Billiard, S. & Alvergne, A. History and Philosophy of the Life Sciences (2018) 40: 9. https://doi.org/10.1007/s40656-017-0173-y
Over the last 40 years or so, there has been an explosion of cultural evolution research in anthropology and archaeology. In each discipline, cultural evolutionists investigate how interactions between individuals translate into group level patterns, with the aim of explaining the diachronic dynamics and diversity of cultural traits. However, while much attention has been given to deterministic processes (e.g. cultural transmission biases), we contend that current evolutionary accounts of cultural change are limited because they do not adopt a systematic stochastic approach (i.e. accounting for the role of chance). First, we show that, in contrast with the intense debates in ecology and population genetics, the importance of stochasticity in evolutionary processes has generated little discussion in the sciences of cultural evolution to date. Second, we speculate on the reasons, both ideological and methodological, why that should be so. Third, we highlight the inadequacy of genetically-inspired stochastic models in the context of cultural evolution modelling, and ask which fundamental stochastic processes might be more relevant to take up. We conclude that the field of cultural evolution would benefit from a stochastic revolution. For that to occur, stochastic models ought to be developed specifically for cultural data and not through a copy-pasting of neutral models from population genetics or ecology.
Evolutionary theory has been applied to the study of culture in various ways for more than 100 years. In the field of cultural evolution, most, if not all, of the approaches developed until the 1970s were narrative-based and interpretive, that is, there were no quantitative predictions for how cultural traits (e.g. behaviours, ideas, artefacts) should vary and be distributed in a population. Contemporary cultural evolution research differs markedly from these previous traditional approaches in that it is built upon a quantitative and mathematical framework. Building on precursors like Gulick (1905) and Binford (1963), a quantitative framework was developed in 1981 by two evolutionary scientists, Cavalli-Sforza and Feldman (1981). They borrowed mathematical tools from population genetics to predict how the social transmission of information between individuals influences the dynamics of culture at the population level. Cavalli-Sforza and Feldman (1981) contributed two main concepts: (1) cultural selection, analogous to natural selection, which describes the differential reproductive success of cultural traits, considered as deterministic, and (2) cultural drift, analogous to genetic drift, which accounts for the role of chance in cultural change.
Following Cavalli-Sforza and Feldman’s seminal work, there has been an explosion of cultural evolution studies in anthropology and archaeology (reviewed in Mesoudi 2011; Lewens 2015), mostly focusing on social transmission mechanisms, for instance the importance of vertical social transmission (i.e. from parents to offspring) as compared with horizontal social transmission (i.e. from peers). Subsequently, the anthropologists Boyd and Richerson (1985) and their students extended Cavalli-Sforza and Feldman’s framework to include various modes of cultural selection, which they coined biased cultural transmission, e.g. prestige-bias and conformist-bias but also payoff-biased social transmission (Henrich and McElreath 2003). Overall, the concept of cultural selection, though arguably more complex than the concept of natural selection, fell on fertile soil in the human sciences as testified to by the intense debates between the schools of cultural (Acerbi and Mesoudi 2015) and cognitive anthropology (Claidière et al. 2014) over what cultural selection actually entails (see also Lewens 2015 for a dispassionate review).
Comparatively to the success encountered by the concept of cultural selection, the concept of cultural drift has been much less popular than that of cultural selection, and it remains little explored. There has undoubtedly been a productive utilisation of neutral models in archaeology and anthropology (neutral models are a special category of stochastic models where population change is not pushed towards a particular direction, “Appendix 1”). To date, however, there have been little controversies over the role of selection versus drift for understanding the evolution of culture and patterns of cultural diversity. This is in sharp contrast with the biological sciences, within which the relative importance of selection and drift for explaining both species diversity and evolutionary change has been intensely debated for years. In population genetics and ecology, it led to a stochastic revolution whereby the cause of change by default is not assumed to be selection anymore but rather stochastic processes, in particular genetic drift and demographic stochasticity (Kimura 1983; Hubbell 2001). Broadly speaking, selection models must now provide more explanatory power than neutral models (models without selection) for the data at hand to be accounted for by selection.
In the field of cultural evolution, such a stochastic revolution has yet to happen. Questioning the role of stochasticity for explaining patterns of cultural diversity is necessary, however, for several reasons. First, there is strong empirical evidence that there is a large population-level variance or noise around the mean value of cultural traits, which is neglected by classic deterministic approaches. Given social systems are subject to stochastic effects, properly defining and modelling random fluctuations around the population mean of cultural traits is key for determining the extent to which cultural diversity is underpinned by adaptive processes. Second, since agents of cultural evolution are discrete (individuals) and populations are finite, stochasticity is necessary to explain some features of the diachronic dynamics of cultural traits. For instance, fluctuations and the extinction of a trait cannot be modelled with classical approaches based on differential equations. Third, stochasticity is a fundamental concept in psychology and neurobiology for making sense of how individuals make decisions (Forstmann et al. 2016). Fourth, in contrast with evolutionary biology and ecology (where the Wright-Fisher’s and Moran’s models are classically used), there is no consensus over which stochastic process is the referent one for cultural evolution models. Finally, since random genetic drift and natural selection are equally important in the neo-Darwinian theory of evolution, the role of stochasticity in driving cultural change and diversity must be considered to adequately evaluate the relevance of the analogy between biological and cultural evolution.
