Cultural evolutionary theory: How culture evolves and why it matters

Many of the first models of cultural evolution drew explicit parallels between culture and genes by modifying concepts from theoretical population genetics and applying them to culture. Cultural patterns of transmission, innovation, random fluctuations, and selection are conceptually analogous to genetic processes of transmission, mutation, drift, and selection, and many of the mathematical techniques used to study genetics can be useful in the study of culture (1, 12). However, these mathematical approaches had to be modified to account for the differences between genetic and cultural transmission. For example, we do not expect cultural transmission to follow the rules of genetic transmission strictly. Indeed, cultural traits are likely to deviate from all three laws of Mendelian inheritance: segregation, independent assortment, and dominance (13).

The simple observation that cultural traits need not conform to Mendelian inheritance is sufficient to produce complex evolutionary dynamics: If children are likely to reject a cultural trait that both of their parents possess, the frequency of that trait in the population may oscillate between generations (4). In addition, if two biological parents have different forms of a cultural trait, their child is not necessarily equally likely to acquire the mother’s or father’s form of that trait (14). Further, a child can acquire cultural traits not only from its parents (vertical transmission) but also from nonparental adults (oblique) and peers (horizontal) (1, 12); thus, the frequency of a cultural trait in the population is relevant beyond just the probability that an individual’s parents had that trait (Fig. 1). In most cases, the more common a cultural trait is in the population, the more likely it is for an individual to have the opportunity to acquire it through social learning (15). However, the size of the population may also influence the continuing transmission, and thus survival, of a cultural trait (16). The relative importance of a population’s size, and its environmental context, for the retention and perhaps expansion of the cultural repertoire constitutes an ongoing debate (16–20).

Many theoretical population genetic studies make the assumption that mating is random within a population. However, in real human populations, this assumption is often violated, as individuals tend to prefer mates with similar phenotypes, such as eye color (66), height, IQ (67), education level (61), and smoking status (68). Cultural evolutionary theory has led to significant advances in our understanding of the effects of nonrandom mating, revealing that the transmission and dynamics of cultural traits can be sensitive to both phenotypic and environmental assorting (41). Assortative mating, leading to an increased correlation between mates for genetic or cultural traits, can increase both genotypic and phenotypic variance in a population (69, 70). In addition, assortative mating (and other forms of homophily) acting on one cultural trait can influence the evolutionary dynamics of other cultural traits, facilitating the spread of rare cultural or genetic variants (71, 72). More generally, assorting can affect not just mate choice but many types of cultural interactions, termed “assortative meeting” (73). Empirical work supports this theoretical finding; for example, beneficial health behaviors spread more readily through a social network when individuals’ social contacts were more similar to themselves (74, 75). Culturally mediated assortment can also lead to biological differences: Partners that are more similar tend to have more offspring (76), thus increasing fitness, and assortative mating within highly homophilic groups affects the average length of homozygous DNA segments (59, 77), leading to the appearance of higher levels of inbreeding than might actually exist. Humans can also assort by language; however, studies of the interactions between language and genetic population structure show that the resulting dynamics can differ by population. For example, in some geographic regions, language boundaries do not seem to act as barriers to gene flow (78–80) whereas, in other places, assorting with respect to language seems to have had a large effect, and genetic similarity is more closely associated with language than with geographic distance (80–83). Assortative mating has had a measurable effect on human genomic architecture, and genetic and phenotypic correlations between partners are substantial (84).

In addition to choosing their mates nonrandomly, individuals can also choose their cultural role models; these cultural transmission biases affect the relationship between a trait’s frequency in the population and its likelihood of transmission (Fig. 3). For example, conformity bias is an exaggerated preference for the cultural variant practiced by the majority of the population, which can lead to an increasingly large majority over time (85, 86). Alternatively, individuals might preferentially seek out novel cultural traits, termed rarity bias or novelty bias (30). These frequency-dependent biases can lead to patterns of cultural diffusion in which the prevalence of a cultural trait can change dramatically over short timescales, producing logistic growth (“S-shaped” curves) of trait frequency over time (87, 88). Examples of cultural traits that are likely to exhibit frequency-dependent transmission are fashion trends (89), career choices (12), and baby names (90). Conformist transmission is likely to dominate when the environment is relatively stable and common cultural traits are well adapted to that environment (86, 91).

