Industrialized, Rich, and Democratic countries like the United States (J. Henrich et al., 2010), using intent to determine wrongness and punishment is so pervasive that the justification for taking all this extra time, effort, and expense to carry out such legal proceedings seems self-evident. But if we look past this cultural lens, it does become a bit more puzzling that in cases like these two, where the harmfulness of the outcomes and the persons who caused these outcomes were not in question, that inferences about these men’s mental states became so vital to judgments about these officers’ actions. Asking why we use mental states to judge behaviour is almost nonsensical in highly individualistic, Western cultures; the assurance that behaviour is indicative of a person’s unseen internal motives and dispositions is so deeply ingrained that it forms a fundamental error in the way that Westerners often explain what causes people to do various things (E. E. Jones & Harris, 77 1967; Ross, 1977). But this is not the case for all cultures. The fundamental error in over-attributing behaviours to dispositions instead of situations is not so fundamental in more community-oriented cultures (Miller, 1984; Lee, Hallahan & Herzog, 1996). More generally, the ways that people think about others’ minds are far less focused on internal states in all but the most individualistic societies, which further influences the ways that people across cultures broadcast their internal mental states through behaviours like emotional displays (Lillard, 1998; Matsumoto et al., 1998; 2008). This cross-cultural variation in approaches taken to other minds also shows up in how young children develop these abilities across cultures (H. C. Barrett, Broesch, Scott, He, Baillargeon, Di Wu, et al., 2013a; Callaghan et al., 2005; Liu et al., 2008; A. Mayer & Trauble, 2012). Children living in more traditional and community-oriented groups often pass false belief tasks at a later age, but exposure to formal Westernized education and even more words for mental states is related to earlier development of mental state reasoning performance as measured by these tasks (Meristo et al., 2007; Pyers & Senghas, 2009; Vinden, 2002). With these cross-cultural differences in how people think about others’ minds, moral judgments that require inferring the actor’s intent also vary. Cultural groups that place greater emphasis on communal and relational values often focus less on intent (H. C. Barrett et al., 2016; V. L. Hamilton et al., 1983; Laurin & Plaks, 2014). This can even be seen across different religious groups within the same broader society. For example, as compared to more communal Catholics and Jews, Protestant Christians are more likely to consider forbidden thoughts as equally bad as forbidden action and are more likely to experience negative affect that must be redirected into productive, creative work after thinking taboo thoughts (A. B. Cohen & Rozin, 2001; T. R. Cohen, Montoya, & Insko, 2006; Cohen, Kim, & Hudson, 2014; Hudson & Cohen, 78 2016). When we consider different kinds of violations that might be moralized, more community-oriented, collectivistic groups also emphasize different moral domains that are more likely to support community in addition to individual integrity (Graham et al., 2011; J. G. Miller & Bersoff, 1992; Rozin et al., 1999; Shweder et al., 1997). The purity domain in particular is often more relevant in collectivistic than individualistic groups (Graham, Haidt, & Nosek, 2009; Haidt, 2012; Haidt & Graham, 2007; Horberg, Oveis, & Keltner, 2009; Inbar, Pizarro, & Bloom, 2009; Inbar, Pizarro, Iyer, & Haidt, 2012) and is also less likely to be judged based upon intent (L. Young & Saxe, 2011; L. Young & Tsoi, 2013). Over time, as people from groups with different levels of collectivism/ individualism mix, lateral cultural transmission via learning from peers may promote more intent-focused stance as they spend more time in the new, individualistic cultural group (Mesoudi, Magid, & Hussain, 2016). This chapter examines how moral judgments, as a sort of applied mental state reasoning task, can give us a window in to the ways that people in different cultural groups think about other minds. We further seek to show that these differences in focus on intent or outcome can be linked back to broader cultural norms that promote or prohibit focus on mental states as the source of individuals’ behaviours. By explaining these cross-cultural differences as a product of broader cultural norms, we can begin to further understand the subtle ways that culture interacts with basic cognitive mechanisms to shape individuals into cultural beings adapted to navigate and perpetuate our culture-specific social worlds. 3.1.1 Moralizing as Applied Mental State Reasoning For most people, thinking about other minds seems as automatic and reflexive as breathing. Despite how effortless this capacity may seem, its long pattern of development throughout early childhood, developmental conditions like autism and psychopathy that result in 79 limitations on these abilities, and the lack of clear evidence for any non-human animals showing the full suite of human cognitive mechanisms to think about other minds make it clear that these mentalizing capacities are fairly complex. Why then might humans have developed such a sophisticated social cognitive toolset? What does thinking about other minds buy us? As a fundamentally social species, this ability to infer other’s goals, motivations, thoughts, and knowledge states can help in coordinating action toward gaining and transmitting shared, accumulated knowledge that is at the core of human cumulative cultural learning (Legare & Nielsen, 2015; Lyons et al., 2011; Marsh et al., 2013; Shipton & Nielsen, 2015; Whiten et al., 2009). However, in order to keep these cooperative learning and teaching systems going, the wider cooperative groups must collectively value cooperation and have systems in place to curb cheaters from making the cooperative system collapse (Boyd & Richerson, 1992; J. Henrich, 2006; B. Herrmann, Thoni, & Gachter, 2008). Shared ideas about right and wrong encoded within a group’s social norms short-cut the cognitive effort needed to determine how to act in a given situation, how to determine if a rule violation occurred, and how to respond to a violation; these norms thus provide a highly effective set of cultural mechanisms for sustaining human social life (Andrighetto et al., 2013; Chudek & Henrich, 2011; Sripada & Stich, 2006). Morality is a subset of these social norms that delineate the group’s core values. Unlike formally codified rules encapsulated in legal systems, moral norms are the informal, often tacit rules that govern social life. Conventions are another set of tacit rules governing behaviour, but tend not to be as vital to the long-term stability of the group’s values. Therefore, breaking conventions tends not to warrant much punishment, but violations against moral norms often receive quick and severe reprisal. Judgments about the severity and response to moral norm violations, in turn, are often calculated based upon the mental state of the perpetrator. This deep 80 linkage between mental state reasoning and moral evaluation implies that mental state attribution may be the very core of moral reasoning (K. Gray et al., 2012). There is some evidence for distinct processes used in judgments about how bad an action was vs. how much punishment it deserves. One process is oriented toward causal attributions of responsibility and degree of violation (big infractions or tiny slip-ups), while the second process is oriented toward mental state information – a “whodunit” process and a “did they mean to” process (Cushman, 2008; 2015; Cushman, Sheketoff, Wharton, & Carey, 2013). For the purposes of isolating moral reasoning as a case of applied mentalizing, the “did they mean to” process is of most interest to the present research. Studies on the neurological basis of such moral decision-making routinely pin-point brain areas associated with inferring others’ mental states (Dodell-Feder, Koster-Hale, Bedny, & Saxe, 2011; Saxe, 2006; Saxe & Wexler, 2005; L. Young & Saxe, 2008; L. Young, Camprodon, Hauser, Pascual-Leone, & Saxe, 2010b; L. Young, Cushman, Hauser, & Saxe, 2007; L. Young, Scholz, & Saxe, 2011). Further, studies with various populations of neurologically atypical populations with specific damages and disorders to mental state reasoning further point to the importance of mentalizing and inferring intent (Koster-Hale, Saxe, Dungan, & Young, 2013; Moran et al., 2011; L. Young et al., 2012; L. Young, Bechara, Tranel, Damasio, Hauser, & Damasio, 2010a; L. Young, Camprodon, Hauser, Pascual-Leone, & Saxe, 2010b). Limited existing cross-cultural data on the neurological basis of moral cognition further indicates that theory of mind processing areas are involved in moral reasoning across cultural contexts, though different areas are emphasized in different cultural groups (R. B. Adams et al., 2010; Kobayashi Frank & Temple, 2009; Kobayashi, Glover, & Temple, 2007; Perner & Aichhorn, 2008). 81 Despite this strong evidence for mentalizing as a fundamental substrate of moral reasoning, specific aspects of violation judgments reveal that moral reasoning includes more than just mental state processing. Judgments of wrongness (in North American samples at least) seem to be scaled almost exclusively by mental state reasoning and focus on intent, but judgments about punishment rely on calculations of intent that are scaled by outcome domain (e.g. whether the perpetrator caused harm or violated purity norms) and scope of the violation (e.g. the difference between a slap and a stab; see: Cushman, 2008; Cushman et al., 2013; J. W. Martin & Cushman, 2015). The result of having two processes involved in these moral reasoning tasks is that mis-matches in intent and outcome, especially in accidents when the intent was neutral or positive but the outcome was negative, can receive more severe reactions than would be expected in a strictly intent-focused system (Costa, 2009; J. W. Martin & Cushman, 2015) This existing research provides a detailed map of moral cognition space in Western cultures. But does this map hold in places were the stakes of social interactions are stacked differently? Resent research in small-scale, traditional societies shows a wide range of variation in how people in these groups use intent to judge actions (H. C. Barrett et al., 2016). However, it is not yet clear whether this variation in intent focus is specifically related to differences in how people think about minds. If mental state reasoning is for coordinating social interactions, and if morality is for sustaining social structures within societies, then it would be reasonable to expect that mental state inference would have a very different emphasis in moral judgments made in different cultural groups. 3.1.2 Moralizing and Mentalizing in the Field: Research Sites in Fiji To examine how cross-cultural differences in thinking about mind influences different patterns of moral judgments, we recruit participation from people living in societies with social 82 norms that either emphasize or restrict interpreting others’ actions through the rubric of inferred mental states. We investigate differences in intent vs. outcome focus in three cultural groups: one WEIRD (North American adults and university students); one with norms that explicitly dissuade people from using anything but observable outer displays to explain behaviour (Indigenous Fijians living on Yasawa Island, Fiji, hereafter Yasawans); and another group that shares wider national institutions with Indigenous Fijians, is more community-oriented than WEIRD cultures, but normatively still values inferring desires and intentions to explain behaviour (or Fijians of Indian descent, hereafter Indo-Fijians). By including these three groups, we aim to demonstrate how wider cultural systems of norms can influence social decision-making by modulating how basic social cognitive processes operate. 220.127.116.11 Field Sites: Yasawa and Lovu Village, Fiji Fiji’s range of cultural diversity, deep traditions that link back to ways of life that were developed before contact with Europeans, and societal organization that places the majority of power and influence in the hands of non-WEIRD (J. Henrich et al., 2010) cultural groups makes Fiji an ideal location to investigate how culture modulates mentalizing and how this mentalizing influences moral decision-making. Of particular interest, Indigenous Fijians emphasize a more relational, group-oriented sense of self (Brison, 2001) and Opacity of Mind norms, or traditional norms that discourage referencing others’ thoughts and motivations to explain behaviour. These Opacity of Mind norms lead to lower tendencies to infer false beliefs and less of a tendency to think about others’ thoughts (but not their emotions or how to interact in social situations; see Chapter 2). Governance in Fiji is still largely in the hands of Indigenous Fijians, which promotes conditions for traditional norms to persist even with growing pressure to modernize and engage with global markets. Formally, Fijian criminal law stipulates distinctions between e.g. murder 83 and manslaughter as separated by perpetrator intent that are similar to the British system introduced in the colonial period (Corrin & Paterson, 2007). However, despite this formal legal observance of intent, previous research shows that Yasawans indeed focus less on intent than even other small-scale, traditional, non-Western groups (H. C. Barrett et al., 2016). Fiji also has a large population of Indo-Fijians, people of Indian descent who were originally brought to Fiji as indentured labourers to work on sugar farms during the period of British Imperial colonization. Comparisons between Indigenous and Indo-Fijians are particularly illuminating in that they share the larger country-level institutional structures, but come from a distinct, non-Western cultural tradition. In this section, we briefly sketch the ethnographic details about the sites (shown in the map in Figure 3.1) where our Indigenous Fijian participants live on Yasawa Island, and where our Indo-Fijian participants live on the main island of Viti Levu. Figure 3.1 Map of Fijian Archipelago showing locations of Yasawa Island and Lovu Village. Turtle IslandYasawaIslandTaveuni IslandMatagi IslandLaucala IslandQamea IslandVanuaLevuLabasaPacific HarbourBeqa IslandSuvaOvalauIslandGau IslandVatuleleIslandLautokaNadiNausoriSigatokaViti Levu50 kmVanua Balavu IslandNorthernLau GroupSouthernLau GroupSavusavuKadavu IslandThe Fiji IslandsCopyright 2001 Under Watercoloursvisit www.underwatercolours.com for purchase & usage information10 3020 40Waya IslandWayasewa IslandNarara IslandKuata IslandNaukacuvuIsland Nanuya Balavu IslandDrawaqa IslandNamaraNatawaYalobiWayalevuSomosomoMarouSosoGunuYageta IslandNaviti IslandMatayalevuNacula IslandTavewa IslandMatacawa IslandNanuya Levu IslandNanuya Lailai IslandVuakeSisiliNaculaYasawa IslandSawa-i-lau IslandNanuya-i-yata IslandTeciVawa IslandYawini IslandYasawairaraBukamaNabukeruNanuya-i-ra IslandYasawa IslandGroupCopyright 2001 Under Watercoloursvisit www.underwatercolours.com for usage informationPacific HarbourBeqa IslandSuvaOvalau IslandVatulele IslandLautokaNadiBa VatukoulaTavuaRakirakiNausoriSigatokaViti LevuCopyright 2001 Under Watercoloursvisit www.underwatercolours.com for usage informationLovu%Village%Western'Vi*'Levu' 84 18.104.22.168.1 Yasawa Island The people of Yasawa, Fiji, live as traditional fisher horticulturalists in small villages of around 70-150 adults. Village life revolves around the traditional Fijian political hierarchy that culminates power in a hereditary chief and structures social networks around kinship ties. These kinship ties organize the vast majority of the cooperative and coordinated efforts that Yasawans rely on for daily survival; these cooperative tasks range from sharing food to building houses (McNamara & Henrich, n.d.; Nayacakalou, 1955; 1957). This kinship hierarchy also defines the traditional practices and norms that foster a more relational, socially-defined sense of self among most Indigenous Fijians (Brison, 2001; A. Rumsey, 2000). Keeping tradition alive is often also associated with basic Fijian political and interpersonal identity (France, 1969; Jolly, 1992). One set of traditional Indigenous Fijian norms that are particularly important for the purposes of this study revolve around the idea that other minds are fundamentally unknowable, invisible behind the opaque barrier of social obligations and improper fodder for conversation as one’s thoughts are an individuals’ private business. This normative stance that minds are unknowable, more generally referred to as the Opacity of Mind, is seen in small-scale societies around the Pacific and in various other places around the world (Duranti, 2015; Groark, 2008; Hollan, 2012; Keane, 2008; A. Mayer, 2013; Robbins & Rumsey, 2008). Traditional Indigenous Fijian norms are further supported by practices that combine Christian and traditional ancestor spirit beliefs. The Kalou-vu (traditional ancestor spirits, “root gods”) are the mythical deified progenitors of the kin groups in a given location and form the spiritual backbone of traditional Fijian social hierarchies. Kalou-vu are believed to care about traditional norms and enforce them by adversely impacting the health and fortune of those who deviate from traditions. Kalou-vu can also be called upon for traditional medicine among those 85 who lead a proper traditional lifestyle (Katz, 1999; Shaver, 2014). As the mythical roots tying people to a particular place, Kalou-vu help solidify kinship networks by looking over their specific progeny, and are often most active in the physical space where each clan originated. These beliefs help foster deep social and spiritual bonds linking individuals to the kin network and the lands they live on/ in. The vanua (the land and the people of that land; the two are deeply interwoven and cannot be separated), is another essential component of Indigenous Fijian identity (Abramson, 2000; Jolly, 1992; Turner, 1988; Williksen-Bakker, 1990). Christianity has also become a major part of what it means to be Fijian. Christianity in Fiji exists in partially syncretic, interconnected tension with these traditional beliefs that tie people to each other and to the land (Newland, 2004; Ryle, 2010; Tomlinson, 2007). Yasawans are primarily members of Wesleyan Methodist and the Assemblies of God Pentecostalist churches, though several other denominations exist around Fiji. 22.214.171.124.2 Lovu Village, Viti Levu Our Indo-Fijian participants provide an important comparison point because they share nation-wide institutions and social structures with Yasawans but live by very different norms about mentalizing to explain behaviour. By including Indo-Fijians in study 1, we can further isolate whether there is something about being in a more collectivistic group, being in Fiji, or specifically being a member of Indigenous Fijian culture that may be causing any differences in intent focus we may find. We primarily recruited our Indo-Fijian participants from around Lovu village, located on the main island of Viti Levu near Lautoka. Several others were recruited from the nearby towns of Nadi and Ba. Today, Indo-Fijians work primarily as wage labourers or sugar cane farmers. The religious landscape among Indo-Fijians is primarily Hindu or Muslim, with a minority of Sikhs and converts to Christianity. Yasawans do have some connections to the 86 Lautoka area, but these groups rarely directly interact with a member of the other sample. We therefore consider these as geographically related but distinct cultural groups. 3.1.3 Overview of Studies In study 1, we map out the moral judgment patterns reported by our Yasawan, Indo-Fijian, and North American participants. As the population with normative proscriptions against thinking about minds, Yasawans are the population of primary interest for the present research. Both Yasawans and Indo-Fijians provide an interesting contrast to the highly individualistic conceptions of self and relationships common in North America (M. M. Gervais, 2013a; Kelly, 1988; Kline et al., 2013). Both emphasize strong family ties that are called upon in times of need. Both also have hierarchically-structured social roles than are common among North Americans. Other studies contrasting highly individualistic American samples against more urbanized but collectivistic people in Japan and India show that these more collectivistic cultures also focus less on intent (V. L. Hamilton et al., 1983; Laurin & Plaks, 2014). We therefore expect that both Yasawans and Indo-Fijians may be less focused on intent than North Americans. On the other hand, Yasawans may strictly favour outcome because of normative tendencies against thinking about motives. Though Yasawans in particular may not favour intent in their moral reasoning to the same extent that other groups do, previous work (Barrett et al. 2016) does not present explicit contrasts between positive intent with negative outcome (accidents) vs. negative intent with positive outcome (failed attempts). In study 2, we further isolate differences in thinking about minds as the cause of cross-cultural differences in thinking about intent. We focus in this study only on Yasawans and North Americans, as these two groups are the extreme ends of our accessible range of variation in focus on thinking about minds. We experimentally manipulate the salience of thoughts or actions by 87 asking participants to list various thoughts or actions God might want or not want them to think/ do. We predict that being primed to think about thoughts will boost intent focus, while thinking about actions should boost outcome focus. Importantly, if observed cross-cultural differences in intent focus are more related to differences in underlying activation rather than ability to metalize, then this priming should make Yasawans focus on intent just as much as other more mentalistically-oriented groups. 3.2 Study 1: Cross-cultural Differences in Moral Judgment Intent/ Outcome Focus In study 1, we seek to first document the pattern of moral judgments Yasawans, Indo-Fijians, and North Americans. We expect intent to be most important for North Americans and least important for Yasawans. 3.2.1 Method Data for this chapter was collected in conjunction with data reported in chapter 2 and two additional projects on separate but related questions: how people think about the minds of supernatural agents, and how do these beliefs about supernatural agents’ minds interact with moral reasoning in different cultural contexts and religious belief systems.6 The analyses in this chapter focus on participants’ own perspective and their own moral judgments (rather than their expectations about supernatural agents). These data were collected simultaneously due to inherent constraints in conducting fieldwork: our Fijian participants were largely recruited in their communities and on a voluntary basis. Similarly, our research time in these fieldsite 6 Participants’ expectations about supernatural agents’ responses to moral norm violations are reported in McNamara, Willard, and Norenzayan (n.d.). Data on how humans perceive the agentic vs. experiential capacities of supernatural and human minds are reported in Willard & McNamara (n.d.). Both of these manuscripts are available from McNamara by request. 88 communities is heavily restricted, often to just a few months at a time. Collecting as much data in one time as possible often ends up being the most effective solution to these fieldwork challenges. 126.96.36.199 Participants We recruited 151 Yasawans (93 women; 18-80 yrs., mean = 43.15; 4-15 yrs. formal education, mean = 9.33) and 219 Indo-Fijians (117 women; 17-76 yrs., mean = 38.03; 0-18 yrs. edu., mean = 10.68) in June-July 2012 and May-July 2013. Yasawans participated over repeated, short (15-20 minute) sessions whenever they were available in order to minimize the burden of participation as they went about their daily lives. Because the time burden of participation was less for Yasawans than in other groups, we did not give our Yasawan participants any direct remuneration for participation in this study. Indo-Fijian participants, on the other hand, were recruited from a more dispersed geographic area that made repeated visits difficult. As a result, Indo-Fijian participants completed all study materials in one (approximately 60-minute) session and were given $1 FJD remuneration for their time. We recruited a further 412 North American participants from September 2012 to June 2013. Among our North American participants, 205 (172 women; 17-32 yrs., mean = 19.85; 12-17 yrs. edu., mean = 12.90) were university students who participated through the university psychology department and were remunerated with course credit. An additional 207 (99 women; 18-72 yrs., mean = 33.87; 10-21 yrs. edu., mean = 14.38) were adults from the United States recruited through Amazon’s Mechanical Turk (Mturk). Mturk participants took approximately 30 minutes to complete all study materials and were remunerated with $1.50 USD. See Table 3.2. for participant numbers by culture and year. 89 188.8.131.52 Materials The vignettes in studies 1 and 2 feature multiple different domains of norm violations: harm, theft, poisoning, food taboos, social taboos, and failed cooperation. Previous vignette studies that focused on the relationship between mentalizing and assessments of blame and punishment have focused on harm (see for example: L. Young et al., 2007; 2011), though other work on different aspects of morality across cultures suggest that cultures vary in how much they emphasize different moral domains (Graham et al., 2011; Haidt, 2012). By including multiple domains of norm violations, we examine the general intent/ outcome balance in judging whether violations deserve punishment. For the sake of brevity, we analyze only the average intent/ outcome focus across domains in this chapter. The vignettes also systematically vary combinations of positive vs. negative intent and positive vs. negative outcome, as summarized in the intent/ outcome matrix shown in Table 3.1. The most important conditions for this study are accidents and failed attempts: endorsements of more severe punishments against failed attempts indicate intent focus, while endorsements of more severe punishments against accidents indicate outcome focus. Outcome Intent Positive Negative Positive No Violation Failed Attempt Negative Accident Successful Attempt/ Intentional Violation Table 3.1 Intent/ Outcome Matrix for Intent conditions. Endorsements of stronger punishments against failed attempts indicate intent focus; stronger punishments of accidental violations indicate outcome focus. 90 Following each vignette, participants are asked to make judgments about the violations based on seven questions, all presented in the following order: 1) How good or bad the violation action was. 2) Whether the violation was on purpose or by accident. 3) How positively or negatively each violation affected the victim. 4) How angered or pleased the victim would be. 5) How positively or negatively other people will think of the perpetrator when they find out about the violation. 6) How deserving of punishment or reward the violation was. 7) An open-ended question asking what the participant thinks about the situation presented in the vignette. For all but the final open-ended question, participants answered on a -2 to +2 likert scale, with -2 indicating the most negative outcome (very bad action, definitely on purpose, victim very negatively affected, victim very angered, other people think very negatively of the perpetrator, very deserving of punishment) and +2 the most positive outcome (very good action, definitely by accident, victim very positively affected, victim very pleased, other people think very positively of the perpetrator, very deserving of reward). Fijian versions for Yasawans and Indo-Fijians have names and specific contexts slightly modified for culturally appropriate content (e.g. food taboo for Yasawans is eating shark, for Indo-Fijians it is a vegetarian eating meat, and for North Americans it is a man who keeps Kosher eating pork). Yasawan materials were again translated into Standard Fijian and back-translated into English by Indigenous Fijian research assistants who were fluent in both languages. Our Indigenous Fijian research assistants were recruited from the Fijian capital city Suva, located on the Eastern side of the main island Viti Levu. None of the research assistants were regular inhabitants of these Yasawan villages. Indo-Fijian data were also collected by Fijian research assistants, though these assistants were recruited near Nadi and Lautoka on the Western end of Viti Levu. The assistants working with Indo-Fijian participants were both Indigenous and Indo- 91 Fijian. Indo-Fijian materials were administered in English, though Hindi translations (translated and back-translated by bilingual Indo-Fijian assistants) were available in case of confusion. 