sign in sign up
<<
Home , Coevolution, Cultural evolution, Dual inheritance theory, Vertical transmission

CHAPTER 38 Dual-inheritance theory: the of human cultural capacities and and Richard McElreath

38.1. Introduction means that a child can have multiple fathers, who share paternity according to the number of times In the early 1930's Wintrop and Luella Kellogg they had sex with the mother prior to birth (1933) began co-rearing their 10.5-month-old (in anthropological parlance, 'partible paternity'). son, Donald, with a 7.5-month-old female In response to this cultural belief, women in chimpanzee named Gua. The Kelloggs expected many of these actively seek out extra­ that Gua, with the chimpanzee's popular reputa­ marital copulations, often to provide their child tion for aping, would acquire numerous behaviors with extra fathers. And, while male jealously and practices via imitation from both Donald from the husband is sometimes a problem, it is and themselves. Unexpectedly, however, while regarded as socially inappropriate and thus Gua did finally acquire a few human patterns suppressed. Detailed statistical analyses from two (e.g. combing his hair), Donald was the one such societies, the Bad of Venezuela (Beckerman who began to imitate the chimpanzee in some et aI., 2002) and Ache of Paraguay (Hill and dramatic ways. Following Gua, Donald acquired Hurtado. 1996), show that the optimal number the habits of knuckle walking (which he contin­ of fathers for a child's survival is more than one. ued well after achieving full bipedality), chewing These 'other fathers' (non-husbands of mom) on shoes, scraping his teeth against interior provide resources, in the form of fish and meat, walls, and hard biting. Donald even adopted to their offspring and the mothers, both during some stereotypical chimpanzee grunts, pregnancy and while the child is growing up. barks and hoots, using a particular bark as the Interestingly, since much of the sex associated word for orange. Thus, it was the human who with 'extra fathers' occurs after conception, did most of the aping. many of these social fathers cannot be the People in many small-scale societies believe genetic fathers. Culturally transmitted beliefs in that a human fetus is formed by many repeated partible paternity have been recorded in various ejaculations of sperm into the womb. This belief linguistically unrelated societies across lowland 556 . CHAPTER 38 Dual-inheritance theory

South America, as well as in New Guinea, are theirs. Evidence indicates that we use a variety by multiple researchers over the last 75 years of cues to identify our kin, including phenotypic (Beckerman and Valentine, 2002). similarity and scent (DeBruine, 2002; Thornhill These examples illustrate two key points about et al., 2003), but humans also apparently use humans. First, while chimpanzees do show some their culturally transmitted beliefs about kinship capacities for imitative learning (Horner and and reproduction. More generally, there is also Whiten, 2005; Whiten et aI., 2005), their cultural evidence that influences our spatial shows substantially lower degrees cognition, perception of visual illusions (Segall of fidelity, frequency and internal motivation. et al., 1966), judgement (Nisbett, 2003), risk Compared to chimpanzees, humans are "imita­ preferences (Henrich and McElreath, 2002), and tion machines" (Tomasello, 1999). More gener­ notions of fairness or preferences for equity ally, while only limited social learning abilities (Henrich et aI., 2004). are found elsewhere in , social learning in Given all this, we think that a proper evolution­ our is high fidelity, frequent, internally ary framework for studying human psychology motivated, often unconscious, and broadly and behaviour needs to reckon with our species' applicable. Humans learn, via observation of heavy reliance on cultural learning and cultural others, everything from motor patterns to goals evolved . In providing such a frame­ and affective responses, in domains ranging work, dual-inheritance theory (Cavalli-Sforza from tool-making and food preferences to altru­ and Feldman, 1981; Boyd and Richerson, 1985; ism and suicide. We will refer to this form ofsocial for similar approaches also see Durham, 1991; learning, which may be particular to humans, Laland et al., 2000) aims to incorporate these as cultural learning. and other aspects of human culture under the The combination of both the high fidelity Darwinian umbrella by focusing on three key and frequency of social learning in our lineage concepts: has generated cumulative cultural evolution, which 1. Cultural capacities as adaptations. Culture, may exist to any significant degree only in our cultural learning and cultural evolution arise lineage: this is the process through which learning from genetically evolved psychological adap­ builds a body of culturally transmitted informa­ tations for acquiring ideas, beliefs, values, tion (behaviour, practices, beliefs, etc.) in a practices, mental models, and strategies from population in such a way that locally adaptive other individuals by observation and infer­ aspects aggregate over time, with the accumula­ ence. Thus, the first step is to use the logic of tion of successful additions and modifications. to theorize about the evolu­ Cumulative cultural evolution builds adaptive tion and operation of our cultural learning practices, tools, technique, and bodies of knowl­ capacities. edge (about behaviour, medicinal , etc.) that no single individual could figure 2. Cultural evolution. Our cultural learning out in their lifetime, and that can only be under­ mechanisms give rise to a robust second stood as products of cultural evolutionary system of inheritance (cultural evolution) processes. Paleoarchaeology suggests that sub­ that operates by different transmission rules stantial cumulative cultural evolution has likely than genetic inheritance, and can thus pro­ been occurring for at least the last 280 000 years duce phenomena not observed in other, less (McBrearty and Brooks, 2000), and is thus a cultural, species. Theorizing abou t this key element in understanding human genetic process requires taking what we know about evolution. human cultural learning and human cogni­ Our second point is illustrated by societies tion, embedding these into evolutionary mod­ with partible paternity: culturally acquired els that included social interaction, and beliefs can shape how we understand the world studying the emergent properties of these in ways that influence decisions, including deci­ models. This approach allows researchers to sions arising from essential aspects ofour evolved cobble up from psychology and individual cognition. To invest in their offspring, for example, decision-making to and population­ males need to figure out which offspring level phenomena. Concept 1: evolved psychological mechanisms for cultural learning . 557

