Cultural evolution: Where we have been and where we are going (maybe)

Abstract

The study of cultural evolution using ideas from population biology began about 50 y ago, with the work of L.L. Cavalli-Sforza, Marcus Feldman, and ourselves. It has grown from this small beginning into a vital field with many publications and its own scientific society. In this essay, we give our perspective on the origins of the field and current unanswered questions.

The idea that culture evolves has been around since the time of Darwin and Spencer (1, 2). In the past it often had a progressive flavor—cultures were seen to evolve from band to chiefdom to state. Beginning in the 1970s, a number of researchers began to construct theories of cultural change based on ideas from population biology (3–6). Culture was conceptualized as information transmitted by social learning, and cultural change was understood in terms of the processes that caused some culturally transmitted variants to increase and others to decline. Like Darwinian evolution, progress was not assumed. The resulting theory motivated empirical work in a diverse range of disciplines including anthropology, psychology, and economics. These days it is a vital enterprise with its own scientific society and much ongoing research.

In this essay, we give our perspective on the field. We begin with a brief account of its origins. Our own work on cultural evolution was motivated by a desire to build a theory of human evolution that incorporated culture as an adaptation and could also explain both our extraordinary ecological success and our many singular, often seemingly maladaptive, behaviors. We then turn to current problems in the field and discuss culture as an ultimate cause of evolutionary change, the unique levels of large-scale cooperation in human groups, the role of attractors in cultural evolution, and cultural macroevolution. We end each section with speculations about the likely direction of future work on these problems.

Sociobiology and the Idea of Cultural Evolution

Since the early 20th century, it has been clear that humans evolved from an ape-like creature that lived in Africa 2 million years ago. During the interim, our species has been radically transformed. Humans occupy every corner of the globe, cooperate in huge numbers, make intricate tools and other artifacts, and live in a highly diverse range of societies. It might seem obvious that our evolutionary history and the processes that shaped it would be relevant to understanding why people are the way that they are, and, consequently, social sciences like economics, sociology, and anthropology would incorporate ideas from evolutionary biology. But this is not the case. Scholars in these disciplines have paid little attention to our evolutionary origins. The reasons for this neglect are undoubtedly complex, and this is not the place to try to recount them all. However, one important factor was that evolutionary biology did not have a satisfactory account of how natural selection shaped social behavior until the 1960s. This all changed with the work of scholars like Hamilton (7–9), Trivers (10, 11) and Maynard Smith (12) who provided a rich trove of ideas about how natural selection shaped social behavior. This new paradigm led to an explosion of empirical work on social behavior in evolutionary biology.

It didn’t take long for these ideas to spread to the human sciences. In 1975, Wilson published Sociobiology (13), a synthetic review of this growing body of work. In the last chapter, he speculated that evolutionary biology would come to play an important role in the social sciences. Sociobiology was well received, but its treatment of human behavior was controversial. Critics accused Wilson of “biological determinism” that neglected the role of culture in shaping human behavior. Wilson went on to write three other books on human evolution. The most scholarly of these, Genes, Mind, and Culture, coauthored with Lumsden, presented a mathematical theory of gene-culture coevolution that claimed to show that allegedly cultural differences between societies were often caused by genetic differences between populations.

About the same time, Alexander (14) published an influential review of the new behavioral biology and extended this body of ideas to explain various aspects of human societies. Alexander had a much greater influence on social sciences, particularly anthropology, than did Wilson. We think that there are two reasons for this: first, Wilson did not fully embrace the new ideas that were energizing evolutionary thinking—he had one foot in the new world, but one foot in the older, less dynamic paradigm. By contrast, Alexander’s paper was replete with interesting, provocative ideas from the new paradigm. Second, Alexander cultivated researchers in anthropology, most notably Napoleon Chagnon, William Irons, and Donald Symons, who trained a cohort of students to do anthropological field work rooted in the new evolutionary theory, including Monique Borgerhoff-Mulder, Mark Flinn, Ray Hames, Paul Turke, Laura Betzig, Nancy Berte, and Paul Bugos. This initial cohort produced significant new ethnographic work, much of it focused on social behavior (15, 16). At the same time, a number of young anthropologists interested in human ecology independently adopted an evolutionary approach often under the mentorship of senior behavioral ecologists like Eric Charnov and Gordon Orians. This group focused more on ecological questions like foraging and life history (17, 18), and included Eric A. Smith, Bruce Winterhalder, Kim Hill, Hillard Kaplan, Kristen Hawkes, and Nicholas Blurton-Jones.

