2.1. Conceptual challenges

Conceptual challenges are typically problems of definitional scope. The importance of establishing clear definitions of terms is obvious from survey results that lament the field's lack of ‘shared conceptual references and vocabulary’ and its need to find ‘common language’ (Brewer et al., Reference Brewer, Gelfand, Jackson, MacDonald, Peregrine and Richerson2017: 3). Meanings of ‘copying’, ‘imitation’ and ‘social learning’ are used equivocally across CES, impeding consistent operationalization in experiments. An influential formal model of cultural transmission purports to show that ‘environmental learning’ does not, but ‘social learning’ does, produce the ‘empirical ‘S-shaped’ cumulative adoption curves that dominate the diffusion of innovations literature’ (Henrich, Reference Henrich2001: 992). (The term ‘S-shaped curve’ refers to a sigmoid curve plotting data about the adoption of innovations characterized by, first, lengthy, laboured growth of early adopters followed by a quick increase in adoption, creating a high slope, and concluded by a slowly increasing phase with low slope and many adopters.) Tim Lewens argues (2015: 114–118) that this model treats its concepts of ‘environmental learning’ and ‘social learning’ as mutually exclusive. Henrich's definition of ‘environmental learning’ reads: ‘individuals acquire and evaluate payoff-relevant information about alternative behavioural options by action and interaction in their local social, economic and ecological environments’ (Henrich, Reference Henrich2001: 992). As defined, the meaning of ‘environmental learning’ overlaps with the meaning of ‘social learning’. That is, the two are not independently defined. This entails that ‘environmental learning’ could also be modelled to yield S-shaped curves. This appears inconsistent with Henrich's conclusions from the model.

If the definitions of ‘social learning’ and ‘environmental learning’ are not retooled, subsequent modelling and experimentation using these definitions are unlikely to overcome challenge (C3). This conceptual misstep appears to inform pointed empirical criticisms of the model's explanation (Henrich, Reference Henrich2004) of the contraction of the Tasmanian toolkit (Read, Reference Read2011; Vaesen et al., Reference Vaesen, Collard, Cosgrove and Roebroeks2016a, Reference Vaesen, Collard, Cosgrove and Roebroeksb; Henrich et al., Reference Henrich, Boyd, Derex, Kline, Mesoudi, Muthukrishna and Thomas2016). This preliminarily demonstrates the value of critical philosophical attention on definitions in CES; see Section 3 below.

2.3. Topical challenges

Sub-disciplines within CES face messy challenges when researchers port coarse-grained focal concepts from one subfield to another. In the Cultural Evolution Society survey (Brewer et al., Reference Brewer, Gelfand, Jackson, MacDonald, Peregrine and Richerson2017: 1–2), the challenge of synthesizing knowledge across diverse research areas was rated as the field's most pressing concern. Consider the status of a prominent CES explanation of ‘cooperation’, which says that cooperation within groups emerged because of psycho-social consequences of belief in punitive gods (Johnson, Reference Johnson2005; Norenzayan, Reference Norenzayan2013). This causal claim was supported by diverse studies ranging from anthropological databases that identified statistical relations between a culture's adoption of a moralistic high god and markers of cooperation (Roes & Raymond, Reference Roes and Raymond2003) to economic games experiments that showed priming with religious concepts raised offers to strangers (Shariff & Norenzayan, Reference Shariff and Norenzayan2007), and from cognitive science of religion research on minimally counterintuitive ideas (Norenzayan et al., Reference Norenzayan, Atran, Faulkner and Schaller2006, although see Purzycki & Willard, Reference Purzycki and Willard2016) to psychological studies of agent-based teleological thinking (Kelemen, Reference Kelemen1999).

Yet CES-inspired study of religion and cooperation can succumb to the aforementioned challenges. As part of a wider problem in the social sciences, religious priming studies have suffered from multiple failures to replicate (e.g. Gomes & McCullough, Reference Gomes and McCullough2015; Miyatake & Higuchi, Reference Miyatake and Higuchi2017). While cross-cultural experimental evidence bolsters the supernatural punishment–cooperation relationship (Lang et al., Reference Lang, Purzycki, Apicella, Atkinson, Bolyanatz, Cohen and Henrich2019), the purported causal relation between supernatural punishment and social complexity (an index of broader cooperation) remains unclear (Watts et al., Reference Watts, Greenhill, Atkinson, Currie, Bulbulia and Gray2015) and more basic conceptual concerns about presumed psychological mechanisms linger (Schloss & Murray, Reference Schloss and Murray2011). Echoing our ‘general problem’, critics argue that the concept of ‘cooperation’ at work in models and experiments about effects of belief in supernatural punishment on cooperation often equivocates because within-group parochiality is confused with cooperation between non-kin (Galen, Reference Galen2016).

