Evolutionary Anthropology 123

ARTICLES

The Evolution of Cultural Evolution

JOSEPH HENRICH AND RICHARD McELREATH

Humans are unique in their range of environments and in the nature and diversity of attempted to glean as much as they their behavioral adaptations. While a variety of local genetic adaptations exist within could from the aboriginals about nar- our species, it seems certain that the same basic genetic endowment produces arctic doo, an aquatic fern bearing spores foraging, tropical horticulture, and desert pastoralism, a constellation that represents they had observed the aboriginals us- a greater range of subsistence behavior than the rest of the Primate Order combined. ing to make bread. Despite traveling The behavioral adaptations that explain the immense success of our species are along a creek and receiving frequent cultural in the sense that they are transmitted among individuals by social learning and gifts of fish from the locals, they were have accumulated over generations. Understanding how and when such culturally unable to figure out how to catch evolved adaptations arise requires understanding of both the evolution of the psycho- them. Two months after departing logical mechanisms that underlie human social learning and the evolutionary (popu- from their base camp, the threesome lation) dynamics of cultural systems. had become entirely dependent on nardoo bread and occasional gifts of fish from the locals. Despite consum- In 1860, aiming to be the first Euro- three men (King, Wills and Gray) ing what seemed to be sufficient calo- peans to travel south to north across from their base camp in Cooper’s ries, all three became increasingly fa- Australia, Robert Burke led an ex- Creek in central Australia with five tigued and suffered from painful tremely well-equipped expedition of fully loaded camels (specially im- bowel movements. Burke and Wills ported) and one horse. Figuring a soon died, poisoned and starved from maximum round trip travel time of eating improperly processed nardoo three months, they carried twelve seeds. Unbeknown to these intrepid weeks of food and supplies. Eight adventurers, nardoo seeds are toxic Joseph Henrich received his Ph.D. in and highly indigestible if not properly 1999 from the University of California, Los weeks later they reached tidal swamps Angeles, and is currently Assistant Pro- on the northern coast and began their processed. The local aboriginals, of fessor of Anthropology at Emory University. return. After about ten weeks their course, possess specialized methods He was recently a fellow in the Society of Scholars at the University of Michigan and supplies ran short and they began eat- for detoxifying and processing these at the Institute for Advanced Study in Berlin. ing their pack animals. After twelve seeds. Fatigued and delusional, King He has conducted ethnographic and exper- wandered off into the desert where he imental research among the Machiguenga weeks in the bush, Gray died of illness of Peru and the Mapuche of southern Chile. and exhaustion, and the group jetti- was rescued by an aboriginal group, His theoretical work has involved construct- soned most of their remaining sup- the Yantruwanta. He recovered and ing formal models of the evolution of cul- tural learning capacities, cultural evolution, plies. A month later, they arrived back lived with the Yantruwanta for several and culture-gene coevolution. E-mail: in their base camp, but found that months until a search party found [email protected] their support crew had recently de- him. Richard McElreath received his Ph.D. in 2001 from the University of California, parted, leaving only limited supplies. The planning for this expedition Los Angeles, and is now Assistant Pro- Still weak, the threesome packed the could not have been more extensive, fessor of Anthropology at the University of California. He was recently a postdoc- available supplies and headed to the and these men were not unprepared toral fellow at the Center for Adaptive Be- nearest outpost of “civilization,” Mt. British schoolboys out on holiday. havior and Cognition, Max Planck Insti- Hopeless, 240km south. In less than a However, despite their big brains, tute for Human Development, Berlin. He conducts ongoing field work to investi- month, their clothing and boots were camels, specialized equipment, train- gate cultural microevolution among sev- beyond repair, their supplies were ing, and seven months of exposure to eral ethnic groups in southwest Tanzania. the desert environment prior to run- E-mail: [email protected] again gone, and they ate mostly camel meat. ning out of supplies, they failed to sur- Faced with living off the land, they vive in the Australian desert. This bit Key words: social learning; human evolution; culture began foraging efforts and tried, un- of history makes a simple point: Hu- and cognition; coevolution; dual inheritance theory successfully, to devise means to trap mans, unlike other animals, are birds and rats. They were impressed heavily reliant on social learning to Evolutionary Anthropology 12:123–135 (2003) by the bountiful bread and fish avail- acquire large and important portions DOI 10.1002/evan.10110 Published online in Wiley InterScience able in aboriginal camps, in contrast of their behavioral repertoire. No (www.interscience.wiley.com). to their own wretched condition. They evolved cognitive modules, “evoked 124 Evolutionary Anthropology ARTICLES culture,” or generalized cost-benefit things, how different modes of cul- Conceptualizing culture as socially calculators delivered to these men the tural inheritance affect rates and out- learned information stored in people’s knowledge of how to detoxify nardoo comes of cultural evolution2 and how brains opens up new sets of evolution- spores or how to make and use rat natural selection acting on genes can ary questions. We will review the re- traps, bird snares, or fishing nets from produce a semi-autonomous inheri- search on five of these: How does so- locally available materials. Unlike so- tance system.7 Like human behavioral cial learning in humans increase cial learning in other animals, human ecology,8 coevolutionary and dual-in- adaptability and thereby allow our cultural abilities generate adaptive heritance theories are concerned with species to successfully occupy such an strategies and bodies of knowledge adaptation. Unlike human behavioral enormous range of environments? If that accumulate over generations. ecology, however, these theories cultural learning mechanisms are so Foraging, as it is known ethnographi- model the proximate mechanisms adaptive, why are such mechanisms cally, would be impossible without that produce adaptations. Like evolu- seemingly rare in nature? What cog- technologies such as kayaks, blow- tionary psychology, these theories nitive processes guide human social guns, bone tools, boomerangs, and share an interest in the design of cog- learning? If cultural variants do not bows. These technological examples nition. Unlike most evolutionary psy- replicate like genes, can culture embody skills and know-how that no chology, however, dual-inheritance evolve? How does the coevolution of single individual could figure out in and gene-culture models are rigor- genes and culture influence human his lifetime. Nonmaterial culture, ously formalized, take account of so- psychology and the histories of hu- such as seed processing techniques, cial learning, and explore population man societies? These five questions tracking abilities, and medicinal plant processes. For many questions, build a natural progression of puzzles, knowledge, reveals similar locally strictly outcome-oriented or culture- from the genetic origins of cultural adaptive accumulations. Interest- free models are sufficient and insight- inheritance to the dynamics of mod- ingly, this adaptive information is of- ful. For many others, however, taking ern cultural and societal evolution. ten embodied in socially learned account of cultural dynamics is essen- rules, techniques, and heuristics that tial. As the Burke and Wills story illus- are applied with little or no under- trates, even hunter-gatherer adapta- WHY IS CULTURAL LEARNING standing of how or why they work. tion is substantially reliant upon ADAPTIVE? Thus, understanding a substantial evolved cultural knowledge and tech- To understand the evolution of so- amount of human adaptation requires nology. To understand adaptation in cial learning, theorists have developed understanding the cultural learning human societies with any time depth formal models to study how temporally processes that assemble our behav- seems very difficult without some at- and spatially changing environmen- ioral repertoires over generations. tempt to account for the evolutionary tal conditions affect the evolutionary This is not, however, a call to separate dynamics that produce such adapta- trade-offs between capacities for in- humans from the rest of nature. A tions. dividual learning (for example, trial productive approach should seat hu- Throughout this paper we will use and error), social learning, and “hard- mans within the broader context of “cultural learning” and “cultural wired” behavioral responses.1,5,7,9–15 mammalian and primate evolution transmission/acquisition” to refer to Most of these models are very abstract while at the same time being able to the subset of social learning capacities and apply to a wide range of animal explain how and why humans are so that allow for cumulative cultural evo- social learning, not just human cul- different in the diversity and nature of lution. We use “culture” to refer to the tural transmission. They show that so- their behavioral adaptations. Our goal information acquired by individuals cial learning is favored throughout a in this paper is to review recent devel- via social learning. Processing nardoo large intermediate range of environ- opments in understanding both the and making arrow poison, for exam- mental fluctuation, especially when evolution of the psychological mecha- ple, are cultural practices because in- environments are highly autocorre- nisms that make cultural evolution dividuals learn them from other mem- lated. The intuition behind these re- possible and the population-level con- bers of their social group. The mental sults is that social learning allows or- sequences of those individually adap- representations that allow individuals ganisms to respond more quickly to tive mechanisms. Most of the relevant to detoxify the fern spores or bring environmental changes than do hard- work occurs within a pair of closely down large game with relatively light- wired responses, but only by exploit- related approaches: gene-culture co- weight bows and arrows do not come ing a body of adaptive knowledge that evolution1–5 and dual-inheritance the- coded in their genes, nor are these is stored in the learned behavioral rep- ory.6,7 These approaches examine the continually relearned by each individ- ertoire of the population. At one ex- interactions between genetic and cul- ual via trial-and-error experimenta- treme, when environments fluctuate tural inheritance systems. In these tion or deduced solely by fitness-ori- on the order of thousands of genera- models, individual phenotypes are ented cost-benefit analysis. Instead, tions, social learning serves no pur- combinations of both genetic and so- such adaptations result from and em- pose since raw natural selection act- cially transmitted characters, which body the cumulative effects of the ef- ing on genes can, on average, do just in turn affect the transmission rates of forts, experiments, errors, insight, and as well without paying for expensive different alleles and cultural variants. interactions of many individuals social learning machinery. At the Early models explored, among other across generations. other extreme, when fluctuations oc- ARTICLES Evolutionary Anthropology 125 cur on the order of single generations, tropical primate to spread so rapidly ous spares herself those potential there is little adaptive knowledge for and successfully into so many habi- costs, provided that the behavior of social learners to exploit. However, tats—from the dry savannahs and others is adaptive. with intermediate rates of change, on tropical forests of equatorial Africa to However, Rogers showed that this the order of tens or hundreds of gen- the Arctic tundra and humid swamps argument is insufficient to explain the erations, social learning mechanisms of New Guinea—while most other adaptive success of our cultural spe- both outpace genetic adaptation and mammals with plausibly well-devel- cies. Using a very simple model, he have sufficient time between environ- oped social learning abilities show proved that sparing individuals the mental changes to accumulate a body comparably restricted ranges? Prior costs of individual learning will not, of adaptive knowledge in the popula- to a clever paper by Rogers,23 several on its own, lead to increased overall tion. researchers had argued that social adaptability in the population—the When viewed alongside a growing learning improves human adaptabil- mean fitness of the population is not pool of empirical evidence, this theo- ity by exempting individuals from the increased. While social learners do retical work suggests that both indi- costs of individual learning.5,7,24 The very well when they are rare, they do vidual and social learning from an in- argument seems cogent enough: Time poorly when they are common. With- tertwined adaptive response to out any individual learners, social increasing amounts of environmental learners cannot track changes in the variability16,17—what Potts18 calls Without any individual environment, and the first individual variability selection. First, there is learner entering a group of social new evidence that increases in brain learners, social learners learners always has higher fitness size relative to body size are corre- cannot track changes in than the others. This means that at lated with both social and individual equilibrium the mean fitness of the learning abilities across species. In the environment, and population as a whole is the same as primates, brain size corrected for the first individual that in a population of purely individ- body size correlates most strongly ual learners. Social learning alone with social learning abilities, but also learner entering a group does not increase adaptability. Box 1 with individual learning (“innova- of social learners always explains this mathematical argument tion”) and tool use, all three of which has higher fitness than in more detail. Boyd and Richerson11 are highly intercorrelated.19 As far as extended Rogers’ result to more com- we know, no similar studies exist for the others. This means plicated models in which social learn- mammals in general, although there that at equilibrium the ers can identify and preferentially are similar findings for birds.20,21 Sec- copy individual learners, the environ- ond, these data suggest that increases mean fitness of the ment varies spatially as well as tempo- in brain size in the paleontological population as a whole is rally, imitation generates errors, and record have been partly driven by in- there are more than two behaviors. creases in social learning abilities. the same as that in a None of these changes alter the result Right up to the present, the record population of purely that the evolution of social learning shows that several mammalian lin- individual learners. does not lead to a more fit population. eages have undergone increases in Cultural capacities, as represented in brain size relative to body size. Fi- Social learning alone these models, do not raise the overall nally, over the same period, ice-core does not increase fitness of the population, so they are data show increasing degrees of cli- unlikely to explain the adaptive success matic variation: Over the last fourteen adaptability. of our species in the last 200,000 years. million years, which is the limit of the In the same paper, however, Boyd time depth of the data, increases in and Richerson11 showed that social climatic variability are mirrored by learning can lead to higher mean fit- increases in brain ratio. This combi- ness provided either that it allows the costs and potential mistakes can make nation of evidence, alongside the for- accumulation of behaviors that no in- individual learning quite expensive. If mal theory that independently impli- dividual learner could acquire in its another individual or group of indi- cates environmental variation with lifetime or improves the efficiency of increases in social learning abilities, viduals has already paid those costs, individual learning. When either is the suggests that human cultural capaci- learning from that behavior may be case, social learning may increase the ties may be a hypertrophied subset of considerably cheaper. Imagine the mean fitness of the population. The a larger class of learning abilities that task of selecting among mushroom first condition is in fact the question have evolved in many species.22 varieties through individual learning. we started the paper with, and we will Yet humans stand out in the num- Because some mushrooms are poi- discuss it at length in the next section. ber and diversity of environments sonous, the price of choosing the The second condition is satisfied if they inhabit. What is the role of social wrong mushroom is quite high. How- learners use individual learning when learning in human adaptability, and ever, an individual who learns from it is cheap and reliable, and switch to how have these abilities permitted a others which mushrooms are poison- social learning when individual learn- 126 Evolutionary Anthropology ARTICLES