This research tradition grew and prospered. The success is evident in the growth of a vigorous professional organization, the Human Behavior and Evolution Society, several new journals devoted to work on human behavior rooted in an evolutionary framework, and the publication of scores of papers, some in high visibility journals. By the mid-1980s, researchers had split into two opposing camps (19, 20): human behavioral ecologists proceeded in much the same way as behavioral ecologists studying other species, relying on ethnographic data to examine how behavior affected reproductive success. Evolutionary psychologists focused on how natural selection shaped the psychological mechanisms that generate behavior. They assumed that natural selection shapes these mechanisms slowly over time, and thus may not maximize fitness in contemporary environments because human environments have changed very rapidly since the advent of agriculture. They formulated hypotheses about the kinds of psychological mechanisms that would have been favored in Pleistocene foraging societies.

Researchers in both camps largely ignored culture or were actively hostile to it. Most thought of culture as only a proximate cause (21). They acknowledged that learning, development, and other forms of phenotypic plasticity allow humans and other animals to respond to environmental contingencies, but in the end such proximate mechanisms are shaped by natural selection. Culture has no independent causal role. Behavioral ecologists believed that these mechanisms could be ignored because however they worked, they would tend to generate fitness maximizing behavior. Evolutionary psychologists believed that the only explanation for the observed characteristics of human minds was adaptive evolution, and any systematic patterns were “either adaptations, byproducts or noise” (22). Having been subjected to frequent, often intemperate and poorly informed, criticism from practitioners of “standard social science,” most researchers in both camps rejected culture as a muddle-headed unscientific concept analogous to phlogiston. The exceptions were a few behavioral ecologists like Winterhalder, Smith, and Borgerhoff-Mulder who utilized models of cultural evolution rooted in population biology, and this approach has spread within human behavioral ecology over the years.

Contemporary theories of cultural evolution derive mainly from the work of Cavalli-Sforza and Feldman (23) and ourselves (24). We do not presume to say what motivated Cavalli-Sforza and Feldman. We were strongly motivated to develop models of culture that could explain the anomalous features of the human species within a Darwinian framework. These are as follows:

• There is much greater variation in behavior among human groups than among groups in other species. Much of this variation has important effects on fitness, persists for generations, and exhibits phylogenetic history, but is not due to underlying genetic variation.

• Humans have complex behaviors that show adaptive design, but cannot be explained as resulting from evolved, genetically transmitted modules. For example, reading has many of the attributes said to identify the output of an evolved module, but until a few hundred years ago, very few people could read.

• Many human behaviors seem maladaptive. For example, some of us climb mountains and have devoted more time and energy to writing scientific papers than raising a fitness-optimizing number of children.

We wanted to construct a theory in which the psychological mechanisms that give rise to cultural variation, and sometimes give rise to maladaptive behaviors, were evolved adaptations shaped by natural selection. We were firmly convinced that there could be no other naturalistic explanation for such a costly adaptive system.

The key to culture’s adaptive power, and its tendency to produce seeming maladaptive behavior, is that it acts like a system of inheritance. To see why, it is useful to first consider an alternative model. In Genes, Mind, and Culture, Lumsden and Wilson model culture like an ordinary system of phenotypic plasticity (25). There are two cultural variants, one better in one environment and a second better in an alternative environment. There are three genotypes, one that has a higher probability of learning the behavior that is adaptive in one environment, a second that is more like to learn the behavior that is adaptive in the second environment, and a “tabula rasa” genotype that learns the two behaviors at random. They show that selection favors the alleles that make it more likely to learn behavior favored in the local environment, and conclude that natural selection will favor genetically transmitted learning rules that cause locally adaptive behavior to be common in each environment and will select against unbiased learning rules.

Now, consider a different kind of blank slate. Individuals have two sources of information about what is the best behavior. They get information from the environment and can use this information to choose how to behave. Moreover, they can estimate how likely this process is to produce locally adaptive behavior. Sometimes it’s a sure thing, but sometimes is unlikely to yield the right answer. Second, learners observe the behavior of individuals from the previous generation and can copy the behavior of a randomly chosen individual. Here is the question: Suppose that most individuals that they observe are doing one thing, but individual learning is telling them to do something different. When should individuals rely on individual learning and when should they copy a random individual? It turns out that the answer depends on their estimate of how good individual learning is and how slowly environments are changing. If you are pretty confident that individual learning is accurate, you should ignore the behavior of others, and rely on your own experience. However, if your confidence is low, it is better to imitate any randomly chosen member of your group (26, 27). If individuals innovate only when they are likely to be right, almost all of the information flowing into the population will be accurate. Most people imitate because accurate individual learning is rare, and as long as the environment changes slowly, accurate information accumulates in the population and can be accessed by randomly imitating others.

Such a system yields a psychology consistent with our desiderata. First, it can maintain variation among local populations because culture evolves rapidly toward local optima. Second, the repeated action of a little bit of adaptive innovation coupled with accurate social learning can lead to the accumulation of complex adaptations that no individual is likely to invent on their own. We call these “cultural adaptations.” Third, the learning rule encourages people to ignore their own experience and motivations when they are uncertain, and this sometimes allows maladaptive behaviors to spread because they copy the behavior of individuals chosen at random from the group (28).