Critics might argue that research on this topic is consistent with scientific progress owing to its self-correcting nature, and as a result, the community does not require theory-driven re-evaluation of the subfield but more detail-oriented diligence. While we could be incorrect, it seems that focus on prosaic aspects of CES research can camouflage hurdles still thwarting progress. Most hypotheses tested in this subfield were not logical derivatives from a principled theory in part because the concept of a ‘high god’ lacks adequate formality and consistency across operationalizations. Some advocates define the term ‘high gods’ as ‘gods who cared about cooperative – and harmony – enhancing behavior’ (Shariff, Norenzayan, & Henrich, Reference Shariff, Norenzayan, Henrich, Schaller, Norenzayan, Heine, Yamagishi and Kamedda2010: 124); others defined ‘high gods’ as deities specifically responsible for creating reality, and who may or may not be ‘specifically supportive of human morality’ (Swanson, Reference Swanson1960). Confusions of this type put researchers’ ability to falsify causal hypotheses about religious beliefs and cooperation at risk since, if a correlation was not discovered in a dataset, the former circular definition warrants the inference that the dataset represented a ‘low’ not a ‘high’ god (Nichols et al. (Reference Nichols, Slingerland, Nielbo, Kirby and Logan2020: 4), and see Table 1 in that paper for similar logical problems in this subfield; see Lightner et al. (Reference Lightner, Bendixen and Purzycki2023), on the empirical implications of these conflations). Below, Section 5 analyses theoretical difficulties as found in the topical areas of non-human culture and language.

2.4. Implications and applications

There are many appreciable consequences of CES findings for public policy, social change, well-being and research ethics. Aside from a few exceptions (e.g. Wilson, Reference Wilson2011) these implications have not been thoroughly studied. Section 5 seeks to open this area with discussion of one application (about norms) and one implication (about human uniqueness).

These case studies provide ample justification for the value of theoretical and philosophical attention given to major challenges to CES. To improve the conceptual clarity and theoretical discernment in ongoing CES research, and because it may be unfamiliar to readers, we next encapsulate extant philosophy of science research devoted to challenges voiced in the survey results above.

4.1. Models in cultural evolution

As cultural evolution developed into a scientific field, innovative mathematical and computational techniques were devised to represent intricate features of cultural processes (Kandler et al., Reference Kandler, Perreault and Steele2012). Broadly speaking, three important modelling frameworks have helped to improve our theoretical understanding of evolutionary processes. The first framework is composed of gene–culture coevolution models, which assume a continuous, bidirectional interaction between genes and culture (Cavalli-Sforza & Feldman, Reference Cavalli-Sforza and Feldman1981; Boyd & Richerson, Reference Boyd and Richerson1985). Next are cultural neutral models, which assume that all variants are equally likely to be adopted, and attempt to understand cultural evolution in the absence of selective forces by assuming that changes in cultural variants are due solely to random drift and innovation (Bentley et al., Reference Bentley, Hahn and Shennan2004; Blythe, Reference Blythe2012; Reali & Griffiths, Reference Reali and Griffiths2009). Finally, the cultural selection framework assumes that some variants spread more rapidly than others, making evolution a process that is not neutral, but subject to biases (e.g. systematic cognitive and transmission patterns deviating from rationality) or selection pressures (e.g. environmental factors that affect fitness of variants) (Boyd & Richerson, Reference Boyd and Richerson1985; Hagen & Hammerstein, Reference Hagen and Hammerstein2006; Henrich & McElreath, Reference Henrich, McElreath, Barrett and Dunbar2007; Vale et al., Reference Vale, Flynn, Kendal, Rawlings, Hopper and Schapiro2017).

While modelling frameworks make distinctive assumptions about the nature of cultural processes, these frameworks can each be instantiated at either of two common scales. Regarding the scale used to model cultural evolution, researchers in the field have distinguished between two general classes (Van Dyke Parunak et al., Reference Parunak, Savit, & Riolo and L1998; Kandler et al., Reference Kandler, Perreault and Steele2012). The first is composed of micro-scale models, which aim to simulate how individual units in a system behave and interact with one another. The best-known type of micro-scale model is the agent-based model, which consists of a system composed of a set of autonomous units called agents, which simulates the decision-making, behaviour and interactions of agents to understand how patterns emerge at the population level. Such models have been used to study segregation dynamics (Schelling, Reference Schelling1971), the emergence of cooperation (Axelrod & Hamilton, Reference Axelrod and Hamilton1981), the evolution of language (Centola et al., Reference Centola, Gonzalez-Avella, Eguiluz and San Miguel2007; Kirby et al., Reference Kirby, Griffiths and Smith2014) and network dynamics (Segovia-Martín et al., Reference Segovia-Martín, Walker, Fay and Tamariz2020), and to adjudicate between competing explanations for biocultural phenomena like footbinding (Chowdhury et al., Reference Chowdhury, Zhang and Nichols2022).