184.108.40.206 Procedure Participants in all cultural groups followed the same basic procedure: they listened to or read a norm violation vignette, then answered questions about the vignette (see section 220.127.116.11). This was repeated for four vignettes. Each vignette depicted a violation of a different moral domain (e.g. harm, theft, taboo). The domains were crossed with intention conditions (such that we have ratings for all domains in all intent conditions) and counterbalanced across participants (Table 3.2) Following each vignette, participants answered the seven judgment listed in section 18.104.22.168, giving answers for both a) their own moral judgment about the violation, then b) how they think God would judge the violation.7 After finishing vignette questions, participants then answered questions about human, Christian God, and local ancestor spirit (Yasawa only) mental capacities as part of a separate study about how people in different cultures think about human and supernatural agents’ minds (Willard & McNamara, n.d.). In 2012, Fijian participants also completed the EQ short (Wakabayashi, Baron-Cohen, Wheelwright, et al., 2006b) following norm violation vignettes. Mentalizing data is reported in Chapter 2. For our Yasawan participants, data for the vignette judgments, mentalizing, and beliefs about supernatural agent minds were collected in three separate sessions, spread out across the entire time in the field for that year. Data for all four vignettes were collected in one session that 7 The questions about what God thinks were dropped for a sub-sample of 64 Yasawans (32 women) who participated in 2013. We dropped the supernatural agent judgment questions to minimize participant fatigue while assessing different taboo-related violation domains of cooperation, social taboo, food taboo, with minor harm as a contrast domain. These participants answered questions about four vignettes depicting four different domains, but each participant answered 1 intentional, 1 accidental, 1 failed attempt, and 1 no violation intent condition vignette. 92 lasted from 30-45 minutes. Mentalizing and supernatural agent/ human mind capacities data were collected in different sessions that lasted between 10-20 minutes each. We use this separated protocol in Yasawa only, and we do so because our work there is part of long-term relationships we maintain with communities. As part of this ongoing community contact, we strive to maintain positive relationships and to express our gratitude for the community’s collaboration in our research through various means – most of which are non-monetary (e.g. offering basic first aid services while in the village, offering guest lectures at the local school, participating in community discussions of village improvements, etc.). By being in these communities, we are also more readily able to follow-up with participants for multiple sessions. We also do these shorter, repeated sessions in Yasawa as part of our responsibility to keep the content of our research as sensitive to the needs of the community as possible and to minimize the time and cognitive burden of participation. While answering a long series of questions may be more relevant to city life or to Western culture, the tasks in these questionnaire-based studies are less compatible with Yasawan culture; this procedure is a better compromise to adapt to this cultural mismatch. We also administer all materials in Yasawa through an Indigenous Fijian research assistant form another part of Fiji who was fluent in both Standard Fijian and English. We rely on these research assistant interviewers due to variable literacy among Yasawans. In July 2012, 61 Yasawan participants answered questions about accidental and intentional violations. In June-July 2013, 116 Yasawans reported their judgments about vignettes depicting failed attempts and situations where no violation occurred (28 of these Yasawan participants also participated in 2012, making for 90 new participants and a grand total of 151 unique Yasawan participants). 93 For our Indo-Fijian and North American participants, we collected data in one sitting and did not follow up for multiple sessions with the same people. Like our Yasawan participants, a Fijian research assistant also interviewed our Indo-Fijian participants in person. However, some of these assistants were Indo-Fijian and some were Indigenous Fijian. Each session with our Indo-Fijian participants took around 60 minutes in 2012 and 45 minutes in 2013. Data for intentional and accidental violations were collected from 100 Indo-Fijian participants in July 2012. An additional 119 Indo-Fijian participants answered questions about no violation and failed attempts vignettes in June 2013. After the questions about the vignettes, participants answered human and God mental capacities questionnaires, EQ short (2012 only), demographics, and religious beliefs questions. Our 205 North American student participants answered questions about intentional and accidental violations, counterbalanced over 4 different domains of violations. Of these 205 students, 47 participated in-lab with an in-person interviewer reading the norm violation vignettes and entering the participants’ answers directly into the computer. After finishing these vignette items, the in-lab student participants answered the remainder of the questionnaires by directly entering their answers into the computer in private. The remaining 158 students participated online. We included the in-lab subsample to check for possible differences in method that might make the data collected via computer different data collected in-person interview by a research assistant. Comparisons between the in-lab and online students did not show any significant differences. Students took approximately 60 minutes to complete all study materials and were remunerated with course credit. The Mturk American sample answered questions about 4 different norm violation domains counterbalanced across no violation and failed attempts intention conditions. After the questions about the vignettes, participants 94 answered human and God mental capacities questionnaires, EQ short, demographics, and religious beliefs questions. Cross-Cultural Adult Moral Intent/ Outcome Focus: Number of Observations and Unique Participants per Cultural Group and Year Intentional & Accidental (2012) Failed Attempts & No Violations (2013) All Intent Conditions (2013; Yas. only) Total Yasawa 236 observations 61 participants 210 obs. 54 part. (28 2012 & 13, 26 new in 2013) 249 obs. 64 part. 695 obs. 151 part. Indo-Fijian 335 obs. 100 part. 476 obs. 119 part. -- 811 obs. 219 part. North America 820 obs. 205 part. (University) 828 obs. 207 part. (Mturk) -- 1648 obs. 412 part. Table 3.2 Number of repeated observations and unique participants in study of baseline cross-cultural differences in intent/ outcome focus. While all participants were asked to judge up to 4 vignettes each, not all participants completed all 4 vignettes. Due to the continued relationships inherent to working with the Yasawan communities in this sample, Yasawan participants were the only group with the chance to participate in multiple years. 28 of the 54 Yasawan participants who answered failed attempts/ no violation conditions in 2013 also answered intent/ accident conditions in 2012, so that the total number of unique Yasawan participants is 151. 3.2.2 Results Previous research on intent focus in moral reasoning suggests that intent may play a slightly different role in judging wrongness vs. punishworthiness (Cushman 2008; 2015). We therefore focus our analysis on participants’ ratings of how good or bad and how worthy of reward or punishment the action was. We use multilevel linear regression, with vignette answers clustered by participant, to account for participants answering questions about multiple vignettes. We add an interaction between sample and intention condition to determine whether participants 95 across cultural groups are responding to the intent conditions differently. We also include controls for violation domain, how much the participant reported they thought the perpetrator intended to perform the action, and whether participants’ data were collected in the intentional/ accidental or no violation/ failed attempts waves. We include demographic variables of sex, age, and years of formal education to all of the models, though these variables never produce significant effects. Multilevel models are fit in R (R Development Core Team, 2008) using the lme4 (Bates et al., 2014) and lmerTest (Kuznetsova et al., 2014) packages. Degrees of freedom are approximated using the Satterthwaite approximation. Full regression tables can be found in Appendix B . 22.214.171.124 Badness and Punishment Across Intent Conditions Participants’ judgments of how good or bad the action was and how worthy of punishment or reward the action was both show that participants responded to each intent condition significantly differently across samples (Good/ Bad: F(6, 2646.08) = 44.05, p < 0.001; Punish/ Reward: F(6, 2620.54) = 48.67, p < 0.001). Both questions also show Yasawans to have a distinct pattern of treating accidental and failed attempts violations as more severe than other cultural groups. Yasawans rate accidental violations as being roughly as bad as both failed attempt and intentional actions (see Figure 3.2). This would suggest that Yasawans are not totally ignoring intention in favour of outcome. The Indo-Fijians, by comparison, also appear to think failed attempts are approximately as bad as intentional violations, but do not rate the accidental violations as nearly as bad as either failed attempts or intentional violations. Our North Americans participants rate the failed attempts as worse than intentional violations. Similarly, Yasawans also rate accidental violations as deserving roughly equal amounts as both failed attempt and intentional actions. Indo-Fijians rate accidents as less punishable than failed 96 attempts, which are less punishable than intentional actions. North Americans again rate failed attempts as the worst offenses, with accidents as less punishable than intentional violations. This again suggests that Yasawans are attending to intent, but are more focused on outcome than other populations. Figure 3.2 Marginal mean ratings of A) how good or bad and B) how worthy of reward or punishment the norm violation was. Error bars indicate standard errors. -2 -1 0 1 2 Yasawa Indo-Fiji North America Marginal Mean Ratings Higher = Better A How Good or Bad was the action? No Violation Accidental Failed Attempt Intentional -2 -1 0 1 2 Yasawa Indo-Fiji North America Marginal Mean Ratings Higher= More Reward B How much should the person be punished? No Violation Accidental Failed Attempt Intentional 97 126.96.36.199 Comparing Intent vs. Outcome To directly compare the effects of intent and outcome in both good/ bad and reward/ punish ratings, we next combine these two questions into a single dependent variable. We track the differences between answers to these questions with a categorical variable coding for question (good/ bad vs. reward/ punish). We further recode the intent conditions with two categorical variables for intent (positive or negative) and outcome (positive or negative – see intent matrix in Table 3.1). For model 1, we include the same predictors as before, but add a categorical variable for question (good/ bad or reward/ punish) and an interaction between intent and outcome. This 4-way interaction allows participants to rate violations differently based upon intent, outcome, culture, and question (sample x intent x outcome x question). This model allows us to compare how participants might change their assessments of outcomes depending upon actor intent across cultures, and whether they moderate their assessments of outcomes based upon intent differently for judging how bad an action is vs. how much punishment it deserves. If participants are changing their assessments of outcomes based upon intent, then we expect to see significant interactions between intent and outcome. We also analyze the categorical variable for intent conditions (no violation, accident, failed attempt, intentional) in model 2, this time placing it in a 3-way interaction with sample and question to allows participants to answers to each intent condition differently based upon culture and question (sample x intent condition x question). Our focal comparison across the intent conditions is between accidents and failed attempts. If participants rate failed attempts as worse and more punishable than accidents, then they are giving relatively more weight to intent than outcome. We take accidents as the comparison condition in model 2; we take good/ bad as the comparison question in both models. We again analyze these models using the lme4 (Bates et al., 2014) and lmerTest (Kuznetsova et al., 2014) 98 R packages. Degrees of freedom are again approximated using the Satterthwaite approximation. The interactions from model 1 and the differences in judgments about accidents vs. failed attempts are shown in Table 3.3. Model 1: Intent x Outcome Interaction Yasawa Indo-Fiji North America Good/ Bad 0.40*** [0.19, 0.61] 0.94*** [0.74, 1.14] 0.76*** [0.62, 0.89] Reward/ Punish 0.05 [-0.16, 0.26] 0.43*** [0.23, 0.63] 0.54*** [0.41, 0.68] Diff G/B vs. R/P 0.35* [0.06, 0.64] 0.50*** [0.23, 0.78] 0.21* [0.02, 0.41] Model 2: Outcome vs. Intent (Accidents – Failed Attempts) Yasawa Indo-Fiji North America Good/ Bad 0.10 [-0.08, 0.28] -0.72*** [-0.94, -0.50] -0.56*** [-0.75, -0.37] Reward/ Punish 0.07 [-0.10, 0.25] -0.42*** [-0.64, -0.21] -0.53*** [-0.71, -0.34] Diff G/B vs. R/P 0.02 [-0.23, 0.18] 0.30** [0.10, 0.49] 0.04 [-0.10, 0.17] Significance codes: ***<0.001, **<0.01, *<0.05, †<0.1 Table 3.3 Unstandardized regression coefficients with 95% CI calculated by profile likelihood estimation in brackets. Model 1: Interactions between intent and outcome for good/ bad ratings and reward/ punish ratings across cultures. Larger values of the difference in interactions between good/ bad vs. reward/ punish indicate that the effect of outcome depends more on intent in good/ bad ratings than reward/ punish ratings. Model 2: differences between Accidents (positive intent with negative outcome) vs. Failed Attempts (negative intent with positive outcome) indicate relative importance of intent or outcome in judgments. Lower values indicate negative intent (failed attempt) is considered worse and punished more strongly than negative outcome. We find that participant responses to the outcomes of each vignette depend upon actor’s intent in different ways for both good/ bad and reward/ punish judgments (intent x outcome x question F(1, 5201) = 21, p < 0.001), though this dependency does not significantly differ across 99 cultures (sample x intent x outcome x question F(2, 5201) = 2, p = 0.20). Across all cultural groups, good/ bad ratings of positive and negative outcomes changed more based upon positive or negative intent than reward/ punishment ratings. Thus, the contrast between good and bad intent vs. good and bad outcomes was more potent for goodness/ badness ratings than whether the action should be punished. This may be due to a general tendency to say that an action is more extremely good or bad on the one hand, but then also be hesitant to say that its goodness or badness actually deserves active response in the world via reward or punishment. The slightly smaller range of ratings on reward/ punish compared to good/ bad suggest this might be the case. Importantly, Yasawans differ from the other groups in their assessments of how much punishment violations deserved based upon either intent or outcome – for our Yasawan participants, a positive intention could not compensate for negative outcome and a positive outcome could not compensate for a negative intention. For model 2, we do find an overall difference in how participants across cultures respond to each intent condition based upon question (sample x intent condition x question F(6, 5449) = 3, p = 0.002). We find that both Indo-Fijians and North Americans weigh intention relatively more than outcome; participants in both cultural groups rated failed attempts as significantly worse and deserving significantly greater punishment than accidents. Though other studies have shown relatively greater emphasis on intent for assessments of how wrong an action is (Cushman, 2008), we find that only our Indo-Fijian participants place significantly more weight on intent when judging how bad the action was (as shown by the significantly smaller difference between accidents and failed attempts, see Table 3.3). North Americans place more weight on intent for both good/ bad and reward/ punish judgments by judging failed attempts more harshly than accidents. However, North American participants place this extra emphasis on failed 100 attempts in approximately the same way for both kinds of judgments. Yasawans, on the other hand, do not show a significant difference in judgments of either how bad or how much punishment accidents and failed attempts deserve, which indicates Yasawan participants are placing roughly equal value on intent and outcome for both judgments. 188.8.131.52 Predicting Intent Perception in Yasawa With the marked difference we find in Yasawans’ additional focus on outcome, we briefly examine what individual differences might predict Yasawans rating violations as more or less intentional across intent conditions. We analyze the by accident/ on purpose question, and predict Yasawans’ answers with age, education, sex, years lived in the village, as well as controls for domain and intent condition (details of these regressions are shown in B.2). Though education, age, and years in the village are highly related variables, we do find age and years in the village producing some significant effects on rating intent in these vignettes. Specifically, our older Yasawan participants tended to report more intent, while those who have lived in the village longer (controlling for age) rated violations as slightly less intentional. This trend for longer time in the local in-group predicting lower ratings of perpetrator intent does suggest that more time in Indigenous Yasawan culture may be a possible source for these cultural differences, but age and time in the village only emerge as significant predictors when we control for both of them in the same model. 3.2.3 Discussion In line with previous work showing that Yasawans habitually focus less on thoughts and motivations than other cultural groups, Yasawans rate accidents – positive intentions with negative outcomes – as both worse and more worthy of punishment than do Indo-Fijians and North Americans. Clearly, however, Yasawans are not strictly evaluating based on outcomes. 101 Yasawans judge failed attempts – negative intentions with positive outcomes – as approximately as bad and as punishable as accidents. North Americans, on the other hand, placed far more emphasis on intent – so much so that our North American participants even rated failed attempts as worse and more punishable than successful violations. This is a striking finding that differs from explicit Western legal code. In this case, it may provide further evidence of the extensively mentalistic focus in North American culture – intent is indeed far more important than outcome. It may also be an artefact of the procedure; presenting the contrast between accidents and intentional violations in one case and failed attempts against non-violations in the other may have inflated ratings of failed attempts (though this did not affect the other cultural groups). It may also be a product of student respondents being less inclined to use the extreme values of the scale than Mturk participants. However, other work among North Americans does provide additional support for the notion that North Americans may be especially focused on intent; for example, in some cases, North Americans may judge perpetrators more harshly for failed attempts if the actual violation was intended but never occurred than if it actually did happen by other means than what the perpetrator intended (Cushman, 2008; 2015). This may be evidence for the Opacity of Mind norms in Yasawan culture de-emphasize intent by way of reducing focus on internal mental states on the one hand, and North American culture, with its heavy emphasis on understanding action as a result of internal states and dispositions, may be boosting intent focus. However, as it stands, we do not specifically pin-point differences in thinking about minds as the source of these cultural differences in intent focus. For study 2, we explicitly manipulate the salience of thinking about thoughts to test whether this underlying cultural difference between Yasawan and North American culture is relevant to the differences in intent focus that emerged in study 1. 102 3.3 Study 2: Priming Intent For this study, we test whether cultural differences in focus on intent for making moral judgments is a result of different underlying tendencies to focus on internal thoughts vs. external behaviours. We expect that, if Yasawans are focusing less on intent because they are habitually thinking less about internal mental states, then explicit reminders of thoughts before moral judgments should boost intent focus. Similarly, if North Americans are habitually not taking situation into account and over-emphasizing mind, then reminders of actions should promote outcome focus. Alternatively, Yasawans may be so habitually focused on situation that extra reminders to think about thoughts may not be sufficient to induce them to focus to intent more than outcome. Similarly, North Americans may be so habitually focused on mind that a simple reminder to think about actions may not be sufficient to weaken their primary focus on intent. If thinking about thoughts or actions has nothing to do with these moral judgments, then we should see no change from the baseline patterns of reasoning across intent conditions seen in study 1. 3.3.1 Method As with study 1, study 2 for data presented in this chapter was also collected in conjunction with another project focusing on how cross-cultural differences in beliefs about supernatural agents’ minds interact with moral reasoning.8 The present analyses focus on participants’ own perspective and their own moral judgments. These data were collected simultaneously to collect the necessary amount of data in our short fieldwork timeframes and minimize strain on the continuing relationships we have with participants in these communities. 8 Participants’ expectations about supernatural agents’ responses to moral norm violations are reported in McNamara, Willard, and Norenzayan (n.d.); this manuscript is available from McNamara by request. 103 184.108.40.206 Participants For study 2, we recruited 72 Yasawans (39 women; 18-80 yrs., mean = 42.18; 4-16 yrs. formal education, mean = 9.36) from May-June 2014. Yasawans participated over repeated, short (15-20 minute) sessions on an entirely voluntary basis whenever they were available in the course of conducting their daily affairs, so we did not give our Yasawan participants any direct remuneration for their participation. Our 132 (94 women; 17-33 yrs., mean = 20.20; 12-20 yrs. formal education, mean = 13.81). North American participants were university students studying psychology from January-June 2015 and were remunerated with course credit. 220.127.116.11 Materials We use the same norm violation vignettes from study 1, though the domains of norm violations covered include only poisoning a water source, theft, violating a social taboo, and failures in cooperation.9 We also introduce a prime to encourage participants to think about thoughts (Thought Prime) or think about actions (Action Prime) before considering each vignette. We couch these primes about thoughts and actions within a question about supernatural agents: What do these agents want and not want people to think or do? In some ways, this question is a theory of mind task in-and-of-itself. However, by keeping this question about what another agent wants constant across both primes, we can determine if the focal content of what the agents want – thoughts or actions – will differentially influence moral reasoning. We use this religious framing in our primes to make them less suspicious and to make them more equivalently culturally relevant to both cultural groups. Especially for our Yasawan 9 As with study 1, the focus of the analyses for study 2 is also on the average intent or outcome focus across domains. 104 participants, our ongoing research with these communities often involves interviews about supernatural and religious beliefs. Because these participants know we are interested in what they believe God or local spirits want, this framing is less likely to arouse suspicion and provides additional information for our separate but related projects on religious belief. Again, even though this religious/ supernatural framing applies to these primes, we focus on the effects they have on participants’ own moral judgments in this project.10 Further, this supernatural agent framework provides a context that is culturally relevant to both Yasawans and North Americans. People in both of these cultural groups often report that supernatural agents can know the contents of a person’s mind (even in Yasawa where Opacity of Mind norms make minds less knowable to humans). Similarly, because supernatural agents care about thoughts and actions, this framework gives us a roughly equivalent way to contrast thoughts and actions. For example, we could have asked participants to do some theory of mind task instead of the thought prime, but an equivalent verbal task for action is harder to come by. Similarly, results in chapter 2 suggest that theory of mind task performance varies in these cultures, which may not provide equivalent priming in both groups. The lower emphasis on mind and on individualism in Yasawa may also make direct questions about what kinds of thoughts are good or bad less relevant for Yasawans than North Americans. Conversely, the religious framework in our primes provides a social context involving an agent that can know another’s mind, which may help equalize these cultural differences. Finally, we rely on explicit priming here because sub-liminal or sub-lingual priming in more urbanized populations often relies upon visual cues of written language. 10 See McNamara, Willard, and Norenzayan (n.d.; available from McNamara upon request) for details on the effects of these primes on participants’ judgments about how supernatural agents will respond to moral norm violations. 105 However, our Yasawan participants are not uniformly literate, so non-literary primes are a more reliable option. For our Thought Prime, we ask participants to list up to 5 examples of thoughts they think God would or would not want them to have, whether God can reward or punish them for these thoughts, and (if they answered yes to possible reward or punishment) what kinds of rewards or punishments they might receive. Our Action Prime used the same wording but asked about what actions God would or would not want them to do, if they could be rewarded or punished, and what those punishments or rewards might be. Because both the Thought and Action primes include positive (desirable/ reward-worthy) and negative (undesirable/ punishment-worthy) elements, we asked about the positive and negative thoughts or actions as separate questions presented in counterbalanced order. For both cultural groups, the primes were administered within subject, such that all participants were primed with both the Thought and Action primes. For example, in a single observation, participant A might answer about thoughts God would want them to have (positive), thoughts God would not want them to have (negative), then consider a norm violation vignette. For the next observation, participant A might then answer about actions God would not want them to do (negative), actions God would want them to do (positive), then consider a norm a second violation vignette. Yasawans answered Thought and Action Prime questions according to their beliefs about either the Christian God and the local ancestor spirits (Kalou-vu, or “root gods” in Fijian) in counterbalanced order. Religious affiliation in Yasawa is relatively uniform and previously known based upon our other work with these communities, so Yasawan participants did not need to directly self-identify as members of a particular religious denomination as a part of this study. For North Americans, we asked them to answer questions based upon whatever God they believe 106 in (or happen to think of in the case of non-religious or non-believer participants). We also asked North Americans to then report their primary religious affiliation and the supernatural agent they think of when they think of God (e.g. Christians might think of the Christian God, some Hindus might think about Krishna, etc.) in demographic questions after they competed the prime and vignette questions. Yasawan materials were translated into Standard Fijian and back-translated into English by Indigenous Fijian research assistants who were fluent in both languages. Our Indigenous Fijian research assistants were recruited from the Fijian capital city Suva, located on the Eastern side of the main island. None of these assistants were regular inhabitants in these Yasawan villages. 18.104.22.168 Procedure The design for the Yasawan data collection featured 4 repeated observations that encompassed all iterations of Christian God/ local ancestor spirits crossed with thoughts or actions. Participants answered questions in 5 phases, each separated by 24 hours.11 For the first 4 data collection visits, Yasawan participants answered questions about what supernatural agents’ want them to think or do, followed by one norm violation vignette. The norm violation vignettes included domains of poisoning a communal water source, theft, violating a social taboo, and failed cooperation. Each participant also answered about 1 intentional, 1 accidental, 1 failed attempt, and 1 no violation vignette. On the 5th data collection visit, participants answered questions about God, Human, and local ancestor spirits mental capacities (Willard & McNamara, 11 Some participants answered questions about vignettes on the same day, but separated by several hours. The research assistant who collected their data left the participants’ house and came back later for each phase of data collection. 