Box 1. Mathematical Argument

As we explain in the text, Rogers’23 model demonstrates that social learning alone will not increase the average fitness of a population of cul- tural organisms. A simple graph can make the argument much clearer. Figure (a) plots the fitnesses of in-

dividual learners (wl, dark line) and social learners (ws, thin line), as well as the population mean fitness (w៮ , dotted line), for all frequencies of so- cial learning in the population (p). When social learners are rare, they do significantly better than individual learners, since most potential models the same as the average population crease as the frequency of social are practicing the correct behavior. fitness when p ϭ 0, when there are learning increases. Now the popula- As social learning becomes more no social learners. tion mean fitness w˜ is greater than the common, however, the population But if we allow the frequency of mean fitness in a population of indi-

lags behind the environment more social learning to reduce the costs of vidual learners, w0. In order for social and more until individual learning individual learning, a new equilibrium learning to increase adaptability, the pays just as well as social learning, at arises at which the population mean mean fitness of the population, it p˜ . Natural selection will stabilize the fitness is greater than that of a popu- must also somehow increase the fit- population at this equilibrium, at lation of individual learners. Figure (b) ness of individual learning. In the text, which both social and individual shows the modified model in which we discuss plausible mechanisms for learners receive fitness w˜ , which is the costs of individual learning de- this effect.

