Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Evolving peculiarly human minds : novel evidence from social and developmental psychology Chudek, Matthew 2009

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


24-ubc_2009_fall_chudek_matthew.pdf [ 1.93MB ]
JSON: 24-1.0070860.json
JSON-LD: 24-1.0070860-ld.json
RDF/XML (Pretty): 24-1.0070860-rdf.xml
RDF/JSON: 24-1.0070860-rdf.json
Turtle: 24-1.0070860-turtle.txt
N-Triples: 24-1.0070860-rdf-ntriples.txt
Original Record: 24-1.0070860-source.json
Full Text

Full Text

Evolving Peculiarly Human Minds:Novel Evidence from Social and DevelopmentalPsychologybyMatthew ChudekB.A.(hons), University of Melbourne, 2003A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFMaster of ArtsinTHE FACULTY OF GRADUATE STUDIES(Psychology)The University Of British Columbia(Vancouver)August 2009c?Matthew Chudek, 2009AbstractAccounts of the evolutionary origins of human psychology can be as intriguing asthey are difficult to test. The dearth of direct evidence of ancient conditions can inpart be alleviated by careful investigation of their consequences for contemporarycognition. Here I report the results of three studies designed to test evolutionaryinferences using modern psychological evidence - that is, trying to gain insight intohow our brains came to be by looking at how they currently function. Two of thesestudies report empirical evidence of novel psychological phenomena, predicted apriori by an evolutionary theory. The third attempts further empirical verification ofa result previously claimed to have evolutionary significance. The inferential logicof such investigations is very different to that typically employed by psychologistsstudying the proximate mechanisms behind the same phenomena. I also considerthe value and difficulties particular to drawing evolutionary inferences from psy-chological evidence and lay out criteria for ensuring that these are reliable.iiTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Document Organisation . . . . . . . . . . . . . . . . . . . . . . . 11.2 The Question . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.3 Memories of an Encephalised Ape . . . . . . . . . . . . . . . . . 41.3.1 Ecological Intelligence Hypotheses (EIH) . . . . . . . . . 61.3.2 Social Intelligence Hypothesis (SIH) . . . . . . . . . . . . 71.3.3 Cultural Intelligence Hypothesis (CIH) . . . . . . . . . . 71.3.4 Psychological Evidence . . . . . . . . . . . . . . . . . . 81.4 Novel Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111.5 Evolutionary Inference from Psychological Evidence . . . . . . . 131.5.1 Why Bother with Large Scale Evolutionary Accounts? . . 141.5.2 The Trouble with Evolutionary Inference . . . . . . . . . 171.5.3 Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . 191.5.4 Good Psychological Observations(For Evaluating Evolutionary Theories) . . . . . . . . . . 221.6 Framing My Contributions . . . . . . . . . . . . . . . . . . . . . 24iii2 Social Biases in Recall: Replication Across Cultures and Materials 252.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262.1.1 The Social Recall Bias . . . . . . . . . . . . . . . . . . . 272.2 Study One . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282.2.1 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 292.2.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 322.2.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 422.3 Study Two . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 432.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 432.3.2 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 462.3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 482.3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 512.4 General Conclusions . . . . . . . . . . . . . . . . . . . . . . . . 513 Differences in Memory Biases for Strategic and Non-Strategic Cul-tural Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . 543.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 543.2 Operationalising the Cultural Intelligence Hypothesis . . . . . . . 563.3 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 573.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 603.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624 Prestige-Biased Learning in Children: Attention from others as aCue for Cultural Transmission . . . . . . . . . . . . . . . . . . . . . 654.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 664.1.1 The Evolution of Prestige-Bias . . . . . . . . . . . . . . . 674.1.2 Predictions . . . . . . . . . . . . . . . . . . . . . . . . . 684.2 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 694.2.1 Participants . . . . . . . . . . . . . . . . . . . . . . . . . 694.2.2 Proceedure . . . . . . . . . . . . . . . . . . . . . . . . . 704.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 714.3.1 Is Children?s Imitation Biased by a Model?s Prestige? . . . 714.3.2 Does Prestige-Bias Operate Across Learning Domains? . . 72iv4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 724.5 Supplemental Material . . . . . . . . . . . . . . . . . . . . . . . 744.5.1 Is Prestige-Bias Strongest in the Domain Where PrestigeWas Observed? . . . . . . . . . . . . . . . . . . . . . . . 744.5.2 Does Prestige-Bias Generalise Across Behavioural Domains? 744.5.3 Is Language Learning a Special Domain? . . . . . . . . . 754.5.4 Are There Gender Effects? What About Other Effects? . . 764.5.5 Do Children Explicitly Report Their Prestige-Bias? . . . . 775 Conclusion: Evaluating My Contributions . . . . . . . . . . . . . . 78Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80A Research Ethics Board, Certificates of Approval . . . . . . . . . . . 89vList of Tables2.1 Main Effects of Type on Recall - Summary . . . . . . . . . . . . 332.2 Main Effects of Type on Recall - Details . . . . . . . . . . . . . . 332.3 Main Effects of Culture on Recall - Summary . . . . . . . . . . . 352.4 Main Effects of Culture on Recall - Details . . . . . . . . . . . . 362.5 Culture-Type Recall Interactions - Summary . . . . . . . . . . . . 392.6 Culture-Type Recall Interactions - Details . . . . . . . . . . . . . 402.7 Discrete, Unrelated Recall Items . . . . . . . . . . . . . . . . . . 472.8 Discrete Recall Model . . . . . . . . . . . . . . . . . . . . . . . 503.1 Strategic Recall Items . . . . . . . . . . . . . . . . . . . . . . . . 593.2 Non-Strategic Recall Items . . . . . . . . . . . . . . . . . . . . . 613.3 Frequencies of Agreement . . . . . . . . . . . . . . . . . . . . . 624.1 Parsimonious Models - Logistic Regression Coefficients . . . . . 754.2 Parsimonious Models - Odds Ratios [with 95% C.I.s] . . . . . . . 764.3 Full Models - Logistic Regression Coefficients . . . . . . . . . . 77viList of Figures2.1 Number of Propositions Recalled, by Type with 95% C.I.s . . . . 342.2 Mean Number of Propositions Recalled, by Condition with 95%C.I.s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372.3 Exact Number of Propositions Recalled, by Type, Condition andGeneration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 414.1 Prestige Cuing and Imitation Testing Stills . . . . . . . . . . . . . 70viiAcknowledgmentsThis work would not have been possible without the help of more people than I canmention. Still, I offer special thanks toJoesph Henrich, my primary advisor, for giving me the room to make my owndiscoveries and the guidance to recover from them.Mika McKinnon, for her love, care, support and harsh editing (and for feedingme!).The faculty who have advised me, especially Susan Birch, Steven Heine, MarkSchaller and Ara Norenzayan, for a very interesting apprenticeship.My lab-mates at our often renamed Lab 2202, for their company, argumentsand friendshipsMy family, who, though they?re far away, are always with me.The community of Green College, for dramas, friendships and adventures.University of British Columbia Graduate Fellowships and the Ron HowardWebster Foundation, for supporting this work.This thesis is dedicated to my father. Thank you for helping shape me intosomeone I?m proud to be.viiiChapter 1IntroductionWhenever a theory appears to you as the only possible one, take thisas a sign that you have neither understood the theory nor the problemwhich it was intended to solve. ? Sir Karl Popper1.1 Document OrganisationAt the core of this thesis are three empirical investigations, which draw on experi-mental and statistical techniques from contemporary psychology to expose proper-ties of modern cognition which have important implications for understanding ourevolutionary history. These studies have been written up in a manner suitable forpublication in psychology journals. The first, which comprises Chapter 2, extendsfindings which suggest a bias in memory favouring social information. The secondstudy, comprising Chapter 3 employs a similar methodology to test a new predic-tion, derived from ?coevolutionary theory?, concerning discontinuities in memorybiases. The final study, Chapter 4, tests another, pre-existing prediction from co-evolutionary theory concerning whom young learners prefer to imitate.Reports typical of psychology journals tend to focus on identifying and ex-plaining specific, interesting empirical effects. Each of the studies reported hereshould meet this standard independently: they either identify new effects or clarifyboundary conditions on known ones. In each paper, this feature has been empha-sised. Together these studies also constitute a contribution to a broader theoretical1project, whose core is actually quite far removed from any of these effects in par-ticular.This thesis seeks to contribute to testing evolutionary theories, rather than ex-plaining particular psychological phenomena. To that end, this introduction buildsthese results into a broader framework. It covers the question I hope to help answer,the relevance of these studies to answering it and the difficulties that accompany in-ferences from very specific psychological phenomena to very general evolutionaryquestions.1.2 The QuestionI don?t exactly remember the day or even where I was, but I clearly remember thefeeling when as a child I first asked myself the question: What am I?My language and culture had a lot of ready answers: a human, a person, anAustralian or maybe a Polak, a child, a boy. At other times these had been perfectlysufficient, but on that day I remember finding them disquieting. Being able to labelsomething, to distinguish it from other things, wasn?t the same as understanding it;it was just the first step.To really answer my question I wanted to know how you?d make somethinglike me: what are the essential ingredients, how do they work and, why had theycome together in my case? Once we knew this, I wondered, could we make anothercreature, an animal or perhaps a machine, that could reason and communicate andask questions about itself just as I was doing?My questions and speculations could just as easily have (and probably did)belong to another child, living a thousand, ten thousand or perhaps many moreyears ago. My ancient counterpart would have had little luck finding anythingbut speculative answers, as likely to be close to the truth as their own whimsy.Since then, by our cumulative endeavour, our species of curious children has madeprogress in carefully observing the world and generating better answers, or at leastones consistent with all the the best evidence we?ve gathered. Though that ancientchild and I wondered similar questions with similar minds, by virtue of being born alittle later, I had the pleasure of learning many of these painstaking derived answersas I grew.2Many pieces of the puzzle have been filled. We now have excellent, well-verified theories about the constituent components of the universe, the origins ofstars, some of the properties of complex organic molecules, the processes by whichthey form and come together in self-replicating, autocatalytic sets, the dynamicswhich cause these self-replicators to change over time, how they become morediverse and complex and why and how they undergo major transitions which bringnew dynamics into play. Many pieces of this story are still missing or in dispute,but one part in particular has always struck me as conspicuously absent.It is far from obvious, given all of our present physical and biological theories,what part of that story in particular was essential to producing something like me- something that can reason, communicate and ask questions like mine. Whatneeds to happen for a primate, not unlike many other primates or even mammals,to start behaving in those peculiar ways? What steps would we need to take ifwe wanted, in the distant future, to have an intelligent conversation with anotherspecies, or perhaps in the less distant future, with a computer? We still lack a clearunderstanding of what made us different, and thus of what precisely we are.Academic careers have been built on the debate over whether humans are some-how qualitatively unique or peculiar only in the magnitude of traits continuous withthose of other species. I hope to side-step this mire1 by retreating to the claim thatwe humans are, in the very least, peculiar: we fly to the moon, read fiction andresearch what we are. Whether these traits are unique or continuous, their ultimatecauses are still poorly understood and likely to be a crucial part of explaining whatwe are. This is my first attempt at contributing to our search for this explanation.I begin by reviewing two classes of explanations of human brain expansion,which we have good reason to suspect is closely linked to the origins of our pe-culiar cognition. Mesoudi, Whiten and Dunbar (2006) began the work of usingpsychological evidence to discriminate between these hypotheses. I carry this for-ward by testing for more proximate explanations for their effect. I also derive anew prediction from a third class of evolutionary explanation and demonstrate twonew empirical effects. When drawing inferences from the present state to the an-cient history of a rapidly evolving system, even the prediction of novel phenomena1Though readers interested in this debate might begin by considering recent arguments fromeither side, such as Tomasello & Rakoczy (2003) and Roth & Dicke (2005)3provides only tenuous evidential support that must be interpreted cautiously. I con-clude by carefully evaluating the inferential process which links these results to thequestion that I and other curious children have been hungering to answer.1.3 Memories of an Encephalised ApeThe advent of human peculiarity coincided with very rapid expansion of the humanbrain between about 600?150 thousand years ago, particularly during the MiddlePleistocene (Ruff et al., 1997). In addition to its temporal coincidence we haveseveral good reasons to suspect that this brain expansion was intimately associatedwith the emergence of human cognitive peculiarity.Firstly, the disproportionately large regions of the human brain vis-a`-vis otherprimates, are those associated with some of the most distinct functional aspects ofour cognitive peculiarity (Schoenemann, 2006); for example the cerebellum (impli-cated in language) which is approximate 2.9 times larger than expected in a primateour size, the temporal lobe (implicated in conceptual understanding and memory)and the prefrontal cortex (implicated in executive functions).Secondly, brains are expensive (Foley et al., 1991; Aiello & Wheeler, 1995):they account for about 20% of our body?s oxygen uptake and energy expenditurebut only 2% of its mass. Large brains would have been outcompeted had they notaccured more benefits to our ancestors than the costs they imposed. These benefitsmust have exceeded both the advantages of increased efficiency (smaller brainedcompetitors required less food to survive and reproduce) and the alternative waysthese metabolic reserves could have been invested (for instance in greater musclemass or digestive tissue). Given the rapid proliferation of modern humans, it isreasonable to suspect that some aspect of human peculiarity was facilitated by brainexpansion and in turn provided these fitness benefits.Lastly, brain expansion has received particular attention from scholars inter-ested in accounting for human uniqueness simply because, unlike the vast bulk ofour evolutionary history, ancient skulls preserve a record of it. Like the drunkardwho stumbles about under the light-post looking for his keys, it is not unreasonableto begin our search in the places where we can see.Thus, at the moment, many of our best leads on the origins of human pecu-4liarity are attempts to explain our rapid encephalisation. These theories come intwo flavours: those that emphasise the constraints preventing brain expansion, andthose that emphasise the pressures driving it. Constraint-oriented theories tend toassume that larger brains are useful and focus on the fact that growing them is hard.These theories tend to account for how the conditions of early humans made it pos-sible for brain expansion to occur, for instance by the availability and incorporationof higher quality foods, the delayed maturation (Foley et al., 1991) or the fatten-ing of infants(Cunnane & Crawford, 2003), or the reduced size of the digestivesystem (Aiello & Wheeler, 1995). Pressure-oriented theories, on the other hand,assume that sufficient selection pressures will, if it is physically possible, eventu-ally overcome constraints. Instead they ask: ?What did brains get that big for??These theories tend to emphasise how the circumstances of early humans providedespecially large fitness benefit to individuals with larger brains.A reasonable account of brain expansion, if it is to account for human pecu-liarity, must explain not only why human brains have grown rapidly but why thebrains of other primates have not. One possibility is that we were simply the firstto overcome some metabolic constraint, that all mammalian, or at least primate,brains are sufficiently complex to display all the features of human peculiarity, ifonly they could grow sufficiently large. On their own, such accounts don?t an-swer the question: why, when these metabolic constraints were lifted, was fitnessmaximised by brain expansion not the expansion of some other system? Though afew theorist argue for an exclusively constraint-oriented approach (see for instanceCunnane & Crawford (2003)), such accounts are typically incomplete without anadditional argument about positive selection pressures. Notice however, that theinverse is not true: the absence of a pressure, or its being small compared to someother pressure, is sufficient on its own to explain the occurrence in some cases (andlack of in others) of phylogenetic changes.Currently three classes of pressure-oriented theories stand out in the literature:ecological, social and cultural.51.3.1 Ecological Intelligence Hypotheses (EIH)Besides crania, our scant archaeological record is most replete with evidence ofhistorical diets: teeth, partially digested remains, bones, shells discarded duringfood preparation, and so on. It is not surprising then that several theories havelinked encephalisation to dietary changes. These theories typically emphasise theadditional cognitive demands that a changing ecology imposed on early humans,rewarding cleverer individuals with more access to food. Among the best expli-cated theories in this category are arguments encephalisation was driven by:? a shift to frugivory (a primarily fruit-based diet) because though fruits arerich in nutrients, they are ephemeral and patchily distributed and thus har-vesting them demands and rewards a larger memory capacities than does fo-livary (subsistence mostly on leaves)(Milton, 1981; Harvey & Krebs, 1990);? a diet that required foraging over an ever larger territory, demanding thecapacity to maintain mental maps on larger scales (Clutton-Brock & Harvey,1980)? a shift to a diet of foods that require extraction - termites from a mound orfruit from a shell, for instance, because it demands and rewards ever moresophisticated cognition (Gibson, 1986).Another theory that fits in this category, though not focused on diet, is RichardPotts? theory of Variability Selection (Potts, 1998). Deep-sea and ice core datasuggest that the rate of climactic variation on our planet has increased dramaticallyover the past six million years. Potts argues that, unlike stable environments whichselect for organisms optimally efficient in their habitat, highly variable environ-ments select for costly structures like big brains, which buffer their bearers againstrapidly changing circumstances.A consistent prediction can be derived from each of these hypotheses: thathuman brains should be well adapted to mastering their ecology, either processinginformation related to foraging and diet or, more generally, quickly developingbehavioural adaptations to diverse environmental conditions.61.3.2 Social Intelligence Hypothesis (SIH)Unlike many mammals, most primates live in relative stable groups of non-kin.Such arrangements have been thoroughly investigated by game theorists, evolu-tionary biologists and social scientists alike (i.e. Boyd & Richerson, 1988c; Hauertet al., 2002) and one thing is clear: they are not easy to maintain. The individ-ual benefits of cooperation are often exceed by those which could be gained byexploitation of other group members. Protecting oneself from exploitation whilemaintaining the benefits of being in a group is a difficult balancing act requiringspecial cognitive skills (for instance: reputation tracking ,explored by Nowak &Sigmund, 2005).Since one peculiar feature of humans is our large-scale sociality despite lowgenetic relatedness (it is not uncommon to see alliances of tens or thousands ofunrelated individuals), several theorists have posited the challenges and benefits ofsociality as the driving pressure behind encephalisation (Humphrey, 1976; Whiten& Byrne, 1988; Dunbar, 1998). These hypotheses are supported by comparativeevidence from across primate species, including correlations between indexes ofbrain size and social group size, frequency of deceptive acts, amount