107 n.d.). All materials were presented in counterbalanced order. Each session lasted approximately 20 minutes. North American participants participated in one, approximately 60-minute session. Each session included only 2 observations of primed norm violations; half of these participants answered the Thought Prime for their first vignette, and half answered the Action Prime for their first vignette. The two Thought/ Action primed vignettes were separated by a distractor task that asked participants to view 8 neutral images and list up to 5 words or phrases they thought would describe each image. After the second thought/ action primed vignette, North American participants answered questions about human and God mental capacities (Willard & McNamara., n.d.), mentalizing abilities (EQ short, Wakabayashi, Baron-Cohen, Wheelwright, et al., 2006b), and demographics. We recruited 61 of our North American participants to complete study materials administered by a fellow university student working as a research assistant in the lab, with another 6 participants answering questions online. The in-lab participants were again included to help control for possible additional social effects of having another person physically present when making these norm violation judgments. 3.3.2 Results Our analysis of study 2’s results parallels study 1, except that we add interactions to test for possible priming effects. We again focus our analysis on participants’ ratings of how good or bad the action was and how worthy of reward or punishment the action in each vignette was. We account for participants answering questions about multiple vignettes using multilevel linear regression, with vignette answers clustered by participant. We include an interaction between sample, intention condition, and prime to determine whether participants across cultural groups are responding to the intent conditions differently based upon the primes. We also include 108 controls for violation domain (centered on poison as the reference domain), how much the participant reported they thought the perpetrator intended to perform the action, and the order each vignette was observed in. Multilevel models are fit in R (R Development Core Team, 2008) using the lme4 (Bates et al., 2014) and lmerTest (Kuznetsova et al., 2014) packages. Degrees of freedom are approximated using the Satterthwaite approximation. 22.214.171.124 Priming Effects on Badness and Punishment Across Intent Conditions Though the religious framing of our primes was intended to minimize cultural mis-match across Yasawan and North American culture, our university participants were substantially less religious than the Yasawans; of the 132 student participants, 58 reported not believing in God. To test for possible impacts this disbelief might have had on the effectiveness of the prime, we included an additional interaction term for prime and belief (measured as “I believe in God” yes/ no). However, this interaction was both very small and not statistically significant (Good/ Bad: F(1, 505) = 0.01, p = 0.93; Reward/ Punish: F(1, 385.84) = 0.19, p = 0.67). Belief also failed to show any significant effects outside of interactions, so we drop this variable in the analysis. Primes had the strongest impact on both good/ bad and reward/ punish ratings in Yasawa, with the major movement happening in Yasawans’ ratings of failed attempts (see Figure 3.3). Yasawans and North Americans rated the badness of violations across intent conditions significantly differently for each prime (Good/ Bad sample x intent condition x prime: F(3, 509) = 2.82, p = 0.04). For reward/ punish ratings, there was a significant overall priming effect (F(1, 389.46) = 9.43, p = 0.002), with ratings becoming significantly more negative (ratings of more punishment) in the Thought Prime. Yasawans and North Americans also responded to intent conditions significantly differently (F(3, 462.74) = 102.33, p < 0.001), with Yasawans generally giving more extreme ratings (both more reward and more punishment). However, the primes did 109 not significantly affect Punish/ Reward ratings across intent conditions and samples (Reward/ Punish sample x intent condition x prime: F(3, 493.63) = 1.67, p = 0.18). 110 Figure 3.3 Marginal mean ratings of A/B) how good or bad and C/D) how worthy of reward or punishment the norm violation was by cultural group and prime. Error bars indicate standard errors. For both DVs, the prime had the largest effect in Yasawans’ ratings of failed attempts. -2 -1 0 1 2 Marginal Means Rating Higher = Better A Yasawa : Good/ Bad Action Thought -2 -1 0 1 2 Marginal Means Rating Higher = Better B North America : Good/ Bad Action Thought -2 -1 0 1 2 Marginal Means Rating Higher = More Reward C Yasawa : Reward/ Punish Action Thought -2 -1 0 1 2 Marginal Means Rating Higher = More Reward D North America : Reward/ Punish Action Thought 111 126.96.36.199 Priming Effects on Intent vs. Outcome We again combine Good/ Bad and Reward/ Punish ratings into a single dependent variable to directly compare their effects using a categorical variable coding for question (good/ bad vs. reward/ punish). We also recode the intent conditions with two categorical variables for positive/ negative intent and positive/ negative outcome. We use the same model 1 as in study 1, but this time we add prime to the larger interaction to allow participants in each culture to rate violations differently based upon intent, outcome, question, and prime (sample x intent x outcome x question x prime). This model allows us to compare how participants in each cultural group might be primed to change their assessments of outcomes depending upon actor intent, and whether the prime might act differently for judgments of badness vs. punishment. We again analyze the categorical variable for intent conditions (no violation, accident, failed attempt, intentional) in model 2, this time placing it in an interaction with sample, question, and prime (sample x intent condition x question x prime). Our focal comparison across the intent conditions is again between accidents and failed attempts. We expect the Thought Prime to increase emphasis on intent, making failed attempts worse and more worthy of punishment than accidents. Similarly, we predict the Action Prime will make accidents worse and more punishable than failed attempts. We again analyze these models using the lme4 (Bates et al., 2014) and lmerTest (Kuznetsova et al., 2014) R packages, with degrees of freedom again approximated using the Satterthwaite approximation. Intent x outcome interactions and the differences between accidents vs. failed attempts are shown in Table 3.4. 112 Intent x Outcome Yasawa Action Prime Thought Prime Action vs. Thought Prime Good/ Bad 2.04*** [1.44, 2.65] 2.20*** [1.62, 2.78] 0.15 [-0.68, 1.00] Reward/ Punish 0.76*** [0.62, 0.89] 1.65*** [1.07, 2.23] 1.00* [0.16, 1.84] Diff G/B vs. R/P -1.39*** [-2.18, -0.60] -0.55 [-1.34, 0.24] 0.84 [-0.28, 1.96] North America Action Prime Thought Prime Action vs. Thought Prime Good/ Bad 1.10*** [0.46, 1.73] 0.86** [0.22, 1.49] -0.24 [-1.14, 0.66] Reward/ Punish 0.55† [-0.09, 1.18] 0.31 [-0.33, 0.95] -0.23 [-1.14, 0.67] Diff G/B vs. R/P -0.55 [-1.41, 0.31] -0.54 [-1.40, 0.32] 0.01 [-1.21, 1.22] Outcome vs. Intent (AC - FA) Yasawa Action Prime Thought Prime Action vs. Thought Prime Good/ Bad 0.73*** [0.32, 1.15] -0.24 [-0.67, 0.18] -0.98** [-1.56, -0.39] Reward/ Punish 0.27 [-0.14, 0.69] -0.49* [-0.91, -0.07] -0.77* [-0.18, -1.35] Diff G/B vs. R/P -0.46 [-1.02, 0.10] -0.24 [-0.81, 0.31] 0.21 [-0.58, 1.00] North America Action Prime Thought Prime Action vs. Thought Prime Good/ Bad -0.47* [-0.93, -0.004] -0.13 [-0.59, 0.32] 0.34 [-0.31, 0.98] Reward/ Punish -0.05 [-0.52, 0.41] -0.21 [-0.66, 0.25] -0.14 [-0.49, 0.79] Diff G/B vs. R/P -0.55 [-1.41, 0.31] -0.07 [-0.68, 0.53] 0.48 [-0.38, 1.36] Significance codes: ***<0.001, **<0.01, *<0.05, †<0.1 Table 3.4 Unstandardized regression coefficients with 95% CI calculated by profile likelihood estimation in brackets. Model 1: Interactions between intent and outcome for good/ bad ratings and reward/ punish ratings across cultures. Model 2: differences between Accidents (positive intent with negative outcome) vs. Failed Attempts (negative intent with positive outcome) indicate relative importance of intent or outcome in judgments. Priming effects were strongest for intent in Yasawa. 113 We find marginally significant evidence that the primes are leading our Yasawan participants to shift their judgments based upon intent differently than North Americans (sample x intent x outcome x prime F(1, 794) = 2.92, p = 0.09), but this cross-cultural priming effect does not differ across good/bad vs. reward/ punish questions (sample x intent x outcome x prime x question F(1, 892.23) = 0.92, p = 0.33). We further find that Yasawans and North Americans are changing their responses to outcomes significantly differently across primes (sample x outcome x prime F(1, 778.73) = 9.67, p = 0.002). As shown in Table 3.4, we again see that interactions are slightly larger for good/ bad than reward/ punish in both cultures, indicating that the range of judgments across positive and negative intents and outcomes is wider for good/ bad than reward/ punish, though this difference is only significant in the Yasawans’ Action Prime. The cross-cultural difference in the priming effect comes from the range of Yasawans’ assessments across positive and negative intent and outcome being larger (seen as larger interactions) in the Thought Prime, while this range is slightly larger in the Action Prime for North Americans. This interaction between intent and outcome is only significantly larger for Thought Prime for Yasawans’ reward/ punish ratings. When we shift to model 2, we see that this priming effect seen in Yasawa is due to greater emphasis on outcome for good/bad ratings in the Action Prime (on average, ratings of accidents are 1.10 points worse than failed attempts) and greater emphasis on intent in especially reward/ punish rating in the Thought Prime (failed attempts are on average rated as 0.49 points more worthy of punishment than accidents). We find an overall difference in how Yasawan vs. North American participants respond to each intent condition based upon prime (sample x intent condition x prime F(3, 815.52) = 4.54, p = 0.004), but this again does not differ by question (sample x intent condition x prime x question F(3, 892.23) = 0.86, p = 0.46). We find only weak 114 evidence for any priming effects in North America; North Americans place more weight on intent in good/ bad assessments in the Action Prime, while they show non-significant increases in weight on intent in reward/ punishment ratings after the Thought Prime. This persistent focus on intent for especially good/ bad ratings may corroborate other North American findings showing that North Americans assess the wrongness of a violation almost strictly based upon an actors’ internal mental states rather than causation or outcomes (Cushman, 2008). 3.3.3 Discussion In study 2, we find that reminders of thoughts do shift Yasawans out of their baseline equivalence on intent and outcome to favour intent. We also find that the Action Prime led to Yasawans judging accidents as significantly worse and slightly more worthy of punishment than failed attempts. We also find that participants in both cultural groups make more extreme good/ bad judgments than reward/ punish judgments, and both kinds of judgments are scaled by outcome and intent. The prime appeared to be less effective in North America, perhaps because emphasis on mental state reasoning is so pervasive that this prime did not impact it (North Americans have been similarly hard to experimentally shift away from dispositional focus when explaining behaviours, see: Choi & Nisbett, 1998). We do see that both primes resulted in North Americans treating completed violations as worse than failed attempts. We also find that North Americans’ assessments of how good or bad the action was continue to be more impacted by intent even in the Action Prime. This may be seen as further evidence for the extra emphasis on intent in wrongness judgments among North Americans (Cushman, 2008). However, North Americans’ ratings of reward/ punishment show equivalent emphasis on outcome and intent in the Action Prime that shifts to slightly more focus on intent in the Thought Prime. 115 3.4 General Discussion In study 1, we showed that our Indigenous Fijian participants living in Yasawa, Fiji, who normatively avoid referencing mental states to explain behaviour, also emphasized intent less than North Americans and non-Indigenous Indo-Fijians. Importantly, our results in study 1 presented explicit contrasts between accidents and failed attempts which allowed us to show that our Yasawan participants were not merely using outcome to make their moral judgments, but were incorporating both outcome and intention in their judgments approximately equally. Both North Americans and Indo-Fijians, on the other hand, clearly favoured intent. Indo-Fijians showed the biggest shifts in judgments of both how good or bad and how worthy of reward or punishment various violations were, and these shifts were based on both outcome and intent. North Americans, on the other hand, focused on intent almost exclusively. In study 2, we found dramatic differences in our Yasawan participants’ judgments about failed attempts vs. accidents. Yasawan participants showed a mild preference for outcome by judging accidents as worse when primed to think of actions, but shifted their emphasis to intent by judging failed attempts as more punishable when they were primed to think about thoughts. These results show that our Yasawan participants’ baseline equivalence of outcome and intent seen in study 1 is likely an effect of Yasawan participants thinking less about thoughts in their everyday lives. We argue then that this result seen among our Yasawan participants is due to Yasawan cultural norms that de-activate cognitive mechanisms for thinking about minds rather than a difference in how mentalizing cognitive mechanisms may be functioning within our Yasawan sample. We did not find as much of a shift in our North American participants’ judgments based upon priming, perhaps due to a restricted range of how severely these participants would judge the actions. Alternatively, our North American participants may have 116 already been so focused on mind that they were harder to shift in their moral evaluations than Yasawans. Other evidence that shows North Americans may be less inclined to take situation into account even when explicitly reminded to do so supports this interpretation that North Americans may be hypermentalizing to such an extent that our Action Prime was not strong enough to noticeably move their judgments (Choi & Nisbett, 1998). It is important to note, however, that the previous work on situational reminders emphasized their effects on making attributions about stable traits, not transient mental states (Malle, 2006a). Whether the Action Prime has similar effects to more explicitly situational reminders and whether these reminders may have differing effects on transient or stable internal states are important areas for further research. Other individual differences in religious belief may also have affected the primes, though we did not find evidence that the primes interacted with a binary yes/ no measure of belief and this binary belief variable did not significantly predict judgment for either dependent variable. This data presents evidence that cultural norms can indirectly shape social decisions by moderating activation of basic cognitive processes. Further research into how these norms may further structure other social interactions may shed light on why these norms are present in some societies but not others. Specifically, the present research indicates that baseline activation of thinking about thoughts may drive cross-cultural differences in focus on intent. From a Western perspective, it is tempting to conclude that this must mean that Yasawan culture is suppressing mentalizing. But, it is equally possible that something about Western culture boosts mentalizing, again making Westerners the WEIRD ones when compared to the average member of any given society around the world. 117 One might also argue that these different patterns of intent vs. outcome focus have less to do with differences in mental state reasoning than differences in extra cognitive load participants experience due to basic unfamiliarity with the task. Specifically, the vignette and questionnaire tasks are the least familiar to Yasawan daily life, somewhat more similar to daily activities many Indo-Fijians might engage in, and so common for university students that the tasks are trivially taxing. We do not find that age or education predict any significant differences in judgements across cultural groups in either study 1 or 2. Further, a specifically difference in cognitive load for Yasawans would not explain the pattern of findings in study 2. One might argue that the cognitive load with the prime would be even greater, but we find strong evidence that Yasawans do focus in on intent to the exclusion of outcome when they are reminded to think about thoughts. On a more basic level, this research suggest that there may be something about the ways that social situations are set up in different cultural groups to emphasize or downplay mentalizing as an accurate source of information about why someone might behave in a certain way. Thinking about minds as a source of behaviour is a reasonable strategy when individuals are given the autonomy to act according to their own desires and goals, as is the case in highly individualistic societies like the U.S. and Canada. However, the desires and goals of the community or the family are often far more important than individual concerns for the majority of the world population (Brison, 2001; Gelfand et al., 2011; Heine, 2001). As a part of living in these more community-oriented societies, intent is often less emphasized in moral judgments (A. B. Cohen, 2003; V. L. Hamilton et al., 1983; Laurin & Plaks, 2014). Another aspect of living in more collectivistic groups is that general deviation from any norm, moral or otherwise, is typically less tolerated (Kim & Markus, 1999; Phelan & Rudman, 2010). The emphasis on 118 maintaining tight group cohesion and distinct group boundaries through adherence to norms may further stem from cultural adaptations to sustain cooperation, especially in situations that pose existential threat from harsh environments, disease, or resource scarcity (Bauer, Cassar, Chytilová, & Henrich, 2014; Botero et al., 2014; Fincher & Thornhill, 2011; Hruschka et al., 2014; D. R. Murray, Trudeau, & Schaller, 2011; Van de Vliert, 2011). As such, one reason why intent may be less important in collectivistic cultures may be that actions do not accurately convey information about mental states – the mind may indeed be opaque because actions are so thoroughly structured by the situation that they give little or no accurate information about what the person doing them really wants. Future research into how situations might limit access to mind through interpreting behaviours may further illuminate how societies structure interactions to emphasize or de-emphasize intent. Another important future direction for this research lies in assessing how these patterns of moral judgment may vary along with psychological development. Studies with infants in North America show that infants as young as eight months use intent to make socio-moral judgments (Hamlin, 2013). Though there is some variation in how old children are when they pass verbal theory of mind tests (H. C. Barrett, Broesch, Scott, He, Baillargeon, Wu, et al., 2013b; Callaghan et al., 2005; Shahaeian et al., 2011), North American infants show some evidence for thinking about beliefs by their second year of life (Onishi & Baillargeon, 2005; Surian, Caldi, & Sperber, 2007; Baillargeon, Scott, & He, 2010). Taken together, these infant and child studies suggests that intentionality reasoning happens early in development and may be a culturally universal aspect of our core cognitive architecture (Woodward, 2009). It remains to be seen whether this focus on intent may also be present in cultures that downplay mentalizing like Yasawa, or whether the equivalent focus on intent and outcome is present early in life. 119 3.5 Conclusion More than just random variation, culturally-transmitted norms that dictate appropriate social behaviour and guide our interpretations of others’ actions may fundamentally shape the ways we see and interact with the social worlds we live in. This chapter examines how cross-cultural differences in norms around thinking about minds leads to different moral evaluations of especially accidental and attempted but unsuccessful violations. This differential focus on outcome vs. intent is further explicitly linked to underlying salience of thoughts when making these moral decisions. Because these cultural norms may have developed to address specific social and/ or ecological pressures societies face, they may have an important mediating role in helping people adapt to various environments around the world. By linking the individual to the collective in this manner, we can begin to see how culture shapes us into cultural beings who are uniquely adapted to build and maintain the social worlds that are essential to being human. 120 Chapter 4. Learning to See (or Unsee) Mind: Culture Modulates Intent vs. Outcome Focus Across Development 4.1 Introduction In the early days of Darwinian evolutionary theorizing, biologists became fascinated with the striking physical similarities in early life development seen across species from vastly different taxa. As classically demonstrated in Ernst Haeckel’s illustration of embryonic development (Figure 4.1), many organisms start life in deeply similar forms that only begin to differentiate into species-typical traits later in development. Much of the research in psychology has been driven by a similar set of goals toward identifying the cognitive mechanisms that psychologically unite us as a species and the processes that make us psychologically diverge as we grow into adulthood. In developmental psychology, a common aim is to discover the earliest developing cognitive mechanisms and processes that are common to all people regardless of cultural learning (Spelke & Kinsler, 2007; Carey, 2011). Similarly, cultural psychological research is often aimed at both finding the psychological mechanisms common across all cultures and determining how specific cultural contexts shape people into different psychologies (Norenzayan & Heine, 2005; Shweder 1990; 1999). This chapter unites these two psychological research approaches to ask how basic cognitive processes that emerge early in development and are common across cultures are shaped throughout the lifespan by specific cultural influences. In so doing, this chapter aims to emphasize both the deep similarities and meaningful differences that arise out of the interplay between cognition and culture at the heart of human nature. 121 Figure 4.1 1892 copy of Ernst Haeckel's embryo drawings. The deep similarity observed among very young embryos across diverse species has long been taken as evidence for shared ancestry. To best test whether a psychological phenomenon is due to universal core cognitive architecture, to culture-specific learning, or to the interplay of the two, psychologists must recruit participation from people in diverse cultures across different phases of the lifespan. If we hypothesize that a given psychological phenomenon is a universal part of human nature, then a good first test is to check for its presence or absence in a culture with norms and/ or practices that may suppress its expression. For example, the case for approximate number inference as a core cognitive function is greatly bolstered by evidence for approximate number cognition even among people living in cultures with no formal mathematics and few words for number above 122 three (Dehaene, Izard, Spelke, & Pica, 2008; Izard, Sann, Spelke, & Streri, 2009; Pica, Lemer, Izard, & Dehaene, 2004; Wynn, 1992). However, if we do find a difference among adults, it still is not clear whether this difference is evidence that the cognitive function is non-universal, or whether the basic cognitive process has been shaped through development into different adult forms. It may well be that some aspects of the cultural environment encourage the expression of underlying tendencies that remain dormant if the appropriate environmental cues are missing (e.g. nearsightedness in the context of literacy, see: Cordain, Eaton, & Miller, 2002; Wong, Coggon, Cruddas, & Hwang, 1993). Alternatively, the cultural environment may provide a novel stimulus that forms the consistent push for development of cognitive processes that would never arise otherwise (e.g. the cognitive shift from logarithmic to integer thinking about number: Dehaene et al., 2008). In either case, if the cultural environment is the cause of the psychological phenomenon’s expression, then we would expect developmental change in the phenomenon to be consistently increasing across the lifespan. However, if the differences observed in adulthood are the result of cultural practices modulating universal cognitive processes, then we would expect to see evidence of more consistent performance among children across cultures that gradually differentiates as members of different cultures grow up in their specific cultural environments. Beyond cataloguing difference, the systematic study of how cultural differences arise across development can provide vital insight into how humans adapt to a variety of ecological niches through the transformative power of culture. This can provide insight into how culture directly shapes the mind, and how culture adapts our minds to help reinforce and perpetuate cultural solutions to social life across generations. This chapter seeks to demonstrate one example of how culture modulates basic social cognition throughout development in the domain 123 of socio-moral decision-making. As a fundamentally social species, the ability to interpret other’s actions and anticipate behaviours is often achieved through inferring others’ mental states. However, the quality of mental state inference from behavioural cues is dependent upon how free actors are to behave in ways that indicate internal mental states – in situations that preference group over individual and constrain behaviour, mental state inference may be challenging if not impossible. 4.1.1 Why Intent? Mental State Reasoning and Moral Decision Making The ability to track the mental states of others is a key cognitive function underlying our ability to build and sustain the social worlds that are the signature of our species. Inferring the goal states of others provides a rubric to fast-track learning by acquiring the skills without the extra time, effort, and luck needed to come upon the same learning individually. Indeed, the focus on intent seems to be the key difference that separates human cultural learning via imitation and overimitation from the emulative social learning seen in our closest primate relatives (Legare et al., 2015; Marsh et al., 2013; Okamoto-Barth & Tomonaga, 2006; Whiten et al., 2009). Similarly, the ability to understand the link between goals and the actions needed to accomplish them may be the cognitive scaffolding needed to successfully coordinate social actions. The ability to anticipate a competitor or collaborate with a teammate simply by referencing what we think they know and want to do might have provided our social ancestors the edge needed to survive. Further, if inferring mental states can help us predict future actions, inferring the valence of that mental state – whether the goals are for ends that benefit or harm us – can be an essential safeguard against exploitation that threatens any would-be cooperator. Therefore, the ability not only to calculate intent to engage in social action, but also the ability to 124 judge the other actors based on the qualities of their intentions, are vital skills for successful social life. As a result, intentionality reasoning may also be at the core of social judgments. However, the ability to infer the future actions of others may also be guided by reference to norms, or culturally-specific systems of rules that shape expectations about how to behave. Just as social learning can provide a short-cut in skill acquisition, reference to shared norms can short-cut the cognitive load of inferring other’s mental states (J. S. B. T. Evans, 2008; J. S. B. T. Evans & Stanovich, 2013; Kimbrough & Vostroknutov, 2013; Sripada & Stich, 2006). Moral norms form one subset of these culturally-transmitted rule sets for governing behaviour. Moral norms are the rules that generally go unwritten. However, moral norms also tend to provide such a fundamental structure guiding social interactions that, if broken, they require a stronger punitive reaction than other rules with less essential social structuring functions. Further, this strong response is often differentiated based upon intentional states of the actor and the kind of violation committed. For those actions that cause the greatest harm, the potential of recidivism makes the malicious intent more important. On the other hand, some acts, like incest and improper handling of potentially disease-causing items, are often considered in and of themselves reprehensible. Thus, intentionality reasoning in moral norm violations varies based upon the domain of the violation – harms tend to be considered more harmful when the perpetrator intended to cause damage than purity violations, which are intrinsically bad and are less readily scaled by intent (L. Young & Saxe, 2011). Similarly, adults’ judgements about how wrong an action is appear to be more sensitive to mental state information than judgments about how much punishment the action deserves, which is more readily moderated by the amount of damage done (Cushman, 2008; J. W. Martin & Cushman, 2015). Therefore, even if the cognitive effort behind making a mental state inference mental state inference may be reduced by 125 referencing norms, some level of intentionality reasoning is still important when choosing how to react to people who fail to meet those normative moral expectations. 