ing is expensive.7,14 We think both are ary puzzle: Why haven’t the social necessary for cumulative cultural evo- at work in human cognition. However, learning capacities that generate cu- lution. Other kinds of social learning the adaptive gains possible through the mulative cultural adaptations repeat- may lead to traditions, but not to the second mechanism alone seem modest edly evolved along with other individ- accumulation of adaptive informa- in comparison to those produced by cu- ual and social learning abilities in tion. Imagine that individuals are ca- mulative cultural evolution. many mammalian lineages over the pable of a modest amount of individ- last fourteen million years? Thus, here ual learning, so that interaction with WHY ARE CAPACITIES FOR we attempt to explain why human-like the environment slowly generates CUMULATIVE CULTURAL cultural capacities should be rare in adaptive behavior. If naive individuals nature, as we believe they are, despite tend to hand around other individu- TRANSMISSION RARE? being extremely adaptive. als, and some of these individuals pre- Several of our colleagues are fond of While an increasing amount of field fer to hang around certain kinds of the “Why not baboons?” stratagem: If evidence suggests that other animals, food sources, because they have indi- an evolutionary scenario is meant to particularly chimpanzees, may main- vidually learned how to exploit those explain some unique (or at least tain traditions that result from social food sources (for example, cracking nearly unique) feature of humans, learning,25–28 there is little reason to nuts or termiting), then naive individ- then it must also be able to explain believe that nonhuman social learning uals would be more likely to devise a why baboons—and many other ani- capacities can generate cumulative means to exploit that resource. This mals—do not fall under the same evo- adaptation.7,29,30 In contrast, accumu- would be social learning, but since in- lutionary logic. We have seen many lated cultural skills and knowledge are dividuals have to reinvent the details clever theories crumble before this in- characteristics of all human societies. of the behavior for themselves, albeit terrogation. The story we outlined While the psychological mechanisms accelerated by proximity to conspecif- earlier is vulnerable to the same criti- that make cumulative culture possible ics, the behavior cannot become more cism. Although human cultural capac- are unclear, there are some promising complex across generations beyond a ities can be seen as part of a more ideas. Tomasello, Kruger, and Rat- certain point. Naive individuals do not general pattern of adaptation for ner29 suggested that true imitation, or get a “head start,” and thus cannot learning in variable environments, observational learning—the direct begin where previous learners left off. their immense adaptiveness and ap- and accurate copying of behaviors, If, instead, individuals acquire their parent uniqueness poses an evolution- strategies or symbolic knowledge—is behavior by directly observing and ARTICLES Evolutionary Anthropology 127 copying the details of others’ tech- netic influences and cultural trans- be learned on one’s own. But, despite niques, then individual learning can mission. Other genes affect an indi- being difficult to get started, once a build atop previous innovations.31 A vidual’s reliance on imitation, but reliance on cultural learning is com- version of this distinction that allows carry an incremental fitness cost. All mon in the population, it is easy to for more continuity with chimpanzees individuals engage in some individual sustain. Provided that the environ- would be that chimps possess modest learning, which moves their pheno- ment is not too variable, the rate of true imitative capacities, but the com- types a small amount toward the cur- accumulation of adaptive behavior plexity of the skills and technologies rent optimum. But individuals with a through cultural learning can easily they can represent and the fidelity of substantial reliance on cultural learn- pay for the cost of the psychological capacities needed to make it possible: their transmission is less than that of ing can acquire phenotypes much Cultural learning mechanisms pro- humans. closer to the optimum, once such phe- vide access to the knowledge accumu- True imitation is probably not the notypes exist in the population. These lated over generations that simple so- whole story, however, at least not in phenotypes are then improved a small the long run. In modern humans, a cial learning does not. However, amount by individual learning. This suite of social learning abilities con- because cultural capacities are not fa- process repeats every generation. tributes to the maintenance and accu- vored when rare, we should not expect mulation of culture. Simpler forms of them to be widespread in nature. A population must traverse a fitness val- observational learning (of physical While an increasing skills, for example) likely provided a ley before the frequency of true imita- foundation for more complex kinds of amount of field tion is high enough to make it individ- social learning and inference, such as evidence suggests that ually advantageous. Because other those associated with symbolic com- forms of social learning are often built munication and language. Symbolic other animals, principally out of individual learning, communication through proverbs, particularly and do not involve inferential recon- stories, and myths allows for a great structions of behaviors and strategies, deal of cultural transmission without chimpanzees, may they do not face this dilemma—but “observation” in the usual sense.31,32 maintain traditions that they also cannot generate cumulative For example, !Kung hunters knew a cultural adaptation. great deal of natural history, includ- result from social Having offered an explanation of ing the fact that porcupines are mo- learning, there is little why cumulative cultural abilities might be rare in nature, we are left nogamous.33 It is hard to imagine that reason to believe that knowledge of this kind is preserved with the question of why it was spe- through observational learning alone. nonhuman social cifically the human ancestral lineage that crossed the cultural threshold. However, Tomasello30 argues that learning capacities can One possibility is that ancestral hu- true imitation, rooted in a genetically mans just happened to drift geneti- evolved capacity for Theory of Mind, generate cumulative cally across the threshold. Random generates both linguistic and nonlin- adaptation. In contrast, events of this kind were likely impor- guistic forms of cultural evolution, tant in the evolutionary histories of and that linguistic symbols (including accumulated cultural many species. However, we think it is grammatical structures) have gradu- skills and knowledge more productive to ask if there was ally accumulated, improved, and are characteristics of all something particular about the hu- adapted through a cultural evolution- man lineage that made it more likely ary process analogous to that ob- human societies. than other species to cross this cul- served in the domain of material cul- tural threshold. Perhaps our evolving ture and technology. cultural capacities depended first Whatever the specific nature of the upon some other adaptation, which mechanisms—be they true imitation might have arisen for another reason or not—it remains puzzling why they Unlike the simpler social learning models discussed in the previous sec- entirely.34 Good answers here are should be so rare. Boyd and Richer- probably a long way off, but specula- son34 constructed a model of the evo- tion, this work demonstrates that a substantial reliance on cultural learn- tion based on the existing information lution of cumulative cultural capaci- will help direct future research. ties designed to explore this puzzle. In ing is unlikely to spread initially, but their model, a population lives in a goes to fixation and is stable once a variable environment in which there critical threshold frequency is sur- WHAT COGNITIVE mounted (Box 2). Natural selection fa- is a unique optimal adaptive value of a MECHANISMS GUIDE CULTURAL quantitative trait. Each generation, vors cultural learning only when the EVOLUTION? there is some probability that the en- costs of developing and maintaining vironment changes so that a new cultural learning mechanisms are Like evolutionary psychology, dual- value of the trait is optimal. Individual smaller than the benefits gained by inheritance theory combines evolu- phenotypes are a combination of ge- acquiring simple behaviors that could tionary theory with empirically 128 Evolutionary Anthropology ARTICLES

Box 2. Cultural Learning

Regions for which cultural learning has an advantage over individual learn- ing, for the Boyd and Richerson34 model. The curves show the internal unstable equilibrium of cultural learners versus individual learners and repre- sent the threshold frequency at which cultural learning becomes favored by selection. This is shown for two values of ␥, the probability that the environ- ment changes each generation and renders a new behavior adaptive. In each case, culture learners will in- crease in the region above the curve and decrease below it. When individual learning is difficult (left side of plot), cultural learning cannot invade the population be- cause too few individuals have the skilled behavior. When cultural learn- ing is rare, the only behaviors that exist in the population are those that are solely the result of individual learning (those that could be figured out by one individual in his lifetime). The cost (for example, in adding met- tive and inferential abilities, as well as essentially everyone acquires highly abolically costly brain tissue) does a strong desire to imitate70). adaptive behavior without paying the not easily outweigh the benefit be- As we move toward slightly easier additional costs of cultural learning cause there is little adaptive informa- problems, cultural learning still can- capacities, so cultural learning is tion contained in the behavior of oth- not invade, but is stable once com- rarely adaptive. ers that the animal cannot figure out mon. Once cultural capacities are In all cases, there are many more on its own. It is important to keep in common, population processes will combinations of parameters for mind that the developmental or fit- begin to assemble complex adapta- which cultural learning is stable when ness costs of true imitation mecha- tions, and individuals who have the common, but cannot invade the pop- nisms may be quite high, even ability to acquire them will do sub- ulation when rare. In the case where though, from our human perspective, stantially better than those who can- ␥ϭ0.4, implying that the optimal be- imitation itself strikes us as being not. Under these conditions, cultural havior changes in 40% of genera- quite easy. It is “easy” because our learning does for genes what they tions, there is no difficulty of individ- cognition is “designed for” imitation cannot do directly for themselves. ual learning for which cultural learning and social learning (for example, in- Looking at the far right side of the invades, but a wide range for which fants show both sophisticated imita- plot, where individual learning is easy, cultural learning is stable.