Biased social learning and teaching can enhance cultural adaptation. Social learning is biased when learners are more likely to learn from some individuals than from others. Several types of bias are possible. “Content bias” occurs when some content is more attractive or easier to learn than other content. Sometimes content bias may be adaptive. For example, Wertz (29) has shown that small children are cautious about touching or eating unfamiliar plants. Because plants often contain toxins, this bias is plausibly the result of genetic adaptation. Other times, content bias may arise from cognitive factors. “Payoff bias” occurs when social learners can estimate the fitness of individuals in the previous generation, and preferentially imitate relatively higher payoff behaviors. This creates a selection-like process that tends to increase the representation of variants that are correlated with fitness. If fitness is difficult to estimate and correlations between particular variants and fitness are weak, most people will imitate, approximately at random. Nonetheless, over time the weak bias will cause the population to evolve toward traits that increase fitness. In many environments, it may be difficult to determine which traits are correlated with fitness and so proxies are used. One useful proxy could be who other people are imitating. This can lead to adaptation but may also lead to the spread of maladaptive traits (30). Imitating the majority may also enhance local adaptation, but increase the variation among local populations. Biased teaching occurs when social models selectively teach one variant in preference to others.

In contrast, modeling culture as an ordinary system of phenotypic plasticity, as Lumsden and Wilson do, does not naturally explain any of the essential features of human culture. It is implausible that individual learning alone could explain the immense variation observed in the human species (31), and the core assumption that learning reflects genetically transmitted motivations is inconsistent with the second desiderata, and not easily reconciled with the systematic deviations from fitness maximizing behavior. This critique also applies to fundamentally similar proposals of the evolutionary psychologists Tooby and Cosmides (32–34).

We based our models in 1985 largely on the extant literatures on human social learning (35), the diffusion of innovations (36), and nonhuman social learning (37). Together this work suggested that humans had a much more varied and sophisticated system for social learning than any other animal. Later, more sophisticated comparative studies underlined this conclusion (38, 39). It does turn out that nonhuman social learning is much more variable, complex, and important than the early work implied, but so far humans are still quantitatively distinctive in our dependence on complex cultural adaptations (40).

We believe that the last several decades of work on cultural evolution provides a satisfying Darwinian explanation for why our species is such a different kind of animal. It explains both why we are able to adapt to a wider range of environments than any other species and at the same time do all kinds of things that don’t make sense as fitness maximizing behavior.

Gene-Culture Coevolution: An Ultimate Role for Culture

Many scholars who take an evolutionary approach to human behavior portray cultural evolution as a proximate system analogous to individual learning; ultimately only genes evolve by natural selection (33). But this is incorrect. If cultural traditions persist for many generations, natural selection can act on cultural variation. As Dawkins (41) observed, natural selection acts on any pattern of heritable variation, and cultural evolution can lead to cases in which specific traits end up closer to a cultural fitness optimum than the genetic one, even when the general capacity for culture is at the genetic optimum [(6), see also ref. 42].

Mayr’s (21) distinction between proximate and ultimate causes is too rigid to account for the coevolution of genetic and cultural variation. Human cultures evolve, creating novel niches that exert selective pressures on genetic variation in those populations. For example, humans and other mammals normally shut down the synthesis of the enzyme lactase which facilitates the digestion of lactose in milk after weaning. But in some dairying populations, mutations that lead to adult lactase persistence have been favored and have reached high frequencies (43). Lactose accounts for about 40% of milk calories, and milk consumption by postweaning children in dairying populations might reduce child mortality, especially during famines. Skin color is another example (44, 45). Humans were originally a tropical species but acquired cultural adaptations that allowed them to live in high-latitude environments with long, dark winters. Natural selection favored lighter skin in these environments, probably to allow more efficient vitamin D photosynthesis. There is by now a fairly long list of other more or less well-documented cases of culture led gene-culture coevolution in which cultural changes came first and led to subsequent genetic changes (40, 46–48).

Humans also directly select some genes to adapt themselves to cultural requirements. Human societies are shaped by institutions and their associated norms (49–51). Institutions and norms can have a long-term impact on human psychology because people who conform to institutional rules are rewarded and those who fail to conform are punished. If the tendency to conform to rules is genetically variable, then it could be subject to selection (52). For example, the strong dominance seeking behavior of chimpanzee males and bonobo females (53, 54) is moderated in humans. Norm-based social selection beginning early in the Homo lineage may have led to the relative “domestication” of our species (55–58).

The rarity of psychopathy may be an important element of our domestication (59–61). A strong drive for dominance is a component of psychopathy that seems to tap into the ancient tendency for primate groups to structure their societies hierarchically. The genetic heritability of the dominance-seeking component of psychopathy is high. It is plausible that as hominin species became more culturally sophisticated, social norms and institutions generated social selection that reduced the fitness of dominance-seeking individuals. Strong dominance-seeking behavior was transformed from a normal component of most individuals’ social psychology to a pathology that disrupted the economically productive cooperation that became a progressively more important element of hominin societies.