Other types of transmission biases reflect not how common a trait is in a population, but the characteristics of the people who have the trait. In the case of prestige bias, individuals attempt to acquire cultural traits that are perceived to be high quality by selectively learning from those individuals with high social rank (92). For example, in an experimental test, children were much more likely to choose an adult cultural role model if they had observed bystanders attending to the potential model rather than ignoring him or her (93); thus, even at a very early age, humans can assess such characteristics as prestige or social standing. Individuals can also use observations of success associated with a cultural trait, such as a fruitful hunt with a certain tool, to develop a preference for cultural role models that are demonstrably successful (30). This bias has been demonstrated experimentally (94, 95); for example, when individuals participated in simulated hunting with virtual arrowheads and then modified their arrowheads either by trial and error or imitation, copying successful individuals gave significantly better results than trial and error (94).

For thousands of generations humans have been carving their existence in the world with cultural tools that have become integral to their livelihoods, thereby shaping their environment at all scales, both intentionally and unintentionally. Attempting to answer the question of what are the extensions of human biology through culture leads to a striking conclusion: There are few aspects of human biology that have not been shaped by our culture. Human culture has also affected the biology, even the survival, of nonhuman species (96). In this section, we review a number of cases for which incorporating culture into models of ecoevolutionary dynamics has proven valuable for the interpretation, prediction, and, in some cases, direction of human ecology and of human impact on the ecosystem.

The growth and age structure of human populations are both affected by norms and beliefs of their members. A predominantly agricultural lifestyle produced higher population growth than the hunting-gathering lifestyle it replaced (148, 149). This increased growth was most likely due to the spread of a complex of cultural traits (150) whose adoption may have created conditions that favored the accumulation of subsequent culturally transmitted behaviors (151, 152). Beginning in the late 19th century, parts of Europe, Asia, the United States, Australia, and New Zealand began to undergo a second demographic transition, which involved a change from a high birth rate, high mortality regime to a lower birth rate, low mortality regime. These changes were due to the spread of fertility-reducing and survival-increasing behaviors that became part of the developed countries’ cultures.

Standard quantitative models of demographic change do not include within-population variation in behaviors that affect fecundity or mortality. Projections usually use fixed values for birth and death rates; however, religious preferences, marriage customs, dietary choices, population subdivision, and mortality profiles may affect fecundity but are usually not part of demographic models. Further, aspects of cultural transmission, such as prestige bias and the choice of nonparental cultural role models, can facilitate the spread of fertility-reducing behaviors (12, 153). Thus, cultural evolutionary approaches should be integrated into demography, especially the processes that have led to fertility decline (154).

Many models for life history analysis of humans divide the lifespan into an ordered series of age classes. These models first define the fertility rates of each age class and the survival rates from one age class to the next. Then, they iterate the number in each age class produced by these parameters to determine the dynamics of the population, including whether the number in each age class approaches a stable equilibrium, termed the stationary age distribution, or whether the population will grow or go extinct and at what rate (155).

Carotenuto et al. proposed a demo-cultural framework for such an age-structured population, in which each individual carried one variant of a dichotomous trait, say H or h, where H represents the presence of a socially learned behavior (for example, fertility control) and h is its absence (156). An individual of type H might also be more likely to survive into the next age class. This integration of demography and culture yields complex dynamics; for example, the trait H can persist in the populations even if it lowers fertility, as long as the cultural transmission of H is reliable enough, or if H also sufficiently increases the chance of survival. Additional learning steps can also be added to age-structured models, such that vertical and horizontal transmission can occur at different rates for different age classes (101). In this case, horizontal learning accelerated the trait’s spread and led to faster population growth than vertical transmission alone.

An important outgrowth of demo-cultural modeling has been its application to the sex-ratio problem. In many places, the sex ratio at birth is strongly biased in favor of males and, in China and parts of India, has resulted in up to 120 male births for every 100 female births (157). This cultural preference for sons can be manifested in sex-selective abortion or withholding of resources from daughters. This bias has both economic and socio-cultural antecedent, as well as important ethical and demographic consequences (158).