The second class is composed of macro-scale models, which attempt to capture the overall behaviour of the system. Macro-scale models may consist of ordinary or partial differential equations that contain one or more functions with their derivatives. In this context, the functions represent quantities (e.g. populations A and B), the derivatives define their rates of change and the equation defines the relationship between both variables. These models have been used to study topics such as language competition (Abrams & Strogatz, Reference Abrams and Strogatz2003), political competition (Misra, Reference Misra2012), the evolution of religiosity (Rowthorn, Reference Rowthorn2011), the distribution of cultural traits (Manrubia & Zanette, Reference Manrubia and Zanette2002) and innovation (Kandler & Laland, Reference Kandler and Laland2009), among other social issues (see Brown, Reference Brown2007). Other macro-models, such as phylogenetic or statistical inference models, are oriented to encompass the broader evolutionary relationships between organisms and groups using existing experimental, historical and archaeological data (Turchin et al., Reference Turchin, Whitehouse, Francois, Slingerland and Collard2012).

An appreciable challenge facing the use of modelling to draw substantive inferences about actual cultural processes is that varied costs involved in selection of a framework and a scale for analysis often limit the inferential range of the model's results. Micro-scale models allow more control over the properties of individual entities in the system, generally at the expense of higher computational cost and greater complexity of formalization, which limits their reproducibility and generalizability. Macro-scale models, on the other hand, allow greater control over population variables and the evaluation and formalization of the dynamical system, although at the cost of a lack of control over the individual entities of the system. From these two types of models emerged a general theoretical concern over how to integrate their results.

Of very special consideration in models of cultural evolution is the development of techniques to represent multivariate causal interactions between phenomena at both micro- and macro-levels, e.g. the ‘ratchet effect’ (Tennie et al., Reference Tennie, Call and Tomasello2009), often regarded as indispensable for explaining cumulative cultural evolution. Recent formal developments enabling the combination of micro- and macro-models are allowing modellers to meet this need and, consequently, they will stimulate high-quality research in CES and assist in overcoming challenge (C3). Micro- and macro-models can be combined by building adaptive dynamics models or with multilevel selection models where both events at the individual level generate effects at the group level and events at the group level affect the individual properties of the system (Mullon & Lehmann, Reference Mullon and Lehmann2017; Fishwick, Reference Fishwick1995; McElreath & Henrich, Reference McElreath, Henrich, Barrett and Dunbar2007; Waring et al., Reference Waring, Goff and Smaldino2017; Segovia-Martin & Tamariz, Reference Segovia-Martin and Tamariz2021). This represents the future of modelling in cultural evolution, but it is not itself without challenges.

Despite numerous advantages, joint micro-macro models are now confronting novel theoretical problems. These include reconciling the role of individual-level behaviours with aggregate-level phenomena; addressing emergent properties that cannot be explained by the actions of individual agents alone; accounting for feedback loops between the micro- and macro-levels; and specifying the content, behavioural influence and population-level effects of ‘regulatory traits’, i.e. traits governing cultural transmission of other traits (Acerbi et al., Reference Acerbi, Ghirlanda, Enquist, Cagnoni, Mirolli and Villani2014). Given their content, progress on these problems would benefit from formal analyses of concepts like collective action, coercion and institutions, which would facilitate operationalization in models.

4.2. Experiments in cultural evolution

Challenge (C3) focuses on the need to overcome several problems facing the effort to model culture as a complex, adaptive system. In their report of survey results, authors wisely distinguish this from a corollary challenge, (C4), which involves the need to develop an evidential bridge between modelling results and research on the psycho-social mechanisms instantiating cultural transmission in the world. Given the ascendant place of modelling in CES, validating models against data from experiments and field studies has remained a persistent challenge. This challenge for CES typically has one of four points of inception: (i) a lack of data, making validation unfeasible; (ii) a conceptual gulf between model and empirical data; (iii) limitations of the model's ability to properly describe new, previously unseen data; and (iv) a lack of parallelism between model mechanisms and actual mechanisms, making models of little use in capturing the causal interactions between real evolutionary processes (see Weisberg, Reference Weisberg2007, on assignment, scope and fidelity). Mechanisms for cultural evolution like social learning biases have of course been proposed (Boyd & Richerson, Reference Boyd and Richerson1985), modelled (Denton et al., Reference Denton, Ram, Liberman and Feldman2020) and experimented upon (Efferson et al., Reference Efferson, Lalive, Richerson, Mcelreath and Lubell2008). Yet CES nonetheless faces several problems integrating theory and modelling with experimentation often owing to the absence of a set of criteria for a measurable outcome that is commensurable across methods.