Where We Are Going.

Unraveling the history of gene–culture coevolution is crucial for understanding human evolution, but it will be difficult. Very few cases of gene-culture coevolution have been worked out in detail. Even the simple case of adult lactase persistence turns out to be surprisingly complex (43). Most likely, there are very many cases to be worked out. The ratio of gene-led versus culture-led cases will tell us about the degree to which culture has played an ultimate role in our evolution. Many important questions are difficult to address with the paleoanthropological record. When did language evolve? When did human social cooperation begin to operate on a scale larger than kin groups? These major human transitions were likely coevolutionary, but the cultural leg is largely invisible. We believe that the history of the genetic component of these changes should become visible with the rapid improvement in paleogenetics.

Large-Scale Cooperation

Large-scale cooperation is another singular human behavior that needs a Darwinian explanation. People in human societies cooperate in large groups to produce collective goods. They construct shared capital facilities like drivelines, fish traps, fortifications, and irrigation works, they risk their lives in large-scale warfare, and they maintain public order (62). People who cooperate in large groups to produce collective goods incur a personal cost, but the effect of their contribution on their own fitness is very small. This kind of behavior is unknown among other vertebrate species. Some like communally nesting birds and chimpanzees cooperate with distantly related individuals in small groups to produce collective goods, but very few vertebrate species do this in larger groups, and the few that do, like African mole rats, live in groups of close kin.

The nonhuman pattern fits with evolutionary accounts of the evolution of social behavior. Hamilton’s (7) theory of kin selection says that when relatives interact, selection acts as if individuals place a positive value on the fitness of others. Because ants and termites can have many siblings, selection can favor collective action in truly immense societies. But the reproductive biology of most vertebrates limits the number of close relatives, and kin selection does not favor large-scale cooperation. Reciprocity (10, 63), along the lines of “I would participate now if others participated in the past” can favor cooperative behavior in small groups, but not in large ones (31). So, the evolution of large-scale human cooperation is a puzzle—something must be added to the usual account.

Many authors have argued that the psychological machinery that supports contemporary large-scale cooperation evolved in Pleistocene foraging societies where cooperation was usually limited to band-sized groups of 20 or 30 closely related people (64–70). Contemporary behavior is a maladaptation that results from the larger, weakly related groups created by agricultural subsistence systems. This kind of explanation is often called the “mismatch hypothesis” because modern human cooperation reflects a mismatch between current social environments and those in which our psychology evolved.

The mismatch account of large-scale cooperation is based on a picture of Pleistocene of hunter-gatherer life that derives from 20th-century ethnographic accounts of several foraging societies. Over the last decade, this picture has come under serious doubt. Quantitative studies of ethnographically known hunter-gatherer societies indicate that relatedness among band members is quite low (71). Band membership is highly fluid and people form close social ties with a much larger network of 500 to 2,000 people (72, 73). As Bird and his colleagues put it, “foragers do not live in small-scale societies” (72). Historical and archaeological data also indicate that Holocene foragers frequently cooperated in large groups to produce collective goods (62). In many parts of the world, hundreds of foragers cooperated in communal hunting and intergroup conflict. Moreover, archaeological evidence suggests that middle Paleolithic peoples participated in large-scale communal hunting of ungulates (62). Thus, the human species may have a long history of large-scale cooperation among weakly related individuals, a fact that is inconsistent with mismatch explanations for human cooperation.

The cultural evolution of social norms provides an alternative route to the evolution of large-scale cooperation. Norms regulate human behavior in every human society (74–80). They specify which kinds of behaviors are desirable and which are forbidden, and reach into every corner of life—who you can marry, how you should dress, how you raise your children, what kinds of contracts are allowed, which god you may worship—the list is long. These requirements are socially enforced. Norm violators may lose status, access to mutual aid, and face monetary or corporal punishment. There is great variation in norm content among societies, and there is evidence that this variation is culturally transmitted (81). For example, linguistically related societies have more similar norms than less closely related societies (78). Moreover, on average, norms support collective action and other forms of cooperative behavior (49).

Norms supporting participation in intergroup conflict among the Turkana illustrate how this works (82–84). The Turkana are pastoralists who live in northern Kenya. They frequently launch large-scale livestock raids against neighboring ethnic groups. Raiding parties average about 300 warriors, and there is about a 1% chance of getting killed on any given raid. Free riding occurs—some men behave with cowardice in battle or take more than their share of the captured livestock. Such free riding is deterred by a graded series of punishments, ranging from negative remarks and fines to severe corporal punishment. These sanctions are administered by the warrior’s local social group, not other members of the raiding party. The Turkana punish because this is also normative—people who tolerate free riding are sanctioned by others (82). If people enforce norms because norm enforcement is the rule, then it follows that the content of norms is not constrained by the enforcement mechanism. Systems of rewards and punishments can stabilize a vast range of outcomes, including noncooperative ones (85).