3. Culture-gene . The second system Such cultural learning mechanisms can be of inheritance created by cultural evolution categorized into (i) content biases and (ii) COlt­ can alter both the social and physical envi­ text biases. Content biases, or what Boyd and ronments faced by evolving genes, leading to Richerson (1985) have called direct biases, cause a process termed culture-gene coevolution. us to more readily acquire certain beliefs, ideas For example, suppose that the practice of or behaviours because some aspect of their con­ cooking meat spread by social learning in tent makes them more appealing (or more likely ancestral human populations. In an environ­ to be inferred from observation). For example, ment of 'cooked meat', natural selection may imagine three practices involving different addi­ favour genes that shorten our energetically tives to popcorn: the first involves putting salt costly intestines and alter our digestive on popcorn, the second favours adding sugar, chemistry. Such a reduction of digestive tis­ and the third involves sprinkling chalk dust sue may have freed up energy for more 'brain on the kernels. Innate content biases that affect building'. In this way, human biology is cultural transmission will guarantee that chalk adapting to culturally transmitted behaviour. dust will likely not be a popular popcorn additive in any human . Both salt and sugar have positive innate content biases for sensible evolu­ 38.2. Concept 1: evolved tionary reasons: with salty or sugary flavours psychological mechanisms were important sources of scarce nutrients and for cultural learning calories in ancestral human environments. Thus. natural selection favoured a bias to acquire a Our approach to understanding culture begins by taste for salty and sweet foods so that we would considering what kinds of cognitive learning abil­ be motivated to acquire and eat them. Of course, ities would have allowed individuals to efficiently if you grew up in a society that only salts its pop­ and effectively extract adaptive ideas, beliefs, corn, you may steadfastly adhere to your salting and practices from their social worlds in the preference even when you find that sugar is the changing environments of our hunter-gatherer standard popcorn seasoning in other societies. ancestors. This approach diverges from main­ Thus, human food preferences are simultane­ stream in its emphasis on ously culturally learned and influenced by innate the costly information hypothesis and on the evolu­ content biases. tion of specialized social learning mechanisms. Content biases may be either reliably develop­ The costly information hypothesis focuses on ing products of our species-shared genetic her­ the evolutionary trade-offs between acquiring itage (i.e. innate) or they may be culture specific. accurate behavioural information at high cost In considering the influences of innate biases and gleaning less accurate information at low (such as those for salty or fatty foods), keep in cost. By formally exploring how the costly infor­ mind that evolutionary products like human mation hypothesis generates trade-offs in the minds are likely to contain accidental by-products evolution of our social learning capacities, we and latent structures that create biases for can generate predictive theories about the details -neutral behaviours, ideas, beliefs and of human cultural psychology. When acquiring values. Boyer (2001) details one kind of information by individual learning is costly, natu­ by-product content bias in his explanation for ral selection will favour cultural learning mecha­ the universality of certain religious concepts nisms that allow individuals to extract adaptive (like ghosts). information-strategies, practices, heuristics and On the cultural side, people may acquire beliefs-from other members of their social beliefs, values and/or mental models that then group at a lower cost than through alternative act as content biases for other aspects of individual mechanisms (like trial-and-error culture. That is, having acquired a particular learning). Human cognition probably contains idea via cultural transmission, a learner may numerous heuristics, directed attentional biases, be more likely to acquire another idea, because and inferential tendencies that facilitate the acqui­ the two 'fit together' in some cognitive or psy­ sition of useful traits from other people. chological sense. For example, believing that 558 . CHAPTER 38 Dual-inheritance theory

a certain ritual in the spring will increase the gleaned, at least partially, through observation. crop harvest in the summer might favour the For example, an aspiring farmer might imitate acquisition of a belief that a similar ritual will the strategies and practices of the most skilful, increase a woman's odds of conception, a healthy successful or prestigious farmers who live around pregnancy, and/or of successfully delivering him. Simply figuring out who obtains the biggest a robust infant. yields per hectare and copying them is a lot easier Context biases, on the other hand, exploit cues than doing all the trial-and-error learning for the from the 'individuals who are being learned immense variety of decisions a farmer (or anyone from' (we term these individuals 'models'), else) has to make. A purely individual learner rather than features of the 'thing being learned', would have to experiment with many types to guide social learning. There is a great deal of of crops, seeds, fertilizers, planting schedules, adaptive information embodied in both who and various plowing techniques. The variety of holds ideas and how common the ideas or prac­ combinations creates a combinatorial explosion tices are. For example, because information is of possibilities, making it virtually impossible costly to acquire, individuals will do better if for an individual to figure out the best farming they preferentially pay attention to, and learn strategy by relying entirely on experimentation. from, people who are highly successful, particu­ This is true of many, if not most, real-world larly skilled, and/or well-respected. Social learn­ decisions. However, along with figuring out who ers who selectively learn from those more likely is the most successful or most skilled, learners to have adaptive skills (that lead to success) can should also be concerned about how the things outcompete those who do not. A large amount they might learn will fit with their own abilities, of mathematical modelling effort has been the expectations of their role or gender, and expended in exploring the conditions under their personal context. Learners should assess which different context biases will evolve, how certain kinds of'similarity'-between themselves they should be constructed psychologically, and and potential models-and weigh this alongside what population patterns will emerge from their assessments of 'skill' and 'success'. Following individuals using such learning mechanisms. this logic, we argue below that learners might Moreover, a vast amount of field and laboratory preferentially learn social norms from individu­ data confirms that these learning biases are als who share their ethnic markers (e.g. their indeed an important part of our cognition, that dialect, language, or dress, see McElreath et aI., they are used by both children and adults, and 2003). that they influence economic decisions, opin­ Figuring out who possesses the adaptive skills, ions, judgements, values, and behaviour. strategies, preferences, and beliefs is often not Our remaining discussion of psychological straightforward. To achieve this, people rely on mechanisms focuses on the theory and evidence a range of cues related to skill (or competence), for two categories of context biases in cultural success, and prestige. For rhetorical purposes, learning: (i) success and prestige biases and this tripartite distinction is helpful because it (ii) conformity bias. captures the continuum of cues from direct observation by the learner (of skill or compe­ tence) to completely indirect assessments based 38.2.1. Selecting good cultural on prestige (defined below). Noting someone's models: success and prestige biases skill or competence, for our purposes, means Once an individual is learning from others, she that one has directly observed and judged their would be wise-in an adaptive evolutionary technique or performance. An apprentice might sense-to be selective about who she pays atten­ watch two craftsmen working side by side, one tion to for the purposes of learning (Henrich hitting all of his marks and gliding right along to and Gil-White, 2001). The idea is that a learner a perfect final product (say a handmade chair) should use cues from, or characteristics of, the while another struggles, cuts himself twice, individuals in their social world to figure out curses a bit, and produces something that only who is most likely to have useful ideas, beliefs, the bravest of his friends would venture to sit on. values, preferences, or strategies that might be Direct observation indicates who the learner Concept 1: evolved psychological mechanisms for cultural learning . 559

should pay attention to for learning to make provides a best guess to the skill ranking until chairs. more information can be accumulated. Figuring Cues of success are less direct and take advan­ out who to learn from, using the distribution tage of easily observable correlates of compe­ of deference, is merely a way of aggregating tence (especially those that are difficult to fake), the information (opinions) that others have as we have defined it. Depending on the domain already gleaned about who is a good person to and society, such cues migh t be measured by learn from. house size, family size, number of wives and/or As part of these deference patterns, people children, number of peer-reviewed publica­ unconsciously cue who they think is a good model tions, costliness of their car, number of tapirs through a series of ethological and behavioural killed, number of heads taken in raids, the phenomena that arise directly from efforts to size of their biggest yam, etc., each of which, in imitate these individuals. These patterns relate particular social contexts, is related to some to attention, eye gaze, verbal tones and rhythms, domain of skill. While these cues provide only and behavioural postures. As learners seem an indirect measure, they are sometimes supe­ keenly attuned to these subtle patterns, it rior to cues of competence. If performances appears that natural selection has favoured are noisy, the observations of a small sample attention to these patterns of deference, as a may lead a learner to misperceive competence. means of assessing whom to pay attention to for Cues of success, in contrast, often average over cultural learning. As we discuss below, a mecha­ many performances, which can help to reduce nism like 'copy the majority' (conformist trans­ the error in the learner's assessment of who to mission) provides an effective way to aggregate learn from. the information gathered by observing and lis­ The evolutionary theory underpinning this tening to others. In this case, conformist trans­ form of model-based cultural learning proposes mission can be used to figure out who to pay that once the psychological machinery that makes attention to for cultural learning. use of competence- and success-based cues for To understand the difference between cues of targeted cultural learning has spread through prestige, success and skill, consider the following the population, highly skilled and successful stylized example of an academic department. individuals will be in high demand, and social A new PhD entering a department and aiming learners will need to compete for access to the at tenure might assess his senior colleagues in most skilled and successful. This creates a new order to figure out who to learn from (with the selection pressure for such learners to pay defer­ goal of getting tenure). Initially, he can glean a ence to those they assess as most valuable measure of people's prestige-deference by listening (those judged most likely to possess adaptive to and observing how people act towards each information) in exchange for preferred access other. If he's really serious, he might pull up and assistance in learning. Deference benefits everyone's CVs and count their publications may take many forms, including coalitional (and divide by their 'years since PhD'). This support, general assistance (helping with labori­ would give a measure of success. Finally, if our ous projects), public praise, caring for the off­ fresh PhD still has not given up all hope offinding spring of the skilled, and gifts (Gurven, 2001). a good model, he might read everyone's papers With the spread of deference for highly skilled (or at least those who rank high in 'success' and individuals, natural selection can take advantage 'prestige') and watch them teach. This would give of the observable patterns of deference to fur­ our learner a measure of skill or competence. ther save on information-gathering costs. Naive Aggregating all these measures, he'd have a decent entrants (say immigrants or children), lacking estimate of who to learn from. detailed information about the relative skill Interestingly, the indirect nature of assessing or success of potential cultural models, may take another person's utility as a cultural model advantage of the existing pattern of deference by (i.e. their possession of adaptive information using the amounts and kinds of deference differ­ that could be useful to the learner) creates an ent models receive as cues of underlying skill. important phenomenon. In a complex world, Assessing differences in deference-received such indirect measures do not tell the learner 560 . CHAPTER 38 Dual-inheritance theory