4.1.2 Development of Intent vs. Outcome Focus in Moral Reasoning The development of intentionality reasoning in moral judgments may be a particularly interesting domain to examine how cultural environments might shape basic cognition across the lifespan. Early research on children’s patterns of moral reasoning show a clear progression from outcome focus to more intent-focused reasoning. Piaget (1932) demonstrated many children focus on the scale of the damage done by a particular violation as the metric for determining naughtiness and the amount of punishment the actor deserved – a child breaking 12 cups on accident is naughtier and deserves more punishment than a child who broke one on purpose or through negligence. Importantly, though, these children can clearly state they understand the actor’s lack of malicious intent, they nonetheless assign more blame and punishment to larger damages regardless of intent (Piaget, 1932, pp. 124-125). Kohlberg (1971) similarly suggests that children’s reasoning about right and wrong starts out with a more concrete calculation of avoiding punishment that gradually shifts into more abstract, individualistic concerns with duty to society and finding justice based on internalized values. More recent work on socio-moral development supports these findings to a degree (see for example: Cushman et al., 2013; Zelazo, Helwig, & Lau, 1996), but do show that children can use mental state reasoning to modulate their assessments of wrongness and punishworthiness when the tasks are structured in a less cognitively demanding way (Heiphetz & Young, 2014; Karniol, 1978; Nobes, Panagiotaki, & Pawson, 2009). The challenge to very young children may be in the extra executive functioning demands of evaluating a behaviour and responding to an adult’s questions about one’s reasoning. When the task is simplified even further to present a mere question of preference – which actor 126 does the child like more – then a preference for characters that helped another character emerges as early as 3 months (Hamlin et al., 2007; Hamlin & Wynn, 2011), and a preference for positive intent even when outcomes are negative emerges as early as 8 months (Hamlin, 2013). These data suggest that intentionality judgments in social actions may have very basic cognitive roots that show up early in development, which would suggest that these processes should be culturally universal. However, the observation that many of the more nuanced applications of intentionality reasoning around particular actions (i.e. purity violations vs. physical harms) that violate moral specific types of norms may suggest that this process is acquired through culturally-specific learning. To investigate whether intentionality reasoning in social judgments may be a core cognitive domain that is directly modulated by cultural learning, we test the social preferences of individuals of all ages in a culture with norms that discourage reference to mental states as attributions for a person’s actions. 4.1.3 Cross-Cultural Differences in Mental State Reasoning: Opacity of Mind Ethnographers working with small-scale societies around the Pacific have reported people in these groups often adopt a normative stance that prohibits inference about minds. Known as the Opacity of Mind, these norms have been cited as a challenge to the universality of mental state reasoning in causal attributions about behaviours and subsequent judgments about moral value of actions (Duranti, 2015; Groark, 2011; Hollan & Throop, 2011; Lillard, 1998; A. Rumsey & Robbins, 2008; Throop & Hollan, 2008; for a review, see: A. Mayer, 2013). Within these societies, the mental contents of others’ thoughts may be described as hidden within an opaque container of their actions; people’s behaviour does not necessarily reflect their true motives or opinions. These Opacity of Mind norms may also indicate that people in these societies do not regard mental states as the fundamental source of action in the way that the 127 Western folk psychological model of minds posits (Lillard, 1998). Though the qualitative, ethnographic approach has been essential in making these cross-cultural variations in approaches to minds known outside of these cultures, it remains unclear whether these norms are strictly evidence of difference in mental state reasoning processes, differences in how the ethnographers are understanding these foreign cultural worlds that do not indicate true differences in psychological experience (Robbins, 2008; Robbins & Rumsey, 2008), or a system of norm structures that shape cognition in a different fashion as children grow up in these cultural environments. The present study seeks to investigate the third possibility. Specifically, we predict that social judgments based upon reasoning about intent develops early, even in cultures that do not explicitly enforce mental state reasoning in adulthood. We predict that, as children grow up in the cultural environment of these differing norms (norms that enforce analyzing behaviours as indicators of internal mental states or suppress mental state reasoning as less indicative of internal state) will produce marked differences in the ways adults approach the same social judgment tasks. 4.1.4 Yasawa, Fiji For this study on how culture shapes mental state reasoning in socio-moral judgments across the lifespan, we enlist the participation of villagers living on Yasawa Island, Fiji. As a group of Indigenous Pacific Islanders, Yasawans hold normative Opacity of Mind practices. Adult mental state reasoning in Yasawa shows clear differences in the ways that Yasawans are approaching the problem of other minds as compared to both North Americans and non-Indigenous Fijians (H. C. Barrett et al., 2016; Chapter 2). Yasawans live in traditional villages of around 70-150 adults, where the primary means of subsistence is through fishing and horticulture. Villages are set up around the traditional Fijian hierarchy that draws upon kinship 128 ties to culminate power in a hereditary chief. The kinship relationships among individuals are the primary normative framework for deciding whom to cooperate with and how to allocate resources for day-to-day tasks ranging from finding and preparing food to building houses (McNamara & Henrich, n.d.; Nayacakalou, 1955; 1957). Traditional practices and normative obligations based on kinship are the primary means of organizing social interactions, creating a more relational than individual sense of self (Brison, 2001; A. Rumsey, 2000). This highly relational social system makes situations more constraining on behaviour and may limit the informational quality of behaviour as an indicator of internal mental states. 4.1.5 Assumptions and Predictions We use simplified moral dilemmas presented as puppet shows that have been successful in highlighting intentionality reasoning in social judgments made be preverbal infants in North America (Hamlin, 2013; Hamlin et al., 2007; Hamlin & Wynn, 2011). We remove the complexity of assigning blame, judging wrongness, or meting out punishments by simply measuring participant’s basic preferences for one character with the prompt, “Who do you like?” We use a forced choice paradigm based on preferential reaching in a non-verbal social scenario to help eliminate constraints of verbal processing both across cultures and across the lifespan. Within these puppet choice tasks, we assume that choosing a puppet in response to the prompt, even among preverbal infants, indicates a basic preference to affiliate with the chosen puppet. Reaching behaviour in preverbal infants is one way of getting around the issue of language by tapping in to infant’s intuitive response to be nearer to/ focus more attention on/ learn more from particular others in their social environments. Our basic assumption is that participants will always prefer actions that promote successful completion of a goal. Specifically, we assume that the basic preference is for the protagonist to complete its goal, such that participants will prefer 129 puppets that interact with the protagonist to aid (or attempt to aid) the protagonist in completing this goal. We test three possible orientations participants can take in making their choices about puppets: 1) intent focus, 2) outcome focus, or 3) success focus. If participants are taking an intentionality orientation, then they should judge puppets based on whether they act in a way that facilitates the protagonist in completing its goal. If this is the case, then participants should choose the puppet that acted in a way that conveys intent to help the protagonist, regardless of whether or not they succeed in helping the protagonist. If participants take an outcome orientation, then participants would instead focus on the outcome that results in the protagonist achieving its goal. If this is the case, then participants should choose the puppet associated with the protagonist completing its goal, regardless of whatever the actor puppets did leading up to the protagonist’s completed goal. If participants are instead taking a success orientation, then they should favour the puppet that does what it intended to do – the puppet that successfully completes its own goal rather than completing the protagonist’s goal. If this is the case, then participants should choose puppets that produce an outcome that matches its intent – even if the intent and outcome are negative for the protagonist. For the three alternatives outlined here, both intent and success orientations require some level of intentionality reasoning; both require that judgments be made based upon inferences about the puppet’s apparent goals. Therefore evidence for either intent or success orientation can be taken as evidence supporting use of mental state reasoning in formulating these social preferences. 130 4.2 Method 4.2.1 Participants Data were collected from participants in Yasawa, Fiji; Vancouver, Canada; and Illinois, USA. In Yasawa, four villages on the northern half of the island participated in June and July 2013 and May and June 2014. For all villages, pre-school aged children from 0.5-3.9 years (n=82, 39 girls, mean age = 2.2 yrs.) were recruited in their homes; school-aged children aged 4-16 yrs. (n=152, 78 girls, mean age = 8.6) were recruited in the village schools; and adults aged 18.3-80.2 yrs. (n=112, 65 women, mean age = 41.7 yrs.) were recruited around the village and in their homes. All Yasawan participants were recruited on a strictly voluntary basis with no direct remuneration. North American data collection took place between December 2014 and July 2015. Children aged 1.5-18 yrs. (n=106, 65 girls, mean age = 6.5 yrs.) were recruited at the UBC Living Lab in Vancouver’s Telus World of Science Center and were remunerated for their participation with stickers. University students aged 19-50 yrs. (n=65, 48 women, mean age = 21.5) participated through the University of British Columbia Psychology Department’s Human Subjects Pool, and were remunerated with extra credit in psychology courses. Adults aged 25-67 yrs. (n=22, 14 women, mean age = 51.7 yrs.) were recruited from a community sample in rural Illinois. Illinois adults participated online using the Qualtrics survey platform; online adult participation was completely voluntary with no direct remuneration. 188.8.131.52 Age Groups We recruited participants of all ages. To get a general sense of the developmental trajectories of intent vs. outcome focus, we divide these ages up into three age categories: Babies are children younger than four years, Children are between 4 and 12 years, adolescents are between 13 and 18 years, and adults are 19 years and older. We categorize the ages of our 131 participants in order to allow potential underlying curvilinearity in the developmental trajectories to emerge from the data despite the small sample sizes that we have access to in Yasawa that necessarily make it difficult to adequately model complex, non-linear relationships. We place the cut-point between babies and children at 4 years because past research suggests most children successfully pass verbal false belief tasks at this age, which would suggest a developmental milestone that could separate younger and older children. We separate children and adolescents at 13 because this a typical age when children begin to enter the early stages of puberty, begin defining themselves more based upon their peers than their immediate family, and – in Yasawa – when children begin to take on more of the traditional adult tasks (Christie & Viner, 2005; Kline, 2013). We place the cut-point for adults at 19 years to match the legal age of majority in the British Columbia (the Canadian province where the primary research institution is based). See Table 4.1 for a breakdown of the age distributions in these groups. Number of total unique participants by age groups North America Yasawa Adults (19 +) 79 (62 women) 19-67 yrs. mean 29.30 112 (65 women) 21.09-80.26 yrs. mean 38.51 Adolescents (13-18) 27 (25 girls) 14.75-18 yrs. mean 17.39 21 (10 girls) 13.01-18.32 yrs. mean 14.40 Children (4-12) 55 (34 girls) 4.13-12.27 yrs. mean 7.25 135 (71 girls) 4.02-12.96 yrs. mean 7.82 Babies (< 4) 25 (7 girls) 1.48-3.87 yrs. mean 3.12 82 (39 girls) 0.45-3.96 yrs. mean 2.22 Table 4.1 Age distribution across age groups 4.2.2 Procedure All participants followed the same basic procedure of watching pre-recorded videos of puppet shows followed by the prompt to choose between puppets with the question, “Who do you like?” Site-specific procedural details are elaborated below: 132 184.108.40.206 Yasawa All study materials were administered by two female researchers - one to present the video and one to present the puppet choice. The experimenter who presented the puppet choice was always an Indigenous Fijian research assistant from another part of Fiji who was fluent in both English and Standard Fijian. Splitting up the video and choice presentations allowed for the experimenters to be blind to conditions to avoid biasing participants’ choices. Pre-school aged children participated by watching videos and choosing puppets while sitting in a caregiver’s lap. Caregivers closed their eyes before the puppets were presented during the choice phase to avoid influencing children’s choices. School children participated alone with two female researchers in a house near the school or in a pop-up tent. Adults participated alone in their houses such that only they could see what the puppets they chose did in the videos. Experimenters presented one video from each condition to pre-school aged children per day, returning on subsequent days to show up to four conditions to each child. This spacing allowed for maximum data collection (as the population of Yasawa is limited) and minimum fatigue for the youngest participants. School-aged children saw puppet shows from two randomly selected conditions in one sitting. This was possible due to older children’s larger attention span and also facilitated maximum data collection with minimum disruption to children’s schooling. Adults watched up to four randomly selected conditions in one sitting. Adults in 2014 also answered a brief follow-up questionnaire about their choices to assess how adults were interpreting the shows after each video. Table 4.2 shows an overview of conditions in 2013 shows and Table 4.3 shows an overview of 2014 shows. Visual details of these stimuli are shown in Figure 4.2 (2013 shows) and Figure 4.3 (2014 shows). Because Yasawan infants and adults were recruited in their homes and Yasawan school children at their schoolhouse, these puppet shows were pre-recorded and presented as videos. 133 The videos were presented to Yasawans on field laptops: 13-inch late 2010 MacBook Pro (2013 and 2014), 12.1-inch Dell Latitude E4200 (2014), or 14-inch Toshiba Satellite C640 (2013).12 All language-based study materials were translated into Standard Fijian dialect by Indigenous Fijian research assistants fluent in both English and Fijian. These translations were also back-translated into English to check for translational accuracy. 220.127.116.11 North America North American children participated alone in a laboratory study room with one female researcher who played videos in a pre-arranged, randomized order, such that she was blind to show condition when presenting puppets for the choice. Children participated in two randomly selected conditions to match the data collection among Yasawan school-aged children. North American children viewed these shows on a 13-inch late 2010 MacBook Pro. All adults watched videos from four-five conditions in a randomized, counter-balanced order from an online survey, even if they answered questions about the videos in person in the lab. This in-lab data condition is an important check on the potential implicit social influence of having another human record answers, as was inevitable in the Yasawa data collection set up. The in-lab adults participated on lab desktop computers with LCD displays while on-line adults participated using computers of their own selection. The adults answered the same short questionnaire about each puppet choice after seeing each video and a brief demographics/ mental state reasoning measurement questionnaire following all videos. Table 4.4 gives an overview of study conditions presented to 12 The specific laptop used to present videos to each participant varied from day to day as each participant was recruited. Because research assistants and the field researcher (McNamara) visited each house at arbitrary intervals based on participant availability and interest, the laptops are not systematically related to participants’ viewings. 134 North American participants. North Americans saw all puppet shows in Figure 2 and the target of help/ hinder control condition of the box show depicted in Figure 1 B. 4.2.3 Stimuli 18.104.22.168 Puppet Shows The puppet show stimuli depict simplified social/ moral dilemmas that present forced choices between combinations of positive/ negative outcomes and positive/ negative intentions. These puppet shows follow the same scenarios presented as live-action puppet shows to infants in North America (Hamlin, 2013; Hamlin et al., 2007). The puppets used in each show are depicted in Figure 4.2 and Figure 4.3, with the specifics of each sample’s forced choice pairings outlined in Table 4.2, Table 4.3, and Table 4.4. For our recruitment, our target sample size for each pairing of puppets (e.g. fox vs. horse, red vs. yellow) was 64 participants total. Due to restrictions on the range of color combinations in puppets and limited attention spans among participants, we presented participants with one of the two possible intent/ outcome pairings listed for each pair of choice puppets (e.g., when participants were shown the Red vs. Yellow show in Table 2/ 3, they were randomly assigned to either Successful Help vs. Successful Hinder or Failed Help vs. Failed Hinder). This data collection design produced a mixed within and between subjects analysis that is described in further detail for each phase of the analysis in section 4.3. Each show features three puppets: a protagonist trying to accomplish a goal (open a box, play with a ball) at center stage, and two actor puppets at either side that interact with the protagonist. These interactions may be attempting to help (positive intent) or hinder (negative intent) the protagonist puppet in completing its goal; the outcome for the protagonist may be a completion of its goal (positive outcome) or failure to achieve its goal (negative outcome). The 135 actor puppets are then presented to the participant to choose between, a choice is taken as an indication of the participant’s preference for various intentions and outcomes (ex: choosing a puppet with a positive intention over a puppet with a negative intention indicates a preference for positive intentions). In control conditions, the participants were asked to choose between puppets that mimicked the intentional motions of test scenarios, but did not directly interact with the protagonist puppet and so served as a neutral intent. The stomper condition mimics the motion of the negative intent but only moves after the protagonist has fallen down on the table and stopped moving; the coordinated actor mimics the up-and-down motions of an attempt to help lift open a box open but does not touch the box. In both stomper and coordinated action scenarios, the protagonist puppet experiences a negative outcome and fails to achieve its goal of opening the box. In another control condition, participants choose between puppets that were the targets of helpful or hindering actions. Actor puppets’ interactions with the Protagonist puppet were presented one at a time in counter balanced order, repeated on a loop. Up to the age of 4 years, these shows were repeated until the child looked away for 10 seconds. School children older than 4 years and adults watched 3 repetitions of choice puppet actions, for a total of 6 repetitions. The total viewing time spent on each puppet show video ranged from 1-5 minutes (depending on each child’s attention). The shows feature two scenarios in which the puppets can help or hinder each other: bouncing a ball and retrieving an object out of a box. In the scenario with the puppets bouncing a ball (Ball show), one puppet is the target of the helpful or hindering behaviour, and two other puppets switch off in interacting with the target puppet after the target puppet drops the ball. The helper puppet rolls the ball back to the target puppet then runs off stage; the hinderer puppet takes the ball off stage. In the box-opening scenario, a target puppet looks into a transparent 136 plastic box, and then attempts to lift its lid to retrieve an item inside (a blue toy truck in 2013; a red toy crab in 2014) three times before the actor puppets intervene. Figure 4.2 2013 Yasawa Puppet shows: A) Ball Show, B) Box Show, C) Failed Attempt A"B"C" 137 Figure 4.3 2014 Yasawa Puppet shows: A) Red vs. Yellow, B) Orange vs. Green, C) Teal vs. Purple, D) Blue vs. Green A"B"C"D" 138 22.214.171.124 Test Conditions We investigate what aspects of actor behaviour – intentions, outcomes, or success in actors achieving their own goals – are most salient in moralistic social judgments across the lifespan. To do so, we systematically vary combinations of intention and outcome and present binary forced choices to participants across four test conditions. The conditions are structured to systematically vary one aspect of the actor’s behaviour at a time, thus allowing us to examine which elements are most important for driving social decision making. 126.96.36.199.1 Successful Attempts To test whether participants exhibit general preference for positivity (in either intent or outcome) by presenting a forced choice between two actor puppets: one exhibiting a positive intent resulting in a positive outcome (successful helper), and another exhibiting a negative intent resulting in a negative outcome (successful hinderer). The successful helper puppet aids the protagonist puppet in opening the box, and the protagonist jumps down and grabs the item in the box while the helper runs off stage; the successful hinderer jumps on the box to close it, then the protagonist puppet jumps onto the stage and lies flat while the hinderer runs off stage. If participants have a preference for the protagonist completing its goal, then they should systematically prefer the successful helper. Though this test case should present the clearest choice to participants regardless of focus on intent, outcome, or success (side puppet producing an outcome that matches its intention, thus achieving its own goal), this condition cannot differentiate which of these possible selection criteria might be most important. The following conditions vary these combinations of potential choice criteria to illuminate their relative salience. 139 188.8.131.52.2 Failed Attempts To test whether participants prefer positive intentions despite negative outcomes (or positive outcomes despite negative intentions), all failed attempt conditions present a forced choice between a positive intention with a negative outcome (unsuccessful helper) vs. a negative intention with a positive outcome (unsuccessful hinderer). The unsuccessful helper puppet attempts to aid the protagonist puppet lift the lid, then struggles with the protagonist before giving up. The protagonist then jumps back onto the stage and lies flat while the failed helper puppet runs off stage. The failed hinderer comes forward to jump on the box lid as before, but the protagonist manages to lift the lid on its own anyway. The target puppet then jumps forward and grabs the item in the box before the failed hinderer runs off stage. If positive intentions are driving preferences, then participants should favour the puppet that shows a helpful intention despite the negative outcome. However, if outcomes are more important for determining preferences, then participants should prefer the failed hinderer despite the negative intention. If outcome and intention are equally important for determining preferences – for example, if an actor puppets’ success at achieving their own helpful or hindering goals are more important than either intent or outcome in isolation – then participants should show no systematic preference. 184.108.40.206.3 Same Outcome, Different Intent To test whether participants will differentiate social actions based upon intent alone, a third condition presents a forced choice in which the outcome the same but the side puppets’ intentions vary. Half of the participants choose between a successful helper and a failed hinderer (both resulting in a positive outcome for the protagonist); the other half choose between a failed helper and a successful hinderer (both resulting in a negative outcome). If positive intentions are driving preferences, then participants should favour the puppet that shows an intention of helping 140 the protagonist open the box regardless of the outcome produced. However, if outcomes are more important for determining preferences, then participants should show no systematic preference for either puppet. 220.127.116.11.4 Same Intent, Different Outcome As a contrast to the same outcome, different intent condition, we test whether participants will focus on outcome when intent does not differentiate actor puppets. Half of the participants choose between a successful vs. failed helper (both have positive intent but only the successful helper produces a positive outcome); the other half chose between a successful vs. failed hinderer (both with negative intent but only the failed helper produces a positive outcome). If participants are focusing primarily on positive outcomes for the protagonist puppet, then they should choose the successful helper and the failed hinderer. If participants are instead focused on choosing actor puppets that successfully complete their own goal, then they should prefer the successful helper and the successful hinderer (despite the negative outcome for the protagonist). 18.104.22.168 Control Conditions Our test conditions do not systematically control for possible low-level mechanisms that may be driving preferences without inference of intent, such as general positivity bias or preference for the visual display of a puppet moving in a particular way. To help eliminate these possible reasons for participants’ preferences, we run a number of control conditions that eliminate the chance of choosing based upon intent. If participants show systematic preferences, then we may consider that the visual aspects of the puppet shows might be driving these preferences instead of inferences of intent. 