Take, for example, research that frames lab- and field-based experiments (see Mesoudi & Whiten, Reference Mesoudi and Whiten2008 and Tamariz, Reference Tamariz2017 for reviews). Cultural transmission has been a focus of laboratory experiments, many of which examine whether or not some of the postulated learning biases exist. Other studies pit multiple learning biases against each other to assess which are more powerful explanations of cultural retention and transmission. While some experiments focus on biased social learning, including, among others, content- (e.g. Boyer & Ramble, Reference Boyer and Ramble2001; Broesch et al., Reference Broesch, Barrett and Henrich2014) and prestige-biased cultural transmission (e.g. Jiménez & Mesoudi, Reference Jiménez and Mesoudi2019) in the retention and recall of stimuli, others utilize sequential transmission chains to examine ostensibly unique properties of human cultural evolution such as cumulative evolution and problem-solving (e.g. Caldwell & Millen, Reference Caldwell and Millen2008; Jiménez & Mesoudi, Reference Jiménez and Mesoudi2020). In step with the classic experiments of Bartlett (Reference Bartlett1932), many of these studies follow the logic and structure of ‘telephone’, a game that exploits distortion of information across chains of multiple learner-teachers. Other experiments stress communication and group coordination through actively and collectively aligning individual behaviours, such as drumming and movement, and examining downstream behaviour (e.g. Tunçgenç & Cohen, Reference Tunçgenç and Cohen2016; Wood et al., Reference Wood, Caldwell-Harris and Stopa2018). Aptly, but unsystematically, the range of participants’ agency varies significantly across cultural transmission experiments. This also applies to studies of cultural evolution in non-human animals, where a variety of evidence stemming from controlled experiments and anchored in social learning models has been observed (Whiten, Reference Whiten2019).

Much of the behavioural experimental work that falls under the rubric of cultural evolution assesses the magnitude and scope of cooperation that is unique to our species. By measuring individual decisions in allocating money and other resources, experiments in cross-cultural behavioural economics suggest that culture and cultural institutions play a significant role in promoting more cooperative strategies (Pisor et al., Reference Pisor, Gervais, Purzycki and Ross2020) or, less appreciated in CES, in reducing cooperation (cf. Akbari et al., Reference Akbari, Bahrami-Rad and Kimbrough2019). Studies that deploy the same experimental game but have culturally salient treatment conditions show that when made culturally relevant, salient beliefs and values can alter the course of interpersonal behaviour (Cronk, Reference Cronk2007; Gerkey, Reference Gerkey2013; Lightner et al., Reference Lightner, Barclay and Hagen2017). In such experiments, the outcome variable is cooperation, not whatever cultural information that is purportedly inducing it, and certainly not the transmission of such cultural information. Yet some paradigms have exploited such experiments, and rather than using beliefs and cultural framings to predict behaviour, some have reversed this structure and examined how game outcomes can predictably shift the content and structure of extant beliefs (Purzycki et al., Reference Purzycki, Stagnaro and Sasaki2020). Neither of these experimental paradigms directly speaks to transmission, however; one seeks to explain the evolution of cooperation and the other seeks to explain the evolution of beliefs as responses to threats to cooperation.

While the ultimate theory that motivates such studies posits that cultural norms can co-evolve with problems of cooperation, there remains a notable deficit in just how little of this work specifically attends to cultural transmission and the co-variation between cooperation and tradition. Experiments about variability of dynamics in transmission across cultures are infrequently motivated by integration with cross-cultural psychology and so, albeit with exceptions (e.g. Mesoudi et al., Reference Mesoudi, Chang, Murray and Lu2014, which found significant cross-cultural variation in the adoption rate of success-based social copying), they generally do not motivate or test explicit hypotheses about cultural differences in transmission. Many such studies might assess the cultural mechanisms that promote or reduce cooperation but do nothing in the way of accounting for the diffusion or maintenance of such beliefs. In other words, while we might see co-varying effects of cultural information and behaviour in the lab, and while this might make certain explanations or theories more plausible, it remains unclear what this has to do with the evolution and persistence of the cultural part of the dynamics. Are particular cultural strains prevalent because they promote (or reduce) cooperation? Do the social benefits of cooperation contribute to the prevalence and maintenance of particular cultural mechanisms? Such questions are ripe for serious experimental treatment, and returning to challenge (C4), the results of such experiments are expected to inform representations of subtle, varied causal interactions in future micro–macro models.