Norms create incentives that stabilize a vast range of behaviors, and these incentives can maintain differences in norms between neighboring groups even though people and ideas move between groups. This means that differences will be stable through time, and when groups split, daughter groups will resemble their parent. When a vast range of outcomes is evolutionarily stable, knowing that a norm is evolutionarily stable isn’t very informative. A satisfying account of human large-scale cooperation must specify the processes that give rise to the particular kinds of norms that are observed—an “equilibrium selection mechanism” in the jargon of evolutionary game theory. We have argued that intergroup competition is one such mechanism (86). Neighboring groups often compete militarily, economically, and for prestige. The norms and values that predominate in a group can affect whether a group survives, whether it expands, whether it is imitated by its neighbors, and whether it attracts immigrants. Norms causing a group to survive will become more common than norms that lead to extinction. Similarly, norms that lead to group expansion, or attract more immigrants, will also increase compared to those that don’t. There is much empirical evidence that these processes, which we collectively term “cultural group selection,” have played a role in shaping the kinds of culturally transmitted norms we observe in human societies (81, 87).

However, cultural group selection can’t be the only important equilibrium selection process. There are many examples of norm change that do not seem likely by themselves to improve the ability of groups to compete including dueling, foot binding, (88) forms of female genital mutilation (89), and changes in footwear fashions (90). Moreover, cultural group selection may not be fast enough to explain the large number of norms that characterize every society. Soltis et al. (91) estimated group extinction rates in New Guinea societies and calculated that there was enough selection to shift one norm every 500 y or so. Genetic data (92) suggests that human populations regularly spread and merge, and so the New Guinea data may underestimate actual cultural extinction rates.

Where We Are Going.

We believe that future research in this area will focus on developing richer models of the evolution of norm content. First, contemporary theories portray norms as simple, often binary traits, that evolve independently of other cultural traits while in reality norms exist as part of complex, interrelated systems that likely affect the evolution of norm content. Second, integrating cultural group selection and processes that act within groups to change norm content will be important. Several such within-group models have recently been proposed. Young (93) argues that random variation within groups leads to norm shifts while Powers and colleagues (66, 94) (see also ref. 95) suggest that decisions about norm content result from within-group bargaining. These models also have weaknesses. Norms are typically shared by large numbers of individuals, and covary with ethnic membership. Existing within-group models are better suited to explaining norms in small groups than on such large, ethnic scales. Bargaining and cultural group selection hypotheses are complementary, and we believe that it is likely that combining them may provide useful models of norm evolution. Finally, we have proposed that tribal scale group selection on cultural variation over a long period of time favored the evolution of gene-based prosocial “instincts” and that these prosocial elements of our psychology influence bargaining over norms on the political time scale (96). We think that such coevolutionary models will be important in future research.

Attraction and Transformation

Much contemporary research on cultural evolution can be grouped into one of two clusters (97). The first descends from the work of Cavalli-Sforza and Feldman (24) and ourselves (23) and has been extended in important ways by many others (e.g., refs. 98–100). We will refer to this cluster as the “population biology tradition.” The second cluster of research began with the work of Sperber (101, 102), and has been developed by subsequent researchers (e.g., refs. 103–105). These authors, mainly drawn from cognitive and evolutionary psychology, philosophy, and linguistics, emphasize that learning from others is a complex process that rarely involves accurate copying or imitation. Instead, cultural learning usually transforms what is learned. We will refer to this cluster as the “attractor” approach.

Attractor researchers believe that social learning depends on inference and ostensive communication, and as a result is both noisy and inaccurate. This means that it is important to explain why cultural items persist instead of being degraded by noise. Authors in this tradition believe that the temporal stability of cultural items is due to convergent transformation, meaning that individuals who experience diverse cultural inputs tend to transform these inputs creating similar outputs. Cultural variants “end up recurring because, for various reasons, they are likely to be transformed in similar ways whenever they are transmitted” (106). In their minds, transformation is distinct from selective forces like payoff bias and content bias and, because cultural transmission is usually inaccurate, biased transmission is unlikely to be important. By contrast, these authors believe that researchers in the population biology tradition think that high-fidelity copying and selective forces are the main engines of cultural stability (107).

Much of the research in the attractor tradition is devoted to demonstrating that transformation can explain patterns of cultural variation and change. For example, Morin (108) shows that the fraction of portraits in which the subject looks directly at the viewer increases during the Renaissance, and presents data indicating that this shift is due to younger painters being more likely to choose more direct gaze than older painters who they learned from. Morin argues that an innate attentional bias favors direct gaze, and this caused the younger painters to transform the style that they learned. Similarly, motor constraints influence the evolution of rhythm (109). Miton (110) provides many other examples in which an empirically known cultural shift is consistent with a cognitive or perceptual bias.