141 22.214.171.124.1 Neutral Intent with Negative Outcome in Coordinated Actor vs. Stomper To test whether participants had a preference for the motions puppets made independent from the interactions they have with the protagonist puppet, we present a forced choice between actions that mimic the puppets’ motions that convey intent, but make intent neutral by showing puppets that are not directly interacting with the protagonist. The protagonist puppet again looks inside a transparent box before attempting to lift its lid three times. The coordinated action puppet is the neutral intent match to the helper – it comes forward to the side of the box and moves up and down with the protagonist puppet as it raises the lid up and down. After three lid lifts, the target puppet stops lifting the lid, jumps onto the stage, and lies flat. Then the coordinated action puppet runs off stage. The stomper is the neutral intent match to the hinderer – the protagonist puppet attempts to lift the box lid as usual, but gives up on its own. A string holds the lid of the box slightly open after the protagonist puppet jumps off to lie flat on the stage floor. Once the protagonist has given up, the stomper runs forward and jumps on the box lid to close it, then runs off stage. By removing the direct interaction with the protagonist and holding the protagonist’s outcome constant, we can isolate whether participants might have a preference for the lower-level physical motions of the puppets without referring to more complicated cognitive functions like mental state reasoning. If participants are focused more on the mentalistic elements of these scenarios, then we expect them to show no systematic preference for either puppet. However, it is possible that some of the physical motions puppets make are more appealing – more dynamic, more amusing – than the other (for example, participants of all ages often laugh at the hinderer/ stomper’s jump onto the box, even if those old enough to speak may explicitly state they think the puppet’s intentions were negative). If there is such a motion preference, this scenario should help reveal it. 142 126.96.36.199.2 Choose Target of Successful Help or Successful Hinder For our second control condition, we flip the scenario and ask participants to choose between the targets of social action rather than the actors themselves. In these shows, the center puppet performs the same action sequence as the successful helper or successful hinderer in the conditions described above while the side puppets come forward one-at-a-time to attempt to complete their goal. Participants then choose between a side puppet that was helped vs. a side puppet that was hindered. Participants’ preferences in this scenario may be interpreted in both mentalistic and non-mentalistic terms. If participants prefer a lower-level focus on positivity outside of the social dimensions of interaction, participants should focus only on the positive outcome and choose the target of helping behaviour. In mentalistic terms, participants may be taking the protagonists’ helpful or harmful actions as information about the target puppets’ social standing. Participants may take the protagonist’s help as a sign of the intrinsic goodness of the help target, which should lead them to prefer the target of helping. On the other hand, participants may instead moralize the protagonists’ actions and take pity upon the target of hindrance by choosing it. However, if participants were strictly choosing between puppets based on moralistic evaluation of the specific puppet’s behaviour (and not on how other puppets treat it), then we would expect to see no systematic preferences. 188.8.131.52.3 Choose Target of Neutral Intent with Negative Outcome Our third control condition asks participants to choose between recipients the neutral intent actions described in our first coordinated action/ stomper control condition. Here, participants cannot differentiate target puppets based on outcome, so the lower level preference for positive outcomes cannot drive these choices. Participants should be less able to take the intentional valence of the protagonist puppet as an indication of the target puppet’s social value, 143 so we do not expect to find any systematic preferences in this scenario; any such preference would most likely be driven by co-incidence (i.e. coinciding with moving up and down vs. jumping onto the box) and have very little to do with the puppets’ social interactions. 144 Side Puppet Actions Outcome by Intent Choosing Show Scenario Puppets Successful help (returns ball) vs. Successful hinder (takes ball off stage) + intent, + outcome vs. - intent, - outcome ACTOR Ball Show Bouncing a Ball Bunny center; Choose Horse vs. Fox Successful help (returns ball) vs. Successful hinder (takes ball off stage) + intent, + outcome vs. - intent, - outcome TARGET Ball Show Bouncing a Ball Bunny center; Choose Horse vs. Fox Successful help (opens box) vs. Successful hinder (closes box) + intent, + outcome vs. - intent, - outcome ACTOR Box Show Opening a box Pig center; Choose Donkey vs. Monkey Successful help (opens box) vs. Successful hinder (closes box) + intent, + outcome vs. - intent, - outcome TARGET Box Show Opening a box Pig center; Choose Donkey vs. Monkey Failed help (cannot open box) vs. Failed hinder (cannot keep box closed) + intent, - outcome vs. - intent, + outcome ACTOR Failed Attempt Opening a box Lamb center; Choose Chicken vs. Parrot Neutral help (Coordinated Action) vs. Neutral Hinder (Stomper) neutral intent, - outcome vs. neutral intent, - outcome ACTOR Failed Attempt Opening a box Lamb center; Choose Chicken vs. Parrot Table 4.2 Overview of Puppet Show stimuli for Yasawan participants in 2013. 145 Side Puppet Actions Outcome by Intent Choosing Show Puppets Successful help (opens box) vs. Successful hinder (closes box) + intent, + outcome vs. - intent, - outcome ACTOR Red vs. Yellow Panda center; Choose red vs. yellow hippo Failed help (cannot open box) vs. Failed hinder (cannot keep box closed) + intent, - outcome vs. - intent, + outcome ACTOR Red vs. Yellow Panda center; Choose red vs. yellow hippo Successful help (opens box) vs. Failed hinder (cannot keep box closed) + intent, + outcome vs. - intent, + outcome ACTOR Orange vs. Green Lamb center; Choose orange vs. green penguin Failed help (cannot open box) vs. Successful hinder (closes box) + intent, - outcome vs. - intent, - outcome ACTOR Orange vs. Green Lamb center; Choose orange vs. green penguin Successful help (opens box) vs. Failed help (cannot open box) + intent, + outcome vs. + intent, - outcome ACTOR Teal vs. Purple Puppy center; Choose teal vs. purple tiger Successful hinder (closes box) vs. Failed hinder (cannot keep box closed) - intent, - outcome vs. - intent, + outcome ACTOR Teal vs. Purple Puppy center; Choose teal vs. purple tiger Neutral help (Coordinated Action) vs. Neutral Hinder (Stomper) neutral intent, - outcome vs. neutral intent, - outcome ACTOR Blue vs. Green Bunny center; Choose blue vs. green pony Neutral help (Coordinated Action) vs. Neutral Hinder (Stomper) neutral intent, - outcome vs. neutral intent, - outcome TARGET Blue vs. Green Bunny center; Choose blue vs. green pony Table 4.3 Overview of puppet show stimuli for Yasawan participants in 2014. All scenarios feature protagonist puppet attempting to open a box. 146 Side Puppet Actions Outcome by Intent Choosing Show Puppets Successful help (opens box) vs. Successful hinder (closes box) + intent, + outcome vs. - intent, - outcome ACTOR Red vs. Yellow Panda center; Choose red vs. yellow hippo Failed help (cannot open box) vs. Failed hinder (cannot keep box closed) + intent, - outcome vs. - intent, + outcome ACTOR Red vs. Yellow Panda center; Choose red vs. yellow hippo Successful help (opens box) vs. Failed hinder (cannot keep box closed) + intent, + outcome vs. - intent, + outcome ACTOR Orange vs. Green Lamb center; Choose orange vs. green penguin Failed help (cannot open box) vs. Successful hinder (closes box) + intent, - outcome vs. - intent, - outcome ACTOR Orange vs. Green Lamb center; Choose orange vs. green penguin Successful help (opens box) vs. Failed help (cannot open box) + intent, + outcome vs. + intent, - outcome ACTOR Teal vs. Purple Puppy center; Choose teal vs. purple tiger Successful hinder (closes box) vs. Failed hinder (cannot keep box closed) - intent, - outcome vs. - intent, + outcome ACTOR Teal vs. Purple Puppy center; Choose teal vs. purple tiger Neutral help (Coordinated Action) vs. Neutral Hinder (Stomper) neutral intent, - outcome vs. neutral intent, - outcome ACTOR Blue vs. Green Bunny center; Choose blue vs. green pony Neutral help (Coordinated Action) vs. Neutral Hinder (Stomper) neutral intent, - outcome vs. neutral intent, - outcome TARGET Blue vs. Green Bunny center; Choose blue vs. green pony Successful help (opens box) vs. Successful hinder (closes box) + intent, + outcome vs. - intent, - outcome TARGET Box Show Pig center; Choose Donkey vs. Monkey Table 4.4 Overview of puppet show stimuli for North American participants. All scenarios feature protagonist puppet attempting to open a box. 147 4.3 Results 4.3.1 Analytical Approach We use binary logistic regression to analyze the probability of choosing puppets based on intent or outcome for the protagonist puppet. We first look for broad patterns of focus on intent or outcome by looking at combined results across shows. We then shift our analysis to specific pairings of intent and outcome that force puppets to choose only on intent, only on outcome, or ignore the mismatch between intent and outcome in failed attempts. 4.3.2 Combined Analysis: Intent vs. Outcome Focus We first look across all shows that allow participants to distinguish choice puppets based upon either intent or outcome. Because participants made choices in multiple conditions, we analyze data with multiple observations of the same individual using regular OLS regression with standard errors adjusted for these repeated observations (cluster-robust standard error corrections). We analyze this data in the R statistical programming environment (R Development Core Team, 2008) with the car (Fox & Weisberg, 2011) package. 184.108.40.206 Combined Test Conditions: When do Participants Choose Positive Intent? For our broadest analysis of when participants prefer positive intent, we combine results from shows that present forced choices that present the option for participants to distinguish puppets based upon intent. These shows include: successful helper vs. successful hinderer conditions from 2013 ball and box shows, red vs. yellow 2014 shows, failed attempt conditions from 2013 and 2014, and same outcome/ different intent shows. Sample sizes of individual choice observations and unique participants are reported in Table 4.5. 148 Positive Intent Across Shows: Number of total observations and unique participants North America Yasawa Adult (19+) 152 obs. (82 pos. outcome) 79 participants 185 obs. (97 pos. outcome) 98 participants Adolescents (13-18) 32 (16) 27 20 (12) 15 Children (4-12) 48 (22) 55 176 (80) 114 Babies (<4) 25 (9) 19 129 (60) 72 Table 4.5 Culture x Age Group sample sizes for individual observations of puppet choices and unique participants who made repeated choices in shows that allowed puppets to be distinguished based on intent. The sample size of the subset of individual puppet choices that included choosing the puppet associated with the positive outcome are reported in parentheses next to the repeated choice observation sample sizes. The total number of unique participants is reported underneath the choice observations samples sizes. We investigate whether participants were focused on intent to determine their choices by predicting the odds of choosing the puppet with the positive intention as our dependent variable. For our independent variables, we look for differences across age groups and test whether any age group differences may also differ by culture with an interaction between our categorical variables for age group and culture. We also added a dummy variable that accounts for whether the outcome was positive or negative to this interaction. This analytical strategy allows us to examine whether any cultural and age differences in odds of choosing the positive intent differ based upon whether the intention resulted in a positive vs. negative outcome. We look at the contrast between these positive and negative outcomes: When participants chose the puppet associated with a positive outcome, what are the odds that the chosen puppet was acting with an intention to help? When participants chose the puppet that’s actions were followed by a negative outcome, what are the odds that the chosen puppet was acting with an intention to help? 149 A Wald chi-squared test calculated with variances adjusted for clustered observations shows a significant difference between the North American and Yasawan developmental trajectories, depending on whether the outcome was positive or negative (Outcome x Culture x Age group !2 (3) = 7.0, p = 0.03). These developmental trajectories are shown in Figure 4.4 as the average log-odds of age groups in each culture choosing positive intent across outcomes. The estimated average odds ratios for each age group by culture and outcome are reported in Table 4.5. Figure 4.4 Log Odds of choosing positive intent across successful attempts, failed attempts, and same outcome/ different intent shows. Results in log odds to preserve linear scale across age and cultural groups. Error bars show robust standard errors corrected for clustered data around multiple observations of individual participants. -2 -1 0 1 2 3 Log Odds Choose Positive Intent Positive Outcome North America -2 -1 0 1 2 3 Negative Outcome Yasawa 150 Positive Outcome Negative Outcome North America Yasawa North America Yasawa Adult (19+) 5.83*** [3.14, 10.84] 1.62* [1.08, 2.44] 6*** [2.89, 12.46] 1.15 [0.72, 1.82] Adolescents (13-18) 7.00** [1.68, 29.24] 3.00 [0.77, 11.66] 4.33* [1.3, 14.48] 1.00 [0.25, 4.06] Children (4-12) 2.14 [0.83, 5.51] 2.08** [1.34, 3.22] 2.71* [1.12, 6.60] 1.04 [0.70, 1.56] Babies (<4) 3.50† [0.86, 14.21] 0.82 [0.5, 1.34] 0.45 [0.16, 1.28] 0.92 [0.56, 1.50] Significance codes: ‘***’ 0.001,‘**’ 0.01, ‘*’ 0.05, ‘†’ 0.1 Table 4.6 Odds ratios with 95% CI calculated using cluster robust adjusted standard errors. Average odds of choosing positive intent across cultures and age groups. 220.127.116.11.1 Positive Outcome We first consider the odds of choosing positive intent when the chosen puppet also was associated with a positive outcome for the protagonist. There are two mechanisms that might boost a preference for positive intent: both intent orientation in evaluating the social/ moral value of actions and success orientation in preferring actors that successfully achieve their own goals. The North American sample shows a stable preference for positive intent across the lifespan with no significant differences across the age ranges. This replicates previous findings of an early-developing positive intent/ outcome focus seen in other North American developmental studies (Hamlin, 2013). Similar to the North American results, when participants chose the puppet associated with a positive outcome, Yasawan adults and children show a significant preference for positive intent. However, the Yasawan sample shows a growth trajectory that significantly differs across age groups (Age group Wald !2 (3) = 10.8, p = 0.01), with increasing odds of choosing positive 151 intent through childhood. Yasawan adults are 1.98 times more likely to choose a positive intent resulting in a positive outcome compared to children under 4 (OR 95.CI [1.05, 3.75], p = 0.04), while children between 4 and 12 are 2.34 times more likely to choose positive intent than their younger counterparts (OR 95.CI [1.31, 4.88], p = 0.005). Interestingly, Yasawan children between 4 and 12 are 1.28 times more likely to choose positive intent than Yasawan adults (OR 95.CI [0.70, 2.33], p = 0.42). Though this difference is not measured with sufficient precision as to show statistical significance, it does suggest a distinct pattern in development compared to North Americans. Further, the North American adults are 3.60 times more likely to choose positive intent than Yasawan adults (OR 95.CI [1.72, 7.54], p = 0.001), while children between 4 and 18 do not significantly differ by culture. This further suggest something may be happening between middle childhood and adulthood to produce the cultural differences in mental state and moral reasoning observed between North American and Yasawan adults. 18.104.22.168.2 Negative Outcome Though a positive intent preference could indicate either intent orientation or success orientation when outcomes were positive, only positive intent orientation will produce a preference for positive intentions when the outcome is negative. Conversely, a success orientation will produce a significant preference for negative intentions when the outcomes are negative. Only North American adults and children continue to significantly prefer positive intent when the outcome for the protagonist puppet is negative. North American children under 4-years-old were 2.2 times more likely to choose the negative intention when the outcome was negative (95.CI [0.78, 6.17], p = 0.13), though this not statistically significant. Older children were 6 times more likely to choose a positive intention (difference between babies and children p = 0.01), adolescents 9.53 times more likely p = 0.005) and adults were 13.2 times more likely to 152 choose a positive intention (difference between babies and adults p < 0.001). However, unlike North Americans, Yasawans across all age groups showed no preference for either positive or negative intent when outcomes were negative. In the cases where Yasawans choose a puppet that was associated with a negative outcome, intent does not appear to be an important factor in influencing their preferences. 22.214.171.124 Combined Test Conditions: When do Participants Choose Positive Outcome? For our next analysis, we turn to the results of conditions that allow participants to distinguish puppets based upon outcome. If participants are adopting an outcome orientation to evaluate puppets, then they should be most likely to pick the positive outcome regardless of intent. This should be especially apparent in the case of negative intent, as positive outcome preference in the case of positive intent may also result from a success orientation. On the other hand, if participants are primarily using a success orientation to select puppets, they should instead be more likely to choose negative outcomes if the puppet also showed negative intent. If participants are using a strict intent orientation, then neither outcome should be favoured for positive or negative intent. We analyze data combined across successful helper vs. successful hinderer conditions from 2013, red vs. yellow 2014 shows, failed attempt conditions from 2013 and 2014, and 2014 same intent/ different outcome shows. Sample sizes of individual choice observations and unique participants are reported in Table 4.7. 153 Positive Outcome Across Shows: Number of total observations and unique participants North America Yasawa Adult (19+) 152 obs. (101 pos. intent) 77 participants 184 obs. (95 pos. intent) 98 participants Adolescents (13-18) 34 (25) 23 20 (11) 15 Children (4-12) 48 (27) 42 180 (82) 117 Babies (<4) 20 (10) 17 121 (60) 70 Table 4.7 Culture x Age Group report of sample sizes for individual observations of puppet choices and unique participants who made repeated choices in shows that made it possible to distinguish puppets based on outcome. The sample size of the subset of individual puppet choices that included choosing the puppet associated with the positive intent are reported in parentheses next to the repeated choice observation sample sizes. The total number of unique participants is reported underneath the choice observations samples sizes. A Wald chi-squared test calculated with variances adjusted for clustered observations again shows the North American and Yasawan developmental trajectories differ significantly, this time depending on whether the intent was positive or negative (Intent x Culture x Age group !2 (3) = 6.4, p = 0.042). These developmental trajectories are shown in Figure 4.5 as the average log-odds of age groups in each culture choosing positive outcome as a result of either positive or negative intent. The estimated average odds of choosing positive outcome for each age group by culture and intent are reported in Table 4.8. 154 Figure 4.5 Log Odds of choosing positive outcome across successful attempts, failed attempts, and same intent/ different outcome shows. Results presented in log odds to preserve linear scale across age group and cultural comparisons. Error bars show cluster robust standard errors corrected for multiple observations. Chose Positive Outcome Chose Negative Outcome North America Yasawa North America Yasawa Adults (19+) 1.59* [1.04, 2.44] 1.16 [0.79, 1.7] 1.55 [0.86, 2.81] 1.07 [0.7, 1.63] Adolescents (13-18) 2.8* [1.09, 7.16] 1.75 [0.45, 6.85] 2.75† [0.91, 8.32] 0.8 [0.25, 2.58] Children (4-12) 1.14 [0.57, 2.3] 1.23 [0.83, 1.82] 1.57 [0.61, 4.02] 0.52** [0.33, 0.8] Babies (<4) 1.5 [0.42, 5.39] 0.82 [0.48, 1.4] 0.25† [0.05, 1.25] 0.85 [0.48, 1.51] Significance codes: ‘***’ 0.001,‘**’ 0.01, ‘*’ 0.05, ‘†’ 0.1 Table 4.8 Odds ratios with 95% CI calculated using cluster robust adjusted standard errors. Average odds of choosing positive outcome across cultures and age groups. -2 -1 0 1 2 Log Odds Choose Positive Outcome Positive Intent North America -2 -1 0 1 2 Negative Intent Yasawa 155 126.96.36.199.1 Positive Intent Unlike the analysis modeling when participants choose positive intent, this analysis for outcome is measured with fewer results that reach statistical significance. North Americans and Yasawans show similar patterns of increasing odds of choosing positive outcome across age groups. We again see that all North American age groups are around 1.5 to 1.6 times more likely to choose the puppet associated with the positive outcome when it also exhibited a positive intent, though only the adults do so significantly more than might be expected by chance. Yasawans adults and children over 4-years also show a trend toward favouring the positive outcome after showing a positive intent, though none are significantly more likely to pick positive outcome than might be expected by chance. 188.8.131.52.2 Negative Intent When we model the odds of participants choosing a puppet associated with a positive outcome even though it expressed a negative intent, however, we find a significantly different odds of choosing the positive outcome marginally significantly differ across North American and Yasawan age groups (Negative intent Culture x Age Group Wald !2 (3) = 5.3, p = 0.07). Though adults show very similar odds of choosing positive outcome regardless of intent, neither is significantly above chance. Children between 4 and 12-years-old, however, show the biggest differences between the two cultures: Yasawan children are 1.93 times more likely to choose the negative outcome (p = 0.003), which is 3.03 times more likely than North American children (OR CI.95 [1.46, 6.28], p = 0.036). North American children under 4 years are also 4 times more likely to choose the negative outcome when the puppet expressed a negative intent, though the preference is not sufficiently precise to reach significance (p = 0.09). These results suggest that, when presented with a situation that leads them to choose a puppet that expressed a negative 156 intent, North Americans older than 4 show an outcome orientation by preferring the puppet that failed in its negative intentions and produced a positive outcome. Yasawan children, on the other hand, may be following a success orientation by choosing the puppet that produced the negative outcome when they are choosing a puppet that also expressed a negative intent. By adulthood, Yasawans no longer show a clear preference for either outcome when intent is negative. Though Yasawan Opacity of Mind norms may suggest that Yasawan adults should be more likely to choose based upon outcome, they do not appear to be using an outcome orientation to distinguish puppets. 4.3.3 Targeted Analyses of Specific Intent/ Outcome Pairings In our pooled results above, it remains somewhat unclear what the exact preferences for intent or outcome are across age groups and cultures. For example, including failed attempts in this pooled analysis may attenuate the apparent preference for outcome or intent because participants could be choosing based upon either. To further delineate what the choice rules that participants might be using to make these socio-moral decisions, we analyze specific pairings of intent/ outcome. When presented a situation that only allows puppets to be distinguished by intent, only allows them to be distinguished by outcome, or forces participants to ignore the mismatch between intent and outcome, we can further highlight how these different inputs are evaluated. We again use the car (Fox & Weisberg, 2011) package in the R statistical programming environment (R Development Core Team, 2008) to analyze these data. The results from shows that force participants to choose based upon intent by holding the outcome constant (section 184.108.40.206.3) or on outcome by holding the intent constant (section 220.127.116.11.4) both have only one observation per participant, so we use OLS logistic regression with unadjusted standard errors 157 and 95% confidence intervals calculated using log likelihood estimation. We have a few repeated observations of forced choices pitting failed attempts against each other in the Yasawan sample (section 18.104.22.168.2), so that analysis again employs a cluster robust standard error adjustment to account for these repeated observations. 22.214.171.124 Same Outcome, Different Intent Our next targeted intent vs. outcome analysis focuses on shows that force the participants to choose based upon intent by holding the outcomes constant between the two choice puppets. This data comes from the positive outcome, positive vs. negative intent (Successful Helper vs. Failed Hinderer) and negative outcome, positive vs. negative intent (Failed Helper vs. Successful Hinderer) shows described in Table 4.3 and Table 4.4. If participants are adopting an intent orientation, they should be most likely to choose the positive intent in both the positive and negative outcome conditions. On the other hand, if participants are adopting a success orientation, they should be more likely to choose the positive intent only in the positive outcome condition while also being more likely to prefer the negative intent in the negative outcome condition. If participants are adopting an outcome orientation, then they should show no particular preference for either intention because the outcomes are the same. The raw numbers of participants who chose positive intent out of all participants in either the positive outcome or negative outcome conditions, plus the total number who chose positive intent out of all participants who participated in this condition, are shown in Table 4.9. 158 Positive Outcome Show Negative Outcome Show Total North America Yasawa North America Yasawa North America Yasawa Adult (19+) 33 pos. intent/ 34 participants 23/ 31 36/ 42 21/ 32 69/ 76 44/ 63 Adolescents (13-18) 8/ 8 4/ 4 9/ 11 1/ 4 17/ 19 5/ 8 Children (4-12) 9/ 14 20/ 31 9/12 19/ 30 18/ 26 39/ 61 Babies (<4) 3/ 4 8/ 26 2/ 8 12/ 25 5/ 12 20/ 51 Table 4.9 Raw counts of participants who chose positive intent when outcomes are held constant and total sample size from each age group across cultures. We do not find a statistically significant difference between North America and Yasawan age trajectories in this forced choice (Wald test with unadjusted variances Intent x Culture x Age group !2 (3) = 3.9, p = 0.14). These age patterns are depicted as average log-odds per cultures’ age group in Figure 4.6. The first obvious impression from reviewing the raw results in Table 4.9 is overwhelming preference for positive intent among both adults and children over 4-years-old in both North American and Yasawan samples. This holds for North American participants over 4-years and Yasawan adults regardless of outcome. 159 Figure 4.6 Average log odds of choosing positive intent when outcomes are the same for positive and negative outcomes across age ranges and cultures. Error bars show standard errors. For the positive outcome condition, adolescents in both cultural groups selected the positive intent every time. This produced undefined estimates for this age group in this condition that are therefore not graphed here. Positive Outcome Show Negative Outcome Show North America Yasawa North America Yasawa Adults (19+) 33.00*** [7.12, 586.94] 2.88* [1.34, 6.86] 6.00*** [2.72, 15.84] 1.91† [0.94, 4.11] Adolescents (13-18) undefined [0, 0] undefined [0, 0] 4.50† [1.16, 29.52] 0.33 [0.02, 2.6] Children (4-12) 1.8 [0.62, 5.86] 1.82 [0.89, 3.93] 3.00† [0.90, 13.52] 1.73 [0.84, 3.75] Babies (<4) 3.00 [0.38, 60.65] 0.44† [0.18, 0.99] 0.33 [0.05, 1.45] 0.92 [0.42, 2.04] Significance codes: ‘***’ 0.001,‘**’ 0.01, ‘*’ 0.05, ‘†’ 0.1 Table 4.10 Odds ratios with 95% CI calculated using profile likelihood and unadjusted standard errors. Average odds of choosing positive intent across cultures and age groups. -2 -1 0 1 2 3 4 5 Log Odds Choose Positive Intent Positive Outcome North America -2 -1 0 1 2 3 4 5 Negative Outcomes Yasawa 160 126.96.36.199.1 Positive Outcome Though the overall increasing odds of choosing positive across age groups does not significantly differ between North Americans and Yasawans (!2 (3) = 3.00, p = 0.21), the North American adults are 11.48 times more likely to choose the positive intent than Yasawan adults (OR CI.95 [1.92, 220.52], p = 0.03). Adolescents in both cultural groups chose positive intent for every one of these positive outcome shows, such that their estimated odds are undefined in this regression. Children show similar odds of choosing positive intent when the outcomes are positive (OR = 1.01, CI.95 [0.26, 3.72], p = 0.99). Yasawan age groups do show an overall difference (!2 (3) = 10.80, p = 0.004) that is driven by the children younger than 4. North Americans all overwhelmingly prefer the positive intent, though the sample size of the youngest North America children is too small to make firm claims about them. The Yasawans under 4-years, on the other hand, do permit some stronger inference; these children are 2.25 times more likely to choose the negative intent rather than positive intent (p = 0.056), while older Yasawan children are 4.91 times more likely to choose positive intent than younger children (OR CI.95 [1.69, 15.41], p = 0.005) and adults are 4.09 times more likely to choose positive intent (OR CI.95 [1.39, 13.00], p = 0.01). Though the North Americans, older Yasawan children, and Yasawan adults may be considered to be showing either an intent or success orientation, this seems not to be the case for the youngest Yasawans. However, given that the youngest Yasawans do not show the same preference for the hinderer in other shows, this may be evidence that, all else being equal, the visual display of the puppet is more appealing to the youngest Yasawans. 