4.3. Field studies

Field studies fall into various kinds. Ethological field studies concern cultural performances beyond our species. Archaeological and palaeoanthropological excavations provide insight into the deep past and into the emergence of cultural evolution of hominins from australopithecines and different Homo species up to anatomically modern humans. Anthropological field studies, often separated into cultural anthropology and evolutionary anthropology, examine social and material interactions in present-day human groups and societies and their cultural products. Although these three fields differ markedly in the nature and acquisition of their basic data as well as in their theoretical presuppositions, they share important commonalities. Key features of all three approaches are: adoption of long-term data collection methods and analysis; attention to facts about the material, ecological and behavioural contexts of cultural behaviours; and interest in identifying a potential variety of contributing elements. Field studies in CES aim to gain a broad overview of cultural performances and their environmental conditions, and to examine the differences between distinct social and ecological settings in order to increase understanding of the historical and biological circumstances of cultural evolution. Field studies in CES incorporate fuzzy evidence about multi-faceted entanglements between agents, environments and behaviours, captured with a variety of data types ranging from anecdotal reports to behaviour coding to structured research protocols, within which meaningful patterns of relations and development can be detected.

Field studies have played and will continue to play the most prominent role in CES responses to challenge (C8), which encompass the need to convince policymakers of the need to conserve animal cultures. Until the second half of the twentieth century, culture had only been ascribed to humans. Since the expansion of the definition of culture to make it applicable to non-human animals (McGrew, Reference McGrew1992), and subsequent studies on behavioural traditions in chimpanzees (Whiten et al., Reference Whiten, Goodall, McGrew, Nishida, Reynolds, Sugiyama and Boesch1999), which combined field research from several groups in search of differences of occurrences of behaviour across sites under similar environmental conditions, ethological field studies with cultural evolutionary scope have been on the rise. The variety of animal species examined as well as evidence of performances with basic cultural features have markedly increased. Ethological research has produced studies on the pace of animal innovations, as found in seasonal changes in songs of humpback whales (Schulze et al., Reference Schulze, Denkinger, Oña, Poole and Garland2022), on cultural transmission in the manufacture of tools, as in New Caledonian crows (Hunt & Gray, Reference Hunt and Gray2003), and on the adoption of new behavioural elements from resident individuals when they must adapt to new habitats after dispersal, as in male orangutans (Mörchen et al., Reference Mörchen, Luhn, Wassmer, Kunz, Kulik, Noordwijk and Schuppli2023). The current state of the field suggests that there is a variety of basic cultural performances in a wide array of animal orders from invertebrates to birds and mammals (Whiten, Reference Whiten2021) and that the number of (unspecified) cases of ‘culture’ is underestimated (Schuppli & Van Schaik, Reference Schuppli and Van Schaik2019).

In contrast to field studies on animal behaviour and ethnography, which include direct observations of performances, primary data from archaeological field studies provide only indirect insights into past human behaviour through trace material evidence. Performances, and thus also group-specific patterns and their development, must be inferred from finds via multiple analytical steps before interpretation (Perrault, Reference Perrault2019). The chronological frame of archaeological field studies spans more than 3 million years. Research focused on distinct spans of time within this epochal stretch has enabled data collection from the material record to respond to quite different questions about cultural evolutionary processes. Inspired by behavioural studies on animals, evolutionary archaeological research focusing on an early phase up to 500,000 years ago (Shennan, Reference Shennan2008) is dominated by the distinction of primate and hominin cultural capacities (Tennie et al., Reference Tennie, Premo, Braun and McPherron2017), long-lasting technical traditions and convergent evolution in technologies (O'Brien et al., Reference O'Brien, Buchanan and Eren2018), and the development of simple stages of communication and cooperation (Gärdenfors & Högberg, Reference Gärdenfors and Högberg2017). For a middle phase between 500,000 and 100,000 years ago, the research focus shifts to the study of evidence for the increasing complexity of technical expressions (Lombard & Haidle, Reference Lombard and Haidle2012) and the necessity of advanced forms of cultural transmission (Stolarczyk & Schmidt, Reference Stolarczyk and Schmidt2018), group structures and the advent of ritual behaviour (Dapschauskas et al., Reference Dapschauskas, Göden, Sommer and Kandel2022). In a late phase from 100,000 years on, relations of techno-complexes and social systems as well as the development of group identities (Vanhaeren & d'Errico, Reference Vanhaeren and d'Errico2006) and belief systems become increasingly important. The analysis of spatial, material and social networks and their dynamics (Mills, Reference Mills2017) aligns with other anthropological approaches to questions about the biocultural evolution of our species.