We believe that disagreements between the two schools arise from different understandings about what is known about the mechanisms that underlie social learning. Authors in the attraction school believe that a specific account of social learning is correct and that it has important implications for our understanding of cultural evolution. In contrast, we are impressed that there is active debate about the psychology of social learning (111–114) and that the views of attractor theorists are not widely shared. Reliable replication of observed behavior is possible and occurs frequently enough to serve as a basis for modeling the population dynamics of cultural change.

Human social learning has been extensively studied in a comparative context (38, 39, 115). Compared to other primates, humans acquire novel traits more rapidly and accurately, and they are also more likely to imitate behaviors that have no obvious function. This makes sense because humans must acquire a very large repertoire of cultural traits, and the function of many elements in complex traits is not obvious to learners. It takes two decades of social learning for humans to become minimally competent economic producers (116, 117). Research on imitation in autistic (neurodiverse) people bears out this inference. The inability to imitate efficiently is one of the major features that define autism (118). An effective compensation method employs reinforcement-based techniques to finesse autistics’ inefficient imitation. Such intervention is quite time-consuming and results in near-typical development only in the least affected individuals (119).

There are some real disagreements about the population consequences of mechanisms of social learning. Here, we believe that there are two important questions:

• Does the population biology tradition have an adequate model of the dynamical effects of transformation in cultural evolution?

• Can selective processes be important when cultural learning is subject to significant transformations?

Does Population Biology Deal with Transformation?

In several papers (23, 26, 27, 99, 120, 121), we have described a process we termed “guided variation.” The basic idea is that people acquire beliefs and skills by cultural learning, and then modify these cultural variants by individual learning. The modified variants are then acquired by the next generation, and the kinds of cultural variants that characterize the population change through time. There has been discussion of whether transformation and guided variation are the same thing. Acerbi and Mesoudi (122) maintain that “Standard cultural evolution models, from the very beginning, have contained transformative processes such as guided variation, that seem to us to be identical to narrow cultural attraction.” They go on to argue that if everyone in a population transforms their cultural input in similar ways, then the population will rapidly converge to the same behavior. Morin (123) demurs, arguing that guided variation as conceptualized in our work (23) models individual learning and cultural transmission as distinct processes that do not interact. Individuals acquire some cultural variant from others and then transform it. The degree and nature of the transformation do not depend on the process of cultural learning. But there are many important kinds of transformation in which the transformation and the cultural acquisition are intimately entwined, and so it is a mistake to reduce all forms of transformation to guided variation.

It seems to us that this is a debate about what words to use, not about how the world works. Morin is certainly correct that the processes that underlie guided variation are only a subset of those that may lead to transformation. People do adaptively modify what they have learned, but many other quite distinct transformative processes are likely to play an important role. For example, errors and misunderstandings may cause social learners to transform complex adaptive traits that they learn in ways that make them less functional (99, 124), a kind of misguided variation that may play a crucial role in impeding cultural adaptation. Morin (103) provides an interesting description of how manuscript copying errors lead to distinctive patterns of change in manuscript content. Different processes work differently, and so different labels may be useful.

On the other hand, models of guided variation capture the population dynamics of convergent transformation as envisioned by authors in the attractor tradition. As evidence for this claim, consider the model presented by Acerbi et al. (107). Many of the authors of this paper are advocates of the view that culture is usually transformed as it is acquired, and so the model reflects their view of transformation. In the “convergent transformation” treatment, this model has almost exactly the same structure as the model of guided variation analyzed in our 1985 book (23), Chapter 4]. In both, there is a continuously varying trait. Naive individuals acquire an initial value, and then this value is transformed on average a fraction of the distance to a target value. The transformed value serves as the initial value for an individual in the next generation. Both models lead to the same qualitative conclusion-convergent transformation can lead to long-run stability.

Is Selection Important When There Is Significant Transformation?

Authors in the attractor tradition frequently assert that selective processes only play an important role in determining evolutionary equilibria when transmission is accurate. This conclusion is both incorrect and inconsistent with the most widely cited model in the cultural attractor tradition. Claidière et al. (106) develop a mathematical structure using “evolutionary causal matrices (ECM)” to model cultural change within the attractor framework. The ECM approach starts with a finite set of n discrete cultural variants. The state variables are the frequencies of each variant. The next step is to write down an n × n matrix. The ijth element of this matrix represents the average impact that each item of type i at time t has on the frequency of type j at time t + 1. So, the diagonal elements give the effect of a trait on its own reproduction and the off-diagonal elements give the effect of other traits.

The ECM formalism is limited when compared to models in population biology tradition. It is a “top down” approach that starts with a specified mathematical structure and this limits the kinds of processes that can be represented. For example, with two traits, the recursions for trait frequency are limited to a linear fractional form and can have at most one stable equilibrium. Many important and interesting processes such as conformist social learning generate multiple stable equilibria and cannot be modeled in the ECM framework. The population biology approach is bottom–up. It begins with a description of the processes that shape cultural variation as it is learned and affects the behavior of individuals and groups, and then this is used to generate a mathematical model. This approach is more flexible and focuses attention on the underlying causal processes.