188.8.131.52.2 Negative Outcome The cultures again do not significantly differ in their overall changes in odds of choosing positive intent across age ranges (!2 (3) = 3.90, p = 0.28). The effect sizes for the negative 161 outcome condition are smaller and measured with less precision than in the positive outcome condition, though the difference between positive and negative outcome conditions never reaches statistical significance for any age group for either culture. This may indicate that the negative outcome makes the positive intention less salient – perhaps indicating a need to suppress an outcome calculation (for dual processes in moral reasoning, see: Cushman, 2008; Cushman et al., 2013) – but does suggest that the positive intent remains the preferred selection principle for most of the participants sampled. Despite the lack of evidence for overall differences across cultures, the North Americans do show a significant overall difference across age groups (!2 (3) = 9.70, p = 0.008). This is again driven by the difference between the children younger that 4, who are significantly less likely to choose the positive intent than older children and adults. However, the small sample size makes strong inference about this difference difficult. North American adults are 3.14 times more likely to choose positive intent in the negative outcome condition than the Yasawan adults (OR CI.95 [1.04, 10.31], p = 0.05). Though the North American children over 4-years are also 1.74 times more likely to choose positive intent than Yasawan children of the same age range, this difference fails to reach statistical significance (OR CI.95 [0.41, 9.08], p = 0.47). Though the sample sizes are too small for strong inferences, adolescents show very different choice patterns; Yasawans prefer the negative intent – which may point to a success orientation. Taken together, these results suggest that children older than 4 years and adults in both cultures do reference an intent orientation when making choices about puppets that produce the same outcome. Further, the focus on positive intent is present among adults in both cultures, but the North Americans appear to show a much stronger focus on positive intent even when the outcome is negative. 162 184.108.40.206 Same Intent, Different Outcome For our next targeted intent/ outcome choice, we investigate how participants select puppets when only the outcome can be used to distinguish them. By holding the intent constant, we can more accurately detect if participants are adopting an outcome orientation or a success orientation. If participants predominantly favour an orientation that focuses on the positive outcome for the protagonist puppet, then they should be most likely to choose the positive outcome regardless of a positive or negative intent. However, if participants are primarily using an intent orientation, then they should show no preference. If participants are adopting a success orientation, then they should prefer the positive outcome only in the positive intent condition; they should be more likely to instead choose the negative outcome in the negative intent condition. The raw counts of individuals who chose the positive intent across cultures and age groups by the two intent conditions are reported in Table 4.11. This forced choice based on outcome was the only condition that showed a potentially significant difference between the university student population and the North American adult community sample (university – community OR = 0.39, CI.95 [0.13, 1.12], p = 0.08), so these sub-sample counts are listed separately. 163 Positive Intent Negative Intent Total North America Yasawa North America Yasawa North America Yasawa Adult (19+) Uni: 21/ 30 Com: 4/ 10 14/ 32 Uni.: 16/ 27 Com: 4/ 9 17/ 30 Uni.: 37/ 57 Com: 8/ 19 31/ 62 Adolescents (13-18) 8/ 9 2/ 3 9/ 12 1/ 5 17/ 21 3/ 8 Children (4-12) 10/ 14 20/ 34 9/ 12 13/ 31 19/ 26 33/ 65 Babies (<4) 2/ 3 8/ 20 1/ 4 13/ 23 3/ 7 21/ 43 Table 4.11 Raw counts of participants who chose positive outcome when intentions are the same. This is the only condition that showed a potentially significant difference between the two sub-samples of North American adults, so this table reports their raw counts separately. Total participants who chose positive outcome across both conditions are reported in the Total column. Uni. = university sample; Com. = community sample. We fail to find sufficient evidence for an overall difference between the changes in odds of choosing positive outcome across North American and Yasawan age groups based on puppets’ intention (Culture x Age Group x Intent !2 (3) = 3.7, p = 0.29). The average log-odds for each age group across cultures and intent conditions are shown in Figure 4.7. The average odds-ratios of each age group across culture in each intent condition are reported in Table 4.12. 164 Figure 4.7 Average log odds of choosing positive outcome when intentions are the same for positive and negative intentions across age ranges and cultures. Error bars show standard errors. North American university and community adults were marginally significantly different, so we separate them here. Positive Intent Show Negative Intent Show North America Yasawa North America Yasawa Adult (19+) Uni.: 2.14 * [1.07, 4.53] Com.: 0.84 [0.30, 2.30] 0.78 [0.38, 1.56] Uni.: 1.59 [0.78, 3.34] Com.: 0.62 [0.21, 1.74] 1.31 [0.64, 2.75] Adolescents (13-18) 8.00* [1.47, 148.42] 2.00 [0.19, 43.04] 3.00† [0.90, 13.52] 0.25 [0.42, 2.04] Children (4-12) 2.50 [0.84, 9.11] 1.43 [0.73, 2.89] 3.00† [0.90, 13.52] 0.72 [0.35, 1.46] Babies (<4) 2.00 [0.19, 43.0] 0.67 [0.26, 1.60] 1.31 [0.64, 2.75] 1.30 [0.57, 3.05] Significance codes: ‘***’ 0.001,‘**’ 0.01, ‘*’ 0.05, ‘†’ 0.1 Table 4.12 Odds ratios with 95% CI calculated using profile likelihood and unadjusted standard errors. Average odds of choosing positive outcome across cultures and age groups. Uni. = university sample; Com. = community sample. -3 -2 -1 0 1 2 3 Log Odds Choose Positive Outcome Positive Intent N. Am (University) -3 -2 -1 0 1 2 3 Negative Intent Yasawa N. Am (Community) 165 220.127.116.11.1 Positive Intent In the positive intent condition, only the university sample of North American adults and the North American adolescents show a significant preference for positive outcomes when both puppets expressed a helpful intent. The university participants were 2.56 times more likely to choose positive outcome than the North American community sample (though this is not statistically significant p = 0.08) and 2.75 times more likely than the Yasawan adults (CI.95 [1.03, 7.7], p = 0.047). The North American community sample and the Yasawan adults, however, do not show significant evidence of difference in choosing positive outcome (N. Am. Community – Yasawa OR = 1.08, CI.95 [0.31, 3.70], p = 0.91). These results are consistent with the North American children and university participants showing either an intent or success orientation. Importantly, outcomes do not appear to be differentiating the puppets in these scenarios for Yasawan adults. This then further suggests that, for Yasawans, both intent and outcome matter – especially in adulthood. Yasawans appear to be choosing based on intent, but the outcome appears to also be a more important factor in their choices than it is for North Americans. 18.104.22.168.2 Negative Intent When both puppets express a negative intent, only North American children and adolescents between 4 and 18-years-old show a marginally significant preference, and that is for the positive outcomes. Because these are mismatched to the intentions, the North American children and adolescents are showing a general positivity bias with an outcome orientation when the intentions are the same. However, given their preferences for positive intentions in other conditions, it would appear that outcome is a secondary orientation when intention cannot distinguish the puppets. This appears to attenuate for adults in both the university and 166 community samples of North Americans. Compared to North American children, the Yasawan children are 4.15 times more likely to pick the negative outcome when the intentions are both negative (OR = CI.95 [1.02, 21.60], p = 0.06). The Yasawan children and adolescents’ preference for negative outcome when the intention is negative do not reach statistical significance (p = 0.37), it does suggest that they are more likely to adopt a success orientation when intentions are indistinguishable than the North Americans of similar ages. 22.214.171.124 Failed Attempts: Preference for Positive Intent Despite Negative Outcome? Our last test condition forces participants to choose between failed attempts. If participants are choosing based upon an intent orientation, they must ignore the negative outcome to select a positive intention. Similarly, a participant adopting an outcome orientation will have to choose a puppet that expressed a negative intent. If participants are adopting a success orientation, participants should not show any preference. The only show with a forced choice between two failed attempts for the North American sample was the failed helper vs. failed hinderer half of the Red vs. Yellow show described in Table 4.4 and shown in Figure 4.3. Yasawan participants made choices about contrasted failed attempts in shows from both the 2013 (see Table 4.2) and 2014 (see Table 4.3) field seasons, so a few Yasawan participants presented multiple observations. Because there are some repeated observations in this data, we again use robust cluster adjusted standard errors to account for the repeated measures in the Yasawan sample. The total number of observations, number of observations of participants choosing positive intent, and the total number of participants across each culture’s age groups are shown in Table 4.13, along with the results of logistic regression predicting the odds of choosing a positive intent despite a negative outcome for each culture’s age groups. The age trends in choosing 167 positive intent for each culture are shown as average log odds for each culture’s age groups in Figure 4.8. Chose Positive Intent Across Failed Attempts Shows: Number of total observations and unique participants Logistic Regression Odds Ratios North America Yasawa North America Yasawa Adult (19+) 35 obs. (24 chose pos.) 35 participants 55 (26) 52 2.18 * [1.22, 3.90] 0.90 [0.50, 1.60] Adolescents (13-18) 6 (4) 6 6 (3) 5 2.00 [1.12, 3.57] 1.00 [0.58, 1.79] Children (4-12) 12 (10) 12 45 (30) 42 5.00* [2.80, 8.93] 2.00* [1.12, 3.57] Babies (<4) 4 (3) 4 36 (21) 33 3.00 [1.68, 5.36] 1.40 [0.78, 2.50] Significance codes: ‘***’ 0.001,‘**’ 0.01, ‘*’ 0.05, ‘†’ 0.1 Table 4.13 Culture x Age Group report of sample sizes for puppet choices and unique participants in shows that present puppets producing outcomes that mismatched their intentions. The numbers of participants who chose positive intent are reported in parentheses next to the repeated choice observation sample sizes; total number of unique participants underneath the choice sample sizes. Average odds of choosing positive intent for each culture’s age groups reported with 95% CI calculated using cluster robust standard errors. 168 Figure 4.8 Average log odds of choosing positive intent between failed helpers and failed hinderers. Error bars show cluster robust standard errors. Both North American children over 4-years and adults show a significant preference for the positive intent despite it’s being paired with a negative outcome. The Yasawan children over 4-years also show this preference, though the Yasawan adults show no preference for either intent. North American adults are 2.29 times more likely to choose the positive intent puppet than Yasawan adults (CI.95 [0.98, 6.04], p = 0.056) and Yasawan children are 2.43 times more likely to choose the positive intent than Yasawan adults (CI.95 [0.80, 7.39], p = 0.06), though the children in both cultures do not significantly differ (p = 0.28). Taken together, these findings not only suggest that Yasawan children do adopt an intent orientation – especially when success is not a distinguishing factor – but also that something happens between middle childhood and adulthood to make Yasawan adults place equal weight on outcomes and intentions such that neither can produce a significant preference when intentions and actions do not align. -2 -1 0 1 2 3 Babies (<4) Children (4-12) Adolescents (13-18) Adults (19+) Log Odds Choose Positive Intent North America Yasawa 169 4.3.4 Control Conditions: Neutral Intentions and Targets of Action Across our test conditions, it is not possible to distinguish a preference for positive intent or outcome from a general positivity bias, nor is it possible to eliminate the possibility that participants simply preferred the visual characteristics of the motions puppets were making without inferring any intent behind their motions. To help eliminate these less interesting possibilities, our final analysis will focus on conditions with neutral intentions or choices between targets of action rather than the actors themselves. If we see a consistent preference for either coordinated actor that is moving up and down (as seen in the positive intent conditions) or stomper that is jumping on the box (as seen in the negative intent conditions), we might infer that this low-level physical characteristic of the visual display may be driving preferences more than the inferred intentions or outcomes of the puppets’ actions. The sample sizes and raw counts of participants choosing various puppets across these shows are reported in Table 4.14. Coordinated Action/ Stomper (neutral intent) Target of Successful Help or Hinder Target of Neutral Intent Coordinated Action/ Stomper North America Yasawa North America Yasawa North America Yasawa Adult (19+) 32 obs. (28 chose Coord. Act.) of 32 participants 50 (34) 50 43 obs. (25 chose Helper Target) of 43 41 (24) 33 41 obs. (20 chose Coord. Act Target) of 41 31 (9) 31 Adolescents (13-18) 8 (7) 8 6 (3) 6 6 (3) 6 4 (3) 4 7 (3) 7 7 (4) 7 Children (4-12) 13 (10) 13 46 (19) 43 13 (6) 13 25 (17) 22 9 (4) 9 28 (15) 28 Babies (<4) 6 (3) 6 34 (12) 32 4 (2) 4 20 (8) 20 8 (3) 8 23 (16) 23 Table 4.14 Raw counts of participants who chose coordinated actor, target of help, or target of coordinated action out of total observations and unique participants. 170 The results of these shows can be seen in Table 4.15. For the coordinated action/ stomper conditions, the results are similar to the same outcome/ different intent test shows described in section 126.96.36.199.3. Both North Americans and Yasawans show a steady increase in their preference for the coordinated actor, the puppet that makes the same motions as the positive intent puppet in test conditions. When asked why they chose these puppets, North American participants often cited an inferred positive intent. The youngest Yasawans show a marginally significant preference for the stomper in this condition, which may suggest a general preference for the jumping motion – many participants laugh at this display. In the absence of more direct social interaction, the visual characteristics of the jumping on the box may be more attractive, especially to younger Yasawans. Alternatively, the negative outcome for the protagonist in this condition may indicate Yasawan children infer negative intent and choose according to a success orientation that gradually shifts to a preference for the coordinated actor in adulthood. Though the neutral intent show may have been interpreted as intentional, the next control conditions eliminate the possibility to choose based upon intent by asking participants to choose between targets of positive, negative, or neutral intentions. When participants were asked to choose between targets of successful helpers vs. successful hinderers, only Yasawan children over 4-years show a preference for the target of help. Both North American and Yasawan adults show the same trend toward also favouring the target of help, but neither reaches significance. When asked to choose between targets of coordinated action or stompers, Yasawan adults are the only group that shows a preference for targets of the stomper. Yasawan children of both age groups trend toward preferring the target of the coordinated actor (unlike the potential preference for the motion of the coordinated actor in our first control condition). Because both the older children and adults in the Yasawan sample did not show a consistent bias toward positive intentions or 171 outcomes across all shows (as the North Americans did), these control results help eliminate the possibility that participants are picking based upon the low-level visual aspects of the shows. Coordinated Action/ Stomper (neutral intent) Target of Successful Help or Hinder Target of Coordinated Action/ Stomper North America Yasawa North America Yasawa North America Yasawa Adult (19+) 5.60*** [3.09, 10.15] 2.12* [1.17, 3.85] 1.39 [0.75, 2.60] 1.41 [0.75, 2.65] 0.95 [0.38, 2.37] 0.41* [0.16, 1.03] Adolescents (13-18) 7.00† [3.86, 12.67] 1.00 [0.55, 1.81] 1.00 [0.19, 5.18] 3.00 [0.29, 30.65] 0.75 [0.29, 1.88] 1.33 [0.53, 3.34] Children (4-12) 3.33† [1.84, 6.04] 0.70 [0.39, 1.28] 0.85 [0.28, 2.63] 2.13† [0.95, 4.77] 0.80 [0.32, 2.00] 1.15 [0.46, 2.89] Babies (<4) 1.00 [0.55, 1.81] 0.55† [0.28, 1.10] 1.00 [0.13, 7.48] 0.67 [0.27, 1.67] 0.60 [0.23, 1.50] 2.29† [0.91, 5.73] Significance codes: ‘***’ 0.001,‘**’ 0.01, ‘*’ 0.05, ‘†’ 0.1 Table 4.15 Average Odds of participants from age groups of both cultures choosing the coordinated actor in the coordinated action/ stomper (neutral intent) show, target of help in the target of helping or hindering show, or target of coordinated action in target of neutral intent show. The neutral intent/ negative outcome and target of helping or hindering shows had repeated measurements of a few participants, so 95% CI (presented in brackets) were calculated using cluster robust standard errors; 95% CI shown in brackets for the target of neutral intent show were calculated using profile likelihood. 4.4 Discussion Across these simplified socio-moral dilemmas, we presented our participants with choices that pitted intent against outcome. By looking across the patterns of their choices, we see the signature of intent processing as the primary choice metric for both North Americans and Yasawans from at least middle childhood on. Though Opacity of Mind norms in Yasawan culture might lead to the expectation that Yasawans should favour an outcome orientation, they do not appear to be making decisions based on outcome when the intentions are the same. However, when intentions can be used to differentiate puppets producing the same outcome, Yasawan adults and children both hone in on positive intent. Yasawan adults deviate from both 172 North American adults and younger Yasawans when the intentions and outcomes do not match; Yasawan adults do not favour positive intentions that lead to a negative outcome (the puppet show equivalent to accidental hindering) over negative intentions that lead to a positive outcome (a failed attempt at hindering). Yasawan children also appear to prefer positive intent even when both outcomes are negative, though this preference is potentially weaker than the preference seen among North American children. However, when Yasawan children are compelled to choose based upon outcomes produced by the same negative intent, they diverge from both North American children and Yasawan adults by showing a tendency toward preferring the puppet that succeeded in carrying out their negative intent to produce a negative outcome for the protagonist. When the option to choose a puppet that succeeds in carrying out it’s own goal is eliminated, however, Yasawan children again favour positive intent. Taken together, these results suggest that intent is the primary metric for judging social actors in both cultures from middle childhood onward. However, when the option to choose based upon intent is removed, North Americans favour positive outcome. Yasawan children may instead prefer the puppet that succeeds in accomplishing its own goals when intentions are the same – a preference that still requires a calculation of the puppet’s intent – but Yasawan adults do not appear to strongly differentiate characters when intent is the same. Therefore, North Americans’ secondary selection strategy appears to be positivity of outcome while younger Yasawans may be selecting based upon ability to achieve one’s own goals. Finally, when the option to choose a successful character is removed, Yasawan children again appear to favour intent in their selection criteria. Interestingly, when Yasawan children differ from North American children, it appears to be in the direction of favouring a successful character rather than necessarily a positive one. Why 173 might these children be following a success orientation? Focus on success may reflect the greater emphasis on informal cultural learning around traditional Fijian practices that are important for successful village life (Kline, 2013; Kline et al., 2013). Because traditional village teaching does not focus as much on explicit pedagogy as much as children’s own abilities to focus on the right teachers and learn by observation, Yasawan children may be more oriented toward finding successful models – even if these models are negative - to boost their success in culturally acquiring traditional practices essential to village life. On the other hand, children in the United States have been shown to use actor benevolence to infer how much an actor knows (Johnston, Mills, & Landrum, 2015; Landrum, Mills, & Johnston, 2013; Landrum, Pflaum, & Mills, 2016). This tendency to see niceness as an indicator of being knowledgeable may further reflect the general positivity orientation seen in the present data, and it may also indicate a different social learning orientation in these two cultures. Though we do find some indication that children four-years-old and older are likely preferring puppets based upon some calculation of intent, our data for younger children show no clear pattern. Rather than being evidence for lack of mental state reasoning among these participants, the null results here are likely a result of the differences between our field set-up for the shows and the more controlled setting in the lab. Our puppet shows were presented on laptops that have many distracting features that draw infants’ attention away from the task at hand. We ran some comparison data with North Americans in a laboratory setting that replicated the field set up as much as was possible, and also found that North American infants were not choosing puppets in any coherent pattern. Other work in laboratory settings suggests that presenting stimuli without additional auditory cues to attend to results in similarly ambiguous results. For future research with infant cognition in field settings, we recommend establishing a 174 set-up that includes as large of a screen display as possible, minimal other features of the display item (i.e. no additional buttons), and inclusion of extra pauses between repetitions of the target display to allow for habituation both within and across stimulus displays. At the other end of our age range, the patterns of choices among our Yasawan adult participants are particularly striking. Despite other research (Chapter 2) showing Yasawan adults respond to direct invocations of mental state reasoning with less intensity than other cultural groups, one might expect Yasawan adults to show a preference for outcome rather than intent. Instead, we find that they appear to be very focused on intent. Their inclusion of intent in their choices is especially clear in their responses to the choices between puppets that expressed the same intent but produced different outcomes. Many participants in both North America and Yasawa stated they either wanted to pick both (in the case of positive intent) or neither (for negative intent) – many participants stated that they had a hard time choosing between the two puppets because the were ‘the same.’ These results show that Yasawan adults are indeed thinking a lot about the intent dimension of these shows North American adults reported similar difficulty deciding between the two puppets showing the same intent. Along the lines of results shown in chapter 3, Yasawan adults appear to also be accounting for outcome more than North Americans – as is most obvious in their responses to the choices between mismatched outcome and intent. For our Yasawan adult participants, a positive intent was not enough to overcome a negative outcome and vice versa, such that there was no overall preference. While these results suggest universality in mental state reasoning for making the most basic moral judgments, more fine-grained detail about how intent and outcome are weighed in more complex reasoning around blame or punishment may still vary even at the youngest ages. Future research toward pinpointing the exact timing of shifts in explicit judgments based upon 175 intent or outcome may further delineate how these processes work in different cultural environments. However, perhaps the most important implication from this research is that culture may be increasing or decreasing a baseline tendency to mentalize in social judgments. Though both cultural groups show evidence of intentionality reasoning, the children in North America and Yasawa were in many cases more similar to each other than the adults. The social constraints on behaviour in the more relational, collectivistic, and interdependent framework of Yasawan culture may reduce the emphasis or necessity for fluency in mental state reasoning by making situation more deterministic of behaviour. However, it may also be that North American culture so heavily emphasizes behaviour as a window into a person’s soul – that people can and should be read like books – that North American children grow up into hyper-mentalizers as adults. Cross-cultural differences in making the Fundamental Attribution Error and dispositionism may further suggest this possibility (E. E. Jones & Harris, 1967; Ross, 1977). North American participants are far more likely to exhibit a tendency to interpret all behaviour as driven by a person’s internal characteristics, even when the situation dictates people must act against their internal preferences. However, people from more collectivistic societies more appropriately cite the situation as causing many behaviours (Choi & Nisbett, 1998; Miller 1984). Our results compliment existing research on how other elements of one’s cultural environment shape cognition. Much of this research has focused on how linguistic differences influence the structure of cognition and perception around categorizing items based on shape vs. material (Lucy & Gaskins, 2003), on how tight or loose the fit of a container is (S. Choi, 2006; S. Choi, McDonough, Bowerman, & Mandler, 1999), and chunking continuous light waves into discrete colour categories (Franklin et al., 2008; Ozgen & Davies, 2002; Robertson, Davies, & Davidoff, 2000; Robertson, Davies, Davidoff, & Shapiro, 2005). These studies provide evidence 176 for a common cognitive orientation seen early in development that is later modulated into language-specific cognitive patterns as learners grow into the language. The present research contributes to and expands upon this body of research by emphasizing how non-linguistic social norms might also modulate cognition in culturally specific ways. One potential difference between cognitive differences transmitted by norms vs. differences transmitted by language could be that the differences that arise from norms may show up even later in development. This may be especially relevant to social norms, as children are often socialized into these norms, and these norms are often more strictly enforced, as children transition into membership in the adult community through middle childhood and adolescence. 4.5 Conclusion Our results suggest that mentalistic interpretation of actions for the purposes of social evaluation do indeed form a core part of human social cognition across cultures, even when the cultural environment might inhibit its expression. We further find that this cognitive process appears to be modulated into different adult forms throughout development. Though Yasawan norms discourage using mental state reasoning to explain behaviour, Yasawan adults appear to be using both intention and outcome to formulate their social preferences in these simplified socio-moral dilemmas. Yasawan children appear to use intent more, and may grow into roughly equivalent focus on intent and outcome later in life. These results further suggest that cross-cultural differences in how cultural norms may define behaviour to such an extent that the actually modulate cognitive processing around mental state reasoning into uniquely cultural adult forms. 