2.1. Individual-Level Explanations of Human Culture

The commonest of all types of cultural evolutionary explanation are those focused on individual-level human culture. Cultural evolutionary work is called evolutionary for multiple reasons. The primary reason is its focus on understanding how culture is transmitted and evolves in various types of human populations. Culture is understood as the effect or output of a biological adaptation (also sometimes treated as “culture”; I will come back to this shortly), a capacity that has the function of permitting humans to acquire phenotypic traits that on the whole raise their fitness without having to go through the lengthy process of ordinary natural selection (Boyd and Richerson Reference Boyd and Richerson1996). One consequence of this is that individual-level models treat culture at once as something primarily possessed by individual humans as “information … capable of affecting [their] behavior” (Richerson and Boyd Reference Huber2005, 5) and something that takes the form it does because those humans nonetheless live in a particular kind of social environment or population in which particular kinds of cultural evolutionary processes occur (Boyd and Richerson Reference Boyd and Richerson1985). Another consequence is that researchers in these various fields describe culture as, or in terms of, certain phenotypic traits of individuals. Occasionally (mostly in informal explanations) individual-level culture is described as a series of psychological traits: consider, for example, the values of the Pennsylvania family farmers described by Richerson and Boyd (Reference Huber2005), the preferences for seeing free riders punished on the behavioral economists’ explanations of strong reciprocity (Bowles and Gintis Reference Bowles and Gintis2011), or the mental representations in Sperber’s (Reference Sperber1996) explanations of reconstruction. Formal models, however, often describe the transmission of behaviors or patterns of behavior, for example, punishment (Henrich and Boyd Reference Henrich and Boyd2001b) or farming (Aoki, Shida, and Shigesada Reference Aoki, Shida and Shigesada1996); even Sperber, writing with a colleague, does this on one simple model of the transmission of smoking behavior (Claidiere and Sperber Reference Claidiere and Sperber2007). In some cases, the models or explanations instead describe the evolution of an artifact (such as a kayak or spear): this is obviously common among paleoanthropologists (e.g., Lycett and von Cramon-Taubadel Reference Lycett and von Cramon-Taubadel2008; Eren, Buchanan, and O’Brien Reference Eren, Buchanan and O’Brien2015) but also occurs in the standard cultural evolution literature in which the case being considered requires using paleoanthropological data (see, e.g., Henrich Reference Henrich2004).

The (on the whole) adaptiveness of human culture is proposed to be primarily due to the nature of social learning. Social learning in the simplest sense means something like “learning from others” rather than “learning by oneself.” However, social learning can involve any degree of social mediation of learning from detailed, high fidelity imitation all the way to forms of learning common in animals (such as goal emulation or stimulus enhancement) whereby environments are constructed by conspecifics in such a way that trial-and-error individual learning by individuals reliably leads to their acquisition of the appropriate trait (Whiten Reference Whiten2000; Heyes Reference Heyes2012). Different social learning methods result in different cultural evolutionary processes (Henrich and Boyd Reference Henrich and Boyd2001a; Henrich and Gil-White Reference Henrich and Gil-White2001). All social learning of any type, however, involves exploiting the social environment and the similarities between self and others to potentially reduce the costs of learning (Aoki and Feldman Reference Aoki and Feldman2013); given complex, cumulative evolution, individuals can thereby acquire much more adaptive behaviors than they could alone (Boyd and Richerson Reference Boyd and Richerson1996). While cultural evolutionists such as the dual inheritance theorists and EoR researchers disagree on what goes into the social learning process in humans—that is, imitative learning for the dual inheritance theorists (Boyd and Richerson Reference Boyd and Richerson1985) and something close to reconstruction of a trait from impoverished evidence for the EoR group (Sperber Reference Sperber and Aunger2000)—both nevertheless accept that social learning is the core feature of cultural transmission.