of socialplay, and the degree to which male dominance does not predict reproductive suc-cess, presumably because of the greater range of effective social wooing strategiesavailable to lower-ranking males (see Dunbar, 1998, for a review).These hypotheses also make a consistent prediction: human brains will be welladapted to the task of thriving in a social environment, including noticing, remem-bering and exploiting relevant information.1.3.3 Cultural Intelligence Hypothesis (CIH)Besides being very social, humans are distinct from other social primates in beinghighly cultural. We rely heavily on complex, encoded information that we acquireduring our lifetimes from our conspecifics. For instance consider our extended ju-venile period: even up to the age of twenty, long after chimps have become calor-ically independent, humans in both modern and traditional societies produce farless than we consume (Kaplan et al., 2007). By the time we?ve paid the metabolicdebt on even our own lives, let alone the costs of reproducing, we are well on our7way to being full-fledged !Kung foragers, Mesopotamian farmers or Canadian aca-demics, phenotypes whose lifetime production dwarf that of our earlier-maturingchimp counter-parts. Most peculiarly, the particular characteristics of these pheno-types that make this amazing feat possible (such as farming or waged labour) arenot at all described by our genotype.Scholars from a variety of disciples (see for instance Cavalli-Sforza, 1981;Boyd & Richerson, 1988b; Durham, 1992) have contributed to a growing bodyof reasoning and formal models which take as their starting point our ability toshare during our lifetimes complex, encoded information that substantially altersthe way we behave as adults. These accounts typically attempt to answer ques-tions like: why would such difficult, costly cultural learning be selected for overindividual learning or even simpler genetic inheritance (Boyd & Richerson, 1995);why have other animals not experienced this transition (Boyd & Richerson, 1996);and what are the evolutionary consequences once it pays to be a cultural learner(Henrich & Boyd, 1998; McElreath et al., 2003; Boyd & Richerson, 1988c). Onthis account larger, more specialised brains were selected to tackle the distinctchallenges of sorting, storing, integrating, utilising and retransmitting the valuablecultural information that accumulated by this process.Direct empirical support for this position is converging from direct tests in dis-parate fields, including cross-species comparisons (Herrmann et al., 2007; Galef &Whiskin, 2007), cross-cultural comparisons (Efferson et al., 2008; Kline & Boyd,forthcoming; but cf. Efferson et al., 2007), spacio-temporal patterns in archaeolog-ical records (Powell et al., 2009) and, recently, developmental psychology (Chap-ter 4).These theories are often explicit in predicting that human brains should befunctionally adapted to extracting and processing cultural information.1.3.4 Psychological EvidenceThese three hypotheses have drastically different implications for what it took togenerate psychologically modern humans. The EIH suggests that the key was im-proving spatial reasoning, memory and other cognitive skills involved in negoti-ating a changing physical environment. The SIH suggests the driving pressure is8improvement in social cognition, that is: better understanding of one?s own andothers? strategic situations, and an improve ability to make the most of this infor-mation. Such advanced strategic thinking is often thought to require a ?theory-of-mind?, an ability to track the world from the perspective of other agents, a skill hu-mans seem to have in abundance relative to other primates (Herrmann et al., 2007).The CIH claims the key piece was something very different: hardware (brains builtfor culture) which facilitated faithful transmission of cognitive software (culturalinformation). The work of producing peculiar, psychologically modern humansis then shared between the relatively rapid evolution of culture, and the sloweradaptation of our neuronal hardware to the regularities produced by this culturalsoftware. Theory-of-mind also plays a key part in CIH accounts, but as a tool forcultural learning by imitating other?s goals and perspectives.Any evidence that can help evaluate these hypotheses is valuable. Unfortu-nately much of the direct evidence has decayed and the remaining archaeologicalrecord is sparse. Fortunately, the pressures that shaped human psychology haveleft behind another trace: modern human psychology. Deriving predictions fromthese distal hypotheses about the properties of modern cognition can allow evi-dence from modern experimental psychology to discriminate between them.This was precisely the strategy pursued by Mesoudi, Whiten & Dunbar in their2005 study. They derived conflicting predictions from EIH and SIH about the kindsof information humans should be particularly skilled at encoding and retrievingfrom memory, and then simply tested human recall. Their results supported theSIH; their participants showed a bias for recalling social over ecological informa-tion. Mesoudi, Whiten & Dunbar warned that theirs was only a preliminary findingin need of replication with different materials and across cultures. Chapter 2 docu-ments such replications. Its conclusions bear reiterating.Ultimate causes are enacted by proximate mechanisms and, as Mesoudi, Whiten& Dunbar point out, a number of the proximate mechanisms which might explaintheir effect (such as emotional impact or attentional salience) are entirely consistentwith the SIH. There are however two proximate mechanisms which are not.Firstly, culture: if people preferentially recall social information because they?reenculturated to as children, then their doing so no longer constitutes clear evidenceof ancient adaptations. To exclude this possibility our research team replicated this9social recall bias, using Mesoudi, Whiten & Dunbar?s original materials, amongparticipants inculcated in two different cultures: East Asians and North Amer-icans. Participants from these two groups have manifested culturally-covaryingcognitive differences in previous investigations (Nisbett, 2003). One of the mostcommonly referenced differences between these groups is their emphasis on indi-vidualism versus collectivism (Triandis, 2001), a broad conceptual continuum thatcaptures, among other things, how much their members need to attend to and in-terpret the subtleties of social context. Given this difference in emphasis on socialcontext, not to mention differences in visual perception of contextual information(Kitayama et al., 2003), it is quite plausible that these groups might also differ inhow much they privilege social information in memory.We also tested a third group, whose parents were culturally East Asian but whohad been reared in North America, to tease out any ecologically evoked differences.We did not find evidence of cultural variation in the the social bias effect, neitherin its presence nor magnitude in any of these groups. Culture seemed to play verylittle role in predicting raw recall rates for information and though it did interactwith information type, this was not in a way clearly consistent with cross-culturalvariation in Mesoudi, Whiten and Dunbar?s effect. Given this result, Mesoudi,Whiten & Dunbar?s inference seems safer from criticism on cultural grounds.A second potentially incompatible proximate explanation is that the materialsMesoudi, Whiten & Dunbar designed were responsible for their result. We identi-fied three ways in which these materials seems to vary unsystematically. The firstwas the degree of redundancy, both logical (i.e. certain facts literally implied morethan once) and superficial (i.e. repeated words, names, themes or other surface de-tails) in the information conveyed. The second was how familiar participants couldbe expected to be with the materials and thus the extent to which their expertiseenhanced their recall. The last was whether the material described real or fictionalpeople and places, and thus whether participants could ascribe it a truth value toagree or disagree with.We designed a new set of materials which controlled for these confounds. Themost critical change was that our materials consisted of discrete, unrelated itemswhile Mesoudi, Whiten and Dunbar?s had been vignettes arranged meaningfullyaround a coherent theme.10Using a sample large enough to detected the original effect, we detected no dif-ferences between social and non-social information using this new material. Thisresult is consistent with a mechanism well known to memory researchers: the ex-pertise effect on recall: experts have far better recall than non-experts for mean-ingfully arranged or related items in the domains they?re more familiar with, butnot for unrelated or randomly arranged items (see introduction to Vicente & Wang,1998 for a recent review, but c.f. Simon & Gobet, 2000).The evolutionary implications of participants having more expertise or famil-iarity with social situations (and thus better recall for it) are more difficult to inter-pret than those of an unconditional recall bias. Neither SIH nor EIH seem obvi-ously better suited to explaining the fact that expertise improves recall, especiallysince it does so across many domains (even visual evaluations of ecological fea-tures Kawamura et al., 2007) and since new expertise can be acquired quite rapidly(Chase & Ericsson, 1982). Both theories can, with some tinkering, explain whyindividuals today would tend to acquire social expertise in particular, but so coulda large number of theories, particularly direct ecological ones.Follow-up work with specialised materials and populations could tease apartthese possibilities. For the time being I conclude that the bias Mesoudi, Whiten& Dunbar observed, though strong and cross-culturally consistent, doesn?t provideclear cut support for SIH over EIH.1.4 Novel TestsIn addition to probing deeper into Mesoudi, Whiten & Dunbar?s findings, I at-tempted to extend their methodology to a test of the CIH.Any model is a simplification; it paints a characture of our complex worldwhich explains only certain dynamics. The SIH paints a world populated by strate-gic actors playing games of conflict and coordination: forming alliances, vying forstatus, excluding defectors and so on. The CIH paints a similar world, but one inwhich the agents are also concerned with learning, recombining and broadcastingthe best quality cultural information they can. Each hypothesis predicts humanswill be well-adapted to the environment it models.There is considerable overlap between these hypotheses ? since cultural in-11formation comes from one?s social peers, to access this valuable information oneat least needs access to their peers. As culturally transmitted skills become morecomplex, more social access is needed. For very complex skills, like ANOVA ormaking stone-tools (Stout, 2002) learners need active facilitation by knowledge-able teachers. While some degree of competence in social cognition is probablya prerequisite for cultural animals, the CIH makes additional predictions over andabove this. In Chapter 3, I operationalise the CIH, deriving from it a predictionabout modern recall biases. Specifically: a mind well-adapted for culture shoulddiscriminate between strategic and non-strategic information, only using estimatesof the accuracy of strategic information to bias its assimilation and recall.Our research team tested this hypothesis by using a similar design to that inChapter 2, where participants rated items before recalling them, but this time dis-tinguished strategic from non-strategic items. A clear pattern emerged: partici-pants? agreement predicted their recall for non-strategic items, but not for strategicitems. In fact, for strategic items, there was a large but non-significant trend inthe opposite direction. This is a new and interesting finding, but like Mesoudi,Whiten & Dunbar?s, it needs to be replicated cross-culturally and with a variety ofmaterials to eliminate incompatible proximate explanations.I want to be very clear - the CIH does not constitute a good proximate expla-nation for this phenomenon. It makes no claims whatsoever about the mechanismsthat generate it, besides their being heritable. Rather, the CIH makes claims aboutancient pressures which, if they can be verified, can compliment more proximateexplanations. As I discuss below, it is exceedingly rare that the investigation of anysingle phenomenon is sufficient constitute such a verification of a distal theory. Theevidential value of studies like this only becomes apparent when they are viewedalongside the emerging corpus of convergent evidence from across the domain ofthe distal theory.I have attempted a second contribution to this corpus: a demonstration of pres-tige bias among children, documented in Chapter 4. Besides biases in how theyprocess different kinds of information, CIH predicts that cultural animals can ben-efit by innate biases in how they treat information sources. Henrich and Gil-White(2001) made detailed predictions about how the relative amount of attention aninformation source receives from other learners should bias how cultural learners12weight it: they should prefer information which comes from sources attended toby others. If the benefits of such a bias were consistent enough to produce ge-netic adaptations, we would expect young children, whose highly plastic brainsare rapidly absorbing cultural information and have had relative little time to ac-quire cultural tactics for discriminating good from bad information, to be particularlikely to rely on it. We tested children aged three and four by showing them twomodels, one who was attended to by third parties and one who wasn?t. Each model(now on her own) demonstrated several preferences and behaviours which contra-dicted the other model?s. Children were given the chance to imitate one of thesedemonstrations and were clearly biased towards imitating the model who?d beenattended to.Again, a number of plausible proximate explanations can account for this ef-fect. Some of these are entirely consistent with an ultimate evolutionary cause, forinstance a combination of eye-tracking and low-level attentional effects. Some areplausible alternatives and need to be excluded experimentally, such as the attentionof third-parties being a cue to the patterns of social relationships rather informationquality.For the local phenomena we observed, none of the studies described here pro-vide proximate explanations. What about for the phenomenon we?re ultimatelyinterested in, the advent of human peculiarity? Can these local pieces togetherhelp solve that more distant puzzle?1.5 Evolutionary Inference from Psychological EvidenceGeologists, faced with the challenge of inferring continuous ore bodies from ascattering of discrete core samples, are well acquainted with what they call theMickey Mouse Problem. These specialists can only directly observe several one-dimensional samples from various locations, long tubes with ore either present orabsent at different depths. Each piece of ore in a sample may be a tiny isolatedstone encountered by chance or it might be part of some larger deposit connectingit to other ore samples. By drawing on their knowledge of the properties of the orethey?re mapping and surrounding materials, these experts attempt to infer an un-observed, underground map that explains their distribution of core samples. Their13efforts are plagued by the fact that, whatever the pattern of core samples they?veuncovered, it is always possible to connect them so that they form a picture ofMickey Mouse.Psychologists face a similar problem. Many disparate psychological phenom-ena may be explained entirely by local mechanisms or could be united under anynumber of grand theories, some of which are doubtlessly no more meaningful thanan animated mouse. A scientist who begins speculating about distal, unifying theo-ries may find their peers suspicious, and should themselves be concerned, that theymay be doing nothing more profound than sketching pictures of Mickey. Worsestill, since the psychological record is so large and diverse and it is not at all obvi-ous which phenomena are related, Mickey Mouse sketchers are free to cherry-pickjust those phenomena, and just those aspects of them, which nicely fit their pic-ture. Many psychologists have avoided this mire by abstaining altogether fromlow-resolution, large scale explanations. Instead they focus their efforts on the dif-ficult and valuable task of developing, by painstakingly detailed experimentation,high-resolution local understandings of precise phenomena.Given my interest in explaining human peculiarity, this thesis has unashamedlybeen an attempt to contribute to the project of developing large-scale accountsrather than local explanations. Neither I, my committee, nor my peers should feelconfident I haven?t spent my time and efforts on elaborate theoretical doodling, un-less I confront and address the problems of drawing evolutionary inferences frompsychological evidence. Below I argue first for the value of such large-scale infer-ences, difficult though they sometimes are; then I make explicit the problems thatplague such inferences and the techniques that can overcome them. Finally I usethese insights to evaluate my own contributions.1.5.1 Why Bother with Large Scale Evolutionary Accounts?The relationship of distal theories to proximate is like that of large-scale, low-resolution maps (like a map of North America) to local, high-resolution maps (likea map of UBC). Given that psychologists are making excellent progress investigat-ing the proximate high-resolution details of psychological phenomena, it is reason-able to ask what the value is of trying to evaluate low-resolution theories directly.14Won?t our developing understanding of the high-resolution details eventually an-swer these questions anyway?Theorising about and testing the regularities that exist on larger, distal scaleshas both practical advantages and an important deeper, logical role in science. Firstthe practical advantages. Low-resolution maps can direct our attention to valu-able high-resolution questions and phenomena that wouldn?t otherwise be appar-ent. Hopefully the novel experimental effects identified in this thesis will do justthis. Distal theories can also be invaluable in linking together seemingly disparateinvestigations under a common theoretical framework, making plain their bearingon each other and vastly improving our inferential power. Lastly, sometimes large-scale distal theories are the only ones capable providing general answers to simplebut compelling question like that with which I began this thesis: what am I?However, in my opinion, the most important reason to concern ourselves withlow-resolution research is that though distal theories can often remain agnostic tohigh-resolution particulars, the reverse is not true. An important twentieth centuryinsight into the enterprise of science was the ?theory-dependence of observation?