grounded assumptions about the en- offs between acquiring accurate be- man cultural psychology. When infor- vironments inhabited by ancestral hu- havioral information at high cost and mation is costly, natural selection will man populations to make predic- obtaining less accurate information at favor cognitive mechanisms that al- tionsa about the details of human low cost. When accurate information low individuals to extract adaptive in- psychology—details that often specify is unavailable or too costly, individu- formation, strategies, practices, heu- cognitive mechanisms people use to als may exploit the information stored ristics, and beliefs from other extract adaptive ideas, beliefs, and in the behavior and experience of members of their social group at a practices from their social environ- other members of their social group. lower cost than through alternative ments. However, the approach di- By exploring how the costly infor- individual mechanisms. Human cog- verges from mainstream evolutionary mation hypothesis generates trade- nition probably contains numerous psychology in emphasizing the costly offs in the evolution of our cognitive heuristics and learning biases that fa- information hypothesis. This hypoth- capacities, we can generate produc- cilitate the acquisition of useful esis focuses on the evolutionary trade- tive theories about the details of hu- knowledge, practices, beliefs, and be- ARTICLES Evolutionary Anthropology 129

Box 3. Cultural Learning Mechanisms

Content biases and heuristics arise from the interaction of human psy- chology and the characteristics or “cues” associated with the thing be- ing transmitted (the idea, representa- tion, or behavior). These biases affect the likelihood of a particular mental representation being transmitted be- cause of the content of the represen- tation. Content effects can take many forms. They may reflect the direct ac- tion of natural selection on our “pre- pared learning” abilities, such as lan- guage, folk biology, and color categories. They may also arise as by-products of cognitive evolution: for either genetically transmitted imitated”) and make the ideas, men- Boyer’s38 approach to cultural phe- cognitive structures (as in Boyer’s tal representations, or behavior of nomena like ghosts and gods is one argument) or culturally acquired their possessor more likely to trans- example. Or they may emerge from mental representations. Context mit than those held by other individ- a kind of more generalized cost- heuristics arise from the learning uals. Other model-based biases benefit calculation: People prefer environment or context. Model- may include age, sex, ethnicity, and steel axes to stone axes because it based biases result from cues or healthful appearance. Frequency bi- is much less work to cut down trees characteristics of the potential ases use the commonality or rarity with steel. Such biases may result model (“an individual who may be of a behavior as a cue.

havior (“cultural traits” or “represen- cussion of them here. However, in exploit features of potential models or tations”). These mechanisms can be thinking about content biases, it is the frequencies of alternative behav- usefully modeled at the algorithmic important to keep in mind a number iors or strategies, rather than features level, much as some cognitive scien- of things. First, jury-rigged evolu- of the alternatives themselves, to tists investigate other kinds of infor- tionary products, like human minds, guide social learning. There is a great mation processing. are likely to contain accidental by- deal of adaptive information embod- Such cultural learning mecha- products and latent structures that ied in both who holds ideas and how nisms, all of which build atop other create biases for fitness-neutral be- common the ideas are. A large social and cultural learning abilities, haviors, ideas, beliefs, and val- amount of modeling effort has been can be categorized into content bi- ues.36,37 Boyer38 detailed one kind of expended in exploring the conditions ases and context biases. Box 3 orga- by-product content bias in his expla- under which different context biases nizes the various forms of cultural nation for the universality of reli- evolve and how strong natural selec- learning mechanisms. Content bi- gious concepts (like ghosts). Second, tion would prefer they be. These mod- ases, or what Boyd and Richerson7 even content biases that arose be- els derive from first principles about called direct biases, exploit informa- cause they led to the adoption of fit- how individual cognitive biases affect tive cues of an idea, belief, or behav- ness-enhancing behavior in ancient both individual fitness (when they ior itself, and thereby influence the environments may now promote the evolve) as well as the patterns of in- likelihood of imitation. An equiva- adoption of quite maladaptive prac- formation in the population (what lent perspective prefers to discuss tices. Third, content biases may be they evolve). Our remaining discus- cultural learning as adaptive infer- either reliably developing products sion of psychological mechanisms fo- ences triggered by content biases for of our species-shared genetic heri- cuses on two categories of context bi- cues provided in the behavior of oth- tage or they may be culture specific. ases in cultural learning: success and ers.35 Many such biases may have People may learn valuable content prestige bias and conformity bias. evolved because they facilitate the cues via cultural learning or, having acquisition of fitness-enhancing cul- acquired one idea or practice via cul- Success and Prestige Bias tural traits.2,4,7 Because content bi- tural transmission, may be more ases are likely numerous and gener- likely to acquire another because the If individuals vary in skills (for ex- ally confined to particular domains two “fit together” in some cognitive ample, tool making), strategies (track- of culture, for space considerations sense. ing techniques), or preferences (for we have omitted any substantial dis- Context biases, on the other hand, example, for foods) in ways that affect 130 Evolutionary Anthropology ARTICLES