Claidière et al. (106) use the ECM approach to show that selection can be seen “as a special case of attraction” (p. 5) that becomes important only when transmission is accurate. In the ECM framework, this occurs when the diagonal entries in the ECM are much larger than the off-diagonal elements. As the relative magnitude of the off-diagonal elements increases, nonselective processes become more important. This is true, but selection still can be important in determining evolutionary outcomes even when transformation is also important. To see why, we use the population biological approach to construct a model with both selection and transformation. Assume there is a population in which people hold one of two variants, labeled 1 and 2. Each generation, naive individuals select a person who they will learn from. Individuals with trait i are chosen with probability w_i. If $w_1\ne w_2$, this will create selection in favor of the variant that is more likely to be chosen. Now we also suppose that during cultural learning trait 1 is transformed into trait 2. The probability that an individual learning from an individual with trait 1 acquires trait 2 is m. Individuals learning from models displaying trait 2 never acquire trait 1. As the value of m increases, transformation becomes more important relative to selection. These assumptions lead to the following Evolutionary Causal Matrix

$$\begin{array}{c}\hfill \left(\begin{array}{cc}(1-m)w_1& 0\\ m{w}_1& w_2\end{array}\right),\end{array}$$ [1]

If $w_2>w_1$, both selection and transformation reduce the frequency of trait 1 and so the equilibrium frequency of that trait is zero. When $w_1>w_2$ transformation and selection are opposed, and the population has a mixed equilibrium as long as $\frac{(w_1-w_2)}{w_1}>m$. When m is small the diagonal elements are much larger than the off-diagonal elements and selection dominates. As m becomes larger, transformation becomes more important, but selection still affects the equilibrium behavior. This exercise illustrates that the “population biology” formalism and the ECM formalism can be equivalent ways of representing the same underlying processes. It also shows that selection can have an important effect on the outcome even when nonselective processes have a substantial effect. Claidière et al. (106) write (p. 6) as if selection is only important when transformation rates are low. But this is incorrect, and authors in the population biology tradition have studied models in which transformation and selection both play important roles (e.g., refs. 23, 27, and 99) because such models are useful for understanding the genetic evolution of the capacities that allow cultural learning.

Where We Are Going.

It seems to us that there is a large range of social learning systems (125). Both transmission and transformation are always involved but to varying degrees and the details are diverse and consequential, and it is important to understand how different kinds of social learning work. There has been a significant uptick in interest in the psychology of social learning recently, but a wide range of opinion remains. Some authors emphasize that the same mechanisms, especially reinforcement learning, are at work in individual and social learning (114), while others argue that other cognitive mechanisms are necessary (113). In recent years, a number of researchers have developed mathematical models and tested these with laboratory experiments (124, 126). Researchers have studied the role of causal reasoning on social learning both in the laboratory (127) and the field (128), and how functional interdependencies among different components of a skill may facilitate social learning (105). We expect that there will be much progress in the future and that this will affect our understanding of the population dynamics of cultural variation.

At the same time, much of this work does not make contact with mathematical models of cultural evolution or the adaptive function of social learning. Human culture needs the support of our large and expensive brain, a long juvenile period, and cooperative breeding to support brains and nonproductive juveniles. Our cultural system must cover these costs. Boyd and Richerson (23); see also refs. 129 and 130) proposed that the enhanced speed of cultural evolution compared to genetic evolution deriving from the transformative forces of guided variation and the various bias forces allowed humans to adapt more quickly to temporally varying environments and more accurately to spatially variable ones. We believe that real progress will require attention to these issues.

Cultural Macroevolution

Macroevolution concerns the large-scale and long-term events in evolutionary history. Important questions include why humans evolved so recently in the history of life, why culture is so important in our evolution, and why the agricultural subsistence systems that support our current massive numbers started to evolve only about 11,000 y ago. Macroevolution is difficult to study because it takes place deep in the past, and the fossil, archaeological, and historical records are seriously incomplete (131).

Simpson (132) suggested macroevolutionary questions come in two flavors, tempo and mode. Tempo refers to the rate and pattern of long-term evolutionary change and mode to the processes that generate such change. The tempo of evolution is comparatively easy to observe in historical, archaeological, and paleontological records. A comparison of archaeological and paleontological data indicates that cultural evolution is typically much faster than genetic evolution (133). Reconstructing the mode of cultural macroevolution is much harder because processes are difficult to reconstruct from historical and archaeological data and because gene-culture coevolution multiplies the number of hypotheses we need to consider.

A major feature of human cultural macroevolution is the massive diversification of human cultures, especially in the Holocene. Even though humans are a species with low genetic diversity (134), our subsistence economies and social organization are extraordinarily diverse, e.g., refs. 135 and 136, suggesting that macroevolutionary processes may have been driven by cultural evolution.