(Quine, 1951; Kuhn, 1970; Gillies, 1998). There is no natural, absolute perspectivefrom which high-resolution observations, least of all psychological ones, can bemade. Researchers? assumptions, often implicit, folk-psychological ones, structurethe questions they ask, the terms and concepts they invoke to answer them and theirsense of what even constitutes an explanation. Distal theories (formal ones at least),have the advantage of making many of these assumptions explicit and thus testable.My own favourite example to explicate this is the Polish concept of przykrosc.Przykrosc loosely translates as the bad feelings that result from not meeting an-other?s social expectations, not having ones expectations met or anticipating thissituation. English does not have an equivalent concept and the reference of przykrosccan only be communicated exactly in English by numerous case examples2. Przykroscis an integral part of Polish folk-psychology and is regularly invoked to explainpeople?s motivations, beliefs and behaviour. Self-esteem on the other hand, whichthe Poles rarely invoke, seems quite central to North American folk-theories ofpsychology.2 Demonstrated by Polish-Australian linguist Anna Wierzbicka (2001)15It can be just as difficult for Poles to grasp that przykrosc might not be an ob-vious, universal, regularly occurring and causally potent state of mind as it is forNorth Americans to grasp that self-esteem might not obviously be a meaningfuluni-dimensional psychological variable. Both przykrosc and self-esteem are the-ories about human minds. The fact that large communities of people naturallyreason and interpret their own internal states in terms of these concepts does notmake them any less conjectures.The influence of such implicit theories on psychological science is not triv-ial. North American psychologists have developed many tools (typically surveyswhose answers correlate across items and individuals) to measure self-esteem, butnone (as far as I know) to measure przykrosc. Countless studies find it perfectlynatural and obvious to ground their investigations in the self-evident meaning ofsuch concepts. Why did our participants behave differently? Because of differ-ences in their self-esteem. What is the result of this manipulation? A change inself-esteem. Such answers are perfectly meaningful to a North American and re-quire no further definition or justification.However, like medieval philosophers who were no less fascinated by the worldthan we are, but happily terminated their explanations when they reached a truthmade self-evident by the bible, investigations which begin, end or proceed by theself-evidence of przykrosc, self-esteem or any other untested theories encoded inour language and culture are liable to slip right past important discoveries. The ad-vantage of distal theories, in particular theories which formally define their ontol-ogy and the relationships between its elements, or build from a broader frameworkwith a formally-defined ontology (like evolutionary or game theory), is that theymake these assumptions explicit and testable.Our language and culture have had thousands of years to learn about the world,especially our own psychology. Compared to this, formal theories are only takingtheir first tentative steps and are likely at first to be terribly simplistic comparedto our folk-intuitions. However they also have advantages: their strict explicitnesspermits them to draw on the language of mathematics to represent the complexrelationships they postulate, making it far easier to see our intuitions and beyondthem.Ultimately the choice to investigate distal, low-resolution theories belongs to16each individual scientist, though the current weight of opinion in psychology seemsto be against it. For those who undertake this project (and undertake it empirically)it is important they be aware of the distinct challenges it presents.1.5.2 The Trouble with Evolutionary InferenceFor any psychological phenomenon, or finite set of them, a logically infinite andpractically very large number of evolutionary stories can be devised. Consistingof claims about events in the distant past, or worse their aggregate statistical prop-erties, these theories can sometimes be entirely untestable. This lack of directevidence makes it all too easy for a clever distal theorists to, when their predictionsfail, make just the right ?calibrations? to their assumptions to bring their theoryback in line with any evidence. This means that distal theories, incautiously de-ployed, can be nigh on unfalsifiable and almost certainly unfalsifiable by observa-tions on the scale with which most high-resolution scientists concern themselves.It gets worse. Such invocation of local excuses to defend distal theories fromlocal inaccuracies is not an invalid move on the part of the distal theorist. Real, gen-uine distal patterns are distorted by local noise and don?t explain high-resolutionphenomena precisely. We should expect that even if we correctly identify an ac-curate distal theory, it won?t fit local data as well as many proximate theories thatmake no distal claims at all. To understand why this is, consider that noise occurson different scales. For evolutionary theories, these scales include individual vari-ation caused by ontogenetic differences, cross-cultural variation, ecological vari-ation and temporal or historical variation. A simple example is height - a traitwith obviously heritable differences whose current mean and variance have verylikely been influenced by evolutionary selection pressures. Nonetheless the localhistory of height is littered with examples of systematic deviations. Though ge-netic differences between modern populations seem relatively unimportant in theirdetermination, height differences between modern populations are stark (Deaton,2007). For instance, childhood nutrition and disease exposure can alter the heightsof entire populations, as has been particularly evident in the universal pattern ofmean heights first rising and then levelling off across the Western world between1950 and 1980, starting with the richest countries. This is true everywhere except,17apparently, Africa where these relationships do not hold . Larger scale temporalvariation is also evident in groups whose heights have been well-documented, likemilitary recruits for example: ?Norwegian recruits of the 1760s were as short asbushmen; today they are as tall as the Dinka of the Sudan? (Floud et al., 1990,pp.5).Though an optimum height may historically have been selected and may stillbe under stabilising selection, due to their cultural history and local ecology, gen-erations and populations deviate from this. By sampling across enough cultures3,environments and historical periods, the random distribution of these differencesmay become apparent and the evolved pattern shine through, just like random sam-pling within these groups is used to reduce the noise between individuals in high-resolution psychological studies. If we?re lucky, such massive investigations maybe possible, but more often than not we are restricted to investigating psychologi-cal phenomena on scales smaller than the noise that distorts evolutionary patterns.In these cases proximate explanations which account for the noise along with thepattern will always produce a better fit to data than the evolutionary explanation.Worse still, there may be local distortions of evolutionary patterns which arenot random on any scale. One school of evolutionary psychology emphasises pre-cisely such situations - our stone aged minds, they claim, are often maladapted tothe modern world (Tooby & Cosmides, 2005). When this is true, even idealisedstudies can?t observe our behaviour serving hypothesised adaptive functions: itno-longer does. Then distal theorists are left with only errors, mistakes and mal-adaptations from which to draw inferences, and must also rely on additional (oftendifficult to test) hypotheses about unobservable ancient conditions, the changesthese underwent and the resulting distortions.Given that a) even accurate distal theories can struggle with precise predic-tions, b) painstakingly investigated details can be effortlessly incorporated with theclaim that they?re consistent local mechanisms, or just as easily dismissed as noise,and c) the threat of spending one?s career pursuing a false lead that happens tobe unfalsifiable, it is easy to understand why researchers might be wary of distal3 Cultures are transmitted and so vary phylogenetically, like species. Even if a random processis distorting cultures around an evolved equilibrium, the shared history of the cultures in any givensample can make this difficult or impossible to discern.18theorising.Still, I believe it is possible to make clear and worthwhile contributions toscience as a distal empiricist - that is, a researcher who uses concrete, replicable,modern evidence to help resolve distal questions. This job is made considerablyeasier by clear and explicit criteria for what sets good distal theories apart frombad ones, what sort of evidence can have bearing on this question and conclusionscan be reliably drawn from it. To complete this thesis I found it necessary to try toestablish at least provisional answers to these questions, which I present below.1.5.3 SolutionsSometimes the ore in a scattering of core samples actually is connected by a some-what regular, continuous deposit. Better still, sometimes the shape of this depositis, by and large, the consequence of the repeated action of a few simple physicalforces and constraints. On these occasions, by understanding these forces we caninfer accurate maps of underground deposits from our scattered samples and thusthe location of the next deposit. When we get it right these maps are incrediblyvaluable.Two related properties make these accurate maps discernible from pictures ofMickey Mouse: their ability to straightforwardly generate distinct, accurate novelpredictions and their ability to straightforwardly account for disparate data. By?straightforwardly? I mean: with little or no ad hoc hypothesising or calibration ofthe theory by incorporating local details.In psychology, we are fortunate to be not constrained to fitting existing data; wecan perform new tests. Though a picture of Mickey Mouse can be redrawn aroundany data points we gather, only an accurate map can correctly predict a large rangeof novel phenomena without needing any redrawing. This property makes accuratedistal theories valuable and eminently testable.Novel predictions alone are insufficient. By filling in the area between obvi-ously related data-points and by capitalising on the intuition and cleverness of thetheorist, sketches of Mikey Mouse can also sometimes find success in novel pre-dictions. Distinct, straightforward novel predictions are sufficient as they are theexclusive domain of accurate theories. Two particularly salient historical examples19of successful novel prediction will help to demonstrate these properties and theirpower.The planet Neptune, too faint to be seen by eye, was discovered in 1846 onlyafter astronomers hunted for a hypothetical planet to account for deviations in ob-servations and physics-based predictions of Uranus? orbit. In 1819, science giantSimon Poisson attempted to discredit a theory by relatively nameless engineer, Au-gustin Jean Fresnel?s. Poisson used Fresnel?s equation describing light as a waveto derive a patently absurd conclusion: the shadow cast by a disc of just the rightsize, at just the right distance from a light will have a bright white spot in the centreas though the disc had a hole. When the unintuitive white-spot was observed, thedistinct straightforward novel prediction helped establish Fresnel?s theory as thenew orthodoxy (Worrall, 1988).Besides often having their historical significance over-emphasised, these salientexamples of novel prediction share three important properties: straightforwardness,distinctness/precision and diversity. It is no surprise that a theory built to explain aphenomenon explains it. Genuinely temporally novel predictions, those made be-fore a phenomena is ever observed, are valuable precisely because they guaranteesuch tinkering could not have occurred. Straightforward novel predictions then arethose derived from theories not designed, either originally nor by the addition ofad hoc hypotheses, to make them.Each example concerns highly distinct, precise predictions: they do not followfrom many other theories. Any man or his dog could have predicted that more plan-ets existed than had been observed, or that some areas of a shadow may be brighterthan others. Picking the exact location of Neptune from the vast sky, or the exactdiameter, distance and brightness involved in the white spot phenomenon requireprecision and sound theory, not intuition and blind luck. The more distinct a novelprediction, the more unique evidential support its fulfilment realises for the the-ory that made it. The evidential support of indistinct predictions is quickly dilutedby competing theories4. This does not mean that only completely distinct novel4An example of local interest is the human proclivity to discriminate conservatively towards themembers and institutions of their own social group when they feel uncertain. Though this phe-nomenon was first observed after a prediction derived from a theory with evolutionary overtones(Rosenblatt et al., 1989), a number of other theories (Navarrete et al., 2004; Heine et al., 2006;Kay et al., 2008), some of them evolutionary, have since straightforwardly predicted the same phe-20predictions can serve as evidence for distal theories, only that the more distinct aprediction, the more the unique evidential weight its observation carries.The final criterion, diversity, is important because good theories don?t just makenovel predictions. Although less dramatically impressive, similar evidential sup-port as that provided by novel prediction can be garnered from previously observedphenomena. Both Fresnel?s and Newton?s theories were eventually superseded,and though their successors hadn?t been around to predict light-spots and planets,they could easily have done so without invoking any additional hypotheses, andwould have been more accurate that the original theories. When a theory couldhave straightforwardly predicted a phenomenon in advance but didn?t by the his-torical accident of human affairs, logically the prediction?s evidential value doesn?tchange, though its salience in the minds of scientists might. Novelty is impres-sive, but what makes accurate distal theories distinct (including Fesnel?s, Newton?sand their successors?) is that they gracefully, without ad hoc additions, account formany phenomena they were not designed to explain, some before and some afterthe fact. The diversity of distinct phenomena distal theories can straightforwardlyexplain provides a metre by which to compare them.Using this metre is not as simple as tallying which theory makes the most cor-rect, distinct, straightforward predictions; human psychology is far fuzzier than thephysics of interference patterns and orbital determinations. In a paper on the prob-lem of picking out real patterns in a noisy world, Daniel Dennet (1991) provided aclear exposition of the possible results of comparing two distal theories (systems)against noisy proximate data:First, there are the occasions where they agree and are right. Bothsystems look good from the vantage point of these successes. Second,there are the occasions where they agree and are wrong. Both chalk itup to noise, take their budgeted loss and move on to the next case. Butthere will also be the occasions where they disagree, where their sys-tems make different predictions, and in these cases sometimes (but notalways) one will win and the other lose ... in many cases they will dis-agree and both be wrong. When one wins and the other loses, it willnomenon and explicitly challenged the original distal story.21look to the myopic observer as if one ?theory? has scored a seriouspoint against the other, but when one recognises the possibility thatboth chalk up such victories, and that there may be no pattern in thevictories which permits either one to improve his theory by making ad-justments, one sees that local triumphs may be insufficient to provideany ground in reality for declaring one account a closer approximationof the truth.Now some might think this situation is always unstable; eventu-ally one interpretation is bound to ramify better to new cases, or bededucible from some larger scheme coving other data, etc. That mightbe true in many cases, but .. it need not be true in all.It is not yet clear in the case of evolutionary theories of human uniqueness,whether the evidence we gather will eventually select out just one best accountor whether we?re condemned to perpetual indeterminacy between theories, eachaccounting for mutually inconsistent portions of the noise. What is clear is thatany single prediction of any single phenomenon will likely not be enough. Theslow tally of successes may eventually be superseded by an outstandingly distinctdiscovery, the psychological Neptune in the sky, but for now our best strategy forevaluating distal theories is deriving from them distinct, novel predictions aboutdiverse phenomena and comparing these to good psychological observations. Inthe next section I turn to the question of what constitutes a good observation forthis purpose.1.5.4 Good Psychological Observations(For Evaluating Evolutionary Theories)Theories especially designed to explain particular phenomenon can and shoulddraw on other empirical results, other proximate theories, folk-psychological in-tuitions and all of our best available data on the phenomenon itself. A distal theorymust do precisely the opposite: explain phenomena it was not designed to. It isunsurprising that our criteria for what constitutes good psychological evidence forproximate theories do not hold for distal ones.A good distal empiricist doesn?t look too deep. For example the theory of evo-22lution is tested by its ability to account for the pattern of physiology and behaviouracross many species, not by mechanistic predictions for any one species and themyriad of maladaptive counter-examples. Similarly a distal theory explains broadtrends in related phenomena, not the exact local mechanism. Local psychologicalphenomena that run counter to a distal theory are, alone, trivial to evaluating anevolutionary hypothesis. Thus, detailed specification and exclusion of all plausiblemechanistic explanations of a phenomenon, the bread and butter of high-resolutionstudies, often provides absolutely no additional information for discriminating dis-tal theories. Instead the efforts of scientists probing distal questions are more pro-ductively applied to generating predictions and tests of functional consequencesacross a range of phenomena.This is not an excuse for sloppy empiricism ? distal empiricists need to lookwide, but they still need to look deep enough. Two key features of proximatemechanisms can spell disaster to distal hypotheses and deserve careful attention:inconsistency and indistinctness.A proximate mechanism is inconsistent with a distal theory when it is not alogical outcome of the theory. To evaluate inconsistency, a distal empiricist shouldask: ?Have I excluded plausible mechanisms which could explain this phenomenonbut could not have been generated by my distal cause?? For evolutionary theo-ries, one core feature proximate mechanisms require is heritability. Non-heritablemechanisms could not have come about by cumulate selection of heritable differ-ences. Besides heritability, the evolutionary empiricist should evaluate if incremen-tal changes (in the right direction) in the local mechanism driving this phenomenonwould give consistent advantages in the environment proposed by the evolutionarytheory.A mechanism is indistinct when too many other distal theories (perhaps evenour foe Mickey) are consistent with it. Evaluating indistinctness is harder in thatthe distal empiricist must evaluate the consistency of alternative distal theories withthe most plausible local mechanisms. Logically there are an infinity of distal the-ories, but practically it is reasonable for empiricists to compare just the available,well-developed theories. A heuristic makes this task easier: the more precise a pre-diction, the more distinct it usually is. A good distal theory predicts not only theoccurrence of a phenomenon but also the circumstance where it won?t. Faced with23an indistinct mechanism, an empiricist may attempt using a distal theory to predictboth the presence and absence of a phenomenon under different circumstances inan effort to acquire more robust results.Therefore a good distal empiricist looks at a wide range of phenomena, ex-amining each just deep enough to evaluate inconsistencies and indistinctness with-out getting so deep as to get lost in trivial counter-examples. A good distal the-orist leaves evaluating every proximate mechanism for a particular phenomenonto local specialists and instead evaluates proximate mechanisms for consistenceand distinctness. We are now in a position to succinctly state the job of a low-resolution empiricist: straightforwardly deriving distinct predictions from distaltheories about a large range of phenomena, and testing each with careful attentionto evaluating whether plausible proximate explanations are inconsistent or indis-tinct.1.6 Framing My ContributionsIn Section 1.5, I tried to define the task facing a researcher evaluating evolution-ary theories with psychological evidence. I am not certain that I identified all thechallenges, all the solutions or even that the solutions I did identify are reliable.Nonetheless this is the frame in which I have undertaken the studies introducedin Section 1.3 and reported below. In the first study, I started from an existingevolutionary inference from psychological evidence and tested its robustness withrespect to inconsistent local explanations. In the second and third studies, I testedevolutionary predictions of novel psychological phenomena.24Chapter 2Social Biases in Recall:Replication Across Cultures andMaterialsRecently psychological evidence has augmented the archaeological record in dis-criminating which pressures drove our species? rapid brain expansion. Our mem-ory biases seem more consistent with social than ecological pressures (Mesoudiet al., 2006). However, a number of inconsistent proximate explanations need to beevaluated before we can be confident what this modern psychological phenomenonindicates about the distant past. Here we evaluate two such proximate alternativesand find that this results is robust with effect to culture but not materials. Althoughmembers of different cultural groups have the same recall biases for coherent nar-ratives, recall for discrete, unrelated propositions does not show the same effect.This newly qualified phenomenon fits well with the effect of expertise typical inmemory research: recall is improved only for meaningfully arranged items. Theimplications of this bias for our evolutionary history are less clear than they origi-nally seemed.252.1 IntroductionThe emergence of our capacity for symbolic communication, reasoning, tool-useand other higher cognitive functions was accompanied or preceded by a periodof rapid brain expansion, particularly during the Middle Pleistocene (Ruff et al.,1997). Understanding the causes of these dramatic changes can provide insight intothe cognitive peculiarity of the human species and possibly help build a commonframework for integrating the presently disparate sub-fields of psychology.Valuable though it would be, progress towards a rigorous understanding ofthese events is still rudimentary. Since direct observation is impossible, we are con-strained to drawing inferences from traces of these events that survived the rigoursof time: fossils (Schwartz et al., 2005), phylogenetic reconstructions from the DNAof contemporary populations (Relethford, 2008) archaeological records (Mcbrearty& Brooks, 2000) and evidence of historical environmental conditions (Vrba et al.,1996; Ash & Gallup, 2007). Evidence is also emerging from cross-sectional com-parisons of the correlates of similar cognitive abilities in related species (see Dun-bar, 1998, for a review). These sources, though valuable, provide only contex-tual clues to the functional changes which occurred in Pleistocene minds and theircauses.Additional insight can be gleaned from the traces left in the very thing theseprocesses produced: modern human psychology. For instance, Herrmann and col-leagues (2007) have made progress towards understanding the historic divergenceof human cognition from that of our nearest primate relatives by comparing presentday performance on battery of tests. Mesoudi, Whiten & Dunbar (2006) attemptedto put the methods of social psychology to the same end. They considered twoclasses of explanations for human brain expansion: Ecological Intelligence