fitness, and at least some components (See Henrich and Gil-White31 for a deference by using the amounts and of those differences can be acquired summary of the laboratory and field kinds of deference different models via cultural learning, then natural se- evidence.) receive as cues of underlying skill. As- lection may favor cognitive capacities However, an additional problem sessing differences in deference pro- that cause individuals to learn prefer- created by using indirect indicators of vides a best guess of the skill ranking entially from more successful individ- successful strategies is that it is often until more information can be accu- uals. The greater the variation in ac- very unclear which of an individual’s mulated. This also means that skilled quirable skills among individuals, and many traits have led to success. Are individuals will prefer deference dis- the more difficult those skills are to people successful because of how they plays that are easily recognized by acquire via individual learning, the tend their farms, cook their food, or others (in public). Thus, along with greater the pressure to preferentially make sacrifices to the spirits, or all the ethological patterns dictated by focus one’s attention on and imitate three? Because of this ambiguity, hu- the requirements for high fidelity so- the most skilled individuals. If indi- mans may have evolved the propen- cial learning (proximity and atten- viduals evaluate potential “cultural sity to copy successful individuals tion), deference displays also include models” (individuals they may learn across a wide range of cultural traits, diminutive body positions and socio- from) along dimensions associated only some of which may actually re- linguistic cues. The end point of this with competence in underlying skills late to the individuals’ success.7,31,43 If process gives us the psychology, soci- (such as hunting returns), and focus information is costly, it turns out that ology, and ethology of “prestige,” their social learning attention on this strategy will be favored by natural which must be distinguished from those who are more successful, they selection even though it may allow those associated with phylogeneti- will be more likely to acquire adaptive neutral and maladaptive traits to cally older “dominance” processes.31 strategies.31 Interestingly, while the hitchhike along with adaptive cultural From this theory, Henrich and Gil- ability to rank individuals by foraging traits. In a world of costly informa- White31 derived twelve predictions success is observed in nonhumans tion, cognitive adaptations do not al- about the interrelationships between (for better scrounging),39 there is no ways produce adaptive behavior from preferential imitation or influence, evidence that individuals in these spe- the point of view of genes, even in deference, and other ethological pat- cies acquire strategies from successful ancestral environments. Nevertheless, terns, individual characteristics (like foragers. With the rise of cultural ca- the theory does allow for predictions age and sex), and memory. A review of pacities in the human lineage, natural about the conditions under which data from psychology, economics, selection needed only to connect these maladaptive cultural traits will and ethnography turned up a sizable learning abilities with preexisting spread. amount of evidence consistent with ranking capacities. The evolution of a success bias may these predictions. A bias of this kind is a standard also be able to explain the formation assumption in evolutionary game the- of prestige hierarchies. Once success- Conformist Bias 40 ory, where a preference for copying biased transmission has spread It is unlikely that success and pres- the strategies of successful individuals through a population, highly skilled tige biases solve all costly information generates an evolutionary dynamic individuals will be at a premium, and problems, however. What do you do that is usually mathematically indis- social learners will need to compete when any observable differences in tinguishable from natural selection for access to the most skilled individ- success and prestige among individu- acting on genes. However, uncertainty uals. This creates a new selection als do not covary with the observable about the payoffs and success of other pressure on success-biased learners to differences in behavior? For example, individuals complicates success-bi- pay deference to those they assess as suppose everyone in your village uses ased learning. Schlag41,42 has ex- highly skilled (those judged most blowguns for hunting except one reg- plored the exact form that such an likely to possess adaptive informa- ular guy who uses a bow and arrow adaptive bias should take in the pres- tion) in exchange for preferred access and obtains fairly average hunting re- ence of noisy feedback about the suc- and assistance in learning. Deference turns. Do you adopt the bow or the cess of other individuals, finding that benefits may take many forms, includ- blowgun? a linear weighting of models by their ing coalitional support, gifts, general One solution for dealing with such observed payoffs may be more adap- assistance (house-building), and car- information-poor dilemmas is to copy tive than simply imitating the individ- ing for offspring.31 the behaviors, beliefs, and strategies ual with the highest observed payoff. With the spread of deference for of the majority.7,14 Termed confor- Another solution is for individuals to high skilled individuals, natural selec- mity bias, this mechanism allows in- use aggregate indirect measures of tion can take advantage of these ob- dividuals to aggregate information success, such as wealth, health, or servable patterns of deference to fur- over the behavior of many individuals. family size, which integrate over ther save on information-gathering Because these behaviors implicitly many instances and smooth out per- costs. Naive entrants (say immigrants contain the effects of each individual’s ceptual and stochastic errors. This or children), who lack detailed infor- experience and learning efforts, con- may explain the widespread observa- mation about the relative skill of po- formist transmission can be the best tion that people copy successful indi- tential cultural models, may take ad- route to adaptation in information- viduals, as defined by local standards. vantage of the existing pattern of poor environments. To see this, sup- ARTICLES Evolutionary Anthropology 131 pose every individual is given a noisy that they decrease their reliance on ory involved tools from population ge- signal (a piece of information) from conformist transmission after recent netics and theoretical evolutionary bi- the environment about what the best fluctuations or increase it after immi- ology, there are good reasons to practice is in the current circum- grating. examine the strength of the analogy stances. This information, for any one Work combining these models with between genes and “memes.” individual, might give them a 60% empirical investigations is growing. Dawkins, in The Extended Pheno- chance of noticing that blowguns Kameda and Nakanishi44 have further type,46 described what he saw to be bring back slightly larger returns than extended the Henrich and Boyd14 the necessary characteristics of any bows. Thus, using individual learning model to predict how human psychol- replicating entity: longevity, fecun- alone, individuals will adopt the more ogy should respond to changes in the dity, and fidelity. The structure of this efficient hunting practice with proba- cost of individual learning and de- argument has been used to support bility 0.6. But, if an individual sam- signed experiments to test their pre- the analogy between genetic and cul- ples the behavior of 10 other individ- dictions. By analyzing the temporal tural (or “memetic”) evolution: Cul- uals, and simply adopts the majority dynamics of historical cases of the dif- tural ideas can be replicators as well, behavior, his chances of adopting the fusion of innovations, Henrich45 has and hence culture may evolve as do superior blowgun technology increase found evidence that is consistent with populations of alleles. Some cognitive to 75%. a strong role for both conformity- and and evolutionary anthropologists, Obviously, if everyone uses only success-biased transmission and in- however, have severely criticized the conformist transmission, no adapta- power of this analogy, arguing that tion or cultural evolution occurs, but cultural ideas are rarely if ever repli- models of interaction among different Obviously, if everyone cated during social learning and that learning mechanisms indicate that culture is substantially transformed natural selection will very often favor uses only conformist by human psychology so that ideas a mix of social and individual learning transmission, no are rarely transmitted intact so there with a substantial reliance on confor- adaptation or cultural are no or few discrete units in cul- mity. Extending Boyd and Richer- ture.35,38,47,48 For these reasons, they son’s7 original model, Henrich and evolution occurs, but argue, cultural variants (“memes” or Boyd14 used simulation to investigate models of interaction “representations”) have little fidelity the interaction and coevolution of ver- and so cannot evolve in a Darwinian tical transmission (parent-offspring among different learning sense. Essentially, if cultural inheri- transmission), individual learning, mechanisms indicate tance involves continuously blending and conformist transmission in spa- (nondiscrete) traits and mutation-like tially and temporally varying environ- that natural selection will processes are powerful, memes will ments. These results confirm that con- very often favor a mix of not fulfill Dawkins’ requirements for a formist transmission is likely to evolve social and individual replicator. Without a replicator, the under a very wide range of conditions. argument goes, there can be no cul- In fact, these results show that the learning with a tural evolution. range of conditions that favor con- substantial reliance on These arguments should be taken formist transmission are wider than seriously. If culture is not an evolving those for vertical transmission alone, conformity. system in the Darwinian sense, then suggesting that if advanced social many coevolutionary theories (and, of learning via vertical transmission course, substantial portions of this pa- evolves at all, we should also expect per) require serious rethinking. Build- to observe a substantial conformist ing on the preceding points, Sper- consistent with a strong role for indi- bias. ber,35 Boyer,48 and Atran47 have vidual learning. We imagine future The model of the combination of argued that many existing models of work will illuminate the complex in- conformity bias with individual learn- cultural evolution are inappropriate, teractions among conformist and ing and vertical transmission leads to transmission cannot explain the per- other social learning biases in envi- three predictions: 1. Individuals will sistence of behavioral variation in hu- ronments in which the costs and qual- prefer conformist transmission over mans, and cultural evolution cannot ities of information vary. vertical transmission, assuming it is produce adaptations. If these argu- possible to access a range of cultural ments are correct, the story we told models at low cost, which is often, but IF CULTURAL VARIANTS DO earlier about culture accumulating not always the case; 2. As the accuracy NOT REPLICATE LIKE GENES, powerful locally adapted skills and of information acquired through indi- technologies is somehow mistaken. CAN CULTURE EVOLVE? vidual learning decreases, reliance on We think the arguments we re- conformist transmission over individ- So far, we have treated the inheri- viewed earlier are valid in this respect, ual learning will increase; 3. Individu- tance of cultural variants as unprob- however. There are good reasons to als should be sensitive to substantial lematic. However, because much of suppose that culture is an evolution- shifts in the relevant environments so the initial work in coevolutionary the- ary system, even if the three claims 132 Evolutionary Anthropology ARTICLES above are true. In two recent articles, either cultural inertia or diffusion of sion is inaccurate. In fact, nineteen of Henrich and coauthors49,50 use three successful variants. The third model the thirty-eight models presented in mathematical models and several combines all the potential problems their book are continuous (nondis- other lines of argument to show that with models of cultural evolution— crete) trait models that allow for an the objections mentioned here do not continuous (nondiscrete) mental arbitrary amount of transmission er- follow from their assumptions. representations, incomplete trans- ror. Similarly, Cavalli-Sforza and Through these analyses, the authors mission, and substantial inferential Feldman2 devoted one of their five demonstrate how Dawkins’ original transformations—and shows not chapters entirely to continuous trait claims about replicators and Darwin- only that adaptive cultural evolution models. These continuous models al- ian evolution were wrong—replica- may still occur under empirically low for substantial error and other tors are sufficient for cumulative evo- plausible conditions, but that it also forms of nonreplication. Similar to lution, but not necessary. predicts when such adaptive evolu- cognitivist critics, Boyd and Richer- In their first model, Henrich and tion will not occur. son also explicitly distinguish public Boyd49 address two complaints: that Many of the insights from these for- representations from mental repre- culturally transmitted ideas are rarely if mal models have been known for sentations (though using different ter- ever discrete and that inferential biases some time but, unlike Dawkins’ repli- minology) throughout their book, and in learning (Sperber’s “strong attrac- repeatedly specify the inferential tors”) swamp the effects of selective transformation between observed be- transmission and prevent Darwinian An understanding of havior and representation formed. adaptation. This model assumes that cultural evolution They also make explicit reference to individuals’ possess mental representa- much research in psychology on the tions (“cultural variants,” beliefs, and requires studying both nature of social learning and propose scripts) that are influenced by selec- the evolved cognitive the following pathway for the trans- tively learning from some individuals mission of cultural variants: Modeled 3 3 (for example, from successful individu- abilities and inferential events Attention Processes Re- 3 als). These mental representations are tention Processes Motor Reproduc- mechanisms that allow 3 3 continuous (nondiscrete or quantita- tion Motivation Processes tive), so each individual may possess a for cultural learning, as Matching. Chapters 4 and 5 in Boyd somewhat different variant of the rep- well as the population and Richerson’s book discuss how cognitive structures—what Sperber35 resentation. There are no “copies” of processes to which they variants, only social “influence.” Fur- would later call “attractors”—bias thermore, in learning these representa- give rise through social cultural change so that some out- tions, individuals use inferential pro- comes are more likely than others, interaction. Culture can and even use some of the same exam- cesses that strongly bias the final form ples as Boyer.48 of the representation. Their analysis have heritable The force of arguments like those of shows that these complaints are deduc- properties and evolve in Sperber, Atran and Boyer seems to be tively invalid. If cognitive inferential in- that cultural learning requires innate, fluences are sufficiently strong relative a Darwinian sense even domain-specific psychological mecha- to selective forces (selective learning), a if it is continuous, error- nisms (we agree!), and therefore that continuous (quantitative) model re- prone, and individually most of the action is in individual psy- duces to a discrete-trait replicator chologies and not in the population dy- model commonly used in population ephemeral. namics. This conclusion is unfounded: models of both culture and genes. In An understanding of cultural evolution fact, the stronger the effects of inferen- requires studying both the evolved cog- tial bias on learning, the better is the