which of the model's behaviours, ideas, practices consumption, beliefs, and dialects. (iii) Consistent and strategies causally contribute to his success with theory, the amount of cultural learning or competence. For example, are people successful observed depends critically on the degree of in farming because of what they , when uncertainty found in the environment. As they plant, how they plant, or how they make uncertainty increases, so does cultural learning. sacrifices to the spirits-or all four? Because of (iv) These learning patterns emerge even when this ambiguity, humans may have evolved the the model's domain of competence, success or propensity to copy successful individuals across prestige is apparently unrelated to the behav­ a wide range of cultural traits, only some of ioural domain in question. (v) Diverse findings which may actually relate to the individuals' from laboratory experiments in both economics success. When information is costly it turns out and psychology, using very different experimental that this strategy will be favoured by natural paradigms, consistently converge-giving us selection even though it may allow neutral and confidence in the findings' robusticity across even somewhat maladaptive traits to hitch-hike experimental contexts. (vi) The patterns of along with adaptive cultural traits. cultural learning observed in the laboratories fit closely with field data-giving us confidence 38.2.1.1. Evidence of selective model-based that the effects observed in the artificial context of cultural learning experiments actually matter in the real world. Evidence for these learning mechanisms is plen­ Below, we swnmarize some ofthe laboratory find­ tiful, and comes from across the social sciences. ings to illustrate points (i)-(v), and then describe A broad spectrum of work shows that both kids a few key field studies that illustrate point (vi). and adults will preferentially learn all kinds of Experimental evidence from Pingle (1995) things from other individuals demonstrating confirms that people (well, university students) particular cues of competence, success and/or imitate the strategies of successful individuals prestige-and there need not be any particular when pay-offs are on the line. Using a series of relationship between domains of prestige or computerized decision situations, participants competence and the things being learned. had to repeatedly select the amount of three Unfortunately, the details don't go much beyond different inputs (e.g. 'fertilizer: 'seed' and 'labour') that. For example, we would like to know how into a production problem for either 21 or 31 different kinds of information are integrated. rounds, depending on the treatment. Before each How important is observed competence com­ decision, i.e. before setting the final amounts pared to prestige? How important is individual (xi' x2 and x3) of the three inputs for a given information when it contradicts the behaviour round, subjects could pay to find out what profit of highly successful people? Having looked at a they would get if they used different sets of wide range of social learning evidence, it is clear inputs (a 'costly experiment'). In the baseline that the tendency to imitate prestigious and treatment, subjects could only learn from their successful people is one of the most powerful own analyses and direct experience (i.e. what they aspects of cultural learning. earned each round from their chosen inputs). In providing a taste of the evidence for success To calculate profit in each round, the subject's and prestige-biased cultural learning, we empha­ inputs were run through a pre-set production size six main points. (i) These imitative patterns function. This function, which was unknown to spontaneously appear in incentivized (where players, had only one set of optima inputs individual's choices influence monetary payoffs (xi, xi, x3); these inputs would make the most or other kinds of returns) and non-incentivized money. In four other treatments, opportunities circumstances, in both non-social and social for imitation were introduced in varying ways and situations, including situations that involve with different costs. Participants in all treatments direct competition among the learners. 'Social faced the same environment (the same produc­ situations' are those in which a person's pay-offs tion function) for rounds 1 to 11 (Block 1). and those of others are jointly influenced by At round 12, the envirorunent shifted and again their choices. (ii) The effects repeatedly emerge remained constant through round 21. For treat­ across a broad range of contexts, including eco­ ments 2-4 and the control, there was also a nomic decisions, opinions, food preferences and 'competitive' environment that commenced in Concept 1: evolved psychological mechanisms for cultural learning . 561

round 22 with an environmental shift that lasted 4. Uncertainty causes a substantial increase in through 31 (Block 3). During this Block, the the reliance on imitation. In rounds 2, 12,22, optimal set of inputs shifted dynamically and when a new environment is first encountered, depended on what other players had done. This rates of imitation are highest. means that participants faced a new environ­ 5. The availability of imitative opportunities, even ment beginning in rounds I, 12 and 22. Blocks I costly ones, improves the average performance and 2 are non-social decisions, while Block 3 of the group. As a group, subjects in imitation provides one type of social interaction. treatments outperformed those of the control. The different treatments manipulated the infor­ mation available for imitation: in treatment I, 6. The 'imitation environment' (treatment) during each round (starting in round 2) partici­ affects the average performance of the group. pants could, at a cost, look at the inputs and out­ Average performance in treatments 3 and 4 put of one other subject who had previously exceeds that of treatments I and 2. Only the played that round. In treatment 2, participants informational environment of treatment 3 could. at a cost, look at a list of inputs and out­ avoids a substantial degradation in group puts for that round for all the subjects who had performance during the Block 3. gone before them. In treatment 3, before the play for each Block commenced, subjects were Other work by economists confirms these given the best outputs and inputs of previous findings. Kroll and Levy (1992) show that players for that Block. In treatment 4, each sub­ individuals readily imitate the investments of ject watched two other subjects complete all successful players, and that adding the possibility 31 rounds before playing themselves. Each treat­ of imitation improves the overall performance ment used different subjects, who were paid real of the group. Offerman and Sonnemans (1998) money according to the profit they earned, show that, not only will people copy economic which was determined by their choices of inputs. choices and investment strategies, but they A comparison of the findings from across the will also preferentially imitate beliefs about treatments highlights several important points the state of the world from successful people. about imitation, all of which have been antici­ Work studying competitive Cournot markets pated by cultural evolutionary models (Boyd and demonstrates the power of this form as imita­ Richerson, 1985, 1988, 1995; Weibull, 1995): tion (Alpesteguia et aI., 2003). Recent studies exemplified by the above J. In non-social situations, participants use imi­ experiment are important because the decision­ tation, often to a substantial degree, even making is incentivized and the available infor­ when decisions are financially motivated and mation is rigorously controlled. Qualitatively cost-benefit analysis is possible (but costly). however, these findings from economics merely The pattern of results across all four experi­ confirm older empirical insights from psychology. ments, vis-a-vis the non-imitation control, Research elsewhere in psychology has shown shows the strength of our propensity for imi­ that individuals preferentially acquire opinions tation: in round 2 oftreatments 1and 2, which from prestigious sources, especially in ambiguous, can be compared directly to round 2 in the no­ uncertain, or difficult situations, and even when imitation control, people imitated 87% and these opinions are not connected to the model's 57% of the time, respectively. domain ofexpertise. See Henrich and Gil-White (200 I) for a review of this evidence. Not only do 2. Imitation tendencies remain strong even in these cultural learning mechanisms operate in competitive social environments. About 43% incentivized decision-making, but they also ofsubjects imitated in round 22 of treatment 2. appear in non-incentivized situations in which 3. People tended to imitate (the inputs) of more behaviour, opinions and preferences shift spon­ successful players (those who got higher out­ taneously and unconsciously. puts). The patterns in the data are only explica­ The same evolved cultural learning mecha­ ble if people are looking at the difference in nisms emerge outside the laboratory, across a performance and using that as a cue about wide range of behavioural domains, including who and when to imitate. two areas that we mention here: (i) the diffusion 562 . CHAPTER 38 Dual-inheritance theory