In this paper, our overarching goal is to evaluate the need for a stochastic revolution in the field of cultural evolution by contrasting the uses and utility of stochastic models in the biological sciences (i.e. population genetics and ecology) and in some of the human sciences engaging with cultural evolution research (i.e. archaeology and anthropology). First, we discuss the role given to stochasticity in the cultural evolution literature. Second, we question why there has only been a few controversies over the role of stochasticity in cultural evolution as compared with the intense debates that neutral theory generated in ecology and population genetics. Finally, we dispute the analogy between cultural and genetic drift and its relevance for the study of cultural evolution. We conclude that a stochastic revolution is much needed in contemporary cultural evolution studies, albeit not in a copy-paste fashion from the biological sciences, but after the sources of stochasticity unique to human culture have been identified. Such a paradigm change from a deterministic to a stochastic view of the world has proven to be fruitful in several scientific disciplines including physics, chemistry, biology and psychology (Gigerenzer et al. 1989; Hacking 1990). We contend that it would also be productive for advancing the field of cultural evolution because making chance a central concept will allow a better description of the processes underpinning cultural evolution and an increased control of uncertainty when interpreting observations.
5 Summary and conclusions
In this article we have reviewed the studies that have engaged with the importance of stochasticity, i.e. chance, in cultural evolution. Drawing on research published in archaeology and anthropology, we focused on the development and use of a ‘neutralist’ theory (i.e. when change is not biased towards one particular direction). We found that in contrast with the stochastic turn that has occurred in population genetics and community ecology, controversies over the importance of stochasticity in cultural evolution have remained scant. Ironically, it is the social sciences that have triggered the stochastic revolution in the natural sciences, even reaching physics (Gigerenzer et al. 1989). Indeed, the field of statistics has been developed since the beginning of the nineteenth century, mainly because of the accumulation of data from social studies, to explain and interpret inter-individual variation. While sociology is now at the forefront of the development of statistical methods for analysing social interactions (e.g. Snijders 2017), in social and cultural anthropology, since the interpretive turn initiated by Geertz and Evans-Pritchard, quantitative approaches to society and culture fell into disrepute.
The general conclusion of our paper is that the question of which fundamental stochastic processes underlie the evolution of cultural traits deserves more attention and critical thinking than it currently receives. While there is now a relative consensus over the form of the most basic stochastic models in population genetics and community ecology, the issue has hardly been raised in the field of cultural evolution. Indeed, following Cavalli-Sforza and Feldman (1981), in analogy with population genetics, cultural drift is defined as the effect of random sampling of cultural parents between generations. However, there are several other potential sources of stochasticity that might impact patterns of cultural diversity: cognitive processes at the individual level, random copy errors during the expression of traits, interactions between individuals in social networks, and random sampling at the population level. Questioning the fundamental differences between cultural and genetic drift might be as important as debating the correspondence between cultural and natural selection for assessing how valid the analogy between cultural and biological evolution is.
There might not be a single answer to the question of which stochastic processes underpin cultural evolution, but working towards answering that question might both facilitate the development of methodologies better suited to the analysis of cultural data and clarify the research program pursued by cultural evolutionists. For instance, the variability observed in cultural data is usually taken as the manifestation of statistical noise. However, this noise can contribute information that ought not to be neglected since it directly arises from the transmission of cultural traits and from demographic and psychological processes, i.e. from the stochastic processes underpinning cultural evolution (see Sect. 4 and “Appendix 1”). In addition, while it is possible to contrast the empirical diversity of cultural traits with predictions from different stochastic models, directly applying methods from population genetics to the analysis of cultural data might produce caveats because of the particularity of the stochastic processes involved in cultural evolution. For instance, different cultural traits do not necessarily share the same demographic and transmission histories, and they might transmit along different routes (e.g. some traits are transmitted from parents, others from peers, teachers or friends; see Boyd et al. 1997 for further discussion). One of the biggest challenges for the development of stochastic models in cultural evolution is to work out how to use the variability of cultural traits as a proxy for inferring the demography or history of populations, disentangling all sources of variability from the level of the individual to that of the population.
Finally, we focused our paper on studies conducted by scholars using an evolutionary and quantitative approach. For cultural evolutionists, it is clear that culture evolves because of processes analogous to those involved in biological evolution, including random drift. However, the opponents of an evolutionary approach to culture argue that other processes underlie cultural diversity. For instance, Ingold (2004, 2007, 2013) believes that the transmission of cultural traits between individuals is only a metaphor (Ingold 2013, p. 15). He accounts for cultural diversity by invoking the constant development of individuals as self-organizing entities. In this view, what individuals know and think and how they behave is due to constant feedback interactions between their development and their ecological and social environments. The idea that cultural diversity has evolved from the self-organizing properties of humans rather than some kind of inheritance mechanism aligns itself well with theories that some biologists have proposed as alternatives to the neo-Darwinian paradigm in biology (De Tiège et al. 2015). Yet, similarly to followers of Cavalli-Sforza and Feldman (1981), the relevance of the analogy between biological and cultural processess is not discussed in the light of the role of stochasticity. Why that is the case remains an open question.