Hy-potheses (EIH) and Social Intelligence Hypotheses (SIH). EIHs emphasise directforaging benefits resulting from better memory (Milton, 1981; Harvey & Krebs,1990), mental mapping (Clutton-Brock & Harvey, 1980) or the capacity for in-creasingly complex extraction techniques (Gibson, 1986). SIHs emphasise thebenefits of ever more skilful, subtle social relations (Humphrey, 1976; Whiten &Byrne, 1988; Dunbar, 1998). From these distal theories they derived conflictingpredictions as to whether modern minds should be biased towards better recall of26social over non-social information, and experimentally demonstrated a bias for re-calling social information.Though their result seems to support SIHs, to be confident in this inference wemust first eliminate plausible but incompatible proximate explanations. Incompati-ble mechanisms would explain the local phenomena but are not a result of the distalevolutionary pressures. In the present paper we carry forward Mesoudi, Whiten &Dunbar?s inference by experimentally testing two incompatible proximate expla-nations: the result is a consequence of a culturally acquired bias or a consequenceof the test materials.2.1.1 The Social Recall BiasTo test whether human brains were biased towards recalling social information,Mesoudi, Whiten & Dunbar composed four vignettes, intended to match in length,complexity and coherence. One vignette was composed for each of four kinds ofinformation:? Social Gossip: concerning intense third-party social relationships; in partic-ular, an illicit affair between a student, Nancy, and her professor.? Social Non-Gossip: concerning every-day third-party social relationships; inparticular, Nancy?s encounters on her way to the swimming pool.? Individual: concerning interactions between a single person and the physicalenvironment; in particular, Nancy?s delayed journey to class one morning.? Physical: concerning interactions and relationships solely within the physi-cal environment; in particular, the climate of Colorado and its relationship tofires and global warming.Mesoudi, Whiten and Dunbar hypothesised that a mind selected for the socialadvantages it granted should recall more of the social and gossipy vignettes thanthe individual or physical ones, and that a really socially-strategic mind shoulddifferentiate within social information to preferentially recall gossipy over non-gossipy information.27Each participant was given a set of vignettes to read, distracted and then askedto recall each vignette in detail. To improve the ecological validity of their studyand check for non-linear effects as information was forgotten and distorted, Mesoudi,Whiten & Dunbar transmitted this information along a four-person chain - what-ever their first participant recalled was then given to the second participant to recall,and so on ? for 10 parallel chains of unique participants.To quantify the amount of meaning recalled, Mesoudi, Whiten & Dunbar cap-italised on evidence that humans abstract and encode the information in text aspropositions, units of meaning consisting of a predicate and set of arguments (Kintsch,1974). Each vignette was constructed of 14 propositions so by re-coding a partic-ipant?s recall in propositional form and comparing it to the original set, they wereable to quantify the amount of the original information that was recalled, abstractedfrom how it was expressed.Their result was unequivocal: throughout a chain, recall for social informationwas far better than for non-social information, but without a recall bias withinsocial information. They did not find any obvious non-linear transmission effects asthe quantity of information recalled seemed to degenerate more or less consistently.Mesoudi, Whiten & Dunbar concluded that they had supported the SIH over theEIH. Looking at their results we propose that this bias may not be the result ofnatural selection for our more socially competent ancestors, but may instead be alesson transmitted epigenetically among members of a cultural species.2.2 Study OneThe historic trend of cross-culturally unrepresentative sampling present a seri-ous hindrance to drawing general inferences from much of the current empiricaldatabase of psychology (Henrich et al., forthcoming). The problem is compoundedwhen those inferences are evolutionary.When evolved cognitive adaptations interact with culture, identifying themclearly requires evidence of heritability, not just a demonstration of their effects.Evidence of substantial cultural variation in the social recall bias would imply theneed for careful experimental partitioning of this variance into genetically herita-ble, cultural, environmental and individual components. Invariance on the other28hand, consistent with a simple cognitive adaptation, would make for much simplerinferences.Cross-cultural psychologists have been uncovering an ever growing class ofcross-culturally variable psychological effects (Heine, 2008; Kitayama & Cohen,2007), where participants? behaviour regularly covaries with the cultural groupamongst whom they spent their formative years. What?s particularly impressiveis that these difference can involve low-level cognitive processes of which partic-ipants are not consciously aware, lending strong support to the claim that culturalenvironments have deep, long-lasting impacts on cognition.Our first study was designed to test for invariance of the social recall bias be-tween two cultural groups, North American and East Asian, between whom Cross-cultural psychologists regularly find cognitive difference (Nisbett, 2003).Vancouver has experienced a recent wave of immigration from the People?s Re-public of China, particularly the Hong Kong Special Administration Area (Statis-tics Canada, 2006). These first and second generation immigrants, along with lo-cals native to Canada for more than two generations, provided a convenient samplefor our purpose.2.2.1 MethodParticipantsWe administered the transmission-chain recall experiment to participants of dis-tinct cultural backgrounds. Forty participants had immigrated to Canada fromChina after reaching puberty, forty had been born in Canada but their parents hadimmigrated from China and forty were from families that had been established inCanada for over two generations and identified themselves as ?Canadian?. Thisfinal group were typically of European descent. All participants were undergradu-ates recruited around the University of British Columbia campus and were recom-pensed for their participation with a candy bar.29Materials and CodingThe participants were organised into thirty four-person chains, each consisting ofparticipants from only one cultural group. The first individual in each chain wasasked to carefully read the original vignettes (available in Mesoudi et al., 2006).Following a short distraction task (completing a simple maze), they were askedto write down as much as they could recall. Their output was given to the nextparticipant in their chain, who completed the same procedure and so on down thechain. Subjects were not aware that they would need to recall the material.Like Mesoudi, Whiten and Dunbar we used Kintsch?s (1974) propositionalframework to code the transmitted data. We experimented with three distinct meth-ods for quantifying changes during transmission.? By tracking changes in the constituent elements of propositions: we con-structed a shared set of uniquely identified predicates and arguments andeach coder reconstructed the propositions in each text from those elements,adding to them as necessary.? By tracking changes to the propositions: each coder began with the 56 (four-teen for each vignette) original propositions and for each recorded whetherit had been lost, distorted or transmitted faithfully by each participant in thechain.? By simply flagging the presence or absence of each original proposition ineach output.Quantifying meaning is complex and each of these methods had its challenges,advantages and disadvantages, yet despite their differences produced qualitativelyidentical results. We report here the final method because it is by far the simplestand thus most direct to analyse and interpret. Using this method the coding optionsavailable for any participant did not dependent on how prior participants in theirchain had been coded, which resulted in an inter-coder reliability of .84, the highestof all three methods.While coding, we discovered that quite often information was implied ratherthan overtly stated. For instance, the gossipy vignette includes the propositions (IS,30PROFESSOR, MARRIED), and (LEAVE, WIFE, PROFESSOR). These encodedwhether participants recalled that the professor was married, and that his wife lefthim. Participants would often omit just the first of these, though our coders feltthat it was still implied by their recalling the second. We felt this was an importantphenomenon and so also coded whenever propositions were implied in this way.This intuitions was vindicated by some participants explicitly recalling proposi-tions implied by their earlier chain members.Method of AnalysisThe dataset that resulted from our study consisted of 56 Propositions partitionedinto four Types, transmitted through 120 Participants partitioned into 30 Chains,each consisting of four Generations (the position of each Participant in the chain),and in turn partitioned into three Cultures. Each Participant either did or did notrecall each Proposition, generating 6720 binary data-points. Here-in we?ll con-tinue to use uncapitalised words (i.e. cultures) to refer to concepts and begin usingcapitalised words (i.e. Cultures) to refer to variables in our analysis1.To make our results directly comparable with Mesoudi, Whiten & Dunbar?s, webegan by replicating their analysis - for each Participant, we summed the numberof Propositions of each Type that they had recalled. On this transformed data, weran a 3x4x1 mixed ANOVA, comparing the main effects and interactions of Cul-ture, Type and Generation. Consistent with the previous stuyd, we treated Chainsas a random effects factor, since observations within a chain were clearly not inde-pendent.Dependence among generations wasn?t symmetrical, later generations were de-pendent on earlier ones. Visual inspection reveals the decline in recall per gener-ation to be roughly linear see Figure 2.3, so we instead modelled generation as alinear predictor. This produced qualitatively identical results to Mesoudi, Whitenand Dunbar?s approach: treating generations as independent levels of a factor.In his detailed investigation of whether texts encoded in memory have a propo-sitional structure Walter Kinstch was careful to point out that ?the difficulty thateach proposition presents is not a constant, but a random variable with an appre-1And strive to avoid sentences beginning with concepts!31ciable variance? (Kintsch, 1974, p.124)2. Since Mesoudi, Whiten and Dunbar werenot faced with the possibility of drawing important inferences from a lack of effectbetween participants (i.e. cultural invariance), it was reasonable for them to sim-plify analysis by discarding information about the variance between Propositionswithin each Type and the order of Generations. In our case, in order to max-imise our chances of detecting effects between cultures, we explicitly modelled thecovariance within Propositions and within Chains simultaneously with a logisticmixed-effects generalised linear model. This effectively let us ask the question:given how difficult a particular proposition is to recall, and which chain they hap-pen to be in, does a participant?s culture change the extent to which a proposition?stype influences their likelihood of recalling it?There are two ways to evaluate the effect of Culture in such a model. The mostdirect is to compare its log-likelihood to that of a model with Culture removedentirely. The second is to consider the ratios of its effects and interactions to theirstandard errors. In our case this is straightforward since the vast bulk of effects inour models fell into two categories: those that were non-significant by conventionalstandards, and those which were significant at p ? 1x10?16, marked? .001. Forsignificance values between great than this extremes, we report the precise valuesfor readers to draw their own conclusion.2.2.2 ResultsDid We Replicate the Main Effect?Yes. A 3x4x1 mixed ANOVA, identified two highly significant predictors of re-call: information Type (F(3,81) = 39.6, p? .001) and Generation (F(1,27) =132.5, p? .001). Neither Culture (F(2,27) = 0.4, p ? .67), the interaction be-tween Culture and Type(F(2,27) = 0.4, p? .67), nor any of the other interactionsin this model had a significant impact on recall.In Figure 2.1 we show the mean number of propositions recalled by each gen-eration, separated by Type. It is plain to see from the confidence interval overlaps(Cumming, 2009) that Social (non-gossip) and Gossipy propositions are recalled2He too ultimately resorted to summing.32Table 2.1: Main Effects of Type on Recall - SummaryType 1 Type 2 Can 2nd 1stIndividual Physical 0.92 0.03 0.58Social *** 0.01 ***Gossip *** *** ***Physical Social *** *** ***Gossip *** *** ***Social Gossip 0.15 0.03 0.22***: p? .001A summary of significant effects of information Type on recall in study 1. Significant valuesrepresent our confidence that the difference between the recall of any two Types is non-zero, forparticipants from a particular culture. Can is Canadian participants, 1st is first generationimmigrants to Canada (i.e. born in East Asia) and 2nd is 2nd generation immigrants (i.e both withinCanada to East Asian parents).Table 2.2: Main Effects of Type on Recall - DetailsType 1 Type 2 Canadian 2nd Generation 1st GenerationIndividual Physical 0.04?0.38, p? 0.92 ?0.83?0.38, p? 0.027 ?0.21?0.38, p? 0.58Social 1.84?0.38, p? 0.001 0.98?0.37, p? 0.008 1.38?0.37, p? 0.001Gossip 2.37?0.38, p? 0.001 1.78?0.37, p? 0.001 1.78?0.37, p? 0.001Physical Social 1.80?0.37, p? 0.001 1.82?0.38, p? 0.001 1.59?0.37, p? 0.001Gossip 2.33?0.38, p? 0.001 2.61?0.38, p? 0.001 2.05?0.38, p? 0.001Social Gossip 0.53?0.37, p? 0.15 0.8?0.38, p? 0.03 0.46?0.37, p? 0.22Details of main effects of information Type on recall in study 1, in the form ??SE, significance.These coefficients represent the difference between the recall of any two Types, for participantsfrom a particular significantly different rates to non-social propositions.Pairwise comparisons are more straightforwardly undertaken in the context ofour logistic mixed-effects model (described in more detail below). Even using verystrict levels of statistical significance to control for our multiple inferences, a veryclear pattern emerges that closely follows the pattern found by Mesoudi, Whittenand Dunbar?s: Social and Gossipy information are not transmitted at significantlydifferent rates from each other, nor are Individual and Physical information - butall other pairs of types are. These results are present fully in Table 2.2.33Figure 2.1: Number of Propositions Recalled, by Type with 95% C.I.sDid We Detect Cultural Variation?Yes, but not in the social recall bias. We constructed a logistic mixed model, withrandom intercepts for both Propositions and Chains. Type, Culture and their in-teraction were linear predictors and Generation as a linear covariate. We dummy-coded the factors Type and Culture and made comparisons among their twelvepossible pairings. We estimated coefficients and their standard errors by likeli-hood maximisation, implemented by the LME4 library of the R statistical pro-gramming language. Our Generation covariate was always a significant predictor,34Table 2.3: Main Effects of Culture on Recall - SummaryCulture 1 Culture 2 Gossip Social Individual PhysicalCanadian 1st Generation 0.64 0.78 0.33 0.762nd Generation 0.22 0.77 0.17 0.751st Generation 2nd Generation 0.44 0.98 0.02 0.99A summary of significant effects of participant Culture on recall in study 1, each cell presents a p-value, representing the the likelihood that the difference in our estimates of the logistic regressioncoefficients for any pair of two culture, for any given type of information was non-zero by samplingerror alone. Terms strictly significant by conventional standards (? = .05) have been underlined, forvisual clarity.? ?SE =?0.73?0.02, p? .001We considered the twelve possible main effects for culture: one for each Type,crossed by each possible pair of Cultures (i.e. 1st/2nd generation immigrants,Canadian/1st, Canadians/2nd). Only one of these twelve comparisons produceda coefficient significantly different from zero, or even approaching this level: thedifference between the recall of 1st and 2nd generation immigrants on Individ-ual information ? ? SE = ?0.87? 0.37, p ? 0.02. Note that since we are mak-ing twelve comparisons, this does not meet a more conservative Bonferroni-Sidakadjusted criterion of significance, ? ? 0.004. These results are presented in Ta-ble 2.3. Given our sample size, participants? culture did not consistently alter theirlikelihood of recalling any particular information type. This is particularly stark incomparison to the clearly and obvious main effect of information Type (Table 2.1).35Table 2.4: Main Effects of Culture on Recall - DetailsCulture 1 Culture 2 Gossip Social Individual PhysicalCanadian 1st Generation 0.17?0.36, p? 0.64 0.1?0.36, p? 0.78 ?0.36?0.37, p? 0.33 ?0.12?0.37, p? 0.762nd Generation 0.45?0.36, p? 0.22 0.11?0.36, p? 0.77 0.51?0.36, p? 0.17 ?0.12?0.75, p? 0.751st Generation 2nd Generation 0.28?0.36, p? 0.44 0.01?0.36, p? 0.98 0.87?0.37, p? 0.02 0?0.37, p? 0.99Details of main effects of participant Culture on recall in study 1, in the form ??SE, significance. Coefficients represent the difference in ourestimates of the logistic regression coefficients for any pair of two culture, for any given type of information. Terms strictly significant by conventionalstandards (? = .05) have had their significance values underlined, for visual clarity.Figure 2.2: Mean Number of Propositions Recalled, by Condition with 95%C.I.sA stricter test of cultural variance in the social recall effect, however, mustconsider the interactions between Culture and Type, asking the question: does thedegree to which social information is favoured in memory vary between cultures?Our first step to answering this question was to construct an identical model withthe interaction predictor removed and comparing the log-likelihood of the two al-ternatives. Modelling the interaction of Culture and Type seemed to significantlyimprove the model: Pr[?2(6) = 25.52] < 0.001.Looking deeper, this picture became less clear-cut. We examined each of the37eighteen possible interaction terms in our model. Each term estimates the changein the coefficient estimating the improvement in recall between one Type and an-other, between one Culture and another (or, symmetrically, vice versa), thus thereis one interaction term for every possible pairing of Types crossed by every pos-sible pairing of Cultures. These are presented in Table 2.6. Given these eighteencomparisons, a conservative ? level to avoid detecting random correlations is about0.002. Concerned that this might still be too liberal given the multiple main effectcomparisons we had conducted, we computed an adjusted ? level given all ourcomparisons, which was closer to 0.001. Fortunately this was a non-issue: anysignificance levels that exceeded the first value also exceeded the second. Theseinstances are marked *** in Table 2.5.There are several notable observations about these results. The vast bulk ofsignificant differences are caused by Individual information, the single informa-tion type in which we detected a main effect of culture. The only significant pair-ing that didn?t involve Individual information was between Gossipy and Physicalinformation, as recalled by Canadians and 2nd generation immigrants, ? ? SE =?0.57? 0.22, p ? 0.01. The interactions with Individual information are just asstrong and significant for Physical information as Social or Gossipy. Finally mostof the significant interactions, the most significant interactions and, indeed, the onlyinteractions when we invoke a conservative criterion of statistical significance, arebetween 1st and 2nd generation immigrants, not either group and Canadians.38Table 2.5: Culture-Type Recall Interactions - SummaryCulture 1 Can 1stCulture 2 1st 2nd 2ndType 1 Type 2Individual Physical ** ***Social * ***Gossip * **Physical SocialGossip *Social Gossip*: p < .05; **: p < .01; ***: p? .001A summary of significant interactions between Culture and Type in study 1. Significant valuesrepresent our confidence that the difference between the coefficients that describe a logistic estimateof the difference in recall between any two Types or Cultures, between any pair of levels the otherfactor. Can is Canadian participants, 1st is first generation immigrants to Canada (i.e. born in EastAsia) and 2nd is 2nd generation immigrants (i.e both within Canada to East Asian parents).39Table 2.6: Culture-Type Recall Interactions - DetailsCulture 1 Canadian 1st GenerationCulture 2 1st Generation 2nd Generation 2nd GenerationType 1 Type 2Individual Physical 0.25?0.23, p? 0.29 ?0.63?0.23, p? 0.005 0.87?0.23, p? .001Social 0.46?0.21, p? 0.03 ?0.4?0.21, p? 0.05 0.86?0.21, p? .001Gossip 0.53?0.22, p? 0.02 ?0.06?0.21, p? 0.78 0.59?0.22, p? 0.006Physical Social 0.21?0.22, p? 0.32 0.23?0.22, p? 0.30 ?0.01?0.22, p? 0.96Gossip 0.29?0.22, p? 0.20 0.57?0.22, p? 0.01 ?0.28?0.22, p? 0.21Social Gossip 0.07?0.20, p? 0.72 0.34?0.20, p? 0.09 0.27?0.2, p? 0.18Details of interactions between Culture and Type in study 1, in the form ??SE, significance. Beta coefficients and the difference between the simpleeffect regression coefficients that describe a logistic estimate of the difference in recall between any two Types or Cultures, between any pair of levelsthe other factor. Terms strictly significant by conventional standards (? = .05) have had their significance values underlined, for visual clarity.The sign on each term is associated with a shift from reference group Type 1, Culture 1 to Type 2, Culture 2. For example, consider the upper-leftmostcell: if we were predicting the recall of Canadians (Culture 1) for Individual (Type 1) information and were interested in how this differed from theirrecall for Physical (Type 2) information, the parameter we?d use in our logistic function would differ by -0.21 logits (top-right cell, Table 2.2).However, if we were considering 1st generation immigrants (Culture 2), we?d add .25 logits to this change (top-left cell, this table), resulting in a netdifference of 0.04 (top-left cell, Table 2.2). However we can see above that given our estimate of the standard error of this interaction, we could not beconfident that this difference had not occurred by sampling error alone.Figure 2.3: Exact Number of Propositions Recalled, by Type, Condition and GenerationFrom this pattern of results indicates that the relative rates at which Individualinformation was recalled differed between cultural groups, especially between 1stand 2nd generation immigrants. The centrality of Individual information to theeffect of culture is corroborated by constructing an identical model of a subset ofour data - just the other kinds of information. The interaction between Culture andType no longer makes a significant contribution to the predictive power of such amodel: Pr[?2(4) = 7.74]? 