nitive abilities and inferential mecha- discrete trait approximation. Moreover, nisms that allow for cultural learning, this means that it is the weak effects of cator argument, have not successfully as well as the population processes to selective transmission that determine spread. While Sperber, Boyer and At- ran’s criticisms apply to the informal which they give rise through social in- the final equilibrium of the system. teraction. Culture can have heritable In the second and third models, the theorizing of some memeticists,46,51,52 they are wide of the mark for much properties and evolve in a Darwinian authors construct systems that allow sense even if it is continuous, error- formal gene-culture coevolutionary for large amounts of transmission error prone, and individually ephemeral. to show that accurate individual-level theory. Continuous trait models go replication of cultural variants is not back to the very beginning of the field. necessary for selective forces to gener- Boyd and Richerson7 argued in 1985 HOW DOES COEVOLUTION that there is no need to assume par- ate either cultural inertia or cumulative INFLUENCE PSYCHOLOGY AND cultural adaptation. The second model ticulate “units” in order to build evo- SOCIETY? shows how conformist transmission lutionary models, in fact showing that can act to drastically reduce the effect blending models best produce herita- A persistent debate in the social sci- of transmission errors and still generate ble variation exactly when transmis- ences is whether the chief causal level ARTICLES Evolutionary Anthropology 133 in social phenomena is the individual low actors to select individuals with vironment, will favor genes that reli- or the social. Instead of arguing that whom to cooperate.64,65 These efforts ably produce this bias since primary causation exists at either fail because, unless some process pre- individuals who prefer to interact level, gene-culture population models vents out-group members from adopt- with those with the same marker are take seriously and treat explicitly ing the same markers, individuals more likely to interact with someone forces at both levels, and sometimes who wear the markers but do not co- with the same norms as themselves, more. From this perspective, classic operate will destroy the signal value of and therefore profit more from social features of human cultures and soci- the symbols.66,67 So the question re- interaction. eties, such as culture being shared by mains: How do such markers arise The model also makes some unex- members of self-ascribed groups, be- and what are their functions? pected predictions about the nature of come results to derive, rather than a In addressing this puzzle, McEl- ethnic marking. While the model re- priori assumptions. These dialectical reath, Boyd, and Richerson54 con- quires spatial variation in norms to models have helped us to understand structed a model of the emergence of evolve the association between norms how interactions between cognition ethnic marking in which markers and markers, once markers are asso- and population processes give rise to function to provide coordination (not ciated with underlying norms, and ethnically marked groups53,54 and eth- cooperation, so there is no free-rider provided other processes permit a nic psychology,55 large-scale coopera- problem) with other individuals who tight linkage between them, spatial tion, prosocial psychologies, and group- share one’s norms. Coordination variation in norms is no longer beneficial cultural norms.7,16,56–60 means that individuals are better off needed to maintain functional ethnic Rather than attempting to summarize when they practice complementary markers. Instead, the “ethnic” groups this large literature, we focus only on behaviors. The familiar example of in the model merge, forming one large one of the most recent models. this occurs in cross-cultural commu- multiethnic community in which indi- nication,68 where different expecta- viduals still coordinate their interac- tions in many aspects of interaction tions based on markers delineating The Coevolution of Ethnically routinely lead to lower payoffs for all distinct ethnic divisions. Since mark- Marked Groups and Ethnic parties. The coaching book market for ers in such a situation allow individu- Psychology international business people attests als to assort nearly perfectly with oth- In almost all ethnographically to the severity of these problems. It is ers who share their norms, members known regions and historical periods, likely that the same phenomenon oc- of smaller norm communities are not humans have organized themselves curs in many other aspects of cultur- at a disadvantage relative to the nor- into self-ascribed groups marked by ally inherited behavior. Having the mative majority. Situations like this arbitrary symbols.61 For example, in same norms about child rearing, bar- resemble in an abstract way modern both historical and modern East Af- ter, marriage, inheritance, and con- multiethnic cities like Los Angeles or 69 rica, different pastoralist groups wear flict resolution can be crucial for suc- Detroit, in which many ethnic differently colored clothing, which cessful social relations. Since the groups live intermixed but preferen- serves as ethnic markers. In one re- number of domains of this kind is tially interact among themselves. gion of modern Tanzania, Maasai likely large and many such rules are The model makes predictions about wear red, Sukuma wear blue, and held unconsciously, the mutual costs both evolved psychological propensi- Taturu wear black. Since no other pri- of interactions between individuals ties and sociological patterns, and ex- mate forms such symbolically marked with different sets of norms can be plicitly links them. Ethnic marking groups, and existing rates of mixing substantial. arises as a side effect of other psycho- among such groups would quickly The model is sketched as follows. logical mechanisms—which them- erode differences of this kind if they First, imitation of the successful and selves have solid individual-level se- were transmitted from parent to off- social interaction produces culturally lective advantages—that happen to spring in any fashion (culturally or ge- differentiated communities. In each generate behaviorally distinct groups. netically), some explanation of their social group, whatever norm is ini- The strategy of using arbitrary sym- formation and maintenance is needed. tially most common leads to the high- bolic markers to choose interactants Prior efforts to explain ethnicity est payoffs, making it more common. then evolves because of features of the have proved theoretically unsound. Then, provided individuals are biased culturally evolved environment. Cul- First, the standard approach to ethnic to interact with people who share the tural transmission mechanisms may actors as strategic manipulators re- same arbitrary symbolic markers as create statistically reliable regularities quires that some other processes gen- themselves, symbolically marked in the selective environments faced by erate and maintain ethnic groups and groups that possess different cultural genes.4,57 Thus, explaining many im- their markings. If ethnicity were norms arise endogenously in the portant aspects of human psychology solely the product of strategic consid- model. Furthermore, even if there is and behavior will require examining eration or a coalitional psychology,62 initially no genetically transmitted how genes under the influence of nat- ethnicity would rapidly disappear as a psychological bias to interact with ural selection responded to the regu- phenomenon and there would not be other individuals who share your larities produced by culture. This anything to manipulate.54,63 A more same marker, natural selection, oper- means that understanding the behav- serious idea is that ethnic markers al- ating in this culturally constructed en- ior of a highly cultural species like 134 Evolutionary Anthropology ARTICLES humans will sometimes demand a cul- 14 Henrich J, Boyd R. 1998. The evolution of 36 Johnstone RA. 1994. Female preference for ture-gene coevolutionary approach. conformist transmission and the emergence of symmetrical males is a by-product of selection between-group differences. Evol Hum Behav 19: for mate recognition. Nature 372:172–175. Phenotypic optimality models and 215–242. 37 Enquist M, Johnstone RA. 1997. Generaliza- models that ignore the population dy- 15 Feldman MW, Aoki K, Kumm J. 1996. Indi- tion and the evolution of symmetry preferences. namics of social learning certainly vidual and social learning: evolutionary analysis Proc R Soc Lond B 264. have their place, and have proven very in a fluctuating environment. Anthropol Sci 104: 38 Boyer P. 1994. Naturalness of religious ideas: 209–232. a cognitive theory of religion. Berkeley: Univer- useful. But satisfying answers to 16 Richerson PJ, Boyd R. 1999. The evolutionary sity of California Press. many important questions concerning dynamics of a crude super organism. Hum Nat 39 Stammbach E. 1988. Group responses to spe- human behavior, from the cultural 10:253–289. cially skilled individuals in a Macacca fascicularis microevolution of foraging adapta- 17 Richerson PJ, Boyd R, Bettinger RL. 2001. group. Behaviour 107:241–266. tions to the coevolution of human psy- Was agriculture impossible during the Pleisto- 40 Gintis H. 2000. Game theory evolving. Prince- cene but mandatory during the Holocene? a cli- ton: Princeton University Press. chology and cultural variation, will re- mate change hypothesis. Am Antiquity 66:287– 41 Schlag KH. 1998. Why imitate, and if so, 411. main elusive unless dual inheritance how? J Econ Theory 78:130–156. 18 Potts R. 1998. Variability selection in hominid or some similar approach is taken se- 42 Schlag KH. 1999. Which one should I imitate? evolution. Evol Anthropol 7:81–96. riously. J Math Econ 31:493–522. 19 Reader SM, Laland KN. 2002. Social intelli- 43 Flinn MV, Alexander RD. 1982. Culture the- gence, innovation, and enhanced brain size in ory: the developing synthesis from biology. Hum primates. Proc Natl Acad Sci USA 99:4436–4441. ACKNOWLEDGMENTS Ecol 10:383–400. 20 Lefebvre L, Gaxiola A, Dawson S, Rozsa L, 44 Kameda T, Nakanishi D. 2002. Cost-benefit Kabai P. 1998. Feeding innovations and fore- analysis of social/cultural learning in a non-sta- This paper coevolved with valuable brain size in Australasian birds. Behaviour 135: tionary uncertain environment: an evolutionary feedback from Robert Boyd, Natalie 1077–1097. simulation and an experiment with human sub- Henrich, Pete Richerson, Eric Smith, 21 Lefebvre L, Whittle P, Lascaris E, Finkelstein jects. Evol Hum Behav 23:373–393. Peter Todd, Annika Wallin, and two A. 1996. Feeding innovations and forebrain size in birds. Anim Behav 53:549–560. 45 Henrich J. 2001. Cultural transmission and anonymous reviewers. the diffusion of innovations: adoption dynamics 22 Box HO, Gibson KR, editors. 1999. Mamma- indicate that biased cultural transmission is the lian social learning: comparative and ecological predominate force in behavioral change. Am An- perspectives. Cambridge: Cambridge University thropol 103:992–1013. REFERENCES Press. 46 Dawkins R. 1982. The extended phenotype. 23 Rogers AR. 1988. Does biology constrain cul- Oxford: Oxford University Press. 1 Cavalli-Sforza LL, Feldman MW. 1973. Cul- ture? Am Anthropol 90:819–831. tural versus biological inheritance: phenotypic 47 Atran S. 2002. The religious landscape. Cam- transmission from parent to children (a theory of 24 Lumsden C, Wilson EO. 1981. Genes, mind, bridge: Cambridge University Press. the effect of parental phenotypes on children’s and culture. Cambridge: Harvard University Press. 48 Boyer P. 1999. Cognitive tracks of cultural phenotype). Am J Hum Genet 25:618–637. inheritance: how evolved intuitive ontology gov- 2 Cavalli-Sforza LL, Feldman MW. 1981. Cul- 25 McGrew WC. 1992. Chimpanzee material cul- erns cultural transmission. Am Anthropol 100: tural transmission and evolution: a quantitative ture: implications for human evolution. Cam- 876–889. approach. Princeton: Princeton University Press. bridge: Cambridge University Press. 49 Henrich J, Boyd R. 2002. On modeling cogni- 3 Feldman MW, Laland KN. 1996. Gene-culture 26 Wrangham RW, McGrew WC, de Waal FBM, tion and culture: why replicators are not necessary coevolutionary theory. Trends Ecol Evol 11:453– Heltne P. 1994. Chimpanzee cultures. Cam- for cultural evolution. J Cogn Culture 2:87–112. 457. bridge: Harvard University Press. 50 Henrich J, Boyd R, Richerson PJ. n.d. Five 4 Durham WH. 1991. Coevolution: genes, cul- 27 Whiten A, Goodall J, McGrew WC, Nishida T, common mistakes in cultural evolution. In: Sper- ture, and human diversity. Stanford: Stanford Reynolds V, Sugiyama Y, Tutin CEG, Wrangham ber D, editor. Epidemiology of ideas: Open Court University Press. R, Boesch C. 1999. Cultures in chimpanzees. Na- Publishing. In press. 5 Pulliam HR, Dunford C. 1980. Programmed to ture 399:682–685. 51 Dennett D. 1995. Darwin’s dangerous idea. learn: an essay on the evolution of culture. New 28 Boesch C, Tomasello M. 1998. Chimpanzee London: Penguin Press. York: Columbia University Press. and human culture. Curr Anthropol 39:591–604. 6 Richerson PJ, Boyd R. 1976. A simple dual 52 Blackmore S. 1999. The meme machine. Ox- 29 Tomasello M, Kruger AC, Ratner HH. 1993. ford: Oxford University Press. inheritance model of the conflict between social Cultural learning. Behav Brain Sci 16:495–552. and biological evolution. Zygon 11:254–262. 53 Boyd R, Richerson PJ. 1987. The evolution of 30 Tomasello M. 2000. The cultural origins of ethnic markers. Cultural Anthropol 2:65–79. 7 Boyd R, Richerson PJ. 1985. Culture and the human cognition. Cambridge: Harvard Univer- evolutionary process. Chicago: University of Chi- sity Press. 54 McElreath R, Boyd R, Richerson PJ. 2003. cago Press. Shared norms and the evolution of ethnic mark- 31 Henrich J, Gil-White FJ. 2001. The evolution 8 Winterhalder B, Smith EA. 2000. Analyzing ers. Curr Anthropol. 44:122–129. of prestige: freely conferred deference as a mech- adaptive strategies: human behavioral ecology at anism for enhancing the benefits of cultural 55 Gil-White FJ. 2001. Are ethnic groups biolog- twenty-five. Evol Anthropol 9:51–72. transmission. Evol Hum Behav 22:165–196. ical “species” to the human brain?: essentialism 9 Boyd R, Richerson PJ. 1988. An evolutionary in our cognition of some social categories. Curr 32 Lancy DF. 1996. Playing on the mother- model of social learning: the effects of spatial and Anthropol 42:515–554. temporal variation. In: Zentall T, Galef BG, edi- ground: cultural routines for children’s develop- ment. New York: The Guilford Press. 56 Boyd R, Richerson PJ. 1990. Culture and co- tors. Social learning: a psychological and biolog- operation. In: Mansbridge J, editor. Against self ical approach. Hillsdale, NJ: Lawrence Erlbaum 33 Blurton-Jones N, Konner MJ. 1976. !Kung interest. Chicago: University of Chicago Press. p Associates p 29–48. knowledge of animal behavior. In: Lee R, De- 111–132. 10 Boyd R, Richerson PJ. 1989. Social learning Vore I, editors. Studies of the !Kung San and 57 Henrich J, Boyd R. 2001. Why people punish as an adaptation. Lect Mathematics Life Sci 20: their neighbors. Cambridge: Harvard Univer- 1–26. sity Press. defectors: weak conformist transmission can sta- bilize costly enforcement of norms in coopera- 11 Boyd R, Richerson PJ. 1995. Why does culture 34 Boyd R, Richerson PJ. 1996. Why culture is tive dilemmas. J Theor Biol 208:79–89. increase human adaptability? Ethol Sociobiol 16: common, but cultural evolution is rare. In: 125–143. Runciman WG, Maynard Smith J, Dunbar, RIM, 58 Boyd R, Richerson PJ. 2002. Group beneficial norms spread rapidly in a structured population. 12 Stephens DW. 1991. Change, regularity, and editors. Evolution of social behaviour patterns in value in the evolution of animal learning. Behav primates and man. Proceedings of The British J Theor Biol 215:287–296. Ecol 2:77–89. Academy, Vol. 88. Oxford: Oxford University 59 Boyd R, Richerson PJ. 1992. Punishment allows 13 Bergman A, Feldman MW. 1995. On the evo- Press. p 77–93. the evolution of cooperation (or anything else) in lution of learning: representation of a stochastic 35 Sperber D. 1996. Explaining culture: a natu- sizable groups. Ethol Sociobiol 13:171–195. environment. Theor Popul Biol 48:251–276. ralistic approach. Oxford: Blackwell. 60 Soltis J, Boyd R, Richerson PJ. 1995. Can BOOK REVIEW Evolutionary Anthropology 135