Ethnographic field studies are a major means for collecting data within cultural anthropological research. Although the method of participant-observation has well-known limitations (Howell, Reference Howell and Stein2018), this is not the only tool in the anthropological tool kit. Complementing participant-observation are long-term and open-ended participatory methods that allow for individuals within the studied population to speak for themselves. Alongside these are tools like qualitative and structured interviews, genealogy and network analyses (Beer & König, Reference Beer and König2020; Seligmann & Estes, Reference Seligmann and Estes2020). Ethnology, the comparative study of peoples, represents another approach to answering canonical CES-inspired questions under the field study umbrella. Although ethnology has also long been subject to critique, it has never been absent from anthropological research (Schnegg, Reference Schnegg2014). More recently, comparative methods have become supercharged by the increased accumulation and availability of cross-cultural data together with rigorous comparative methods (Fox & Gingrich, Reference Fox and Gingrich2002). Such methods are now applied to a broad anthropological domain, expanded by inclusion of urban settings, the digital world and specific groups (e.g. professional, religious, gender) within larger communities. With a growing interest in the universality of human rights (Nagengast, Reference Nagengast1997), relations of performance and cognition (Antweiler, Reference Antweiler, Meusburger, Freytag and Suarsana2016), and the origins and evolution of broader concepts such as religion and morality (Purzycki et al., Reference Purzycki, Henrich, Apicella, Atkinson, Baimel, Cohen and Norenzayan2018), more universal perspectives have been taken.

Contributions of evolutionary anthropology (EA) and its sister, human behavioural ecology (HBE) (differences between which are left aside here) prize coherence with other methods of knowledge discovery and prioritize theoretical relations to natural selection. Unlike some areas of evolutionary science, work in EA and HBE generally avoids optimality assumptions that often characterize evolutionary psychology (about this see Fessler et al., Reference Fessler, Clark, Clint and Buss2016: 1030). The frameworks of EA and HBE occupy an unusually high-traffic crossroads within CES and, to extend the analogy, these crossroads are located near wide bridges traversing the waters between cultural and biological evolution. Despite their pluralistic use of methods and their ecumenical approach to research questions, EA and HBE are unified by the assumption, put roughly, that concepts from evolutionary biology have considerable utility in explanations of gene–culture co-evolutionary phenomena. Evolutionary anthropologists might study modifications of the hip, pelvis and spine revealed in the hominin archaeological record (Lovejoy, Reference Lovejoy2005a, Reference Lovejoyb), or information about ancestral pathogen loads across regions in relation to present-day, culturally variable social psychological traits (Fincher et al., Reference Fincher, Thornhill, Murray and Schaller2008; Faulkner et al., Reference Faulkner, Schaller, Park and Duncan2004). Modifications of the hominin body plan supporting bipedality, and cross-cultural variation in social-psychological factors in relation to pathogen load, are hypothesized to result, in the end, from their consequences on biological fitness. The ready explanatory pluralism of HBE and EA has probably contributed to their nuanced awareness of evolutionary trade-offs at play across human life history, e.g. between current and future reproduction and quantity and quality of offspring (Hill & Kaplan, Reference Hill and Kaplan1999); their attention to both uniformity and difference between primate ‘proto-shame’ and human shame systems (Fessler, Reference Fessler, Tracy, Robins and Tangney2007); and their pleasing, simultaneous concern with multiple dimensions of a single scientific explanation, dimensions such as physical realization, causation and phylogeny, which are often treated autonomously elsewhere in CES.

Working with results of field studies in the context of findings from elsewhere in CES makes a special contribution to challenges about knowledge synthesis and interdisciplinary integration (C5). This is not least because it is in the form of field studies that the bulk of research on non-human culture occurs. Operationalizing the culture concept across archaeology, biological anthropology and ethology, and applying it to behaviourally modern humans, earlier humans and non-humans, each requires improvements to this common but coarse-grained concept. Without more unanimity on the culture concept, survey results suggest that ‘siloing’ will continue (on the problems of disciplinary ‘siloing’ see Brewer et al., Reference Brewer, Gelfand, Jackson, MacDonald, Peregrine and Richerson2017: 2). Perhaps, for example, a law-like systematicity links the roles that social learning plays in the transmission of behavioural traits within different animal species and within hominin groups, but the underlying theory and methods issues have yet to be adequately resolved to support law-like cross-species generalizations about necessary conditions for the ontogeny of culture.

5.1. Culture in other animals

The existence of behavioural traditions passing (not necessarily replicating) between generations through social learning in animals – a working characterization of ’culture’ – is now uncontroversial. As a result, questions of interest to researchers have moved to focus on the mechanisms that support cultural transmission in animals, and questions about the methods with which animal cultures are studied. Answers to these questions can tell us whether human forms of culture – not least cumulative culture – are likely to be found in the animal kingdom.