0.1.This pattern does not seem to reflect a clear cultural difference in recall bias forsocial and non-social information. The main effects of culture on recall and incon-sistent and the interactions do not follow a clear pattern distinguishing Canadiansfrom East Asians. In fact difference between 1st and 2nd generation immigrantsare often clear than those between either group and Canadians. What cultural ef-fects exist seem to be driven by individual information which differs as much fromother non-social as it does from social information. The social recall effect wasclearly evident in each cultural group: each recalled far more social and gossipyinformation than individual or physical; and the magnitude of this bias did notdiffer between cultures on a scale we could detect.Were There Differences in the Amount of Information Implied?Yes. Among the 1680 instance of propositions of each type being either recalled ornot recalled, 129 gossipy propositions, 64 social proposition, 14 individual propo-sitions and 4 physical proposition were implied rather than stated directly. Codingthe meaning implied by any speakers utterances is imprecise and implications willvary depending on each listener?s background knowledge and assumptions. Thesenumbers provide, at best, a rough estimate of our coder?s intuitions of the differ-ence in the degree to which redundant information didn?t need to be stated for eachinformation type.2.2.3 DiscussionWe replicated Mesoudi, Whiten & Dunbar?s social bias in recall using their originalmaterial with a larger, cross-cultural sample. Our participants recalled social infor-mation far better than either individual or physical information. We did not detect a42consistent difference in the magnitude, existence or direction of this effect betweencultures that could have posed a serious challenge to Mesoudi, Whiten & Dunbar?sclaim that this result was meaningful in discriminating evolutionary theories. Sincethese conclusions hinge on a null result, we can only say confidently that if the biasfor recalling social information is influenced by culture, this influence either doesnot differ for children inculcated in Canada and those inculcated in China - thesecultures may both have converged on the same optimal behaviour - or this varia-tion is too subtle to be detected by our instrument, which was powerful enough toclearly detect differences between recall for different types of information.Another serious confound remains: all our participants read the same materialin English. If properties of the material generated this effect, our result so far wouldtell us more about language than either cultural differences or evolutionary history.Given the material we transmitted, we find this is a highly plausible concern.2.3 Study Two2.3.1 IntroductionDrawing on his experience with tens of carefully conducted memory studies, Wal-ter Kintsch warned thatNot only must the number of propositions in a text base be controlled,but also their structural complexity. We do not even know, as yet,precisely what is involved in structural complexity, or what other vari-ables might be important. - (Kintsch, 1974, p.179)Though Mesoudi, Whiten & Dunbar ensured that each of their vignettes con-tained the same number propositions, they nonetheless contained systematic dif-ferences.3 During our many rounds of coding and recoding, we noticed three di-mensions on which the vignettes seemed to differ systematically.3 Mesoudi, Whiten & Dunbar verified their vignette?s invariance on dimensions other than thosebeing manipulated by asking 10 additional participants to rate their coherence, familiarity and real-ism on a 7-point scale and found no significant differences between these ratings. However, withoutsystematic controls nor any explicit theory of how naive participant?s intuitions map onto real struc-tural differences in memorability, Kintsch?s concerns remain relevant.43RedundancyRelated or redundant information between or within vignettes unsystematicallydistorts recall. Superficial similarities, both lexical (repeated words or names) orsemantic (related concepts), can improve recall by priming related material, a well-established phenomenon for which several models have been proposed (see McNa-mara, 2005, for a review). Substantive redundancy, where information is literallypresented or implied more than once, changes the amount of unique information ineach vignette.The most obvious example of superficial redundancy within the vignettes isthat three are about the character Nancy, while one is about Colarado. More sub-tle examples also exist: Nancy?s studenthood is implicated both in her having anaffair with her professor (gossip) and missing her class (individual). Substantiveredundancy is also evident within vignettes: the professor?s marriage is expressedexplicitly and also implied by descriptions of his extra-marital affair and the reac-tions of his wife.One way to glean insight into the degree to which such redundancy was asym-metrical between vignettes is to consider the number of times our coders flaggedpropositions that, though they weren?t explicitly stated, they believed participantshad implied obviously enough that any intelligent adult could infer them. Thisoccurred roughly ten times more often in the social vignettes.To avoid redundancy confounds in the revised study we abandon coherent nar-ratives entirely and instead tested participants? recall of discrete, unrelated state-ments. These statements were precisely syntactically balanced. To gain insightinto the effects of redundancy and its interaction with these information types, wesystematically varied the redundancy between the elements of these statements.For example, our simplest statements took the form (Subject - Verb - Object) andthe subject (the cup, or the farmer) was not obviously related to the object (theposter, or the priest). In higher redundancy statements these syntatic argumentsinstead formed related sets, for example (school-girl, tutor, study, diligently). SeeTable 2.7 for full details.44Familiarity / Expertise / ComplexityRedundancy is not just an objective property of information, it depends on an in-dividual?s background knowledge. A pair of chess-masters can, by describing thepositions of a few key pieces, share what an ordinary person would glean only fromten minutes of step-by-step demonstration. One of the best established results inmemory research is that experts are better able to recall material from domains withwhich they are familiar. Kintsch and his colleague Ericsson (1999; 1995) demon-strated this effect for recall for cohesive texts, theorising that expertise in a domainallows semantic links to long-term memory to augment the capacity of workingmemory.Mesoudi, Whiten and Dunbar?s gossipy, social non-gossipy and individual vi-gnettes concerned causal relationships very familiar to most individuals - a brokenalarm causing someone to hurry and be late, an affair causing someone to get upsetand leave their partner - while the relationships in the physical information mayhave been less familiar - carbon monoxide increasing global warming, dry veg-etation fuelling forest fires. It is quite plausible that participants? recall differedbecause of their close familiarity with the material, not because their brains areany more evolved for dealing with social information than a chess-master?s are fordealing with chess.Without knowing the detailed background of our participants, it is difficult toprecisely control familiarity. We attempted to avoid this confound by ensuring theelements of all our statements were incredibly simple, common words and con-cepts. As a rough proxy for the effects of familiarity, we systematically increasedthe syntactic complexity of our statements. This at least allowed us to analysethe effect of increasing strain on working memory, the opposite pressure to whatKintsch and Ericsson?s investigations suggest expertise exerts.Truth-ValueIn another study (Chapter 3) we found that people?s agreement with a statementinfluences their recall. For non-strategic information (the type in this study) peopleare less likely to recall statements they agree disagree with. Mesoudi, Whitenand Dunbar?s physical vignette concerned real physical relationships (i.e. fires45produce carbon monoxide) in a real place (Colorado), which participants couldhave judged as either true or false. Their other vignettes concerned the interactionsof fictitious characters, no more true or false than any work of fiction. Though wehave no data on participants? agreement with Mesoudi?s vignettes, it is plausiblethat this systematic difference in whether propositions could have a truth-valuecaused recall differences.We explicitly controlled for this in our study by systematically varying whetherstatements concerned purely fictitious, general characters and places or real, spe-cific ones familiar to most people (Alert Einstein, Canada, etc).2.3.2 Method46Table 2.7: Discrete, Unrelated Recall ItemsCategories TypesT.V.? Redundancy Complexity Gossip Social Individual PhysicalNoLessLessThe farmer black-mailed the priestThe pilot intro-duced the bakerThe authordropped the tapeThe cup hit theposterMoreThe driver reck-lessly stole fromthe house-keeperThe teacher pa-tiently waited forher friendThe scientistquietly sat on therockThe tree slowlyfell over near thelakeMoreLessThe girl deceivedher loverThe lawyer ad-vised her clientThe player kickedthe ballThe breeze blewthe leafMoreThe fan shame-lessly flirted withher heroThe school-girldiligently studiedwith her tutorThe diver care-fully swamthrough the waterThe river flowedrapidly to theoceanJustifiedThe businessmancheated on hiswife because hewas boredThe tourist talkedto the local be-cause he waslostThe officerwalked aroundthe vehicle be-cause he wassuspiciousThe weatherdried-out theflowers because itwas hotYesLessJulius Caesarspied on hisfriendsMichael Jacksonlistened to hisbrotherSteven Harperdrank from thecupThe Eiffel Towersparkled in theeveningMoreBrad Pitt divorcedhis wife foranother womanBarrak Obamaconsults his staffabout policyAlbert Einsteinwrote equationsin his bookSnow coversCanada in thewinterItems recalled in study 2. They are organised by whether a) they are about real people and places, and thus might be assigned a truth value (T.V.?);b) their arguments are in related set (i.e. [river,ocean]; [officer, vehicle, suspicious]) and thus have more redundancy; and c) their syntactic and logicalcomplexity, ranging from (subject, verb, object) to structures with adverbs, compliments, conjunctions and clauses justifying the main action.Non-linear effects were not detected in either our nor Mesoudi and Dunbar?stransmission design, so in this study we maximised our statistical-power-per-subjectby reverting to a simple cross-sectional design.We recruited 30 participants, equivalent to the number of chains in our previousstudy, by the same procedure.We generated 28 items for participants to recall, listed in Table 2.7, balancedas described above. Participants were presented with all 28 items in a random or-dered and asked to rate each item on a 5 point scale labelled: ?Completely Agree?,?Somewhat Agree?, ?Neither Agree nor Disagree?, ?Completely Disagree?, ?Some-what Disagree?. Participants were not aware that they would need to recall thematerial.Participants were then asked to recall as many of those items as they could.This resulted in a matrix of binary values: for each item, for each participant, theydid or did not recall it. The probability of any item being remembered covariedwithin each participant, some participants just had better recall.2.3.3 Results30 Participants either did or did not recall 28 Items, generating a total of 840 di-chotomous data points, clustered by Participant and Item. We again constructeda logistic mixed effects model with random intercepts for both Participants andItems, effectively asking the question: given each participant?s mean rate of re-call and each item?s mean rate of being recalled, what effect did our predictorshave? Our predictors were information Type, participant?s Agreement with eachstatement, coded as a factor of discrete, ordered levels, and the Order in which par-ticipants saw each item. The details of this model are presented in Table 2.8 andreviewed below.Variation in information Type had almost no effect whatsoever on recall (Pr[?2(3)=0.45]? 0.93).We considered a linear prediction of the influence of presentation order: an in-teger encoding the randomised sequence in which each participant saw each item.We also mean-centred and squared this vector to produce a quadratic Order pre-dictor. Visual inspection of the effects of Order suggested the well-known primacy48and recency effects: participants seemed to recall items they saw earlier and laterbetter than items presented in between. This was confirmed statistically by thehighly significant contribution the quadratic Order term made to the model?s pre-dictive power (Pr[?2(1) = 23.96]? 0.001), and the inability of the linear Orderterm to significantly improve this fit (Pr[?2(1) = 2.71]? 0.1).Participants? Agreement with statements significantly predicted their recall ofthose statements (Pr[?2(4) = 18.49] ? 0.001). The details of the complex setof 25 interactions between agreement levels that produce this effect do not bearenough relevance to the conclusions of this analysis to warrant a full discussion.Their pattern can be summarised by breaking this predictor into two effects - adichotomous effect of having an opinion at all (i.e. not rating something as no-opinion) and a linear effect of increasing agreement. Having an opinion at all con-siderably improved participants? recall (? ? SE = 1.41? 0.41, p < 0.001), whilea linear approximation of increasing agreement predicted slightly worse recall(? ?SE =?0.21?0.1, p? 0.05).To ensure our results did not depend on the confounding influences we dis-cussed above, we partitioned our items according along the dimensions evident inTable 2.7 and considered the effects of this partition on our model. Neither the ex-tra redundancy we added between the arguments in our statements (p? 0.28), norslight differences in syntactic complexity(p ? 0.58), nor whether statements hada truth value or were purely fictitious(p ? 0.11), nor their simultaneous inclusion(Pr[?2(2) = 2.67]? 0.26) had enough effect on participants? recall to significantlyimprove the predictive power of our model. Notice that complexity could eitherbe encoded as discrete, separate levels (less, more, justified; see Table 2.7) or asa linear approximation of their increasing complexity (1,2,3). This decision didn?tqualitatively change the analysis, so we report the linear approximation here andin Table 2.8 as it generates an easily interpretable coefficient.49Table 2.8: Discrete Recall ModelModel: Y = Intercept + Type + Agreement + Order + Order2 + Complexity + Redundancy + Truth Value? Coefficient -1.2 0.02 0.007 0.20 -0.42 0.60Standard Error 0.24 0.01 0.001 0.30 0.39 0.42p *** 0.09 *** 0.52 0.29 0.15LLRT 0.93 0.001 0.1 *** 0.58 0.28 0.11***: p? .001A summary of the model developed in study 1, including coefficients, their standard errors and the likelihood that coefficients? difference from zero isnot the result of random sampling alone (p). The resultant value, Y, represents the natural logarithm of the odds of recalling a particular statement.Predictors to the left of the vertical bars constitute the core statistical model in this study, predictors to their right are potential confounds. The rowLLRT reports the significance value resulting from a Log Likelihood Ratio Test of a model containing this predictor against one without it, i.e. thelikelihood that the difference in the fit of these two models is greater than would be expected by random sampling alone. LLRTs on predictors in thecore model (those left of the bars) were performed by removing them from the core model, while the confounds (those to the right) were added to themodel. A simultaneous LLRT of all the confounds was also not significant, p? 0.26. Factorial predictors (Type and Agreement) cannot be reported assingle coefficients, but rather dummy coded to generated a coefficient for each possible pairing of levels. However the likelihood that each factor?scoefficients differ from zero can be simultaneously evaluated by their LLRT results.2.3.4 DiscussionUsing a sample large enough to detect difference in recall between social and non-social information using the original coherent vignettes, we could not detect suchdifferences when subjects were asked to recall discrete, unrelated statements.On our reading, the SIH does not straightforwardly predict that recall for socialinformation should only be greater for cohesive passages and not unrelated items.A proximate mechanism well established in memory research: expertise effects onrecall (see Vicente & Wang, 1998 for a review, but c.f. Simon & Gobet, 2000),can account for this boundary condition. Experts in a domain have dramaticallybetter recall for meaningfully organised items in that domain, but this effect is con-siderably reduced or vanishes for random or unrelated items (see Kawamura et al.,2007, for a recent replication). Such domain specific expertise has been shownto be readily acquirable (200-300 hours practice can improve recall for strings ofrandom digits by a factor of 10 (Chase & Ericsson, 1982)).Our results suggest the social recall bias effect can be explained by two proper-ties of Mesoudi, Whiten and Dunbar?s materials. Firstly, their social propositionsconcerned topics participants had expert familiarity with. Secondly they weremeaningfully arranged, that is there was enough semantic redundancy betweenthem to activate established semantic clusters in participants? long term memory,augmenting their short term recall.The expertise effect on recall does not explain why participants should be ex-perts on just social rather than individual or physical information.2.4 General ConclusionsMesoudi, Whiten & Dunbar?s discovery of a bias for recalling social informationseemed to support social theories of brain expansion over ecological ones. Ourfindings here help secure this evolutionary inference against criticisms on culturalgrounds, individuals inculcated in two very different cultures seem to show thesame effect. However we also found that the effect does not apply to recall forunrelated statements. This mirrors a common finding in memory research: expertshave better recall than non-experts in their domain of expertise for meaningfullyrelated items, but not unrelated ones. The implications of these new boundary51conditions for discriminating the SIH from the EIH are not as clear as those of asimple, universal memory bias.It does not seem to us that the SIH straightforwardly implies that social recallshould be better for cohesive stories but not unrelated pieces of information. Nordoes it seem to be a better explanation of the existence of the expertise effect onrecall itself. The advantages of capitalising on familiarity to enhance short-termrecall are so general, as are the domains in which this effect has been observed,that almost any evolutionary hypothesis could account for their emergence.The SIH can account for participants may have been experts at social but notindividual or physical information. Such an account might claim that people reg-ularly become social experts because a bias for attending to the social world, wasgenetically selected for the social benefits it granted our ancestors. One of thesebenefits may have been improved recall for cohesive social information, an exap-tation of an pre-existing memory mechanism.The EIH can also be made more consistent with this newly qualified effect byad hoc hypotheses. For example, an adaptation for extracting food (Gibson, 1986)from a variable environment (Potts, 1998), in combination with an urban ecologycould dispose people to gradually gain expertise in observing social interactions.Since expertise can be developed quite rapidly, neither of these evolutionary hy-potheses are necessary to explain this effect. Our participants? local ecology mayjust happen to demand a lot of attention to social interactions - they were universitystudents living in a modern city. Compared to an invariant, general bias for recall-ing social information, the ontogenetic acquisition of expertise is far less distinctlyimplied by SIH vis-a-vis other hypotheses.Our study and Mesoudi, Whiten & Dunbar?s were designed to test for a simple,direct memory bias for social or ecological information. They do not constitute afair test between these more sophisticated, indirect hypothesis about biased exper-tise acquisition. A test designed for such hypotheses might, for instance, includereplication