group functional behaviors evolve by cultural alists, what remains of the circumstantialist/pri- 68 Gumperz JJ. 1982. Discourse strategies. Cam- group selection? an empirical test. Curr An- mordialist controversy? Ethnic Racial Stud 22: bridge: Cambridge University Press. thropol 36:473–494. 789–820. 69 Smith N. 2001. Ethnicity, reciprocity, reputa- 61 Barth F. 1969. Introduction. In: Barth F, edi- 64 Nettle D, Dunbar RIM. 1997. Social markers tion and punishment: an ethnoexperimental tor. Ethnic groups and boundaries. Boston: Little and the evolution of reciprocal exchange. Curr study of cooperation among the Chaldeans and Brown. Anthropol 38:93–99. Hmong of Detroit, Michigan. PhD dissertation. 62 Kurzban R, Tooby J, Cosmides L. 2001. Can 65 Van den Berghe PL. 1981. The ethnic phe- Department of Anthropology, University of Cali- race be erased? coalitional computation and so- nomenon. Westport, CT: Praeger Publishers. fornia Los Angeles. cial categorization. Proc Natl Acad Sci USA 98: 66 Grafen A. 1990. Do animals really recognize 70 Gergely G, Bekkering H, Kira´ly I. 2002. Ratio- 15387–15392. kin? Anim Behav 39:42–54. nal imitation in preverbal infants. Nature 415: 63 Gil-White FJ. 1999. How thick is blood? the 67 Hamilton WD. 1964. The genetical evolution 755–756. plot thickens...:ifethnic actors are primordi- of social behaviour. J Theor Biol 7:17–52. © 2003 Wiley-Liss, Inc. BOOK REVIEW