of innovation and (ii) the of showing patterns similar to those observed for suicide. In his massive review of the literature on diseases, novel cultural practices, and innovations. the Diffusion ofInnovations, Rogers (I995, p. 18) In Micronesia (Rubinstein, 1983), beginning in summarizes some of the lessons from 50 years 1960 and lasting for at least 25 years, a suicide epi­ of research as follows: demic spread through certain island populations. This case is particularly stark because the sui­ Instead, most people depend mainly upon a subjec­ cides are geographically patterned and distinc­ tive evaluation of an innovation that is conveyed tively stereotyped. The typical victim was a to them from other individuals like themselves young male between 15 and 24 (modal age of 18) who have previously adopted the innovation. This years who still lived at home with his parents. dependence on the experience of near peers sug­ After a disagreement with his parents or girl­ gests that the heart of the diffusion process consists friend. the victim was visited in a vision by past of the modelling and imitation by potential adopters suicide victims who 'called him to them' (we of their network partners who have adopted know this from parasuicides). Heeding the call, previously. the victim performed a 'lean hanging' from Rogers devotes an entire chapter to explaining either a standing or sitting position, usually how the diffusion of new ideas, technologies, in an abandoned house, until he died of anoxia, and practices is strongly influenced by 'local or was accidentally discovered. In 75% of the cases opinion leaders'. Compiling findings from many there was no prior hint of suicide or depression. diffusion studies, Rogers describes these individ­ These suicides occur sporadically in local uals as: (i) locally high in social status (e.g. high outbreaks among socially interconnected male status within the village or village cluster); adolescents who ethnically identify as from Truk (ii) well respected (indicating prestige); (iii) widely or the Marshals (matching on sex and ethnicity), connected; and (iv) effective social models for and can sometimes be traced to the precipi­ others. Rogers' insights are particularly important tating suicides of prominent sons from wealthy here because they confirm that success and pres­ families (associated with prestige). tige-biased cultural learning are important for the Prestige bias also appears in studies of linguistic spread of novel technologies and practices. change (Labov, 1972, 1980), the transmission The theory derived from the logic of selective of managerial styles (Weiss, 1977; Weiss et al., model-based cultural learning even illuminates 1999) and in naturalistic studies of jaywalking some of the robust patterns observed in studies manipulation (Mullen et aI., 1990). It also been of suicide. Data from industrialized societies repeatedly observed by ethnographers in an show that committing suicide, including the immense variety of contexts (Berreman. 1972, methods (poisoning, gun, hanging. burning. etc.), p. 141; Dove, 1993; Boyd, 2001; Rao, 2001). are imitated according to prestige and self­ similarity (Wasserman et aI., 1994; Stack, 1996). For prestige, many studies in the USA, Japan 38.2.3. Conformist transmission and Germany show that suicide rates spike more As an adaptive learner, what do you do when after celebrity suicides than non-celebrity any observable differences in skill, success, and suicides (Stack, 1987; Kessler et aI., 1988), even prestige among individuals do not covary with once media coverage is controlled for (Stack, the observable differences in behaviour, beliefs, 1990, 1996; Jonas, 1992). For similarity, the practices. or values? For example, suppose every­ results show that the individuals who kill them­ one in your village uses blowguns for hunting, selves after celebrity suicides tend to match their except one regular guy who uses a bow and models on age, sex and ethnicity. Finally, the time arrow, and obtains fairly average hunting returns. trends of these suicides do not show regression to Do you adopt the bow or the blowgun? One the mean during the subsequent month, indicat­ solution for dealing with such information­ ing that these were not individuals who would poor dilemmas is to copy the behaviours. have committed suicide in the near future. beliefs and strategies of the majority (Boyd and Because suicide is strongly influenced by Richerson, 1985; Henrich and Boyd, 1998). imitation, it can spread in epidemic fashion, Termed conformist transmission, this mechanism Concept 1: evolved psychological mechanisms for cultural learning' 563

allows individuals to aggregate information over 53.3% of the 'confused' individuals adopting the behaviour of many individuals. Because these bows. For example, of the confused 50,40 might behaviours implicitly contain the effects of each adopt bows, while 10 still decide to go with individual's own experience and learning efforts, blowguns. After all of the transmission this gen­ conformist transmission can be the best route to eration, 120 hunters will use the more adaptive in information-poor environments. bow, while only 80 use blowguns. If the older To see this, suppose every individual is given ('experienced') generation dies, 80% of the new a noisy signal (a piece of information) from the generation wiU use bows (compared to only environment about what the best practice is 40% of the now dead cohort). in the current circumstances. This information, This kind of verbal reasoning has been rigor­ for anyone individual, might give them a 60% ously tested in both analytical models (Boyd and chance of noticing that blowguns bring back Richerson 1985, Chapter 7) and extended to slightly larger returns than bows. Thus, using more complex environments using evolutionary individual learning alone, learners will adopt simulations (Henrich and Boyd, 1998; Kameda the more efficient hunting practice with proba­ and Nakanishi, 2002). In their computer simula­ bility 0.60. But, if an individual samples the tion, Henrich and Boyd investigated the interac­ behaviour of 10 other individuals, and simply tion and coevolution of adopts the majority behaviour, his chances of (parent~ffspringtransmission), individual learn­ adopting the superior blowgun technology ing, and conformist transmission in spatiaUy and increase to 75%. temporally varying environments. The results The same logic can be applied to aggregate confirm that conformist transmission is likely and improve the imperfect information about to evolve under a very wide range of conditions. the relative success of others, who may be useful In fact, these results show that the range of as cultural models. Some individuals may obtain conditions that favour conformist transmission accurate information that allows them effectively are broader than those for vertical transmission to select and copy the most successful individuals, alone-suggesting that if true imitation (via while others may receive noisy (inaccurate) parent-<:hild transmission) evolves at all, we information about relative success, which should also expect to observe a substantial con­ prevents them from effectively distinguishing formist component. Taken together, this work differences. This second group can still take leads to several specific predictions about human advantage of the more accurate information psychology. First, this model predicts that learners received by the first group by adopting the traits will prefer conformist transmission over vertical adopted by the majority. To see this more clearly, transmission, assuming it is possible to access a imagine a group of 200 individuals, wherein range of cultural models at low cost (which is 100 are experienced hunters and 100 are novices often but not always the case). While a direct test who need to figure out which technology to invest of this prediction is lacking, we note that a in learning. Of the 100 experienced individuals, substantial amount of research in behavioural suppose that 40 used bows and 60 use blowguns genetics indicates that parents actuaUy transmit for hunting. In their current environment very little culturally to their offspring-once (which recently changed), however, bows obtain genetic transmission is accounted for, vertical a more efficient return, although the difference cultural transmission often accounts for less is small and hunting returns in general are highly than 5% of the variation among individuals variable. Nevertheless, using the returns of the (Harris, 1995, 1998; Plomin et al., 2000). Those experienced hunters, 40 of the 100 novices familiar with earlier work on cultural transmission selected a bow hunter to learn from, 50 were left might recall high correlations between parents confused, and 10 picked a blowgun hunter to and offspring, suggesting an important role learn from (they got bad information due to the for vertical cultural transmission (e.g. Cavalli­ noise in hunting returns). In their confusion, Sforza et aI., 1982). This work neglected the sim­ the 50 decide to use conformist transmission, ilarity between parents and offspring created by where now 80 hunters use bows (40 + 40) and genetic transmission. Once the influence of 70 use blowguns. This will result in more than genetic transmission is accounted for, the effect 564 . CHAPTER 38 Dual-inheritance theory