with participants from small scale societies for whom foraging involvesextracting food from a physical rather than social environment, and thus might beexpected to be more experienced with attending to physical information.The conclusions discussed here are based on null results. It is entirely plausiblethat a more powerful test would find a social bias in recall for discrete, independent52propositions, but it is not obvious that a very small effect would be evidence of amind adapted for processing social information, rather than merely an incidentalby-product of other cognitive processes or expertise.The social recall bias for cohesive information seems to be stable across cul-tures but not materials: it doesn?t extend to unrelated items. Without follow-upwork, particularly with individuals in small scale societies, the implications of thiseffect for evolutionary theories of human brain expansion are not clear.53Chapter 3Differences in Memory Biases forStrategic and Non-StrategicCultural InformationWritten and formatted to the specifications of a short Research Report in the journal Evo-lution and Human Behaviour.Predicting hitherto unobserved phenomena can reveal more about the accu-racy of evolutionary accounts of modern psychology than their agreement withexisting data. The coevolutaionary or cultural intelligence hypothesis, an accountof evolutionary dynamics that influenced the emergence of our species, seems tostraightforwardly account for many known aspects of our contemporary psychol-ogy and sociology. Here we derive from it a novel prediction about content biasesin memory: agreement predicts recall only for non-strategic information, and re-port an empirical verification. Though follow-up investigation is required to estab-lish the proximate mechanisms that generate this novel effect, its straightforwardtemporally-precedented prediction has important evidential value for coevolution-ary theory.3.1 IntroductionEvolutionary theories, especially those related to human psychology, have beencriticised for relying on untestable speculation about historical conditions (Gould54& Lewontin, 1979). It can sometimes seem that for a bright, creative mind, fittingdistal theories to existing data can be all too easy; for instance at least twenty dis-tinct adaptive hypotheses have been proposed for the female orgasm (Lloyd, 2005).The phenomenon of confabulation (Hirstein, 2004) reveals just how skilled humanminds are at rapidly fitting a plausible story to any set of constraining evidence.Fortunately a stricter test of evolutionary theories exists: their ability to accuratelypredict previously unknown phenomena.Proximate explanations for particular phenomena, the stock in trade of exper-imental psychology, can and should be whittled and tweaked towards accuracyby painstaking experimental investigation of their details. This technique doesn?twork for distal accounts. Real evolutionary patterns, though valuable in explana-tion and prediction, are distorted by the noise of local detail: individual, ecological,cultural and so on. The details of any particular phenomenon, no matter how wellunderstood, are usually insufficient to verify, falsify or discriminate between distalhypotheses (Dennett, 1991). A better criterion is accurate prediction of diverse phe-nomena distal theories were not designed to explain (Worrall, 1988). Completelynovel observations guarantee this lack of deliberate calibration1. The project ofevaluating distal, evolutionary theories then, consists in the derivation of empiricalpredictions across phenomena and the accurate reporting of both their successesand failures.In a recent study, Mesoudi, Whitten and Dunbar (2006) derived predictionsabout hitherto unobserved biases in human memory from evolutionary hypotheseswhich emphasise the role of sociality in the recent rapid increase in human brainsize (Humphrey, 1976; Whiten & Byrne, 1988; Dunbar, 1998). These predictionsbore out experimentally, but may have hinged on linguistic properties of the mate-rials being recalled2.We extend this technique to derive and test a prediction from a related butdistinct evolutionary theory: coevolutionary theory or Cultural Intelligence Hy-pothesis. Several groups of theorists from different traditions have contributed todeveloping this position (see for instance Cavalli-Sforza, 1981; Boyd & Richerson,1988b; Durham, 1992). It is rooted in the claim that the course of human evolution1For a more detailed discussion of criteria for evaluating distal theories see Section 1.5.2See Chapter 2.55was drastically altered by our capacity to accurately share phenotype-specifying in-formation during a lifetime, by high-fidelity cultural learning. A number of formalmodels derive the consequences of the concurrent coevolution of cultural and ge-netic information on the subsequent phenotypes of cultural animals, including sen-sitivity to prestige (Chapter 4; Henrich & Gil-White, 2001), ethnicity (McElreathet al., 2003) and informational conformity (Henrich & Boyd, 1998). Any modelis a simplification, painting a characture of our complex world which emphasisesonly certain dynamics. The Cultural Intelligence Hypothesis paints a world where,in addition to navigating their ecology and social environment, individuals thriveby their ability to acquire, combine, rebroadcast and otherwise skilfully interactwith cultural information.3.2 Operationalising the Cultural Intelligence HypothesisTo derive predictions about content-biases in human memory, we need to ask whatstable, fitness-relevant differences there are in the information that cultural animalsregularly encounter.In considering the abstract properties of interactions between individuals, gametheorists distinguish between strategic and non-strategic interactions. A strategicinteraction is defined as one where an individual?s outcomes depend on another?sactions. This distinction is important, strategic interactions have vastly differentdynamics and consequences for those involved.Cultural information can also be meaningfully partitioned in this way. Imag-ine a highly simplified cultural situation: a population where any individual canhold one of two beliefs, individuals can observe what others believe and what anindividual believes effects them in ways relevant their evolutionary fitness. If thesebeliefs concern non-strategic information (for instance, how often to brush one?steeth), only an individual?s own beliefs significantly affect their outcomes. Withstrategic information (for instance, beliefs about whether that individual is honest),other?s beliefs also have important fitness consequences.The focus of most cultural inheritance models has been on modelling the evo-lution of capacities for acquiring non-strategic information. A number of importantresults have been derived, including the circumstances under which learners will56copy the majority view (Henrich & Boyd, 1998) and whom they will prefer to learnfrom (Henrich & Gil-White, 2001). Our understanding of strategic information hasemerged from a different source - models of the role information plays in strategicsituations. Two important types of strategic information influence how an individ-ual?s peers treat them in fitness relevant ways: reputations (Nowak & Sigmund,2005) and norms (Henrich & Boyd, 2001; McElreath et al., 2003).We can make a clear prediction about minds adapted to cultural information.Since non-strategic information is usually only valuable if accurate 3, culturalminds should weight it by their estimation of its accuracy, internalising informa-tion less likely to be accurate more superficially, to reduce its potential harm. Con-versely, it is worth knowing other?s beliefs about strategic topics, like reputationsand norms, regardless their accuracy since they predict others? fitness-impactingbehaviours, help determine and exploit local strategic equilibria and evaluate thecosts and benefits of changing others? minds. Cultural animals should have a keeninterest and memory for strategic information, regardless their estimation of itsaccuracy.3.3 MethodWe tested participants? recall for strategic and non-strategic information, contin-gent on their agreement with any particular piece of information, predicting thatagreement would be a significant predictor of recall for non-strategic information,but not for strategic information.We recruited 40 undergraduates from around UBC campus and recompensedthem with a candy bar. They were shown a list of sentences and asked to ratetheir agreement with each on a 5 point scale marked ?Completely Agree?, ?Some-what Agree?, ?Neither Agree nor Disagree?, ?Completely Disagree?, ?SomewhatDisagree?. Following this participants were distracted with a simple letter-namingtask and then asked to recall as many of the items they had rated as they could.These items (listed in Table 3.1) were designed to be more or less strategic.How strategic any piece of information is can vary with circumstance: third parties?3I.e. for improving decisions or garnering prestige when retransmitted (Henrich & Gil-White,2001).57beliefs about the price of rice in China are inconsequential for many but massivein consequence for some. There is no clear-cut distinction between strategic andnon-strategic information, just shifting a continuum from more to less strategic.However just as there are no distinct colours besides a continuous spectrum offrequencies, yet we can still identify large-enough differences like that between redand green, it is possible to distinguish information as ?clearly more strategic? and?clearly less strategic?. We used this distinction to generate our study materials.58Table 3.1: Strategic Recall ItemsReputationalTarget ValenceSpeakerGood Psychology researchers are among the most intelligent people.Bad The people who wrote this study aren?t very attractive.ListenerGood People who participate in psychology studies tend to have above average emotional intelligence.Bad People who fill out our surveys are generally dislikable.In-groupGood The residents of British Columbia are generally well-respected.Bad Canadians, on average, are a pretty unimpressive group.Out-groupGood Easter-Islanders are by far the world?s best lovers.Bad Babylonians are very unpleasant people.NormativeDirection Incentivised?PrescriptiveYes People who give to charity often get rewarded in other ways.No People should be polite and kind to strangers.ProscriptiveYes People who don?t give up their seat on the bus usually get punished.No People shouldn?t over-exploit the public health-care system.Participants saw two types of strategic information: reputations and norms.Participants in psychology experiments intuitively interpret surveys as a dialogue(Schwarz, 1996). Capitalising on this, our reputational claims were evaluationsof participants, their dialogue partner (us), groups both we and the participant arelikely to belong to (Canadians, residents of our province), and third parties. Nor-mative items instructed participants to abide by local norms (i.e. giving up theirseat on a bus), either with or without stipulating the consequences.Since each participant did or did not remembered each item, we fitted thesedata to a logistic regression model, using robust standard errors (White, 1980) tocompensate for dependence within subjects.3.4 ResultsOur logistic regression model had five predictors: a dichotomy of whether informa-tion was strategic; a linear representation of participants? agreement with a state-ment, coded from -2 (Completely Disagree) to +2 (Completely Agree); the ran-domised order in which items were presented, broken down into both a linear andquadratic effect; and the interaction between whether information was strategic andparticipant?s agreement with it.There was a significant interaction between participants? agreement with aproposition and whether that proposition was strategic (??SE =?0.31?0.13, p?0.01), removing this interaction significantly diminished the log-likelihood of themodel (Pr[?2(1) = 4.8] ? 0.03). Participants? agreement significantly improvedtheir recall for non-strategic items (? ?SE = 0.15?0.08, p? 0.05), but worsenedtheir recall for strategic items (? ?SE =?0.18?0.11, p? 0.09), though this wasnot significantly different from no effect within conventional bounds of statisticalconfidence.Participants were more likely to recall strategic information, but not signifi-cantly so (? ?SE = 0.23?0.21, p? 0.27).Visual inspection and the statistical significance of a quadratic representationof the order in which items were presented (? ? SE = 0.004? 0.002, p ? 0.05)suggest participants experienced the well-established recency and primacy effects,finding it easier to remember items presented at either early or late. A linear order60Table 3.2: Non-Strategic Recall ItemsLess Fitness RelevantVaries in Valence DomainIntuitiveness Intuitive Social Ethnic clothes and markings reveal alot about people?s beliefs.Biological Cats prey on smaller species.Physical Electricity is a substance that flows through metal wires.Psychological You can tell what people want from what they do.Counter-Intuitive Social Most Germans are more like the French than other Germans.Biological There?s a species of mollusc which cannot die.Physical Quarks, a type of physical particle, can move through other substances.Psychological Chairs want to be sat on and believe they will be.Non-Intuitive Social People with short thumbs tend to resist change.Biological Ribbon worms are unlikely pass through glass.Physical Heat is just the movement of unthinkably many atoms.Psychological People?s behaviour is just a consequence of nerve activation.Likelihood of Agreement Least Diamonds are soft and easily bendable.. Beer is best brewed with air-borne yeasts.. Norway is bigger than Finland.. Forest fires release smoke which contains carbon monoxide.. Colarado is hot in the summer.. Canada is bigger in land mass than the USA.. Australia is south of Japan.Most The Earth is round, not flat.More Fitness RelevantValence Domain Actionable?Positive Resources Yes There?s free pizza at the UBC Student Union Building.No A store in Marrakech, Morocco is giving away furniture for free.Strategies Yes Quiet music can improve memory when studying for an exam.No Daily prayer improves the focus of cloistered monks.Negative Disease Yes One of the symptoms of anthrax is large facial sores with black bumps.No Rubella, a contagious disease, often doesn?t produce any symptoms.Danger Yes Some birds looks harmless but in nesting season cause serious injuries to people.No A large enough meteor could destroy all life on Earth.term remained just outside the conventional bounds of significance along-side thisquadratic effect (??SE = 0.025?0.013, p? 0.06), suggesting subjects had betterrecall for items presented later over those presented earlier.Participants? frequencies of agreement did not differ significantly between strate-gic and non-strategic information (?2(16) = 20, p? 0.22), see Table 3.3.Table 3.3: Frequencies of AgreementCompletelyAgreeSomewhatAgreeNo Opinion SomewhatDisagreeCompletelyDisagreeStrategic 0.12 0.23 0.38 0.13 0.14Non-strategic 0.19 0.23 0.34 0.11 0.13?2(16) = 20, p? 0.223.5 DiscussionWe observed the predicted disparity between strategic and non-strategic informa-tion in the effect of a participant?s agreement with a statement on their likelihoodof recalling it. Participants were significantly less likely to recall non-strategic in-formation if they disagreed with it. Though the effect of participants? agreementon their recall of strategic statements was not significantly different from zero, thenon-significant trend was for participants to be more likely to remember a strategicstatement the more they disagreed with it.The Cultural Intelligence Hypothesis claims that the accumulation of culturedrove human brain expansion and produced functional adaptations in the humanmind for processing cultural information. Our prediction hinges on these specificclaims: strategic/non-strategic is a fitness-relevant distinction in the informationthat cultural animals regularly encounter; non-strategic information is usually valu-able only if accurate while strategic information is valuable regardless; and a mindadapted for culture should exploit this difference.These results are suggestive, but far from conclusive. Survey-based results,particularly memory studies, can be very sensitive to subtle design differences62(Chapter 2; Schwarz, 1996). Our confidence in the the robustness of this effectwill only emerge by its replication with a variety of materials and participants.A deeper logical concern shadows this empirical challenge. If there are ben-efits to be gleaned from discriminating strategic from non-strategic information,individual or cultural learning could be responsible for exploiting them indepen-dently or in conjunction with genetic adaptation. Temporal precedence in predict-ing an empirical fact does lend any more empirical support to a theory than to anyhypothesis that can straightforwardly account for it post-hoc, without needing ad-hoc conjectures. Fortunately these accounts are not completely overlapping, theirunique properties can tease them apart. Evolutionary accounts require heritabledifferences while cultural accounts predict cross-culturally meaningful pattern ofvariance. These could be exploited by future investigations to begin separating outthe plausible explanations for this effect.Though follow-up testing is required to confirm that the Cultural IntelligenceHypothesis is the best explanation of the effect we observed, this study nonethelessmakes an important empirical contribution to verifying this distal theory. Whenconsidering a noisy, complicated phenomenon like human psychology, it is veryrare that general distal causes can be verified by their correspondence to a singlephenomenon. Evolved biases, though they cause real effects across individuals,groups and phenomena, are distorted in particular cases by proximate influences,such as ecology and history. For any given phenomenon it is usually possible togenerate a proximate explanation that fits the local data better that an accurate evo-lutionary account. Good evolutionary accounts can nonetheless contribute some-thing valuable to the theoretical corpus of psychology: parsimonious explanationsthat connect many phenomena that would otherwise be investigated independently.The value of such accounts should be judged against the breadth of phenomenathey accurately predict and the simplicity with which they do it.The Cultural Intelligence Hypothesis, building from the claim that humans areadapted to be cultural animals, accounts for an impressive range of phenomena,including: conformity (Henrich & Boyd, 1998), celebrity, prestige and the patternsof imitation of young children (Chapter 4; Henrich & Gil-White, 2001), ethnicity,racism and in-group bias (McElreath et al., 2003), and the rates at which innova-tions spread (Henrich, 2001). Our confidence in these explanations may ultimately63hinge as much on the theory?s success at straightforwardly predicting unrelated,novel phenomena as does on its fit to the local data of those effects. Our resultis an important contribution to the set of novel predictions made by the CulturalIntelligence.64Chapter 4Prestige-Biased Learning inChildren: Attention from othersas a Cue for CulturalTransmissionWritten and formatted to the specifications of a Psychological Science Research Report.Much of the emerging developmental research on young children?s patterns ofselective imitation supports earlier theoretical work on the coevolutionary foun-dations of cultural learning. Here, after reviewing the intersection between thesedisciplines, we present the first direct test of prestige-bias, a preference for learn-ing from those to whom others have attended, in children. Our results show that anadult model at whom two bystanders preferentially looked for 10 seconds, with-out giving any ostensive cues of endorsement, was over twice as likely to have hersubsequent behaviours imitated by young children, as a model who did not receivebystander attention. This preferential learning was strongest in the same domainin which the model?s ?prestige? was established, where the prestigious model wasabout 8 times more likely to be imitated.654.1 IntroductionSelection pressures have adapted the human species to a unique niche - we are acultural species. Coevolutionary theories reconstruct this transformation by mod-elling the evolutionary dynamics facing an emerging cultural species, in particularthe interaction of genetic and cultural inheritance systems (Boyd & Richerson,1985). By bringing together empirical evidence of human evolutionary history andevolutionary models focused on understanding our capacities for cultural learningas cognitive adaptations to uncertain environments, these theories have derived pre-dictions supported by a wide range of evidence from social psychology, economics,and field studies (Richerson & Boyd, 2005; Powell et al., 2009). These approachessuggest that learners should (1) rely more heavily on cultural learning in uncertainsituations (Boyd & Richerson, 1988a), (2) use the frequency of behaviour amongthose observed as a cue about whether to adopt it (Henrich & Boyd, 1998), and (3)be selective about who they attend to for the purpose of cultural learning (Henrich& Gil-White, 2001).This last prediction involves a suite of hypotheses that learners should at-tend to cues of skill (competence), success, prestige, age, sex, dialect and health(among others) in choosing from whom to learn. Evidence for these predictionshas emerged from work in social psychology (Henrich & Gil-White, 2001), be-havioural economics (Pingle & Day, 1996), and experimental anthropology (Effer-son et al., 2008), as well as numerous field studies (Rogers, 1995) and recent workwith non-human animals (Galef & Whiskin, 2007). One of the advantages of thisapproach is that hypotheses about cultural learning can be readily modelled to gen-erate predictions about higher-level sociological phenomena, such