Historically, animal cultures have been studied via the ‘method of exclusion’ (Whiten et al., Reference Whiten, Goodall, McGrew, Nishida, Reynolds, Sugiyama and Boesch1999; later refined by Langergraber et al., Reference Langergraber, Boesch, Inoue, Inoue-Murayama, Mitani, Nishida and Vigilant2011). This identifies behavioural patterns as cultural when they differ between geographically diverse groups of the same species. If genetic differences between these groups can be ruled out (e.g. because of inter-group migration), and behavioural differences cannot be attributed to local ecological conditions, then behavioural patterns are likely to be products of social learning – and so are indicative of variant cultural traditions. While influential, this method has been criticized as insufficiently informative, and as resting on a conceptual problem, specifically a false dichotomy between nature and culture (Rogoff, Reference Rogoff2003). First, it probably undercounts culture, where behaviours are universal (and so do not differ between populations) but are nonetheless socially learned. Second, it gives no insight into the kinds of social learning mechanisms that support cultural transmission (Sterelny, Reference Sterelny, Laland and Galef2009; Acerbi et al., Reference Acerbi, Snyder and Tennie2022). This criticism is important because some higher fidelity social learning processes, such as imitation (aka ‘process copying’ Buskell & Tennie, Reference Buskell and Tennie2022, or ‘know-how copying’ Tennie, Reference Tennie2023), are thought to play an important (if not always necessary) role in the emergence of cumulative culture (Tennie et al., Reference Tennie, Call and Tomasello2009; Moore, Reference Moore2013). These may be missing in other species, or too inconsequential to produce and maintain similar cultures.

Tennie and colleagues have developed a series of paradigms designed to test whether putatively cultural behaviours in our nearest relatives, the great apes, are a product of imitation. Early paradigms established that chimpanzees do not acquire behaviours whose transmission requires high-fidelity transmission chains (Tennie et al., Reference Tennie, Call and Tomasello2009). More recent studies show that many great ape behaviours can be explained as products of serial reinvention (often through trial-and-error learning) (e.g. Bandini et al., Reference Bandini, Grossmann, Funk, Albiach-Serrano and Tennie2021). Since naïve apes spontaneously reinvent these behaviours when left alone with the necessary apparatus, they cannot depend on social learning. Tennie and colleagues call these behaviours ‘latent solutions’. They do not deny that some forms of social learning support the transmission of great ape behaviours (Tennie et al., Reference Tennie, Bandini, Van Schaik and Hopper2020). However, they argue that great apes do not copy new forms of ‘know-how’ from peers. Rather, for the most part, great ape social learning occurs because features of the environment are made salient through the actions of others. On this account, the presence of others in a learning environment is conceived of as a ‘probabilistic difference maker’ (Scarantino, Reference Scarantino2015): a non-essential but nonetheless causally significant factor. It may facilitate the acquisition of behaviours without being essential for their development.

While it should now be uncontroversial that great apes do not, in their natural habitat, engage in high-fidelity know-how copying, the finding that great apes can learn to imitate (e.g. Pope et al., Reference Pope, Taglialatela and Hopkins2017) raises the question of whether their lack of imitation is itself a product of their developmental environment, and so potentially a culturally evolved ‘cognitive gadget’ (Heyes, Reference Heyes2018). Nonetheless, even trained apes seem unable to retain cultural inventions over successive generations (Tennie, Reference Tennie2019). This means that chimpanzees must often reinvent their ancestors’ technologies in ways that human communities need not, and explains why cumulative culture is rare, at best, in great ape communities. Tentative reports of cumulative culture in whales and monkeys suggest that further empirical research is needed to determine whether and how animal cultures can gain in complexity over time. Theoretical research specifying what would constitute evidence of cumulative culture (and how ‘acultural’ explanations from natural selection would be ruled out) must precede or at least accompany additional field studies. A valuable contribution by theoreticians here would be to assist in fixing the social and cognitive conditions that are necessary and sufficient for cumulative culture, and by developing hypotheses about the kinds of mechanistic changes that support the development of higher fidelity social learning in non-human animals (Berio & Moore, Reference Berio and Moore2023).

A final question asks whether non-human animals, and especially apes, are capable of understanding norms. Norm cognition is taken to be a prerequisite of many of our culturally evolved social institutions (e.g. institutions of justice). While some are sceptical that norm cognition is present in great apes (Schlingloff & Moore, Reference Schlingloff and Moore2017), others have argued that this scepticism is driven by an intellectualized account of the cognition required for normativity, suggesting a need for minimal models of norm cognition (Westra & Andrews, Reference Westra and Andrews2022). A further obstacle remains for investigating norm cognition in animals. The best method for identifying norms is through third party enforcement of perceived norm violations – and this is strikingly absent from great apes (Schlingloff & Moore, Reference Schlingloff and Moore2017). The theoretical formulation of alternative methods for identifying normative behaviours would thus constitute a valuable contribution to the field.