of vertical cultural transmission in creating par­ crops change. After each round, each farmer ent-offspring correlations largely evaporates. learned the yield realized in that year for his Certainly there may be cases in which parents field, and could choose to look at the decisions are the only viable models, and so have a large (crop A or B, but not the yields) of other farmers role, such as in early language acquisition or in the past year. At the end of the 20 rou nds family recipes. But that does not indicate that players were paid according to their total yield people prefer to imitate their parents, nor that over the 20 seasons, making between $4 and $8. parents have a large effect in general. Second, the Consistent with theoretical predictions. McElreath model predicts that as the accuracy of information et aI.'s analysis confirms that (i) people increase acquired through individual learning decreases, their appetite for social information when crop a learner's reliance on conformist transmission variance is high and decrease it in temporally (over individual learning) will increase. Third, fluctuating environments, and (ii) a simple con­ as the proportion of models-in the learner's formist learning rule (copy the majority) seems sample of models-displaying a trait increases, to capture an important part of decision-making the strength of the conformist effect should in this problem, although there is quite a bit of increase non-linearly as well. We address the individual heterogeneity. second and third predictions below. A naturalistic experiment using non­ A substantial amount of empirical research incentivized behaviour further confirms these from psychology shows that people conform in conformist effects by showing the non-linear a wide range of circumstances, particularly influence of the frequency of a behaviour when problems are complex or difficult to solve (Coultas, 2004) on its adoption. Here, subjects on one's own. This work reveals that humans entered a computer laboratory one-by-one, not have two different types of conforrnity that realizing they were in an experiment. and operate in different contexts (Baron et aI., 1996). observed a 'rare behaviour' that involved placing The first, often called informational conformity, the keyboard cover on top of the monitor. matches the theoretical expectations from In pre-testing, the experimenters confirmed that models of conformist transmission and is used no one, without modelling, ever put the cover to figure out difficult or ambiguous problems. on top of the monitor-so without modelling Informational conformity results in people actu­ the expected frequency of placing the cover on ally altering their private opinions and beliefs the monitor is zero. The experimenters were about something. The second, often called able to manipulate the number of individuals normative conformity, is conformity for the placing the cover on the monitor by silently purposes of going along with the group to avoid giving explicit instructions to some few through appearing deviant. Under this type of conformity, their computer monitors. Others, not receiving people alter their superficial behaviour, but these instructions, were observed to see if often do not change their underlying opinions, they placed the cover on top of the monitor. preferences or beliefs. Figure 38.1 summarizes the results by showing Experimental work shows that conformist how the frequency of models performing the transmission is important in individual decision­ cover placement affected a subject's likelihood making situations (non-social circumstances). of making the same placement. The horizontal In an experimental design that parallels the afore­ axis gives the percentage of individuals already mentioned simulation constructed by Henrich present in the room who had their keyboard et al. (2004), McElreath et al. (2005) had under­ covers on top of their monitor as the subject graduate subjects repeatedly face an economic entered. The vertical axis gives the probability choice between two options, A or B, for 20 rounds. that the subject would then place his keyboard This was posed as a 'farming decision' in which cover on top of his monitor. As predicted, the A and B were different crops with different likelihood of performing this behaviour, which yields and yield variances. Players did not know is not otherwise performed, increases non-lin­ the mean yields or yield variances for the two early as the percentage of models performing crops, but were told that the local environment the behaviour rises above 50%. One problem might fluctuate such that the mean yields of the with this experiment is that it does not carefully Concept 2 and 3: cultural evolution and culture-gene coevolution . 565

0.7 n =73 l:: 0.6 J...... ---.-­ 0 E 0.5"­ 'E 04 =o I n =81 ~ 0.3 j ~ ~ 0.2- n =101 n =81 £ 0.1 . n =30 0 n n 0-20 21-40 41-60 61-80 81-100 Percentage of models performing the action Fig.38.1 The figure shows how the percentage of models performing the 'covers on the monitor' behavior influences the likelihood of others performing the same behavior. The n values above each bar gives the number of individuals observed for that bar-e.g., 73 subjects entered a room in which between 81 and 100 percent of the people in the room had their keyboard covers on their monitors; about 60% of these subjects then put their covers on the monitor.

distinguish informational from normative dual-inheritance theory can generate a wide conformity. range of higher-level theories about the cultural As with prestige-biased transmission, con­ evolutionary and culture-gene coevolutionary formist transmission is also important in social origins of sociological phenomena (e.g. ethnic situations, including cooperative interactions. groups). Instead of arguing that unidirectional Conformity effects have also been observed in causation exists at either the individual or society experimental situations involving opportunities level, dual-inheritance theory explicitly models for cooperation and . Players in these individuals with evolved or evolving psycholo­ games were willing to use conformist learning gies in interactions with other individuals to for acquiring cooperative behaviour, selfish behav­ understand more precisely how cultural learn­ iour. the costly pWlishment of non-cooperators, ing mechanisms give rise to cultural evolution, and even the costly punishment of those who and how this might feed back on genetic evolu­ refuse to punish non-cooperators (Carpenter, tion (Cavalli-Sforza and Feldman, 1981; Boyd 2004; Denant-Boemont eta!', 2005). In general, and Richerson, 1985). Here we discuss theory, the powerful effects of cultural learning on rooted in formal modelling efforts, applied to cooperation and altruism are empirically well­ the following questions: (i) why do cultural evo­ established (for summary see Henrich and lutionary rates and degrees of adaptation vary Henrich in press, Chapter 2). among populations (Shennan, 2001; Henrich, 2004) and how might this have influenced the basic cognitive abilities ofdifferent hwnan groups 38.3. Concepts 2 and 3: and (ii) why do ethnically-marked groups emerge cultural evolution and and how did our 'ethnic psychology' develop culture-gene coevolution (Gil-White, 2001; McElreath eta!', 2003). Other work in this area examines how adaptive cultural By combining these kinds ofworking hypotheses learning can sometimes give rise to the otherwise about the nature of our evolved individual-level puzzling patterns of maladaptive cultural prac­ adaptive learning mechanisms (e.g. prestige­ tices, such as the biased and conformist-biased transmission) (Richerson and Boyd, 2005), cooperation as well with formal models of population processes (see as Wliquely human forms psychology (McElreath McElreath and Henrich, this volume, Chapter 39) and Henrich, this volwne, Chapter 39). 566 . CHAPTER 38 Dual-inheritance theory