Data on cultural transmission of son preference can be incorporated into formal demographic analysis (159), linking these data to real-world policy applications (160). Theoretical models can also aid in predicting the effects of policies: For example, one such model tracked the cultural transmission of the perceived present value of a son relative to a daughter, the sex ratio at birth, and their effects on demographic change (161). The results of this model suggest that interventions focused on peer-to-peer cultural transmission of a perceived higher value of daughters might complement existing economic incentives to support and educate daughters, with the goal of mitigating the effects of son preference. The literature on the interaction between cultural transmission and formal demography is quite sparse. Given the large variety of customs that relate to birth and death rates in different human societies, population projections for the future needs of diverse populations should incorporate more cultural dynamics than is currently standard practice.

With the extensive body of theoretical and empirical literature on cultural evolution, researchers in this field are now combining information from multiple disciplines and integrating disparate approaches. Part of this new frontier involves more fully bridging the divide between theory and data, as well as developing mathematical models than can aid in the interpretation of anthropological and archaeological information. In addition to aiding our understanding of human history, the study of cultural transmission and evolution is extremely relevant in the modern era. Insights from cultural evolution and the diffusion of innovations have been coopted in advertising and social media to quantify the viral spread of information (e.g., ref. 162). How can these cultural evolutionary insights be better used for positive action and public health? In addition, how can we better use knowledge about cultural evolution to more fully understand patterns of human genetic variation and population structure? As we continue to understand more about the human genome, it becomes increasingly important to consider environmental and cultural contexts as well as genetic variation; however, in the study of gene-culture interactions, faulty logic or racial biases about “causes” of human differences may be used and must be cautiously guarded against (reviewed in ref. 163).

In this paper, we have reviewed aspects of human cultural evolutionary theory, focusing on those that are most closely linked to the extension of biology through culture. With this focus, we could not do adequate justice to many important domains of cultural evolutionary theory. In brief, many models of cultural evolution focus primarily on the transmission of cultural traits and not on their interactions with genes or fitness. These models include, but are not limited to, models of social learning (e.g., refs. 164–167), models of language evolution (e.g., refs. 168–171), empirically driven verbal models of human evolution based on patterns in material culture (e.g., refs. 172–174), and models of cultural dynamics within and between groups (e.g., refs. 86 and 175–178)). In addition, we focused on human studies, although cultural processes are present in many other species. For example, social learning has been extensively studied in nonhuman animals, in which behavioral strategies, such as producer and scrounger, and cultural trajectories can be more clearly defined than in humans (166, 179). Cultural transmission also has large-scale evolutionary implications for some nonhuman animals: For example, theoretical studies suggest that nonrandom mating in birds based on culturally transmitted songs could accelerate speciation (180, 181) and that sexual selection on learned songs could influence evolution of the neural underpinnings of learning (182). Recently, studies in a range of animal species have shown that cultural practices can emerge, spread, and change over time, potentially influencing individuals’ fitness (183–187). Tool use among chimpanzees and capuchins (188–190) is one such example, which also provides insight regarding the possible origins of the early phases of our own species’ adaptation to the “cultural niche” (191, 192).

In recent years, models that are used for decision making in various fields, such as economics and public health, have begun to take cultural evolution into account, and a growing number also incorporate the modeling—verbal or mathematical—of the human ecosystem’s expected coevolution with the spread of cultural practices. These models play a prominent role in planned strategies related to climate change and reduction of carbon emissions (193), in predicting global food shortages and requirements (194), and in assessing the distribution of new practices and technologies in agriculture (195, 196). In addition, models in epidemiology have begun to integrate cultural transmission of health practices and pathogen ecological dynamics with regard to drug distribution and combating epidemics (e.g., ref. 197).

Deeper analysis of how human culture, human ecology, and the human environment coevolve is necessary for understanding historical and present dynamics, and for predicting future trends. These analyses will provide much-needed tools for the planning and direction of such dynamics. Humans’ worldwide well-being and that of the ecosystem we live in depend on our ability to make such predictions and act accordingly.

We thank the John Templeton Foundation and Stanford Center for Computational, Evolutionary, and Human Genomics for funding.