as the diffusionof innovations (Henrich, 2001), the emergence of ethnic groups (McElreath et al.,2003), or the properties of modern religions (Boyer, 2001; Henrich, 2009). Suchhigh-level accounts rely on assumptions about individual-level cultural learning bi-ases, which can be tested directly. Developmental evidence is particularly critical:observing a bias in young children can strengthen or refute claims of its generalityand evolved character. In this paper we re-examine evidence from developmentalpsychology in light of these coevolutionary hypotheses, then provide the first de-velopmental test of a learning bias predicted to be common in our highly cultural66species: prestige-bias.4.1.1 The Evolution of Prestige-BiasCoevolutionary theories model which learning strategies give learners the most ro-bust fitness benefits under different circumstances; those concerning from whomto learn are termed model-biases in cultural learning. ?Model? refers to the in-dividuals being learned from; ?cultural learning? refers to any social transfer ofinformation (e.g., techniques for using artifacts, food preferences, morals, etc.).Some individuals are just better in certain domains - e.g., they bring home morefood or maintain better relationships - and learning from them pays (Henrich &Gil-White, 2001). Skill-bias refers to learners selecting models by direct percep-tion of their skill or competence. Success-bias refers to identifying the best modelsby the cumulative fruits of their success: bigger houses, better jobs, more expensivecars, fatter (or thinner) bodies (Marlowe & Wetsman, 2001), fancier ornamentation(Malinowski, 1922), or longer yams (Kaberry, 1941).Complementing the substantial multi-disciplinary adult evidence (Henrich &Henrich, 2007), support for model-bias in children is now rapidly emerging. Youngchildren track the history of accuracy of potential models and preferentially learnfrom more accurate individuals (Birch et al., 2008; Corriveau & Harris, 2009;Koenig & Harris, 2005; Clement et al., 2004). Furthermore, although childrenshow a capacity to identify skill differences - trusting makers of artifacts for infor-mation about them (Jaswal, 2006), and generally adults over children - they takethe word of previously accurate children over previously inaccurate adults (Jaswal& Neely, 2006), suggesting they assign relative weights to cues of both age andcompetence. Children also seem well-attuned to evaluate indirect signals of skill,such as confidence; they prefer to learn from confident models (Birch et al., 2009;Jaswal & Malone, 2007). This suggests that in the absence of direct informationon relative skill or success, children may be vigilant for indirect cues of a potentialmodel?s competence.Sometimes a learner needs to make learning decisions in the absence of reliableinformation about skill or success, particularly since skill differences can be subtleand proxies for success can vary dramatically between communities. In such cases67the best models can often be identified by a property they all share: everyone elsewants to learn from them too; they have prestige - the greatest number of learnersobserving and imitating them.Coevolutionary theory predicts that children should display innate dispositionsfor using fast, reliable heuristics for imitating the bearers of higher-quality culturalinformation. A prestige-bias would facilitate more accurate and rapid learning byallowing learners to capitalize on others? beliefs about who is worthy of attention.Such a learning bias might be domain-general (i.e., imitate everything a prestigiousmodel does), task-specific (i.e., imitate only the type of information being demon-strated when the model received bystander attention), domain-specific (i.e., imitateonly in conceptually similar domains) or somewhere in between. Since the atten-tion of third parties doesn?t readily reveal what it is about a prestigious model thatis responsible for their success, Henrich and Gil-White (2001) predicted that learn-ers should initially imitate prestigious models across domains, but give priority todomains most closely related to those in which their prestige is most salient.4.1.2 PredictionsTo date, the best evidence from the developmental literature of prestige-bias comesfrom Fusaro and Harris (2008). Their four-year-old participants saw two modelslabelling the same object differently while bystanders either endorsed or deniedthe model?s claims non-verbally. Children preferentially imitated the model whohad received endorsement - even on subsequent imitative tests in which bystandersweren?t present. However, Fusaro and Harris did not set out to test coevolution-ary theories, thus their design does not provide a direct test of prestige-bias fortwo reasons. First, bystanders gave the model their assent (e.g., nodding, smiling)rather than just their attention, potentially endorsing the model?s message ratherthan the model. A critic could claim this is evidence of informational conformity-bias (Richerson & Boyd, 2005) rather than prestige-bias, or due to other factors likeconcerns about punishment (Henrich & Henrich, 2007). Second, both psycholo-gists and evolutionary theorists have suggested that language acquisition may bea special domain of learning (e.g. Pinker, 1994). From our perspective, learninglanguage may be special because it is about coordinating with one?s group and not68about adapting to the non-social environment. While it is really the case that somefungi are better (and safer) to eat than others (and people can learn this culturally),whether one calls those fungi mushrooms or kinoko depends only on what othersin one?s linguistic community call it. Because Fusaro and Harris? design only em-ployed label learning, a critic of prestige-bias might doubt that their effect wouldgeneralise to non-social domains, like food preferences or artifact-use techniques,and thus not constitute the more general learning bias postulated by coevolutionarytheory. Critics might also argue that Fusaro and Harris? participants did not infermodels? prestige from bystander assent but rather their language or group mem-bership - valuable information for young learners, which has been predicted usingevolutionary modelling (McElreath et al., 2003) and shown (Kinzler et al., 2007)to bias children?s behaviour.In this paper we test prestige-bias directly. We predict children will:1. Preferentially attend to a learn from those models to whom bystanders paidattention. Importantly, unlike Fusaro and Harris? design, our bystanders onlyattended to the models, rather than explicitly endorsing their statements.2. Imitation the behaviours and preferences of prestigious models, even poten-tially costly ones like food and drink preferences, not just word-learning.3. Generalise prestige beyond the domain in which a model was observed (i.e.,the activity the model was doing when they received bystander attention),but be most strongly influenced to imitate a model in that domain and relatedones.4.2 Method4.2.1 ParticipantsWe measured the selective imitation of 23 children (mean age = 50.4 months,SD=5.8 months; 12 girls) recruited from a participant database at a public uni-versity. Data from one boy who did not complete the experiment was excluded.The majority of participants were Caucasian or Asian; all were from householdswhere English was the main language spoken.694.2.2 ProceedureParticipants watched a ?cuing scene? and four 10 second clips of two modelsdemonstrating different behaviours/preferences/labels towards objects. In the ?cu-ing scene? two bystanders stood between the models, attending to only one ofthem - the ?prestigious model?. In all other scenes models individually demon-strated their preference towards an object; then participants? preferences towardthose stimuli were recorded. The order in which models appeared and the identityof the prestigious model were counter-balanced across participants. Participantsobserved the scenes in the following order:Food Choice: Models made a disgusted face at either large round or smallsquare crackers and happily sampled the other. Participants saw this scene beforethe cuing scene but only later, during ?free play?, were offered a choice betweenthe two crackers.Figure 4.1: Prestige Cuing and Imitation Testing StillsPrestige Cuing Scene: Each model played with a toy in different ways, onopposite sides of a room. Two observers entered and for 10 seconds stood betweenand slightly behind the models, angled toward and watching the prestigious model.Artifact Use: Models interacted with a novel toy by delightedly using eithercoloured balls or blocks. Participants were presented with the same apparatus,offered a choice of the balls or blocks and asked ?Can you show me how to playwith this??Beverage Choice: Models made a disgusted face at either a cup of dyed blueor yellow water and drank from the other. Participants were offered the same andasked ?Would you like a drink?? or if unwilling to drink: ?Which do you think isbetter to drink??Novel Label Preference: Each model labelled a different object with the samename - a ?stroop?. Participants were presented with both objects and asked ?Canyou give me the stroop??70Playmate Preference and Popularity Selection: After a few minutes of ?freeplay? participants were shown photos of the models and asked our explicit mea-sures: ?Who would you rather play with? and ?Who do you think is more popular,who has more friends??4.3 ResultsSince participants always chose between two options, each associated with a model,we analysed our data by logistic regression, treating each choice as an endorsementof one actor or the other as the preferred cultural model. Since each participantmade multiple, related binary choices, we compensated for this non-independenceof observations by calculating clustered robust standard errors (clustering on indi-viduals), using standard techniques (White, 1980).When comparing logistic mod-els, we considered the ?2 distributed ratio of their log-likelihoods. To evaluate themodels? goodness of fit we conducted log likelihood ratio tests1; the significanceof coefficients was judged by their Z-distributed ratio to their standard errors. Ef-fect sizes are reported as logistic regression coefficients and odds ratios for easeof interpretation. Odds ratios represent how much more likely a model was to beimitated if she was the prestigious model (i.e. the one attended to by bystanders).4.3.1 Is Children?s Imitation Biased by a Model?s Prestige?Yes. Regressing who participants imitated onto (1) which model was prestigious,(2) the order in which models appeared, (3) participant?s sex and 4) age, producedonly one significant predictor: prestige (p= 0.01). Removal of the non-significantpredictors did not significantly diminish the model?s predictive power (P[ ?2(3) ?2.1] = 0.55). In this parsimonious single-predictor regression model (pLLRT = 0.03),a cultural model was 2.37 times (CI.95 = [1.22, 4.58]) more likely to be imitated ifshe was the prestigious model (? = .86; SE = .33; p = 0.01). Participant?s proclivityto prestige-bias was not confounded with their age (p = 0.32), sex (p = 0.7), theorder in which models appeared (p = 0.11), nor which person was the prestigious1That is, we compared the ?2 distributed ratio of each model?s log-likelihood to that of a logis-tic model with only a constant predictor. The result of these tests will be identified herein by thesubscript pLLRT . Values indicate the probability that a null model would produce as good a fit byrandom sampling alone.71model (p = 0.48). See Supplemental Material for further details.4.3.2 Does Prestige-Bias Operate Across Learning Domains?Since prestige was a significant predictor of imitation across all measures, we nextexamined theoretically-interesting subsets. As predicted, model-prestige predictedimitation most strongly in the domain in which her prestige was established (i.e.,artifact-use), where the model was 8.25 (CI.95 = [1.15, 59.0]) times more likely tohave her artifact-use technique imitated if she was the high-prestige model (? =2.11; SE = 1.0; p =0.04). It warrants emphasis that the artifact used in the imitationtest was not the same artifact used during the prestige cue, thus the effect is not task-specific. Since the artifact test was measured immediately after the prestige cue,one may wonder whether the strength of this effect is attributable to the salienceof the cue in memory. However, the pattern of effect sizes on subsequent measuresis inconsistent with decaying salience (i.e., the next strongest effect was in thefood test which was measured last), leading us to suspect that the domain in whichprestige is established plays an important role.Prestige also biased imitation in the fitness-relevant domain of food and drinkpreferences where the model was 4.09 (CI.95 = [1.02, 16.38]) times more likelyto have her preference imitated if she was the prestigious model. The prestigeeffect, however, did not extend to language learning, which was the only imitationmeasure for which prestige was not a significant predictor (? = -0.57; SE = 1.07; p= 0.59). Model prestige also did not seem to influence children?s answers to explicitquestions about which model they or others would prefer to play with (pLLRT = 0.8).4.4 DiscussionOur study tested whether children use the attention of others (prestige cues) todecide from whom to learn. Our findings indicate that they do: on novel taskswith bystanders no longer present, our participants were more than twice as likelyto imitate a prestigious model. To be precise, children more often imitated anadult who?d received anonymous bystanders? mere attention for just ten secondsover an adult who hadn?t. During this prestige cue adults demonstrate artifact-usetechniques, later when they demonstrated a technique for using a different artifact72without bystanders present, the prestigious adult was eight times more likely tobe imitated. Consistent with Henrich and Gil-White?s (2001) predictions, we alsowitnessed biased learning in other behavioural domains, including the potentiallycostly domain of diet decisions. This strong effect from a minimal manipulationsuggests that prestige-biased transmission is likely a potent force in cultural evolu-tion.Our research extends important work by Fusaro and Harris, who used a similardesign in the domain of language learning, in important ways. First, our approachderives from an explicit evolutionary theory about cultural learning and links manydifferent aspects of learning under one theoretical umbrella. Second, we showthat prestige-bias also influences potentially costly learning about the non-socialenvironment, like artifact-use techniques or food-preferences. Moreover, prestigeand may generalise least from these domains (i.e., artifact, food) to domains likelanguage learning which primarily involve social coordination. Finally, we showthat neutral preferential attention, without nodding or other explicit endorsement,is sufficient to bias cultural learning.An evolved prestige-bias, properly understood, can make important contribu-tions to explaining modern social phenomena. For example, why does the suicideof famous individuals spark increases in the suicide rates in populations of individ-uals of similar age and ethnicity, with people even copying the method (Fu & Yip,2007; Stack, 1987)? Why do celebrity endorsements of products entirely unrelatedto the cause of their fame result in increased sales (Silvera & Austad, 2004)? Fu-ture investigations which experimentally manipulate the domain in which prestigecues appear, the delay before testing, the age and ethnicity of models and learners,and put different learning biases into competition with one another will help clarifythe nature of prestige-bias and its implications.In sum, this research constitutes the first evidence of prestige-bias in children,an evolved learning bias predicted by coevolutionary theory to be endemic in anyhighly-cultural species, thus providing the much-needed foundation for further in-vestigations to clarify its mechanisms, contingencies, developmental trajectory,and role in broader social phenomena.734.5 Supplemental MaterialThis section outlines further the details of the findings reported in the main text.4.5.1 Is Prestige-Bias Strongest in the Domain Where Prestige WasObserved?We suspect so. Imitation in the domain of artifacts, operationalised as toy-usetechniques, registered by far the strongest effects. Regressing the artifact measureon all predictors yielded one significant coefficient: prestige (p=.04). Removingnon-significant predictors did not significantly diminish the predictive power ofthis model (P[ ?2(3) > 5.01] = 0.17). In the resulting parsimonious model, anadult was 8.25 times more likely (CI.95 = [1.15, 59.0]) to have her artifact-usetechnique imitated if she was the high prestige model (? = 2.11; SE = 1.0; p =0.04). However, because the artifact-use test appeared immediately after the cuingscene - the strength of this effect may in part be a consequence of the greatersalience of the prestige cue in memory. Nonetheless, the pattern of effect sizes onsubsequent tests is more consistent with domain-specificity than decaying salience- the food choice measure registered stronger effects than the novel label preferencemeasure despite being measured after it. Given this, we suspect the greater effectof prestige-bias on the artifact-use measure was driven at least in part by domain-specificity.4.5.2 Does Prestige-Bias Generalise Across Behavioural Domains?Yes, but as anticipated, it generalizes to decreasing degrees for domains that areconceptually more dissimilar from the domain of the prestige cue. Our principalinterest was in the directly fitness-relevant domain of food and drink preferences.Regressing just the food and drink measures on prestige, order, sex, and age pro-duced one significant and one marginally significant predictor - prestige (p = 0.03)and sex (p = 0.07). Removal of non-significant predictors did not significantly di-minish the model (P[ ?2(2) > 0.48] = .79). In the resulting parsimonious model(pLLRT = 0.04) an adult was 4.09 (CI.95 = [1.02, 16.38]) times more likely to havetheir food and drink preferences imitated if they were the prestigious model(? =1.4; SE = 0.7; p = 0.05). The actor in the pink shirt was 4.55 (CI.95 = [1.13, 18.31])74Table 4.1 Parsimonious Models - Logistic Regression CoefficientsPredictors/Models All Behaviours Artifact Food/Drink LabellingPrestige .86 (.33)** 2.11 (1.0)* 1.4 (.7)*Sex 1.5 (.7)* -1.84 (.98)??pLLRT .03 .02 .04 .04n 100 23 44 23??: p < .06 ; ? : p < .05 ; ?? : p < .01Significant logistic regression coefficients and their (standard errors). All statistical models regresswhich actor participants imitated onto the listed predictors. Parsimonious models were developedby removing only those non-significant predictors from the full models whose absence did notsignificantly diminish the model?s log likelihood. Prestige encodes which actor was prestigious, andSex the sex of the participant. pLLRT is the result of a log-likelihood ratio test of the model.n is thenumber of observations on which the statistical inference was based, whose non-independence wascompensated for by Huber-White clustered robust standard errors. Odds ratios and their confidenceintervals can be computed by exponentiating e to these efficients ? interesting multiples of theirnormally distributed errors.times more likely to be imitated if the participant was a girl (? = 1.5; SE = 0.71; p= 0.03).4.5.3 Is Language Learning a Special Domain?It seems so. Of all our measures, the only one which was non-significant forprestige-bias was the labelling task (? = -0.57; SE = 1.07; p =0.59). We describein the main text why we suspected that a critic might be concerned that prestigewould generalise least between skill-based domains (like artifact use, where weestablished prestige in this study) and convention-based domains (like language).Though Fusaro and Harris (2008) observed an effect (that arguably could be con-ceived of as a prestige effect) purely within the domain of language learning, ourresults seem to suggest a prestige effect, cued in a skill-based domain, does notgeneralise to language learning. There may exist an effect of skill-cued prestigeon language learning that requires a more powerful test to detect, but we suspectthat it would not be as large as the effects in the artifact and food domains. Inter-estingly, sex was a marginally significant predictor in this analysis (? = -1.84; SE=0.98; p =0.06). Removal of non-significant predictors did not significantly dimin-ish the predictive power of this model (P[ ?2(3) > 3.15] = 0.37), but did produce a75Table 4.2 Parsimonious Models - Odds Ratios [with 95% C.I.s]Predictors/Models All Behaviours Artifact Food/Drink LabellingPrestige 2.37 [1.22, 4.58] 8.25 [1.15, 59.0] 4.09 [1.02, 16.38]Sex 4.55 [1.13, 18.31] .16 [.02, 1.08]pLLRT .03 .02 .04 .04n 100 23 44 23Significant odds ratios and their [.95 confidence intervals]. All statistical modelsregress which actor participants imitated onto those predictors with non-emptytable cells. Parsimonious models were developed by removing only thosenon-significant predictors from the full models whose absence did notsignificantly diminish the model?s log likelihood. Prestige encoded which actorwas cued as prestigious, and Sex the sex of the participant. pLLRT is the result of alog-likelihood ratio test of the model. n is the number of observations on whichthe statistical inference was based, whose non-independence was accounted for byHuber-White clustered, robust standard errors.significant model (pLLRT = 0.05) with sex the only predictor.4.5.4 Are There Gender Effects? What About Other Effects?In the analyses described in the main text, we regressed which model was imitatedon our predictors, (including which model was prestigious) - that is, model prestige(and the other variables) predicted who was imitated. To investigate confounds weinstead regressed whether the imitated model was prestigious on our predictors -that