del Fuego. The Australian mainlanders pos­ 38.3.1. Demography and cultural sessed the entire Tasmanian toolkit plus hun­ evolutionary rates dreds of additional specialized tools including In the last 10000 years, the rate of cumulative multi-pronged fish spears, spear throwers, cultural adaptation has accelerated many times boomerangs, mounted adzes, sewn bark canoes, over, but the distribution in rates has been very ground edge axes, string bags, composite tools, uneven across the continents (Diamond, 1997). and a variety of nets for , fish and wallabies While much of this variation is likely to be (Henrich,2004). explained by historical particulars, we suspect With this puzzle in mind, Henrich (2004) several important general processes are also at constructed a model in which individuals pref­ work (Turchin, 2003). While difference between erentially imitate highly skilled individuals. Unlike continents is probably the most significant pattern previous models, however, Henrich's model left in human history, evolutionary approaches, at open the possibility that transmission was both least those devoid of an explicit appreciation of noisy (highly variable) and negatively biased cultural transmission, have remarkably little to (copies are usually worse than the originals)­ say about it. To illustrate, we briefly discuss one both plausible assumptions, especially for com­ cultural evolutionary model that explores the plex technological skills and areas of knowledge. interaction between demographic conditions The analytical results show two things worthy of and cultural evolutionary rates of adaptation note: (i) the rate of adaptive evolution depends (e.g. in technology, skills, knowledge) that may on the natural logarithm of the effective popula­ help explain both variable rates of cultural adap­ tion size (effective population size incorporates tation in different places and peculiar cases of absolute size and degree of interconnectedness: maladaptive cultural and technological losses. the size of the pool of interacting social learners); As the most extreme and archaeologically and (ii) if a well-adapted large population best-documented case of maladaptive techno­ suddenly shrank, it could enter a regime of logical loss, Tasmania provides an intriguing gradual maladaptive deterioration, as it moved puzzle, and good point of departure for an towards a new, less-well-adapted, equilibrium. inquiry. Humans first arrived in Tasmania about Empirically, the intervening time-period between 34 000 years ago and were subsequently cut off the two equilibria would show a gradual loss of from mainland between 12 000 and complex skills and knowledge (easy-to-learn 10 000 years ago by rising seas that filled the skills would not be affected). Effective population 200 km stretch of land linking Tasmania to size influences the evolutionary rate by making Victoria. At the time of European discovery, 'positive' errors-those that result in a more Tasmania had the simplest technology of any adaptive practice-more likely. This, along with population ever encountered. A combination a few other nuances in the archaeological of ethno-historical and archaeological data record, indicates that the Tasmania pattern of suggests that, over the 10 000 year period deterioration may have been ignited by the after being cut off from mainland Australia, interaction between the dynamics of cultural Tasmanians likely lost, or never evolved, the ability transmission and the sudden drop in effective to make bone tools, fitted cold-weather clothing, population size created when rising oceans sev­ hafted tools, fishing spears, barbed spears, nets, ered the link to the social learning networks of and boomerangs. Bone sewing needles, of the southern Australia. Overall, besides revealing kind used ethnographically by Australian aborig­ the possibility of maladaptive deteriorations inals to make fitted clothing, are clearly present in when networked populations are cut off, this Tasmania before the seas rose. To hunt and fight, simple model also shows that larger, more inter­ Tasmanian men used only one-piece spears, rocks connected, populations can evolve both more and throwing clubs. In all, the entire Tasmanian rapidly and to a better-adapted equilibrium toolkit consisted of only about 24 items, and than smaller, or less well interconnected, contrasts starkly with both their contemporary populations. This may provide an evolutionary aboriginal cousins just across the Bass Strait explanation why Diamond's observation that in southern Australia and other cold-climate rates of technological evolution proceeded at foragers such as the Ona and Yahgan of Tierra different rates on different continents. Concept 2 and 3: cultural evolution and culture-gene coevolution . 567

With the adaptive nature of cultural evolution cooperation and collective action (Henrich and in mind, it is important not to underestimate Henrich, in press, Chapter 9), as well as out­ the degree to which culture, and cumulative group hatred (LeVine and Campbell, 1972). cultural evolution, can influence basic facets of Many social scientists hold the opinion that these human cognition. Consider two aspects of our groups and their markings form out of collec­ psychology: (i) spatial cognition and (ii) numeri­ tive interest alone, or that they are the result of cal conceptions of quantity. strategic switching and signalling on the part of Spatial cognition. At most, human languages political actors. V\'hile we think this is partly possess three different systems for describing true, the existence of strategic ethnic manipula­ spatial position: (a) absolute: the ball is north of tion makes the maintenance of these arbitrarily the tree; (b) object-centered: the ball is on marked groups problematic. If individuals can Richard's left [as an object, Richard inherently merely choose their ethnicity at any time, then has a (culturally defined) left, right, front, etc.]; why should anyone pay attention to the cheap and (c) relative: the ball is to the left of the tree labels at all? Models of ethnic markers as signals (here an imaginary line is drawn from the of cooperative intentions in fact show that the speaker or other reference point to the tree, thus process is unlikely to work (Nettle and Dunbar, creating a 'left' for the tree), However, not all 1997; Roberts and Sherratt, 2002; see McElreath languages have evolved all three systems, with et al., 2003 for more discussion of analogous some and languages lacking the relative biological models). system, and relying heavily on the absolute To explore the relation between social norms, system. Cognitively, speakers of these languages symbolic markers, and cultural learning, (i) possess incredible dead-reckoning abilities McElreath et al. (2003) constructed a mathe­ and seem to have a constantly running mental matical model to study the claim that arbitrary compass, and (ii) perform very differently in and easily acquired 'ethnic markers' (e.g. dialect, non-linguistic tests of spatial memory (Levinson, dress, hairstyle) may function to signal hidden, 2003). It seems that the cultural evolution important norms of behaviour that differ among of linguistic system, and associated cultural population subdivisions. routines, for discussing and dealing with space The model assumes that tlle population is sub­ and orientation influences our non-linguistic divided spatially into 'groups' linked by migration. spatial cognition. Groups are large and each individual is charac­ Numerical conceptions of quantity. Number terized by one of two norms. Norm differences systems are an aspect of culture and language arise in the model because the authors assumed that varies substantially among societies. Many that the norms solve coordination problems, such societies, for example, only have ordinal numbers that individuals with locally common behaviours up to 3. Recent work using experiments from are at an advantage in terms of individual success cognitive science among two Amazonian groups (locally if everyone pays bride price, not dowry, demonstrates tllat growing up with such number one should also pay bride price, to coordinate systems greatly influences people's abilities in with others). The model assumes that these non-linguistic tasks that involve memory and behaviours are not observable, because many matching, in dealing with quantity and number norms are unconscious and not easy to anticipate (Pierre et al., 2004; Gordon, 2005). Thus, the (Nave, 2000; Gil-V\'hite, 2001). Each individual cultural evolution of a number system influ­ also adopts one of two visible markers. These ences the brains of those who grow up using it. markers are costless, but may be observed prior to interacting based upon the hidden norms. 38.3.3. The coevolution of Individuals may preferentially interact with those with the same markers as themselves, but ethnically marked groups this tendency is allowed to evolve within the and ethnic psychology model. A curious feature ofhuman societies is their sub­ Na'ive individuals acquire both norms and division into self-ascribed arbitrarily marked markers by imitating successful individuals groups, sometimes called 'ethnic groups' (Barth, (as discussed above). With some chance, they 1969). These groups are sometimes the loci of acquire both from the same individual, which 568 . CHAPTER 38 Dual-inheritance theory