5.2. Languages as culturally evolving systems

Until not long ago, discussion of language ‘evolution’ was somewhat taboo. As recent as 1994, textbooks observed that ‘Evolution … has become a “dirty word” in modern linguistic theory’ (McMahon, Reference McMahon1994: 314). This is because through the twentieth century the nomenclature of evolution risked some allusion to outdated, often racist, views about how some languages – and by extension some social groups – might be ‘more evolved’ than others. It is now uncontroversial within linguistics that no language is inherently ‘better’ or ‘more useful’ than others. So, while vigilance must be maintained, the socio-political risks associated with evolutionary talk have begun to fade, and in the past 30 or so years evolutionary approaches to languages have grown dramatically in prominence (Tallerman & Gibson, Reference Tallerman and Gibson2011; Dediu & de Boer, Reference Dediu and de Boer2016).

Today, the expression ‘language evolution’ is widely used to describe at least two distinct phenomena and one methodological advance. The two phenomena are: the biological process by which humans, and apparently only humans, became a ’language-ready’ species; and the set of cultural – or ‘cultural evolutionary’ – processes by which relatively simple and unstructured systems become highly structured, and hence acquire some of the common, characteristic properties of languages, such as symbolism, compositionality and duality of patterning. The methodological advance is the application of phylogenetic tools, derived from population genetics, to study language history and language change. All three of these literatures are now very large, comprising thousands of papers each. Multiple past papers review, summarize and synthesize them (e.g. Mace & Holden, Reference Mace and Holden2005; Tomasello & Call, Reference Tomasello and Call2019; Greenhill et al., Reference Greenhill, Heggarty and Gray2020; Haspelmath, Reference Haspelmath2020; Roberts & Sneller, Reference Roberts and Sneller2020; Scott-Phillips & Heintz, Reference Scott-Phillips and Heintz2023; Scott-Phillips & Kirby, Reference Scott-Phillips and Kirby2010). Here we focus on a philosophical issue arising from this growing influence of evolutionary thinking.

The challenge, put simply, is: what exactly does the evolutionary perspective bring that other approaches do not? The biology of language, language emergence and language change have all been important topics for language science for a long time, and a great deal of what has been uncovered by evolutionary approaches has previously been investigated and described in other terms. Research adopting an evolutionary perspective has enriched our understanding, provided new methods and added many new findings, but does it fundamentally alter our understanding of what the empirical phenomenon is and how it works?

One answer might be that a relatively faithful transposition of the Darwinian model – where variation, selection and inheritance combine to generate natural selection and hence the appearance of design in nature – is possible, and brings with it new insights, explanations and tools. With languages, there is something like variation (linguistic items vary enormously), there is something like selection (some items become more common than others) and there is something like inheritance (we learn from the previous generation). Some linguists indeed propose that linguistic items, of some sort or another, could be identified as units of selection closely analogous to genes (e.g. Croft, Reference Croft2008; Ritt, Reference Ritt2004; Tamariz, Reference Tamariz2019). However, deeper analysis raises difficult questions, and the issue is contentious: a decade ago, a group paper on language as a culturally evolving system observed that, ‘Various scholars have proposed that concepts, cultural behaviors, or artifacts may function as replicators. It remains to be seen whether any, all, or some combination of these entities are reasonable candidates for cultural replicators’ (Dediu et al., Reference Dediu, Cysouw, Levinson, Baronchelli, Christiansen, Croft and Lieven2013: 314). The situation has not changed substantially since.

Another response to this question would be to adopt only specific parts of the Darwinian toolkit. That is, one could reduce the role of an analogy between variation in language and variation in genes while retaining other parts of Darwinism (Claidière et al., Reference Claidière, Scott-Phillips and Sperber2014). For instance, there might be scope for perspectives that adopt population thinking and some commitment to selectionism, but without the stronger commitment to replicators.

Languages are arguably the cultural domain par excellence for population thinking because languages are, more than any other cultural phenomena, especially amenable to ‘atomization’ – to being studied in terms of discrete items – and hence to identification and analysis of the types and their changing frequencies over time. As such, an emergent and pressing theoretical question is: how best can linguistic types and tokens be precisely identified to facilitate cultural evolutionary study and modelling? Constructionist approaches to grammar, which describe grammars as networks of form–function pairs (Goldberg, Reference Goldberg2003; Hilpert, Reference Hilpert2014, inter alia), may be especially suitable for this goal. However, because there is little research developing this idea, treating constructions as the targets of population thinking (partial exceptions are Kirby, Reference Kirby1999; Steels, Reference Steels2016) remains necessary.