Books Received

• Pereira, M.E. and Fairbanks, bridge University Press. ISBN • Schwartz, J.H. and Tattersall, I. L.A. (Eds.) (2003) Juvenile Pri- 0-521-80851-0 (cloth) $75.00. (2003) The Human Fossil mates: Life History, Develop- • Burkhardt, F., Porter, D.M., Record. Volume Two. Cranio- ment, and Behavior. xxxiii ϩ 428 Dean, S.A., Evans, S., Innes, S., dental Morphology of Genus pp. Chicago: University of Chi- Pearn, A.M., Sclater, A., White, Homo (Africa and Asia). New cago Press. ISBN 0-22665622-5 P. and Wilmot, S. (eds.) (2003) York: Wiley-Liss. ISBN 0-471- (paper) $30.00. The Correspondence of Charles 31928-7 (cloth) $150.00. • Coe, K. (2003) The Ancestress Darwin. Volume 13: 1865. Sup- • Robbins, M.M., Sicotte, P., and Hypothesis. xiv ϩ 214 pp. New plement to the Correspondence Stewart, K.J. (2001) Mountain Brunswick: Rutgers University 1822–1864. New York: Cam- Gorillas: Three Decades of Re- Press. ISBN 0-8135-3132-2 (pa- bridge University Press. ISBN search at Karisoke. New York: 0-521-82413-3 (cloth) $90.00. per) $29.00. Cambridge University Press. • McKee, J.K. (2003) Sparing Na- • Kappeler, P.M. and Pereira, ISBN 0-521-78004-7 (cloth) ture: The Conflict Between Hu- M.E. (Eds.) (2003) Primate $85.00. man Population Growth and Life History and Socioecology. • Lancaster, R.N. (2003) The Earth’s Biodiversity. New ϩ xxiii 395 pp. Chicago: Univer- Brunswick: Rutgers University Trouble with Nature: Sex in Na- sity of Chicago Press. ISBN Press. ISBN 0-8135-3141-1 ture and Population Culture. 0-226-42464-2 (paper) $30.00. (cloth) $28.00. Berkeley: University of Califor- • Minelli, A. (2003) The Develop- • Drayson, N. (2003) Confessing A nia Press. ISBN 0-520-23620-3 ment of Animal Form: Ontogeny, Murder. New York: W.W. Nor- (paper) $21.95. Morphology, and Evolution. ton & Co. ISBN 0-393-32444-3 xviii ϩ 323 pp. New York: Cam- (paper) $13.95.