may generate covariance between markers and transfers this population covariance into behaviours. This bundled imitation is also allowed within-group covariance. to evolve in the model. 2. Direct selective processes favouring common The central question addressed by the model is behaviours create indirect selection on mark­ not whether stable norm differences can evolve, ers, proportional to the covariance between but rather whether, given stable norm differ­ behaviours and markers. This increases mark­ ences, stable regions of ethnic marking will arise ers associated with common behaviours, within that covary with norm boundaries. That is, do each local group. adaptive cultural learning mechanisms some­ 3. Natural selection favours a psychological bias times give rise to ethnic groups, as an emergent for interacting with those with the same by-product? And, if they do, how does this emer­ marker as oneself, because there is always gent social environment influence the genetic some covariance between markers and evolutionary processes that shaped human behaviours, due to migration. As this interac­ psychology? It is important to realize that in a tion bias increases, selection increases the purely genetic model ethnic groups would be covariance further, because then makers and unlikely to emerge, as migration between subdi­ behaviours form co-adapted pairs. visions would normally swamp selection. Yet, empirically we know that ethnic groups manage 4. While migration may be needed to get the to maintain apparent norm differences despite process going, if it is too strong, it swamps migration rates of one sex approaching I: along the selective forces above, leading to the Vaupes river linguistic exogamy means that unmarked groups, sometimes even if behav­ people must marry someone from a group who ioural (norm) differences remain. This is does not speak their language (Jackson, 1983). where the plausibility of weak migration rela­ Only extreme selection could maintain geneti­ tive to the strong forces of our cultural learn­ cally based behaviour differences under such ing psychology is crucial to the model. If migration. However, as mentioned in the preced­ individuals are not strongly disposed to learn ing sections, both social interactions and mecha­ from group members with higher pay-offs, nisms like conformist transmission can maintain then mixing will erode differences between differences between social groups, even when neighbouring groups. interaction and the physical movement of bodies Once regions of norms and ethnic markers is common. Likewise, selective cultural-learning exist, selection on genes favours an increased processes can be strong even when the direct predisposition to interact with those who look pay-off differences among behaviours are like oneself (share one's markers). It also favours small (see McElreath and Henrich, this volume, acquiring bundles of traits, norms and markers Chapter 39). together, from the same individual during social A feedback loop generates and maintains ethnic learning (this may further enhance the tendency marking, as long as migration exists but is not too of individuals to learn things from successful strong relative to the selective processes created models that do not directly relate to their domain by success-biased cultural learning (that arise of success or expertise). Importantly, the from the need to coordinate social interactions). cultural evolution of behaviourally distinct The model works as follows. groups and their markers leads to natural selec­ 1. Migration creates small amounts of covari­ tion on aspects of psychology. This is the kind ance between specific markers and behav­ of culture-gene coevolution that we think is iours within each local group. This occurs common in . even if there is initially no covariance within In conclusion, if we are right, then constraining each group. The reason is subtle. If local ourselves to purely genetic models of human groups differ at all in their frequencies of evolution will handicap our attempts to under­ markers and behaviours, then there is covari­ stand important domains of human behaviour, ance at the population level. Population because the crucial selective forces that may structure is represented by the covariance account for some of our psychological adapta­ across groups. Migration among local groups tions arose first through the evolution of culture. References . 569

This is not to say that humans may have in any Cavalli-Sforza, L. L. and Feldman, M. (1981) Cultural sense 'transcended' natural selection, any more Transmission and Evolution. Princeton University Press, than domesticated have. Rather, the Princeton. Cavalli-Sforza, L. L., Feldman, M. w., Chen, K. H. sources of our selection pressures may often be and Dornbusch, S. M. (1982) Theory and observation different in important respects from those of in cultural transmission. Science, 218 (4567): 19-27. closely related species, because of our evolved Coultas, j. (2004) When in Rome ... an evolutionary capacities for cultural transmission. Our bet, perspective on conformity. Group Processes and bolstered now by more than two decades of Intergroup Relations, 7: 317-331. DeBruine, L. (2002) Facial resemblance enhances trust. fonnal models of culture-gene coevolution and Proceedings ojthe Royal Society ojLondon, B, substantial evidence from laboratory and field 269: 1307-1312. sciences, is that it will prove very hard in the Denant-Boemont, L., Masclet, D. and Noussair, C. (2005) long run to understand the structure of human Anonymity in Punishment, Revenge and Cooperation: psychology without reference to the dynamic A Public Goods Experiment. Atlanta. Diamond, j. M. (1997) Guns, Germs, and Steel: The Fates population processes that help to construct our ojHuman Societies. Norton, New York. selection pressures. Dove, M. (1993) Uncertainty, humility and adaptation in the tropical forest: the agricultural augury of the Kamu. References , 32: 145-167. Durham, W. H. (1991) Coevolution: Genes, Culture, and Baron, R., Vandello, j. and Brunsman, B. (1996) The Human Diversity. Stanford University Press, Stanford, CA. forgotten variable in conformity research: impact Gil-White, F. (200 [) Are ethnic groups biological 'species' of task importance on social influence. Journal oj to the human brain? Essentialism in our cognition of Personality and , 71: 915-927. some social categories. Current Anthropology, 42: 515-554. Barth, F. (1969) Introduction. In Ethnic Groups and Gordon, P. (2005) Numerical cognition without words: Boundaries, pp. 9-38. Little, Brown & Co., Boston. evidence from Amazonia. Sciellce, 306: 496-499. Beckerman, S., Lizarralde, R.. Lizarralde, M. et at. (2002) Gurven, M. (2001) Does market exposure affect The Bari Partible Paternity Projeer, phase One. In S. economic game behavior? the ultimatum game Beckerman and P. Valentine (eds) Cultures ojMultiple and the public goods game among the Tsimane' of Fathers: The Theory and Practice ojPartible Paternity Bolivia. In ). Henrich, R. Boyd, S. Bowler, C. Camerer, in Lowland South America, pp. 27-41. The University E. Fehr and H. Gintis (eds.) Cooperation, Reciprocity and Press of Florida, Gainesville. Punishment: Experiments in Fifteen Small-scale Societies, Beckerman, S. and Valentine, P. (2002) Introduction: the pp. 194-231, Oxford University Press, Oxford. concept of partible paternity among native South Harris,). R. (1995) Where is the child's environment? Americans. In S. Beckerman and P. Valentine (eds) A group theory of development. Cultures ojMultipLe Father': The Theory and Practice Psychological Review, 102: 458-489. ojPartible Paternity in Lowland South America, Harris, j. R. (1998) The Nurture Assumption: Why Children pp. 1-13. University of Florida, Gainesville. Turn Out the Way They Do. Touchstone, New York. Berreman, G. D. (1972) Hindus oJ the Himalayas: Henrich, j. (2004) Demography and cultural evolution: and Change. University of California why adaptive cultural processes produced maladaptive Press, Berkeley. losses in Tasmania. American Antiquity, 69: 197-214. Boyd, D. (2001) without pigs: recent subsistence Henrich, j. and Boyd, R. (1998) The evolution of conformist changes among the lrilia Awa, Papua New Guinea. transmission and the emergence of between-group Human , 29: 259-281. differences. Evolution and Humall Behavior, 19: 215-242. Boyd, R. and Richerson, P. j. (1985) Cul/ure and the Henrich, j. and Gil-White, F. (2001) The evolution of EvoLutumary Process. University of Chicago Press, prestige: freely conferred deference as a mechanism Chicago. for enhancing the benefits of cultural transmission. Boyd, R. and Richerson, P. j. (1988) An evolutionary Evolution and , 22: 165-196. model of social learning: the effects of spatial and Henrich, N. S. and Henrich, j. (in press) Why Humans temporal variation. [n T. R. Zentall and B. G. Galef Cooperate: A Cultural and Evolutionary Explanation. (eds) Social Learning: Psychological alld Biological Oxford University Press, Oxford. Perspectives, pp. 29-48. Lawrence Erlbaum, Hillsdale, Nj. Henrich, j. and McElreath, R. (2002) Are peasants risk­ Boyd, R. and Richerson, P. j. (1995) Why does culture averse decision makers? Current Anthropology, 43: 172-181. increase adaptability? and , 16: Henrich, j., Boyd, R., Bowles, S., Camerer, c., Fehr, E. and 125-143. Gintis, H. (eds) (2004) Foundations ojHuman : Boyer, P. (2001) Religion Explained: The Evolutionary Economic Experiments and Ethllographic Evidence From Origins ojReligious Thought. New York, Basic Books. Fifteen Small-Scale Societies. Oxford, Oxford University Carpenter, J. (2004) When in Rome: conformity and the Press. provision of public goods. Journal ojSocio-Economics, Hill, K. and Hurtado, A. M. (1996) Ache LiJe History. 4: 395--408. Aldine de Gruyter, New York. 570 . CHAPTER 38 Dual-inheritance theory