Evolutionary change is driven by two classes of processes, those that are internal to the evolutionary process itself and those that are driven by changes in the external environment (137). An example of an internal process is the role of chance in the origins of anatomical, physiological, and cultural traits. For example, high-altitude Andean and Tibetan populations have evolved very different physiological adaptations to hypoxia (138). The external factors play a crucial role when adaptive forces of cultural evolution produce convergent patterns of evolution (139). For example, culturally unrelated peoples in Africa, Asia, and the Americas living in wet tropical forests usually farm using long-fallow swidden cultivation (140). Heavy rainfall and high soil temperature produce weathered soils with low nutrient-holding capacity. Cutting and burning patches of forest to tap nutrients stored in the trees at decades-long intervals is a sustainable production system on such soils.

The simple internal-external dichotomy does not cleanly cover cases of coevolution and niche construction (141) in which relevant parts of the environment are part of the evolutionary process. Plant and animal domestication is an important macroevolutionary example. Domesticates transformed the subsistence foundation of human cultures as much as humans changed their genomes by artificial selection. Grain-rich diets led to the evolution of our digestive physiology. Domestic animal diseases often jump to humans (e.g., smallpox). The higher human population densities made possible by agriculture led to the cultural evolution of increased division of labor and hierarchical governance systems. Human niche construction involved not only the substitution of domesticated plants and animals for wild species, eventually at massive scale, but also the physical modification of environments. For example, the construction of sophisticated shelters allowed a tropical ape to live above the Arctic Circle.

Commonly, human evolutionary history is portrayed as a steady progressive modernization of the human phenotype over the Pleistocene, as if internal processes dominate this pattern (e.g., ref. 142). In fact, human evolution has taken place against a backdrop of dramatic external environmental changes. The Plio-Pleistocene transition involved the earth becoming cooler, drier, and much more variable as the alternation of glacial and interglacial climates took hold (143). Many macroevolutionary events in human evolutionary history may have had external causes, perhaps in combination with internal and coevolutionary processes, including the Pleistocene origin of humans and the origin of agriculture after the end of the last ice age (144). On the other hand, the internal positive feedbacks between the evolution of technology and human demography create unstable dynamics under plausible assumptions (89).

Methods for studying cultural macroevolution have included several approaches. One of the more popular is the use of formal phylogenetic methods borrowed from biology as a framework for other analyses (145). Even the proponents of this approach recognize that cross-lineage borrowing is a common phenomenon and endorse other comparative methods. Some interesting studies focus on particular artifacts such as looms or other forms of technology (146, 147). The objective of these studies is to connect the microevolutionary details of social structure, patterns of socialization, and the ecology of adaptation to macroevolutionary patterns. Where the relevant ethnographic information is available, such studies can be very elegant. A study of within- and between-group variation in music found a quite weak phylogenetic signal. Rzeszutek et al. (148) argue that musical elements are very easily borrowed and recombined to produce syncretic traditions such as Afro-American musical traditions combining European melodic and African rhythmic elements.

Where We Are Going.

The study of human macroevolution needs to move from story telling to hypothesis testing (149). Improving paleoecological data (150), a slowly improving conventional paleoanthropological record, and our increasing ability to extract relevant data from conventional and ancient DNA (92, 151) make future progress possible. The creation of synthetic historical and archaeological databases capable of sustaining formal model testing is an important and ongoing methodological innovation (152).

We expect that cultural macroevolutionary studies will be an area of rapid growth in the next decades. Many patterns of paleolithic human macroevolution are still quite obscure and the creation of quantitative databases and the use of ancient DNA to reconstruct population structure are proving their worth but have only scratched the surface. The historical information in the genomes of contemporary humans also promises to shed light on the timing of important macroevolutionary events in our lineage (153). For example, a recent study of language and genes on Vanuatu showed that the languages are Austronesian but the genes almost entirely Papuan (154).

Conclusion

Over the last 50 y, the population approach to studying cultural evolution has grown from the work of a few investigators to a commonplace idea used in a wide range of disciplines served by two scientific societies, The Cultural Evolution Society, and the European Human Behavior and Evolution Association. It helps developmental psychologists link their study of how children acquire ideas to larger social patterns. Economists have adopted the perspective to help understand why variation in behavior can persist over centuries. Paleoanthropologists make use of theoretical results linking population size to cultural repertoires to understand patterns in the archaeological record. We believe that such applications work because the population approach naturally links processes at the individual level that depend on individual economics, psychology, and decision making to larger-scale social and historical patterns. In the longer run, we hope that such linkage may help to reduce the barriers between different, seemingly incompatible, natural and social sciences. The conceptual, empirical, and mathematical techniques of cultural evolution and gene-culture coevolution should be part of the toolkit of all human scientists, biologists, and social scientists alike. Much has been accomplished in the past half-century, but there is much more left to do.

Data, Materials, and Software Availability

There are no data underlying this work.