177 Chapter 5. Conclusion: How Culture Interacts with Cognition to Shape Mind Throughout this dissertation, I have investigated how our abilities to think about other people in terms of inner mental states like goals, intentions, desires, and emotions are shaped by group-level cultural differences. Much of the existing cross-cultural research in psychology has emphasized the contrast between individual or community-oriented values. I take this a step further by focusing the present research on the case of Indigenous Fijians living in traditional villages on Yasawa Island, Fiji. As a small-scale society that is still highly connected to family, tradition, and the ancestral lands they live in, Yasawan culture provides an interesting point of comparison to the Western, Educated, Industrial, Rich, and Democratic cultural groups that form the bulk of existing psychological research on how we think about minds (WEIRD people, J. Henrich et al., 2010). Like other groups that have been the focus of existing cross-cultural psychological research, Yasawan culture emphasizes community values. However, Yasawans add a further layer of cultural structuring around how people should think about other minds through specific norms about how one should think about others’ actions in terms of mental states. Yasawan culture includes a normative stance, otherwise referred to as Opacity of Mind, that restricts thinking about thinking as the cause of actions and instead promotes focus on external cues to explain others’ behaviours. As such, research conducted with the participation of Yasawan villagers provides a window into how particular cultural norms have direct impacts on psychological functioning. Because norms are transmitted through cultural learning and are subject to change over time in the context of specific environmental challenges, this research may further illuminate the ways that culture acts as an extra layer of adaptation to help mould individual psychology to suit the demands of particular social worlds. 178 The results of chapter 2 demonstrate how different facets of mentalizing processing are affected by culture. The first study of chapter 2 focuses on the theory of mind components of mentalizing via responses to an adult-appropriate false belief task. In this task, Yasawans show less focus on a character’s false beliefs and are more influenced by additional information about an item’s true location than North American university students. Study 2 of chapter 2 expands upon the results of study 1 to examine emotional, empathic processing and comfort in social situations in addition to the thought/ belief orientation in theory of mind processing. Study 2 focuses on self-reported results measured by the Empathy Quotient-short (EQ) to examine how Yasawans, non-Indigenous Indo-Fijians, and North American adults (both university students and adults outside of university) differ in self-reported levels mentalizing abilities across multiple facets of social cognition. The included contrast with Indo-Fijians is especially illuminating in that Indo-Fijians share national institutional structures with Indigenous Fijians and are also more community-oriented than the average, individualistic North American participant. However, Indo-Fijians do not share a traditional normative stance that proscribes thinking about thoughts as the explanation for behaviour. Study 2 results indicate that the EQ taps into three facets of mentalizing and social cognition in approximately the same way across all participating cultural groups. These facets include items that measure theory of mind processing around thoughts and internal mental states; empathic processing around the emotional lives of others; and ease with navigating social situations. Because Yasawan cultural norms specifically focus on thinking about thoughts, only the theory of mind/ internal mental states EQ items should be affected by these norms. Further, because the cultural norms around thinking about minds should be causing this difference, Indo-Fijians (whose cultural norms do not proscribe thinking about thoughts) should not show a reduction in self-reported fluency in 179 thinking about internal mental states. The results of chapter 2 study 2 support these predictions; the tendency to think less about beliefs that emerged among Yasawan participants in study 1 is further supported by a significantly lower emphasis on self-reported fluency in thinking about internal mental states. However, Yasawan participants did not show any less interest in emotions or any more difficulty with understanding social situations than other groups. This difference in lower self-reported tendency to think about internal mental states was specific to Yasawans; Indo-Fijians reported as much fluency in thinking about mental states as North Americans. Further, the difference between Yasawan and North American participants in false belief responses found in chapter 2 study 1 can be explained by this specific difference in self-reported focus in EQ mental states items. Therefore, chapter 2 shows preliminary evidence that Yasawans are less oriented toward theory of mind processing, but not empathic processing, and that this difference may be the result of specific Yasawan cultural norms above and beyond national identity and communal values. A major limitation to the two studies in chapter 2 is that they rely upon self-reports and somewhat abstract hypothetical situations that may be less culturally meaningful outside of North American contexts. For example, the finding that Yasawan participants were more swayed by additional information about an item’s new location in a false belief task may be reflective of actual differences in mental state reasoning, or it may reflect deference to the authority of the experimenter. If our Yasawan participants were merely trying to be respectful of the experimenter, then they might boost their self-reported expectations that the character would look in this incorrect false belief location (even though they are fully aware that the character should not look there first) simply because the experimenter mentioned this new information. One way to get around this issue is to extend the logic implied by the results in chapter 2 to 180 examine how underlying cultural differences in mental state focus might affect social judgments that require mental state reasoning. Chapter 3 expands upon chapter 2’s apparent cross-cultural differences in mentalizing by investigating cross-cultural patterns of intent vs. outcome focus in moral judgments. By examining moral judgments as a case of applied mental state reasoning, we gain perspective on how focus on internal mental states (via intent) or focus on external situational factors (via outcome) might impact social decisions in a way that may be less impacted by cultural influences that have nothing to do with mental state processing. Study 1 of chapter 3 presents baseline differences in focus on intent or outcome across Yasawans, non-Indigenous Indo-Fijians, and North American adults. Differences in focus on intent or outcome should be most apparent in the contrast between violations that have a negative intent with a positive outcome (failed attempts) and violations that have a positive intent with a negative outcome (accidents). If participants are more focused on internal mental states, then negative intent should be judged as more wrong and receive harsher punishments regardless of outcome. Conversely, if participants are more focused on situational factors, then negative outcomes should be judged as more wrong and receive harsher punishment regardless of intent. The results of study 1 in chapter 3 support the conclusion that Yasawans are indeed less reflexively focused on internal mental states; they rate accidents as a bit worse than failed attempts and a bit more deserving of punishment than failed attempts. This is in sharp contrast to Indo-Fijians and North Americans; Indo-Fijians favour intent over outcome in their judgments overall, and North Americans privilege intent to such an extent that failed attempts may be even worse and more punishable than successful intentional violations. Study 2 of chapter 3 attempts to further isolate this pattern in moral reasoning as a result of underlying differences mental state processing activation. If thinking 181 about thoughts is just not as salient to Yasawans as it is to North Americans, then being induced to think about thoughts before judging moral norm violations should make Yasawans focus more on intent than is apparent without these reminders. Chapter 3 study 2 results show that Yasawan participants do indeed place more emphasis on intent when induced to think about thoughts. This supports the conclusion that the differences seen in study 1 are a result of differing chronic activation of intent focus. In other words, though the results of chapter 2 might imply that Yasawans may be less capable of thinking about thoughts, results in chapter 3 show that ability is likely not an issue – mental state reasoning applied to interpret perpetrator intent is still an important part of moral reasoning in Yasawan culture. Instead, the reflexive emphasis in moral judgments within Yasawan culture includes additional weight on outcome that is not seen in other cultural groups. What remains to be seen in investigating the cultural differences in baseline focus on intent demonstrated in chapter 3 is whether these differences are present across the lifespan, or whether they may be the product of becoming a fully encultrated member of adult society. Chapter 4 investigates how cultural norms shape applied mentalizing in moral reasoning across the lifespan by comparing how children and adults in Yasawa and North America respond to simplified moral dilemmas. These simplified moral dilemmas present scenarios that compel participants to choose characters based upon different combinations of intent and outcome. These explicit contrasts between positive and negative outcomes and intentions allow for even more fine-grained insight into how intent and outcome are weighed in adjudicating socio-moral situations. Additionally, these simplified moral scenarios are presented as puppet shows that do not rely upon language comprehension; this avoids potential issues of translation, makes the tasks more accessible to even very young children, and may tap into implicit mentalizing 182 abilities that do not require the extra executive control processing necessary for verbal mentalizing and moral reasoning tasks. Even with the additional layers of complexity removed, chapter 4’s results suggest that intent focus is stronger among Yasawan children than adults. Further, North American adults appear to be more intent focused than children in either culture. This similarity in childhood and differentiation in adulthood supports the conclusion that the differences we see in adults’ mental state focus applied to moral reasoning is indeed a result of culture shaping minds as people are socialized into different cultural groups throughout development. 5.1 Psychological Universals, Core Cognition, and Enculturated Social Beings The work in this dissertation brings together two analytical strategies to uncover basic cognition mechanisms common to all humans: psychological universals seen across cultures and core cognition seen across development. In cultural psychology, much of the work on variation in cognition across cultural groups is aimed at identifying common mental attributes, or psychological universals, that are present regardless of cultural differences (Norenzayan & Heine, 2005). In developmental psychology, one major theoretical approach to documenting cognition as it unfolds across early life is to find the basic set of cognitive mechanisms that underlie all human cognition across the lifespan (Spelke & Kinzler, 2007). Both approaches share the same underlying theoretical proposition that all humans are more psychologically similar than they are different. This psychic unity of humanity approach is appealing both empirically13 and politically.14 The present research supports the hypothesis that certain social 13 For example, at a genetic level, human populations are generally more similar across than within groups, which makes the proposition that there is a genetic basis for race untenable; see for example (Maglo, Mersha, & Martin, 2016). 183 cognitive mechanisms oriented toward thinking about minds are indeed common across cultures, even when cultural circumstances may make them less likely to be evident. This research further supports the hypothesis that these particular mechanisms may be active early in life and further modulated by cultural influences across development. One implication of the present research is further support for the emerging consensus that mentalizing processes involve multiple distinct ways of thinking about minds: thinking about thoughts, knowledge, and belief through theory of mind processing and thinking about emotions through empathic processing (Saxe et al., 2004; Schaafsma et al., 2015). In chapter 2, mentalizing as thinking about beliefs, thoughts, and intentions emerged as distinctly less prominent in Yasawans’ self reported thinking about minds, though the overall interest in the emotional lives of others did not differ across cultural groups. This research also highlights the theory of mind (rather than empathic) component of intentionality reasoning in moral judgments. Though reactions to moral norm violations may be heavily influenced by one’s own emotional response (Haidt, 2001; Laham, Chopra, & Lalljee, 2010; Nelissen & Zeelenberg, 2009), the process of determining how bad an action is and how much it should be punished may be more driven by focus on mental states (Cushman, 2008; 2015). Because the internal mental state focus in theory of mind processing seems to be specifically less chronically active in the Yasawan cultural context, theory of mind processing is a strong candidate for the cognitive root behind the difference in baseline intent vs. outcome focus in chapter 3 study 1. 14 The proposition that all psychologies are basically the same has a strong claim against the hierarchical supposition behind movements like eugenics, which seek to direct human population change to make the average ever more like an ideal type (Mead, 1964). 184 Chapter 3 study 2 further isolates this thinking about thoughts difference by experimentally inducing greater salience in thoughts or actions before making moral judgments. Because Yasawan participants can be induced to focus more on intent, the difference in thinking about thoughts that emerged in chapter 2 and in chapter 3 study 1 can perhaps best be interpreted as an issue of access rather than function. The mentalizing processes underlying intent focus may well be performing the same functions in Yasawan culture and in other cultural groups, but the baseline salience and access to thought focus and intent orientation may not be the same. These results further imply that priming thoughts about thoughts may also boost mental state focus in other theory of mind domains like false belief. Further research with other populations that are less inclined to reflexively think about thoughts, beliefs, and intentions like very young children or people with autism spectrum diagnoses may reveal other ways that theory of mind processing is involved in these social decisions. If the differences seen across cultures are specifically related to differences in activation rather than function, then we might expect adults with autism-spectrum disorders to respond to experimental manipulations increasing thought salience in proportion to the extent of their theory of mind impairment. If increasing the salience of thought makes accessing theory of mind processing easier, this may also result in children performing better on false belief tasks at younger ages. Finally, the results in chapter 4 highlight similarity in intent focus in childhood that diverges according to cultural influences into adulthood. Importantly, children in both Yasawan and North American cultural contexts show approximately similar intent focus that becomes more extreme among North American adults and less extreme among Yasawan adults. The early cross-cultural similarity supports the hypothesis that theory of mind reasoning is a core cognitive process because it shows up earlier in life regardless of cultural influences that may suppress its 185 expression. These results hone in on the interplay between basic cognition common to all humanity and cultural influences that shape psychology into culturally-appropriate orientations as children grow up. Interestingly, the most noticeable difference between Yasawan and North American children came not in outcome focus, but in preference for choosing a based upon a success orientation or positivity bias when they choose a character that acted with a negative intent. If North American children choose a puppet that acted with negative intent, they tended to select one that produced a positive outcome for the protagonist. Conversely, Yasawan children tended to choose the puppet that produced a negative outcome – the puppet that succeeded in carrying out its own goal – if that puppet also acted with a negative intent. This difference in childhood may hint at the different functions theory of mind processing and intentionality reasoning might perform in North American vs. Yasawan culture. Specifically, the orientation toward successful characters might be related to the higher emphasis on children learning traditional practices through observational learning rather than direct instruction. Though characters that produce negative outcomes appear not to be Yasawans’ primary preference, if a positive intent is not chosen, then Yasawans may be more oriented toward finding and attending to successful others in the social environment. This success bias may benefit Yasawan children in particular in that they can direct more of their attentional resources toward learning from those who best know those aspects of village life that they may not be directly taught about otherwise. 5.2 Where Do We Go From Here? Across the five studies, I demonstrate that cognitive mechanisms dedicated to thinking about other minds can be meaningfully shaped by cultural influences. By focusing on how mental state reasoning in Yasawa, Fiji, compares to non-Indigenous Indo-Fijians and North Americans, I further argue that traditional Yasawan norms are the proximate cause of the 186 observed differences in reasoning about false beliefs, emotions, social situations, and intentions. However, identifying a cultural factor that may produce differences in cognition is just the beginning of a long line of questions. At a broader scale, is Yasawan culture suppressing mentalizing, or is North American culture enhancing it – or both? How might norms that promote thinking about motive or intent help guide people through social interactions when the situation does not provide many cues to determine how to interpret and respond to another’s actions? How might other cultural influences enhance or suppress focus on mental states by making situations more or less predictive of behaviour? In particular, how might cultural forms like politeness norms, rituals, or different ways of structuring learning environments modulate mentalizing? I highlight Opacity of Mind norms as one source for variation in how people in different cultural environments may activate cognitive processes dedicated to thinking about minds. It is tempting to think of WEIRD patterns of mentalizing as prototypical, which would imply that Yasawan mentalizing is lower than the baseline found in WEIRD culture. However, if we consider other errors in reasoning about behaviour common in WEIRD culture, the Western prevalence of phenomena like the fundamental attribution error (error in over-emphasizing internal dispositions as the cause of behaviour without taking adequate account of situational factors), dispositionism (overall tendency to prefer thinking about dispositions rather than situations as the cause of behaviour), and correspondence bias (basic assumption that behaviours are indicative of dispositions; see: Krull, 2001) suggest that the focus on mind in the West may in fact be a case of Western culture over-emphasizing mentalizing rather than other groups necessarily underemphasizing it. It is important to note, however, that the over-attribution of internal stable traits on the basis of ephemeral external behaviour is not exactly the same as over- 187 attributing mental states in general. As Malle (2006a) notes, there is a difference between internal transient states – like beliefs about an item’s location, an actor’s momentary intent, or a momentary goal – and more stable internal characteristics like extraversion or a characteristic tendency toward breaking rules. So, while differences in using false belief inference to predict behaviour and tendency to infer stable traits from observed behaviour are clearly not the same process, the combined emphasis on using inferred mental state information to predict behaviour that itself is thought to stem from internal mental characteristics may be part of the wider folk psychological model of mind in Western, individualistic cultures. Further, while people in collectivist cultures do show evidence for some baseline presumption that behaviours can be used to infer dispositions, people in individualistic cultures remain less sensitive to additional information about how situations constrain behaviour (Choi & Nisbett, 1998; Krull et al., 1999; Norenzayan, Choi, & Nisbett, 1999). The apparent greater baseline sensitivity to situational influences on behaviour results in people in collectivist cultures showing less of a tendency toward committing the fundamental attribution error and less emphasis on dispositionism overall (F. Lee, Hallahan, & Herzog, 1996; J. G. Miller, 1984; Schuster, Forsterlung, & Weiner, 1989). Other research on how Western samples attribute mind to groups and individuals does suggest that, as a collective entity (like a corporation or a sports team) is seen as more cohesive, the group itself is assigned more mind, more blame, and more agency than the individual members of that group (Waytz & Young, 2012). This tendency to attribute mind to groups may be an extension of the Western ‘mind as source of action’ model of mind applied to the overlapping selves of people in highly cohesive groups, or it may be a more general aspect of mind attribution that could be more common in more collectivistic groups. Further research on whether these more collectivistic society members would also attribute mind 188 to groups but not individuals may help determine whether this is a more general trend in mind attribution or a more specific form derived from Western cultural norms. Further, though the dispositonism orientation in individualistic cultures is often framed as a reasoning error, there is genuine benefit to using mental states to decode behaviour in cases where the situation does not provide strong determinism of one’s actions. From the perspective of an individual observer who must determine the quality of communication presented by potential social partners, the evolutionary arms race between communicators who might wish to deceive to their own benefit and listeners who need to distinguish good and bad quality information may push cognitive systems to develop a correspondence bias (Andrews, 2001). Similarly, in a social environment where situation provides minimal cues to infer behavioural expectations, a learner’s best strategy may be to reference mental states and dispositions as the cause of behaviour (Walker, Smith, & Vul, 2015). However, as mentioned in chapter 1’s literature review on theory of mind in primates, a common solution to dealing with cooperation and competition among social primates is to form dominance hierarchies. The social structuring provided by knowing one’s own and others’ relative positions in the hierarchy, as well as the behavioural privileges and obligations that come from this positioning, may off-load much of the behavioural information from thinking about minds to thinking about situations. Given that Western, individualistic cultures tend to structure social interactions around individuals as autonomous units while collectivist cultures often reference social groups of known others and hierarchies, the additional emphasis on disposition in individualistic cultures vs. the sensitivity to situation in collectivistic cultures may be a feature rather than a flaw of either cultural context (Brison, 2001; Gelfand et al., 2006; 2011; Hofstede, 1983; Leung & Cohen, 2011). 189 If cultures differ in the extent to which norms structure situations to be more or less deterministic of behaviour, a logical next step in this research program might be to investigate how certain situations might downplay the importance of mentalizing. Two interesting candidates for culturally-transmitted situational constraints that might reduce reliance on mental state reasoning are politeness norms and ritualized behavioural performances. Some of the ethnographic literature on Opacity of Mind highlights a dilemma people in these societies face: they know that people might be saying or doing one thing but really be thinking or intending another (Groark, 2011; Hollan, 2012; A. Mayer, 2013). In large part, this dilemma stems from the knowledge that people are socially expected to behave in certain ways that are independent from their personal interest, and these behavioural expectations largely come down to expectations around politeness. If a person shakes your hand, makes firm eye-contact with you, and smiles at you, but you know that this is just a script one follows to convey polite social behaviour, then you cannot refer to this positive external display as evidence that the person does not secretly want to stab you in the back. There is some evidence to suggest that people in groups with low relational mobility (structured in a way that makes it harder for people to move across different social groups) also tend to tell others less about themselves (Schug, Yuki, & Maddux, 2010). This difficulty in moving into new social groups in the event of a falling out with someone else may also reduce the tendency to display and recognize displays of negative emotion (Fernandez et al., 2000; Matsumoto et al., 2008). The sense that one has enemies, or the idea that there are group members who actively desire your failure, may also be heightened when people feel they are continually enmeshed in the same social networks even if there has been friction or disagreement among people in the group (G. Adams, 2005; G. Adams & Plaut, 2003). Therefore, living in a society that focuses on maintaining existing relationships rather than 190 making new ones may promote greater observance of politeness norms instead of self-disclosure, and it may lead some people within these groups to suspect that positive behavioural displays may not reflect (or may even actively hide) true intentions and motivations. This may be sufficient to cause people in these cultural contexts to believe that actions are not reliable indicators of mind, and thus reduce reliance on mentalizing because one’s true thoughts can never really be known. In addition to politeness norms, ritualized displays are another common feature of highly connected, communal societies. One common element of ritualized performance is behavioural synchrony across participants. Action sequences that convey a conventional or ritual stance (rather than a causal/ instrumental stance) also encourage overimitation. As discussed in the chapter 1 review of literature on synchrony leading to participants showing greater focus on each others’ mental states (Baimel, 2015; Baimel et al., 2015; Newman-Norlund et al., 2007; Sebanz et al., 2006) and theory of mind focus in normative overimitation (Ben Kenward et al., 2011; Kenward, 2012; Marsh et al., 2013), we should expect that ritualized performances will emphasize mental state reasoning to some extent. In the case of synchrony, it appears that this mental state focus is more specifically related to theory of mind/ internal mental state focus (Baimel et al., 2015). However, it is possible that the influence joint and imitative action have on mentalizing is more restricted to synching up the neural activation of participants within the ritual, such that those who are active in these practices do not become better at mentalizing overall as much as better at rapidly and intuitively gauging what the other ritual participants are thinking. This may facilitate cooperative coordinated actions among those who performed the ritual together while their minds are in this connected, synchronized state. However, that does not necessarily mean that they will also be better at inferring the mental states of others outside 191 of the ritual. If the mentalizing effects of ritualized actions are indeed a result of similar activation patterns in participants’ mirror neural pathways, then participants might almost literally be seeing out of each other’s eyes. This then would increase perspective taking abilities across the synchronized group, but it would not necessarily extend to inferring mental states that are not accessible to the individual participants. This may be a stronger case of simulation theory of mind, wherein participants are taking on the perspective of other ritual performers only in so far as they are imputing their own mental states onto others in the ritual. If this is the case, then the behavioural structuring around the ritual performance would constrict the information quality actions have for determining others’ mental states that are not already the observer’s own mental state. If members of a group frequently engage in ritual together, it is possible that this ritual experience may make members more attuned to each other but less generally prone to metalizing by thinking about what thoughts, knowledge, or beliefs others might have that are different to their own. Related to the link between ritualized performance and overimitation, group-level focus on different ways of learning socially may also have some impact on how mentalizing processes are activated. If a learner is overimitating the action sequences of a model, there are two main motivation orientations that might be driving this urge to overimitate: causal reasoning, which is especially relevant when learning action sequences that are causally opaque; and social affiliation, which is especially relevant to learning norms and is increasingly evident as children grow into adulthood (Kenward et al., 2011; Legare et al., 2015; Legare & Nielsen, 2015; Over & Carpenter, 2013). Similarly, the orientation toward causal learning may be more relevant to innovation, while orientation toward social affiliation may be more relevant to maintaining community and tradition. These different learning motivations may also be reflected in the extent 192 to which education is focused on learning a way of thinking vs. learning a body of knowledge that must be memorized. Western-style formal education is especially focused on training students to think in particular ways, often as independent entities who approach problems creatively and critically. Conversely, traditional settings tend to emphasize learning not just content but how to be an appropriate member of a communal society (Lancy & Grove, 2010). As such, learning how to be an appropriate member of society often involves learning deference to authority, knowing and observing the roles conferred by one’s place (rather than explicit ambition to be better than others), and initiative to do things in the manner deemed traditionally appropriate – rather than challenging convention. Handing down tradition in this manner that emphasizes convention rather than innovation may also have community benefits in maintaining closer ties that can be relied upon to distribute risk in a fundamentally uncertain, unstable environments (Botero et al., 2014; Fincher & Thornhill, 2011; Gelfand et al., 2006; Hruschka et al., 2014; Rubenstein & Wrangham, 1986). In a more concrete sense, societies that emphasize science as the primary way of knowing (including WEIRD cultures), may be more inclined to focus on intent as a reflection if individual, internal orientation toward an object or process. Scientism, worldview that science is the most appropriate way of knowing, may also be more directly related to culture-level differences in focus on analytical thinking and value on reason over intuition (Choi, Koo, & Jong An Choi, 2007; N. Lee & Johnson-Laird, 2006; Nisbett, Peng, Choi, & Norenzayan, 2001; Sorell, 1991; Uskul, Kitayama, & Nisbett, 2008). Analytical reasoning in particular tends to push perception toward a sense of self as separated from the environments we inhabit, while intuitive processing tends to promote a sense of expanded self and increased self-other overlap with both other social entities and other elements of the environment (Glöckner & Witteman, 2010; 193 Luhrmann, 2011; Peters, Hammond, & Summers, 1974; Rudd, Vohs, & Aaker, 2012; Wild, Kuiken, & Schopflocher, 1995). As such, learners who are enculturated into a scientistic approach to knowing may also favour logical, analytical thought that focuses on identifying internal characteristics and unseen mental states, rather than social environmental influences, as the causal factors behind behaviour. Conversely, societies that value other ways of knowing like religion/ spiritual experience or Indigenous/ traditional knowledge handed down from elders may instead focus on learning as affiliation first and causal understanding second. As such, referencing the social context and situational cues may be more appropriate for understanding and responding to others in these cultural contexts that favour tradition, affiliation, and intuition over analytical logic, individual ambition, and innovation. 