Horner, V. and Whiten, A. (2005) Causal knowledge Pierre, P., Lerner, C, Izard, V. and Dehaene, S. (2004) Exact and imitation/emulation switching in chimpanzees and approximate arithmetic in an Amazonian Indigene (Pan trogiodytes) and children (Homo sapiens). group. Science, 306: 499-503. Animal Cognition, 8: 164-181. Pingle, M. (1995) Imitation vs. rationality: an experimental Jackson, J. E. (1983) The Fish People: Linguistic Exogamy perspective on decision-making. Journal ofSocio­ and Tukanoan Identity in Northwest Amazonia. Economics, 24: 281-315. Cambridge University Press, Cambridge. Plomin, R., Defries, J. and McLearn, G. E. (2000) Behavioral Jonas, K. (1992) Modeling and suicide: a test of the Werther Genetics. W. H. Freeman, New York. effect. British JOllrnal ofSocial Psychology, 31: 295--306. Rao, V. (2001) Poverty and public celebrations in rural Kameda, T. and Nakanishi, D. (2002) Cost-benefit analysis India. Annals ofthe American Academy ofPolitical and of social/cultural learning in a non-stationary uncertain Social Science, 573: 85-104. environment: an evolutionary simulation and an Richerson, P. and Boyd, R. (2005) Not by Genes Alone: experiment with human subjects. Evolution and How ClIlture Transformed Human Evolution. Human Behavior, 23: 373-393. University ofChicago Press, Chicago. Kellogg, W. and Kellogg, L. (1933) The Ape and the Child: Roberts, G. and Sherratt, T. (2002) Does similarity breed A Stlldy ofEnvironmental Influence Upon Early Behavior. cooperation? Nature, 418: 499-500. McGraw-Hill, New York. Rogers, E. M. (1995) Diffusion ofInnovations. Free Press, Kessler, R. C, Downey, G. and Stipp, H. (1988) Clustering New York. of teenage suicides after television news stories about Rubinstein, D. H. (1983) Epidemic suicide among suicide: a reconsideration. American Journal of Micronesian adolescents. Social Science Medicine, Psychiatry, 145: 1379-83. 17: 657--665. Kroll, Y, and Levy, H. (1992) Further tests of the separation Segall, M., Campbell, D. and Herskovits, M. J. (1966) The theorem and the capital asset pricing model. American Influence ofCulture on Visllal Perception. Bobbs-Merrill, Economic Review, 82: 664-670. New York. Labov, W. (1972) Sociolinguistic Patterns. Philadelphia, Shennan, S. (2001) Demography and cultural innovation: University of Pennsylvania Press. a model and its implications for the emergence of Labov, W. (1980) The social origins of sound change. modern human culture. Cambridge Archaeology In W. Labov (ed.) Locating Language in Time JOllrnal, II: 5-16. and Space, pp. 251-265. Academic Press, New York. Stack, S. (1987) Celebrities and suicide: a and Laland, K. N., Odling-Smee, f., and Feldman, M.W. (2000) analysis, 1948-1983. American Sociological Review, , biological evolution, and cultural 52: 401-412. change. Behavioral and Brain ScienJ:es, 23: 131-175. Stack, S. (I990) Divorce, suicide, and the mass media: leVine, R. A. and Campbell, D. T. (1972) . an analysis of differential identification, 1948-1980. Wiley, New York. Journal ofMarriage and the Family, 52: 553-560. Levinson, S. C (2003) Space in Language and Cognition. Stack, S. (I 996) The effect of the media on suicide: Cambridge University Press, Cambridge. evidence from Japan, 1955-1985. Suicide and McBrearty, S. and Brooks, A. (2000) The revolution that Life- Threatening Behavior, 26: 132-142. wasn't: a new interpretation of the origin of modern Thornhill, R., Gangestad, S. w., Miller, R., Scheyd, G., human behavior. Journal ofHlIman Evolution, 39: McCollough, G. and Franldin, M. (2003) Major 453-563. histocompatibility complex genes, symmetry, McElreath, R., Boyd, R. and Richerson, P. J. (2003) Shared and body scent attractiveness in men and women. nOrms and the evolution of ethnic markers. ClIrrent , 15: 668--678. Anthropology, 44: 122-129. Tomasello, M. (1999) The Cullural Origins ofHlIman McElreath, R., Lubell, M., Richerson, P. et al. (2005) Cognition. Harvard University Press, Cambridge, MA. Applying evolutionary models to the laboratory study Turchin, P. (2003) Historical Dynamics: Why States Rise of social learning. Evolution and Human Behavior, and Fall. Princeton University Press, Princeton, NJ. 26: 483-508. Wasserman, I. M., Stack, S. and Reeves, J. L. (1994) Suicide Mullen, B., Cooper, C and Driskell,). E. (1990) Jaywalking and the media: The New York Times's presentation of as a function of model behavior. Personality and Social front-page suicide stories between 1910 and 1920. Psychology Bulletin, 16: 320-330. JOllrnal ofCommunicalion, 44: 64-83. Nave, A. (2000) Marriage and the maintenance of ethnic Weibull, J. W. (1995) . group boundaries: the case of Mauritius. Ethnic and MIT Press, Cambridge, MA. Racial Stlldies, 23: 329-52. Weiss, H. (1977) Subordinate imitation of supervisor Nettle, D. and Dunbar, R. (1997) Social markers and the behavior: the role of modeling in organizational evolution of reciprocal exchange. Current Anthropology, socialization. Organizational Behavior and Human 38: 93-99. Decision Processes, 19: 89-105. Nisbett, R. E. (2003) The Geography of Thought: Weiss, H. M., Suckow, K. and Rakestraw, T. L. (1999) How Asians and Westerners Think Differently .. Influence of modeling on self-set goals: direct and and Why. Free Press, New York. mediated effects. Human Performance, 12: 89-114. Offerman, T. and Sonnemans, J. (1998) Learning by Whiten, A., Horner, V. and de Waal, F. B. M. (2005) experience and learning by imitating others. JOllrnal of Conformity to cultural norms of tool use in Economic Behavior and Organization, 34: 559-575. chimpanzees. Nature, 437(7059): 737-740.