is, the other variables (including which actor happened to have been cued asprestigious) predicted whether participants imitated whichever was the prestigiousmodel - their proclivity to prestige-bias. Participants? proclivity to prestige-bias onthe behavioural measures was not predicted (pLLRT =0.44) by participants? sex (p= 0.7), their age (p = 0.32), by which actor was the prestigious model (p = 0.48),nor whether they appeared first in videos (p = 0.11).Participants? sex did however predict their preference for one actor over an-other (in the domains of diet and language), though in different directions in eachcase. We think this may have stemmed from girls? significant preference for theactor in the pink shirt, evident from a binomial test of the proportion of times they76imitated each model (pbinom < 0.001), since boys did not show such a preference(pbinom = 0.41). We suspect gender was tapping the self-similarity imitation heuris-tic (Henrich & Gil-White, 2001). That is, girls may have had a competing influenceon their imitative decisions: an affinity between themselves and the model basedon their shirt colour. The hypothesis that children were weighing competing cuesis consistent with our observed pattern of results (see Table 4.1): children?s sexpredicted patterns of imitation in domains where prestige-bias was weaker.Table 4.3 Full Models - Logistic Regression CoefficientsPredictors/Models All Behaviours Artifact Food/Drink LabellingPrestige .90 (.37)** 2.60 (1.30)* 1.42 (.67)* -.57 (1.07)Sex .0 (.34) -1.33 (1.43) 1.30 (.73)? -2.90 (1.48)*Order -.56 (.33)? -2.49 (1.50)? -.42 (.57) -1.94 (1.36)Age -.02 (.03) -.08 ( .14) .03 (.06) .01 ( .10)pLLRT .16 .04 .14 .13n 100 23 44 23?: p < .1 ; ? : p < .05 ; ?? : p < .01Logistic regression coefficients and their (standard errors). All statistical models regress whichactor participants imitated onto the listed predictors. Prestige encodes which actor was prestigious,Sex the sex of the participant, Order encodes which actor appeared first and Age the participant?sage in months. pLLRT is the result of a log-likelihood ratio test of the model.n is the number ofobservations on which the statistical inference was based, whose non-independence wascompensated for by Huber-White clustered robust standard errors. Odds ratios and their confidenceintervals can be computed by exponentiating e to these efficients ? interesting multiples of theirnormally distributed errors.4.5.5 Do Children Explicitly Report Their Prestige-Bias?Regressing all predictors on our explicit measures produced no significant coeffi-cients and a highly non-significant model (pLLRT = 0.8). This is consistent withwork on adults in which prestige-bias often emerges unconsciously (for a review,see Henrich & Gil-White, 2001). If there is an effect of prestige-bias on children?sexplicit awareness of which model they like or which they think others like, it wasnot strong enough for our tests to detect.77Chapter 5Conclusion: Evaluating MyContributionsHave I assisted in the project I set out to? Have I derived predictions about distinctphenomena and tested them with careful attention to whether plausible proximateexplanations are inconsistent or indistinct?I derived one new prediction, I tested an existing novel prediction and I at-tempted to evaluate existing evidence against inconsistent local explanations.My extended test of Mesoudi, Whiten and Dunbar?s inference did not find clearinconsistency with local mechanisms, but did find previously unknown bound-ary conditions for the phenomenon they?d discovered. This newly defined phe-nomenon seemed consistent with a well-established finding in memory research:expertise effects on recall. I re-evaluated, assuming this proximal mechanism, thedistinctness of the Social Intelligence Hypothesis? prediction. I found it predictedthis newly qualified effect less distinctly. Additional hypotheses were required tomake the Social Intelligence Hypotheses entail this new mechanism and the Eco-logical Intelligence Hypothesis could plausibly claim similar additions. Further-more, local ecology alone could also plausibly interact with expertise effects toproduce the observed phenomenon.I did not evaluate the distinctness of this effect against other plausible evolu-tionary alternatives, notably the Cultural Intelligence Hypothesis. This is a seriousgap; the evaluation and testing of plausible alternatives is integral to good distal78empiricism. To fill it, in the future I hope to derive a unique prediction from theCultural Intelligence Hypothesis concerning a memory bias for social informationthat will let me test this other candidate against the same domain.A more proximate concern also plagues my findings - like Mesoudi, Whitenand Dunbar, I only attained this result with one set of materials and can?t be con-fident that my theorising won?t be inconsistent with direct linguistic determinantsof the effect.As for the new effects I observed, they were both derived straightforwardlyand a priori from an evolutionary theory consistent with the effects I observed. Thejob of evaluating these effects against plausible, inconsistent local mechanismsand deriving ever more precise predictions to separate this theory from other distalaccounts has not even started. These new findings stand where Mesoudi, Whitenand Dunbar?s did: a beginning and direction for further research.By my own criteria then, I have taken steps in the right direction, but theyare not very large. Still, I hope each step will contribute to an answer that futurecurious children will find obvious, when they wonder: what am I?79BibliographyAiello, L. C., & Wheeler, P. (1995). The expensive-tissue hypothesis: The brainand the digestive system in human and primate evolution. CurrentAnthropology, 36(2), 199?221. ? pages 4, 5Ash, J., & Gallup, G. (2007). Paleoclimatic variation and brain expansion duringhuman evolution. Human Nature, 18(2), 109 ? 124. ? pages 26Birch, S. A. J., Akmal, N., & Frampton, K. L. (2009). Two-year-olds are vigilantof others? nonverbal cues to credibility. ? pages 67Birch, S. A. J., Vauthier, S. A., & Bloom, P. (2008). Three-and four-year-oldsspontaneously use others? past performance to guide their learning. Cognition,107(3), 1018?1034. ? pages 67Boyd, R., & Richerson, P. (1985). Culture and the Evolutionary Process.Chicago: University Of Chicago Press. ? pages 66Boyd, R., & Richerson, P. (1988a). An evolutionary model of social learning: Theeffects of spatial and temporal variation. In T. Zentall, & B. G. Galef (Eds.)Social Learning: A Psychological and Biological Approaches , LawrenceErlbaum Assoc, (pp. 29?48). NJ: Hillsdale. ? pages 66Boyd, R., & Richerson, P. J. (1988b). Culture and the Evolutionary Process.University Of Chicago Press. ? pages 8, 55Boyd, R., & Richerson, P. J. (1988c). The evolution of reciprocity in sizablegroups. Journal of Theoretical Biology, 132(3), 337 ? 356. ? pages 7, 8Boyd, R., & Richerson, P. J. (1995). Why does culture increase humanadaptability? Ethology and Sociobiology, 16(2), 125 ? 143. ? pages 8Boyd, R., & Richerson, P. J. (1996). Why culture is common, but culturalevolution is rare. Proceedings of the British Academy, 88(300), 77?93. ?pages 880Boyer, P. (2001). Religion explained: The human instincts that fashion gods,spirits and ancestors. London: William Heinnemann. ? pages 66Cavalli-Sforza, M., L.L an Feldman (1981). Cultural Transmission andEvolution: A Quantative Approach. Princeton University Press. ? pages 8, 55Chase, W. G., & Ericsson, K. A. (1982). Skill and working memory. In G. H.Bower (Ed.) The psychology of learning and motivation, vol. 16, (pp. 1?58).New York: Academic Press. ? pages 11, 51Clement, F., Koenig, M., & Harris, P. (2004). The ontogenesis of trust. Mind &Language, 19(4), 360?379. ? pages 67Clutton-Brock, T., & Harvey, P. (1980). Primates, brains and ecology. Journal ofZoology, 190, 309?323. ? pages 6, 26Corriveau, K., & Harris, P. (2009). Preschoolers continue to trust a more accurateinformant 1 week after exposure to accuracy information. DevelopmentalScience, 12(1), 188?193. ? pages 67Cumming, G. (2009). Inference by eye: Reading the overlap of independentconfidence intervals. Statistics in Medicine, 28, 205?220. ? pages 32Cunnane, C., & Crawford, M. (2003). Survival of the fattest: fat babies were thekey to evolution of the large human brain. Comparative Biochemistry andPhysiology - Part A: Molecular & Integrative Physiology, 136(1), 17 ? 26. ?pages 5Deaton, A. (2007). Height, health, and development. Proceedings of the NationalAcademy of Sciences, 104(33), 13232?13237. ? pages 17Dennett, D. C. (1991). Real patterns. The Journal of Philosophy, 88(1), 27?51.? pages 21, 55Dunbar, R. (1998). The social brain hypothesis. Evolutionary Anthropology, 6,178?190. ? pages 7, 26, 55Durham, W. (1992). Coevolution: Genes, culture, and human diversity. StanfordUniversity Press. ? pages 8, 55Efferson, C., Lalive, R., Richerson, P. J., McElreath, R., & Lubell, M. (2008).Conformists and mavericks: the empirics of frequency-dependent culturaltransmission. Evolution and Human Behavior, 29(1), 56?64. ? pages 8, 6681Efferson, C., Richerson, P. J., McElreath, R., Lubell, M., Edsten, E., Waring,T. M., Paciotti, B., & Baum, W. (2007). Learning, productivity, noise: Anexperimental study of cultural transmission on the Bolivian Altiplano.Evolution and Human Behavior, 28, 11?17. ? pages 8Ericsson, K. A., & Kintsch, W. (1995). Long-term working memory.Psychological Review, 102(2), 211 ? 245. ? pages 45Floud, R., Wachter, K., & Gregory, A. (1990). Height, health and history.Cambridge University Press. ? pages 18Foley, R. A., Lee, P. C., Widdowson, E. M., Knight, C. D., & Jonxis, J. H. P.(1991). Ecology and energetics of encephalization in hominid evolution [anddiscussion]. Philosophical Transactions: Biological Sciences, 334(1270),223?232. ? pages 4, 5Fu, K., & Yip, P. S. F. (2007). Long-term impact of celebrity suicide on suicidalideation: Results from a population-based study. British Medical Journal,61(6), 540?546. ? pages 73Fusaro, M., & Harris, P. L. (2008). Children assess informant reliability usingbystanders? non-verbal cues. Developmental Science, 11(5), 771?777. ? pages68, 75Galef, B. G., & Whiskin, E. E. (2007). Use of social information by sodium- andprotein-deficient rats: test of a prediction (boyd & richerson 1988). AnimalBehaviour, 75(2), 627?630. ? pages 8, 66Gibson, K. R. (1986). Integrated interactive computing systems. In P. C. Else, J.G. & Lee (Ed.) Primate ontogeny, cognition and social behaviour, (pp.93?105). Cambridge University Press. ? pages 6, 26, 52Gillies, D. (1998). The duhem thesis and the quine thesis. In M. Curd, & J. Cover(Eds.) Philosophy of Science: The Central Issues, (pp. 302?319). New York:Norton. ? pages 15Gould, S. J., & Lewontin, R. C. (1979). The spandrels of san marco and thepanglossian paradigm: A critique of the adaptationist programme. Proceedingsof the Royal Society of London. Series B, Biological Sciences, 205(1161),581?598.URL ? pages 54Harvey, P. H., & Krebs, J. R. (1990). Comparing brains. Science, 249(4965),140?146. ? pages 6, 2682Hauert, C., De Monte, S., Hofbauer, J., & Sigmund, K. (2002). Volunteering asRed Queen Mechanism for Cooperation in Public Goods Games. Science,296(5570), 1129?1132. ? pages 7Heine, S. (2008). Handbook of Cultural Psychology. W. W. Norton & Co. ?pages 29Heine, S. J., Proulx, T., & Vohs, K. D. (2006). The Meaning Maintenance Model:On the Coherence of Social Motivations. Personality and Social PsychologyReview, 10(2), 88?110. ? pages 20Henrich, J. (2001). Cultural transmission and the diffusion of innovations:Adoption dynamics indicate that biased cultural transmission is thepredominate force in behavioral change and much of sociocultural evolution.American Anthropologist, 103, 992?1013. ? pages 63, 66Henrich, J. (2009). The evolution of costly displays, cooperation, andreligion:credibility enhancing displays and their implications for culturalevolution. Evolution and Human Behaviour, 30, 244?260. ? pages 66Henrich, J., & Boyd, R. (1998). The evolution of conformist transmission and theemergence of between-group differences. Evolution and Human Behavior,19(4), 215 ? 241. ? pages 8, 56, 57, 63, 66Henrich, J., & Boyd, R. (2001). Why people punish defectors. weak conformisttransmission can stabilize costly enforcement of norms in cooperativedilemmas. J Theor Biol, 208(1), 79?89. ? pages 57Henrich, J., & Gil-White, F. (2001). The evolution of prestige: Freely conferreddeference as a mechanism for enhancing the benefits of cultural transmission.Evolution and Human Behavior, 22(3), 165?196. ? pages 12, 56, 57, 63, 66,67, 68Henrich, J., Heine, S., & Norenzayan, A. (forthcoming). The weirdest people inthe world. ? pages 28Henrich, N. S., & Henrich, J. (2007). Why Humans Cooperate: A Cultural andEvolutionary Explanation. Oxford: Oxford University Press. ? pages 67, 68Herrmann, E., Call, J., Hernandez-Lloreda, M. V., Hare, B., & Tomasello, M.(2007). Humans Have Evolved Specialized Skills of Social Cognition: TheCultural Intelligence Hypothesis. Science, 317(5843), 1360?1366. ? pages 8,9, 2683Hirstein, W. (2004). Brain Fiction: Self-Deception and the Riddle ofConfabulation. MIT Press. ? pages 55Humphrey, N. (1976). The social function of intellect. In P. Bateson, & H. R.A.(Eds.) Growing Points in Ethology, (pp. 303?317). Cambridge: CambridgeUniversity Press. ? pages 7, 26, 55Jaswal, V. K. (2006). Preschoolers favor the creator?s label when reasoning aboutan artifact?s function. Cognition, 99(3), 83?92. ? pages 67Jaswal, V. K., & Malone, L. S. (2007). Turning believers into skeptics:3-year-olds? sensitivity to cues to speaker credibility. Journal of Cognition andDevelopment, 8(3), 263?283. ? pages 67Jaswal, V. K., & Neely, L. A. (2006). Adults don?t always know best:Preschoolers use past reliability over age when learning new words.Psychological Science, 17(9), 757?758. ? pages 67Kaberry, P. M. (1941). The abelam tribe, sepik district, new guinea: Apreliminary report. Oceania, 11(3), 233?258. ? pages 67Kaplan, H., Gangestad, S., Gurven, M., Lancaster, J., Mueller, T., & Robson, A.(2007). The evolution of diet, brain and life history among primates andhumans. In W. Roebroek (Ed.) Guts and Brains: an Integrative Approach to theHominin Record, (pp. 47?8). Leiden: Leiden University Press. ? pages 7Kawamura, S., Suzuki, S., & Morikawa, K. (2007). Short report: The effect ofexpertise in hiking on recognition memory for mountain scenes. Memory,15(7), 768?775. ? pages 11, 51Kay, A. C., Gaucher, D., Napier, J. L., Laurin, K., & Callan, M. J. (2008). Godand the government: Testing a compensatory control mechanism for thesupport of external systems. Journal of Personality and Social Psychology,95(1), 18 ? 35. ? pages 20Kintsch, W. (1974). The representation of meaning in memory. LawrenceErlbaum Assoc Inc. ? pages 28, 30, 32, 43Kintsch, W., & Ericsson, K. A. (1999). The role of long-term working memory intext comprehension. Psychologia, 42, 186?198. ? pages 45Kinzler, K. D., Dupoux, E., & Spelke, E. S. (2007). The native language of socialcognition. Proceedings of the National Academy of Sciences of the UnitedStates of America, 104(30), 12577?12580. ? pages 6984Kitayama, S., & Cohen, D. (2007). Handbook of Cultural Psychology. GuilfordPress. ? pages 29Kitayama, S., Duffy, S., Kawamura, T., & Larsen, J. (2003). Perceiving an objectand its context in different cultures: a cultural look at new look. PsychologicalScience: A Journal Of The American Psychological Society / APS, 14(3), 201 ?206. ? pages 10Kline, M., & Boyd, R. (forthcoming). How treadmills and drift create culturalloss in oceania. Presented at Human Behaviour and Evolution Society 21stAnnual Conference. ? pages 8Koenig, M. A., & Harris, P. L. (2005). Preschoolers mistrust ignorant andinaccurate speakers. Child Development, 76(6), 1261?1277. ? pages 67Kuhn, T. S. (1970). The Structure of Scientific Revolutions. University of ChicagoPress. ? pages 15Lloyd, E. (2005). The Case of the Female Orgasm: Bias in the Science ofEvolution. Harvard University Press. ? pages 55Malinowski, B. (1922). Argonauts of the western pacific. London: Routledge. ?pages 67Marlowe, F., & Wetsman, A. (2001). Preferred waist-to-hip ratio and ecology.Personality and Individual Differences, 30(3), 481?489. ? pages 67Mcbrearty, S., & Brooks, A. S. (2000). The revolution that wasn?t: a newinterpretation of the origin of modern human behavior. Journal of HumanEvolution, 39(5), 453 ? 563. ? pages 26McElreath, R., Boyd, R., & Richerson, P. J. (2003). Shared norms and theevolution of ethnic markers. Current Anthropology, 44(1), 122 ? 129. ? pages8, 56, 57, 63, 66, 69McNamara, T. (2005). Semantic Priming: Perspectives from Memory and WordRecognition. Psychology Press. ? pages 44Mesoudi, A., Whiten, A., & Dunbar, R. (2006). A bias for social information inhuman cultural transmission. British Journal of Psychology, 97, 405?423(19).? pages 3, 25, 26, 30, 55Milton, K. (1981). Distribution patterns of tropical plant foods as an evolutionarystimulus to primate mental development. American Anthropologist, 83(3),534?548. ? pages 6, 2685Navarrete, C. D., Kurzban, R., Fessler, D. M. T., & Kirkpatrick, L. A. (2004).Anxiety and Intergroup Bias: Terror Management or Coalitional Psychology?Group Processes Intergroup Relations, 7(4), 370?397. ? pages 20Nisbett, R. (2003). The Geography of Thought: How Asians and WesternersThink Differently... And Why. Free Press. ? pages 10, 29Nowak, M. A., & Sigmund, K. (2005). Evolution of indirect reciprocity. Nature,437(7063), 1291 ? 1298. ? pages 7, 57Pingle, M., & Day, R. H. (1996). Modes of economizing behavior: Experimentalevidence. Journal of Economic Behavior & Organization, 29, 191?209. ?pages 66Pinker, S. (1994). The Language Instinct: The New Science of Language andMind. London: Penguin Press. ? pages 68Potts, R. (1998). Variability selection in hominid evolution. EvolutionaryAnthropology: Issues, News, and Reviews, 7, 81?96. ? pages 6, 52Powell, A., Shennan, S., & Thomas, M. (2009). Late pleistocene demography andthe appearance of modern human behavior. Science, 324(5932), 1298?1301. ?pages 8, 66Quine, W. V. (1951). Main trends in recent philosophy: Two dogmas ofempiricism. The Philosophical Review, 60(1), 20?43. ? pages 15Relethford, J. H. (2008). Genetic evidence and the modern human origins debate.Heredity, 100(6), 555?563. ? pages 26Richerson, P. J., & Boyd, R. (2005). Not by Genes Alone: How CultureTransformed Human Evolution. Chicago: University of Chicago Press. ?pages 66, 68Rogers, E. M. (1995). Diffusion of innovations. New York: Free Press. ? pages66Rosenblatt, A., Greenberg, J., Solomon, S., Pyszczynski, T., & Lyon, D. (1989).Evidence for terror management theory: I. the effects of mortality salience onreactions to those who violate or uphold cultural values. Journal of Personalityand Social Psychology, 57(4), 681 ? 690. ? pages 20Roth, G., & Dicke, U. (2005). Evolution of the brain and intelligence. Trends inCognitive Sciences, 9(5), 250 ? 257. ? pages 386Ruff, C., Trinkaus, E., & Holliday, T. (1997). Body mass and encephalization inpleistocene homo. Nature, 387, 173?176. ? pages 4, 26Schoenemann, P. T. (2006). Evolution of the size and functional areas of thehuman brain. Annual Review of Anthropology, 35(1), 379?406. ? pages 4Schwartz, J. H., Tattersall, I., Holloway, R. L., Broadfield, D. C., & Yuan, M. S.(2005). The Human Fossil Record. Wiley-Liss. ? pages 26Schwarz, N. (1996). Cognition and Communication: Judgmental Biases,Research Methods, and the Logic of Conversation. Lawrence Erlbaum. ?pages 60, 63Silvera, D. H., & Austad, B. (2004). Factors predicting the effectiveness ofcelebrity endorsement advertisements. European Journal of Marketing, 38(11),1509?1526. ? pages 73Simon, H. A., & Gobet, F. (2000). Expertise effects in memory recall: A reply tovicente and wang. Psychological Review, 107, 593?60. ? pages 11, 51Stack, S. (1987). Celebrities and suicide: A taxonomy and analysis, 1948-1983.American Sociological Review, 52(3), 401?412. ? pages 73Statistics Canada (2006). Immigration in canada: A portrait of the foreign-bornpopulation, 2006 census: Portraits of major metropolitan centres - vancouver:Canada?s immigrant gateway in the west. Tech. rep. ?Available at pages 29Stout, D. (2002). Skill and cognition in stone tool production: An ethnographiccase study from irian jaya. Current Anthropology, 43(5), 693?722. ? pages 12Tomasello, M., & Rakoczy, H. (2003). What makes human cognition unique?from individual to shared to collective intentionality. Mind & Language, 18(2),121 ? 147. ? pages 3Tooby, J., & Cosmides, L. (2005). Conceptual foundations of evolutionarypsychology. In D. M. Buss (Ed.) The Handbook of Evolutionary Psychology,(pp. 5?67). New Jersey: Wiley. ? pages 18Triandis, H. C. (2001). Individualism-collectivism and personality. Journal ofPersonality, 69(6), 907 ? 924. ? pages 10Vicente, K. J., & Wang, J. H. (1998). An ecological theory of expertise effects inmemory recall. Psychological Review, 105(1), 33 ? 57. ? pages 11, 5187Vrba, E. S., Denton, G. H., Partridge, T. C., & Burckle, L. H. (1996).Paleoclimate and Evolution, with Emphasis on Human Origins. YaleUniversity Press. ? pages 26White, H. (1980). A heteroskedasticity-consistent covariance matrix estimatorand a direct test for heteroskedasticity. Econometrica: Journal of theEconometric Society, 48(4), 817?838. ? pages 60, 71Whiten, A., & Byrne, R. (Eds.) (1988). Machiavellian Intelligence. Oxford:Oxford University Press. ? pages 7, 26, 55Wierzbicka, A. (2001). A culturally salient polish emotion: Przykro [?pshickro].International Journal of Group Tensions, 30(1), 3?27. ? pages 15Worrall, J. (1988). Fresnel, poisson and the white spot: The role of successfulpredictions in the acceptance of scientific theories. In D. Gooding, T. Pinch, &S. Schaffer (Eds.) The uses of experiment: Studies of experimentation in naturalscience, (p. 135157). London: Cambridge University Press. ? pages 20, 5588Appendix AResearch Ethics Board,Certificates of Approval89


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items