5.3 What Does Learning About Yasawa Teach Us About Minds? A student once told me that cultural psychology was a dying area. Quite to the contrary, my aim is to demonstrate that cultural psychology has yet to hit its stride. I hope to situate the present research as a step toward a multilevel, interdisciplinary approach that treats culture not as noise to be factored out, but as the mechanism our species relies upon to survive and thrive in a biological world. There needs to be a place in our science for understanding how brains develop into social minds ensconced within particular environments. Culture is the key to this understanding. In chapter 1, I emphasize in my review of literature on mentalizing that human cognitive mechanisms for thinking about minds gain their greatest strength when brought to bear in group-level phenomena like cultural transmission, religion, and morality. In this dissertation, I put that claim into action by investigating how Yasawan (and North American) cultural norms around thinking about minds produce differences in social cognitive processing. Though there is a 194 growing body of psychological literature that incorporates cultural perspectives outside of Western, individualistic, large-scale societies, it is tempting to consider this Western perspective as the baseline for human reality rather than a story about WEIRD culture. This temptation is especially strong when our research is targeted at discovering those parts of human cognition that unite us as a species. While searching for the common threads behind the psychic unity of humanity, the perspective lent by laboratory-based psychological research is especially prone to the assumption that whatever we find in the lab is the prototypical baseline. Along this line of reasoning, cultural differences are easily seen as incidental variation around this panhuman prototype. There is strong, convincing evidence that humans are more alike (at least genetically) than we are different (see for example: Maglo et al., 2016). From an evolutionary perspective, the individual is often the most sensible unit of analysis because any differentiation that might arise from genetic influences should be transmitted through individuals. At the same time, humans are set apart from the rest of the biological world by our profoundly social nature. By working with (and against) each other, humans have successfully spread to every terrestrial environment on the planet, visited most marine environments, and even begun to explore to extraterrestrial environments. The secret to this profound success is that we learn from each other across generations – through culture. At the time of this writing, researchers have become more amenable to taking this cultural level of analysis seriously. Quantitative, “big data” approaches via giant databases of documented cultural variation have garnered much attention. While this big data approach is vital to understanding the broad strokes of cultural change throughout human history, focusing solely upon the largest scale still risks missing the interplay between cognition and culture. We are both individual and cultural beings; the reciprocal influence between culture and cognition is an inseparable part of what makes us human. 195 In his attempts to define what cultural psychology is and should be, Shweder (1990; 1999) emphasizes the anthropological take on culture as intentional worlds of meaning systems that result in “ethnic divergences in mind, self, and emotion” (Shweder 1990, pp. 1). The body of research I present in this dissertation comes closest to Shweder’s call to investigate how cultural forces shape us into particular cultural beings across development. The empirical toolkit of scientific psychology is especially suited to investigating general trends and bio-physiological influences on individual cognition, while the cultural nuances and localistic particulars of specific groups have generally been the purview of ethnography conducted by socio-cultural anthropologists. Because cultural psychology must account for both individual and cultural levels of analysis, it is a fundamentally interdisciplinary enterprise. However, from Shweder’s psychological anthropology/ cultural anthropology perspective, the influences cultural forms exert on individual psychology are primarily an effect of expressions and interpretations of meaning. Taking this perspective on culture implies a profoundly anthropocentric theoretical and analytical stance. It runs the risk of getting too close to documenting variation for variation’s sake without keeping an eye to the bigger picture of why cultures vary in particular ways and why cultures shape psychologies into different forms. I diverge from the anthropocentric meaning orientation in culture to instead extend our theoretical reach out to yet another level of analysis within the wider socio-ecological pressures cultural groups face. These connections among land, culture, and mind are at least in part why it is so important for cultural psychology to explicitly include perspectives from small-scale, rural, and Indigenous communities. Life in dense, globally-mobile, cosmopolitan cities gives us one viewpoint; from this perspective, the world seems small and highly interconnected, humanity’s dominion over and separation from nature seems obvious, and the nuances of place and history 196 seem unimportant. But this urbane viewpoint makes it easy to overlook the connections to history, ancestry, and knowing through tradition that forms another major current in the human experience. The inherent value of connections to community, tradition, and local identity becomes easier to overlook. Often, this inherent value of specific cultural forms is related to the demands of living in particular socio-ecological contexts. Living on, in, and with the land is often a necessity for survival and a can form key part of Indigenous identity that is more clearly visible from field than from the city center. In the course of including these perspectives into our broader understanding of how minds work through cultures within environments, we should make efforts to send institutionally-trained researchers into the field. But we should also make efforts to bring Indigenously-trained researchers into the institution. In psychology’s earliest forays into culture in the 1980s and 1990s, the North American institution of academic psychology became far more accessible to people of East-Asian descent and cultural backgrounds. In his 2011 address to the Society for Personality and Social Psychology, Haidt argued for the need to balance the liberal cultural frame within psychology by including more political conservatives in the institution (Haidt, 2011). In making a psychology that includes Indigenous perspectives, we must similarly make space for Indigenous scholars to enter in to the institution and push the institution in new directions that more accurately reflect the full scope of human experience. 197 References Abramson, A. (2000). Bounding the Unbounded: Ancestral Land and Jural Relations in the Interior of Eastern Fiji. In A. Abramson & D. Theodossopoulos, Land, law and environment: Mythical land, legal boundaries (pp. 191–210). London: Pluto Press. Adams, G. (2005). The cultural grounding of personal relationship: enemyship in North American and West African worlds. Journal of Personality and Social Psychology, 88(6), 948–968. doi:10.1037/0022-35188.8.131.528 Adams, G., & Plaut, V. C. (2003). The cultural grounding of personal relationship: Friendship in North American and West African worlds. Personal Relationships, 10(3), 333–347. doi:10.1111/1475-6811.00053 Adams, R. B., Rule, N. O., Franklin, R. G., Wang, E., Stevenson, M. T., Yoshikawa, S., et al. (2010). Cross-cultural reading the mind in the eyes: an fMRI investigation. Journal of Cognitive Neuroscience, 22(1), 97–108. doi:10.1162/jocn.2009.21187 Adolphs, R. (2009). The Social Brain: Neural Basis of Social Knowledge. Annual Review of Psychology, 60(1), 693–716. doi:10.1146/annurev.psych.60.110707.163514 Aiken, L. S., & West, S. G. (1991). Multiple Regression: Testing and Interpreting Interactions (pp. 1–26). Thousand Oaks, CA: SAGE Publications. Ali, F., & Chamorro-Premuzic, T. (2010). Investigating Theory of Mind deficits in nonclinical psychopathy and Machiavellianism. Personality and Individual Differences, 49(3), 169–174. doi:10.1016/j.paid.2010.03.027 Ali, F., Amorim, I. S., & Chamorro-Premuzic, T. (2009). Empathy deficits and trait emotional intelligence in psychopathy and Machiavellianism. Personality and Individual Differences, 47(7), 758–762. doi:10.1016/j.paid.2009.06.016 Allison, C., Baron-Cohen, S., Wheelwright, S. J., Stone, M. H., & Muncer, S. J. (2011). Psychometric analysis of the Empathy Quotient (EQ). Personality and Individual Differences, 51(7), 829–835. doi:10.1016/j.paid.2011.07.005 Anae, M. (2010). Research for better Pacific schooling in New Zealand: Teu le va – a Samoan perspective. MAI Review, 1, 1–24. Andrews, P. W. (2001). The psychology of social chess and the evolution of attribution mechanisms: explaining the fundamental attribution error. Evolution and Human Behavior, 22(1), 11–29. Andrighetto, G., Brandts, J., Conte, R., Sabater-Mir, J., Solaz, H., & Villatoro, D. (2013). Punish and Voice: Punishment Enhances Cooperation when Combined with Norm-Signalling. PLoS ONE, 8(6), e64941. doi:10.1371/journal.pone.0064941 Apperly, I. A. (2008). Beyond Simulation-Theory and Theory-Theory: why social cognitive neuroscience should use its own concepts to study "theory of mind". Cognition, 107(1), 266–283. doi:10.1016/j.cognition.2007.07.019 Apperly, I. A., & Butterfill, S. A. (2009). Do humans have two systems to track beliefs and belief-like states? Psychological Review, 116(4), 953–970. doi:10.1037/a0016923 198 Atkinson, Q. D., & Bourrat, P. (2011). Beliefs about God, the afterlife and morality support the role of supernatural policing in human cooperation. Evolution and Human Behavior, 32(1), 41–49. doi:10.1016/j.evolhumbehav.2010.07.008 Back, E., & Apperly, I. A. (2010). Two sources of evidence on the non-automaticity of true and false belief ascription. Cognition, 115(1), 54–70. doi:10.1016/j.cognition.2009.11.008 Baillargeon, R., Scott, R. M., & He, Z. (2010). False-belief understanding in infants. Trends in Cognitive Sciences, 14(3), 110–118. doi:10.1016/j.tics.2009.12.006 Baimel, A. (2015). From keeping together in time, to keeping together in mind: behavioral synchrony and theory of mind. University of British Columbia, Vancouver, BC. Baimel, A., Severson, R. L., & Baron, A. S. (2015). Enhancing “theory of mind” through behavioral synchrony. Frontiers in Psychology, 6(870), 1–6. Baird, A. D., Scheffer, I. E., & Wilson, S. J. (2011). Mirror neuron system involvement in empathy: a critical look at the evidence. Social Neuroscience, 6(4), 327–335. Bar-On, R. (1996). The Emotional Quotient Inventory (EQ-i): A test of emotional intelligence. Toronto: Multi-Health Systems. Baron-Cohen, S. (1995a). Mindblindness: An essay on autism and theory of mind. Cambridge MA: Bradford Books. Baron-Cohen, S. (1995b). The eye direction detector (EDD) and the shared attention mechanism (SAM): Two cases for evolutionary psychology. (C. Moore & P. J. Dunham)psycnet.apa.org (pp. 41–59). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc. Baron-Cohen, S. (1996). Reading the Mind in the Face: A Cross-cultural and Developmental Study. Visual Cognition, 3(1), 39–60. doi:10.1080/713756728 Baron-Cohen, S. (2002). The extreme male brain theory of autism. Trends in Cognitive Sciences, 6(6), 248–254. Baron-Cohen, S., & Wheelwright, S. (2004). The empathy quotient: an investigation of adults with Asperger syndrome or high functioning autism, and normal sex differences. Journal of Autism and Developmental Disorders, 34(2), 163–175. Baron-Cohen, S., Jolliffe, T., Mortimore, C., & Robertson, M. (1997). Another advanced test of theory of mind: evidence from very high functioning adults with autism or asperger syndrome. Journal of Child Psychology and Psychiatry and Allied Disciplines, 38(7), 813–822. doi:10.1111/j.1469-7610.1997.tb01599.x Baron-Cohen, S., Leslie, A. M., & Frith, U. (1985). Does the autistic child have a "theory of mind"? Cognition, 21(1), 37–46. Baron-Cohen, S., O'Riordan, M., Stone, V., Jones, R., & Plaisted, K. (1999). Recognition of Faux Pas by Normally Developing Children and Children with Asperger Syndrome or High-Functioning Autism. Journal of Autism and Developmental Disorders, 29(5), 407–418. doi:10.1023/A:1023035012436 199 Baron-Cohen, S., Wheelwright, S., Skinner, R., Martin, J., & Clubley, E. (2001). The autism-spectrum quotient (AQ): evidence from Asperger syndrome/high-functioning autism, males and females, scientists and mathematicians. Journal of Autism and Developmental Disorders, 31(1), 5–17. Barrett, H. C., & Broesch, J. (2008). Cross-Cultural Adult False Belief Tasks: Curse of Knowledge and Baillargeon Binder Test. AHRC Culture and the Mind Project. [Unpublished raw data]. Barrett, H. C., & Kurzban, R. (2006). Modularity in Cognition: Framing the Debate. Psychological Review, 113(3), 628–647. doi:10.1037/0033-295X.113.3.628 Barrett, H. C., Bolyanatz, A., Crittenden, A. N., Fessler, D. M. T., Fitzpatrick, S., Gurven, M., et al. (2016). Small-scale Societies Exhibit Fundamental Variation in the Role of Intentions in Moral Judgment. PNAS. [Forthcoming]. Barrett, H. C., Broesch, T., Scott, R. M., He, Z., Baillargeon, R., Di Wu, et al. (2013a). Early false-belief understanding in traditional non-Western societies. Proceedings: Biological Sciences, 280(1755), 1–6. doi:10.2307/23353353?ref=no-x-route:b7979c19dd7877929795e4aa940d02e0 Barrett, H. C., Broesch, T., Scott, R. M., He, Z., Baillargeon, R., Wu, D., et al. (2013b). Early false-belief understanding in traditional non-Western societies. Proceedings of the Royal Society B: Biological Sciences, 280(1755), 20122654–20122654. doi:10.2307/1131213 Barrett, J. L., & Johnson, A. H. (2003). The role of control in attributing intentional agency to inanimate objects. Journal of Cognition and Culture, 3(3), 208–217. Barrett, J. L., & Keil, F. C. (1996). Conceptualizing a nonnatural entity: anthropomorphism in God concepts. Cognitive Psychology, 31(3), 219–247. doi:10.1006/cogp.1996.0017 Bates, D. M. (2010). lme4: Mixed-effects modeling with R (pp. 1–145). New York: Springer. Retrieved from URL http://lme4. r-forge. r-project. org/book. Bates, D., Maechler, M., Ben Bolker, Walker, S., Christensen, R. H. B., & Singmann, H. (2014). Package “lme4.” CRAN. Bauer, M., Cassar, A., Chytilová, J., & Henrich, J. (2011). Warfare Experience during Ontogeny Increases Egalitarian and Parochial Motivations (pp. 1–60). Presented at the United Nations University, Maastricht. Bauer, M., Cassar, A., Chytilová, J., & Henrich, J. (2014). War's enduring effects on the development of egalitarian motivations and in-group biases. Psychological Science, 25(1), 47–57. doi:10.1177/0956797613493444 Beauducel, A., & Herzberg, P. Y. (2006). On the Performance of Maximum Likelihood Versus Means and Variance Adjusted Weighted Least Squares Estimation in CFA. Structural Equation Modeling: a Multidisciplinary Journal, 13(2), 186–203. doi:10.1207/s15328007sem1302_2 Bentler, P. M., & Chou, C. P. (1992). Some new covariance structure model improvement statistics. Sociological Methods & Research, 21(2), 259–282. 200 Berg-Cross, L. G. (1975). Intentionality, Degree of Damage, and Moral Judgments. Child Development, 46(4), 970. doi:10.2307/1128406 Bering, J. M. (2006). The folk psychology of souls. Behavioral and Brain Sciences, 29(5), 453–498. Bering, J. M., & Bjorklund, D. F. (2004). The Natural Emergence of Reasoning About the Afterlife as a Developmental Regularity. Developmental Psychology, 40(2), 217–233. doi:10.1037/0012-16184.108.40.206 Berthoz, S., Wessa, M., Kedia, G., Wicker, B., & Grèzes, J. (2008). Cross-cultural validation of the empathy quotient in a French-speaking sample. Canadian Journal of Psychiatry, 53(7), 469–477. Birch, S. A. J., & Bloom, P. (2007). The curse of knowledge in reasoning about false beliefs. Psychological Science, 18(5), 382–386. doi:10.1111/j.1467-9280.2007.01909.x Blair, R. (2008). Fine cuts of empathy and the amygdala: dissociable deficits in psychopathy and autism. The Quarterly Journal of Experimental Psychology, 61(1), 157–170. Blair, R. J. R. (2002). Theory of Mind, Autism, and Emotional Intelligence. In L. F. Barrett, P. Salovey, & J. D. Mayer, Wisdom in Feeling: Psychological Processes in Emotional Intelligence (pp. 406–434). New York: Guilford Press. Bloom, P. (2004). Descartes' Baby: How the Science of Child Development Explains What Makes Us Human. New York: Basic Books. Boiger, M., Güngör, D., Karasawa, M., & Mesquita, B. (2014). Defending honour, keeping face: Interpersonal affordances of anger and shame in Turkey and Japan. Cognition & Emotion, 1–15. doi:10.1080/02699931.2014.881324 Borke, H. (1973). The development of empathy in Chinese and American children between three and six years of age: A cross-cultural study. Developmental Psychology, 9(1), 102–108. Botero, C. A., Gardner, B., Kirby, K. R., Bulbulia, J., Gavin, M. C., & Gray, R. D. (2014). The ecology of religious beliefs. Proceedings of the National Academy of Sciences, 111(47), 16784–16789. doi:10.1073/pnas.1408701111 Bowler, D. M. (1992). “ Theory of mind” in Asperger's syndrome. Journal of Child Psychology and Psychiatry, 33(5), 877–893. Boyd, R., & Richerson, P. J. (1992). Punishment allows the evolution of cooperation (or anything else) in sizable groups. Ethology and Sociobiology, 13, 171–195. Boyd, R., Richerson, P. J., & Henrich, J. (2011). Colloquium Paper: The cultural niche: Why social learning is essential for human adaptation. Proceedings of the National Academy of Sciences, 108(Supplement_2), 10918–10925. doi:10.1073/pnas.1100290108 Boyer, P. (2001). Religion Explained: The Evolutionary Origins of Religious Thought. New York: Basic Books. Bretherton, I., & Beeghly, M. (1982). Talking about internal states: The acquisition of an explicit theory of mind. Developmental Psychology, 18(6), 906–921. Brison, K. J. (2001). Crafting Sociocentric Selves in Religious Discourse in Rural Fiji. Ethos, 29(4), 453–474. 201 Brosnan, S. F., Henrich, J., Mareno, M. C., Lambeth, S. P., & Schapiro, S. J. (2009). Chimpanzees (Pan troglodytes) do not develop contingent reciprocity in an experimental task. Animal Cognition, 12(4), 587–597. doi:10.1007/s10071-009-0218-z Brown, T. (2006). Personhood as a Tool to Reflect upon Koinonia. Anglican Theological Review, 88(2), 163–179. Brüne, M., & Brüne-Cohrs, U. (2006). Theory of mind--evolution, ontogeny, brain mechanisms and psychopathology. Neuroscience and Biobehavioral Reviews, 30(4), 437–455. doi:10.1016/j.neubiorev.2005.08.001 Byrne, R., & Whiten, A. (1989). Machiavellian Intelligence : Social Expertise and the Evolution of Intellect in Monkeys, Apes, and Humans. Oxford, UK: Oxford University Press. Call, J., & Tomasello, M. (2008). Does the chimpanzee have a theory of mind? 30 years later. Trends in Cognitive Sciences, 12(5), 187–192. doi:10.1016/j.tics.2008.02.010 Callaghan, T., Rochat, P., Lillard, A., Claux, M. L., Odden, H., Itakura, S., et al. (2005). Synchrony in the onset of mental-state reasoning: evidence from five cultures. Psychological Science, 16(5), 378–384. doi:10.1111/j.0956-7976.2005.01544.x Carey, S. (2011). Continuity and Discontinuity in Cognitive Development. In E. E. Smith & D. N. Osherson, Thinking: An Invitation to Cognitive Science (2nd ed., Vol. 3, pp. 101–129). Cambridge, MA: MIT Press. Carey, S., & Spelke, E. S. (1994). Domain-specific knowledge and conceptual change. In L. A. Hirschfeld & S. A. Gelman, Mapping the mind: Domain specificity in cognition and culture (pp. 169–200). New York: Cambridge University Press. Carpendale, J. I. M., & Lewis, C. (2004). Constructing an understanding of mind: The development of children's social understanding within social interaction. Behavioral and Brain Sciences, 27, 79–151. Carpenter, M., Nagell, K., Tomasello, M., Butterworth, G., & Moore, C. (1998). Social Cognition, Joint Attention, and Communicative Competence from 9 to 15 Months of Age. Monographs of the Society for Research in Child Development, 63(4), 1-174. Carruthers, P. (2008). Precis of The Architecture of the Mind: Massive Modularity and the Flexibility of Thought. Mind \& Language, 23(3), 257–262. doi:10.1111/j.1468-0017.2008.00340.x Carsten, J. (1995). The substance of kinship and the heat of the hearth: Feeding, personhood, and relatedness among Malays in Pulau Langkawi. American Ethnologist, 22(2), 223–241. doi:10.1525/ae.1995.22.2.02a00010 Casler, K., Terziyan, T., & Greene, K. (2009). Toddlers view artifact function normatively. Cognitive Development, 24(3), 240–247. doi:10.1016/j.cogdev.2009.03.005 Chang, H., & Cheung, H. (2016). Competition enhances mentalizing performance in autism. Research in Autism Spectrum Disorders, 22, 45–54. doi:10.1016/j.rasd.2015.11.009 Choi, I., & Nisbett, R. E. (1998). Situational Salience and Cultural Differences in the Correspondence Bias and Actor-Observer Bias. Personality and Social Psychology Bulletin, 24(9), 949–960. doi:10.1177/0146167298249003 202 Choi, I., Koo, M., & Jong An Choi. (2007). Individual differences in analytic versus holistic thinking. Personality and Social Psychology Bulletin, 33(5), 691–705. doi:10.1177/0146167206298568 Choi, S. (2006). Influence of language-specific input on spatial cognition: Categories of containment. First Language. Choi, S., McDonough, L., Bowerman, M., & Mandler, J. M. (1999). Early sensitivity to language-specific spatial categories in English and Korean. Cognitive Development. Chudek, M., & Henrich, J. (2011). Culture-gene coevolution, norm-psychology and the emergence of human prosociality. Trends in Cognitive Sciences, 15(5), 218–226. doi:10.1016/j.tics.2011.03.003 Cialdini, R. B., Brown, S. L., Lewis, B. P., Luce, C. L., & Neuberg, S. L. (1997). Reinterpreting the empathy-altruism relationship: when one into one equals oneness. Journal of Personality and Social Psychology, 73(3), 481–494. Cohen, A. B. (2003). RESEARCH: Religion, Likelihood of Action, and the Morality of Mentality. International Journal for the Psychology of Religion, 13(4), 273–285. Cohen, A. B., & Hill, P. C. (2007). Religion as culture: religious individualism and collectivism among american catholics, jews, and protestants. Journal of Personality, 75(4), 709–742. doi:10.1111/j.1467-6494.2007.00454.x Cohen, A. B., & Rozin, P. (2001). Religion and the morality of mentality. Journal of Personality and Social Psychology, 81(4), 697–710. Cohen, D., Kim, E., & Hudson, N. W. (2014). Religion, the Forbidden, and Sublimation. Current Directions in Psychological Science, 23(3), 208–214. doi:10.1177/0963721414531436 Cohen, S. B., Wheelwright, S., Hill, J., Raste, Y., & Plumb, I. (2001). The “Reading the Mind in the Eyes” test revised version: A study with normal adults, and adults with Asperger syndrome or high‐functioning autism. Journal of Child Psychology and Psychiatry and Allied Disciplines, 42(2), 241–251. Cohen, T. R., Montoya, R. M., & Insko, C. A. (2006). Group morality and intergroup relations: cross-cultural and experimental evidence. Personality and Social Psychology Bulletin, 32(11), 1559–1572. doi:10.1177/0146167206291673 Corcoran, R., Mercer, G., & Frith, C. D. (1995). Schizophrenia, symptomatology and social inference: Investigating “theory of mind” in people with schizophrenia. Schizophrenia Research, 17(1), 5–13. doi:10.1016/0920-9964(95)00024-G Cordain, L., Eaton, S. B., & Miller, J. B. (2002). An evolutionary analysis of the aetiology and pathogenesis of juvenile‐onset myopia. Acta Ophthalmologica Scandinavia, 80, 125–135. Corrin, J., & Paterson, D. (2007). Introduction to South Pacific Law (2nd ed.). New York: Routledge-Cavendish. Costa, J. (2009). On The Role of Consequences in Punitive and Remunerative Judgments: Testing an Ultimate Explanation. Harvard University, Department of Anthropology, Cambridge, MA. 203 Cote, S., & Hideg, I. (2011). The ability to influence others via emotion displays: A new dimension of emotional intelligence. Organizational Psychology Review, 1(1), 53–71. doi:10.1177/2041386610379257 Cote, S., DeCelles, K. A., McCarthy, J. M., Van Kleef, G. A., & Hideg, I. (2011). The Jekyll and Hyde of Emotional Intelligence: Emotion-Regulation Knowledge Facilitates Both Prosocial and Interpersonally Deviant Behavior. Psychological Science, 22(8), 1073–1080. doi:10.1177/0956797611416251 Christie, D., & Viner, R. (2005). Adolescent development. BMJ (Clinical Research Ed.), 330(7486), 301–304. doi:10.1136/bmj.330.7486.301 Csibra, G., & Gergely, G. (1998). The teleological origins of mentalistic action explanations: A developmental hypothesis. Developmental Science, 1(2), 255–259. Csibra, G., & Gergely, G. (2006). Social learning and social cognition: The case for pedagogy. In Y. Munakata & M. H. Johnson, Processes of Change in Brain and Cognitive Development. Attention and Performance (Vol. 21, pp. 249–274). Oxford, UK: Oxford University Press. Csibra, G., & Gergely, G. (2011a). Natural pedagogy as evolutionary adaptation. Philosophical Transactions of the Royal Society B: Biological Sciences, 366(1567), 1149–1157. doi:10.1098/rstb.2010.0319 Csibra, G., & Gergely, G. (2011b). Teleological understanding of actions. (M. R. Banaji & S. A. Gelman). Oxford UP. Cummins, D. D. (1996). Dominance Hierarchies and the Evolution of Human Reasoning. Minds and Machines, 6, 463–480. Cushman, F. (2008). Crime and punishment: distinguishing the roles of causal and intentional analyses in moral judgment. Cognition, 108(2), 353–380. doi:10.1016/j.cognition.2008.03.006 Cushman, F. (2015). Deconstructing intent to reconstruct morality. Current Opinion in Psychology, 6, 97–103. Cushman, F., Sheketoff, R., Wharton, S., & Carey, S. (2013). The development of intent-based moral judgment. Cognition, 127(1), 6–21. doi:10.1016/j.cognition.2012.11.008 Danchin, É., Charmantier, A., Champagne, F. A., Mesoudi, A., Pujol, B., & Blanchet, S. (2011). Beyond DNA: integrating inclusive inheritance into an extend
UBC Theses and Dissertations
Morality when the mind is opaque : intent vs. outcome across the lifespan in Yasawa, Fiji McNamara, Rita Anne 2016
Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.
- 24-ubc_2016_september_mcnamara_rita.pdf [ 11.5MB ]
- JSON: 24-1.0307352.json
- JSON-LD: 24-1.0307352-ld.json
- RDF/XML (Pretty): 24-1.0307352-rdf.xml
- RDF/JSON: 24-1.0307352-rdf.json
- Turtle: 24-1.0307352-turtle.txt
- N-Triples: 24-1.0307352-rdf-ntriples.txt
- Original Record: 24-1.0307352-source.json
- Full Text