Finally, perhaps the feature of culture on these models that makes them genuinely evolutionary is that the consequence of social learning is that a successful cultural trait (whether it be a behavior, artifact, or mental representation) is repeatedly copied or reconstructed (although not necessarily very precisely—more on this later) such that, generally, tokens of the same type of trait form a lineage or set of lineages in a population.Footnote ³ Processes of cultural evolution can then generally lead to the differential creation or survival of lineages in a population via a variety of different cultural processes (Boyd and Richerson Reference Boyd and Richerson1985).

2.2. Individual-Level Explanations of Animal Culture

Another type of phenomenon on which cultural models focus is individual-level culture as found in nonhuman animals, especially close relatives of humans such as chimpanzees. The animal culture described in these models is primarily also composed of individual-level, socially learned, phenotypic traits that form lineages and spread through populations. For example, Whiten and colleagues identify 65 behavior patterns that appear to be present in a variety of chimpanzee groups (Reference Whiten, Goodall, McGrew, Nishida, Reynolds, Sugiyama, Tutin, Wrangham and Boesch2001). However, nonhuman animal culture also has some significant differences from the parallel phenomenon in humans. For one, chimpanzee culture relies more than human culture on socially facilitated trial-and-error learning such as goal emulation or stimulus enhancement. Both types of cases require the chimpanzees to engage in significant learning on their own in order to acquire the appropriate behavior (Whiten Reference Whiten2000).

While nonhuman animal culture is of significant interest in its own right, a major reason for pursuing this work is to understand the relationships between human and nonhuman animal culture (Whiten Reference Whiten2000; Laland and Hoppitt Reference Laland and Hoppitt2003). Chimpanzee culture, for example, as a capacity, is often treated as a potential homolog of human culture, and culture in less closely related animals as an analog of human culture (Galef Reference Galef1992). In this sense then, “individual-level culture” in these types of explanations is taken to refer to something that both chimpanzees and humans can have in common, and this is not just the presence of individual-level socially transmitted phenotypic states but possibly some of the competences that underlie them (I return to this shortly).

2.3. Niche Construction Models

Another phenomenon on which cultural evolutionary models have focused is the phenomenon of niche construction, the idea that animals can modify the selection pressures acting on them and their lineage by modifying the environment and then transmitting that modified environment to their offspring or members of the next generation in their population or group. While these models are primarily aimed at understanding how the transmission of environments can modify the transmission of genes, similar effects can occur with cultural traits, and the cultural evolutionists have developed corresponding cultural evolutionary and gene-culture niche construction models (Laland, Odling-Smee, and Feldman Reference Odling-Smee and Feldman2001; Odling-Smee, Laland, and Feldman Reference Odling-Smee, Laland and Feldman2003). Even in some fairly classic gene-culture models, the “culture” is also partly composed by the environment; for example, in the famous sickle cell case, it is agricultural changes to the environment that led to mosquito proliferation and the changes in the distribution of genes for sickle cell anemia (Feldman and Laland Reference Feldman and Laland1996). These models lead to a relatively natural understanding of these modified environments as culture in themselves. For example, Odling-Smee and Laland, taking a “broad understanding of culture,” argue that “cultural inheritance does not comprise only the transmission of brain-based knowledge through teaching and social learning. … Generations of humans also inherit culturally modified rural and urban environments—fields of crops, terraces, artificial lakes, canals, roads, schools, hospitals, factories, police forces, electricity grids, waste products, and pollution” (Reference Odling-Smee and Laland2011, 227). Similarly, Richerson and Boyd, for example, accept that the information they think comprises culture can be in environmental changes and in artifacts (Reference Huber2005, 61).

However, this “culture as modified environment” is not just a simple “extended phenotype” caused by the mental states and behaviors making up ordinary individual-level culture. It can be the consequence of group behavior, it can persist and affect the fitness of future generations, and it is not always replicated. What is more, a variety of authors accept that modified environments also act as part of the social learning system in their own right: for example, Richerson and Boyd (Reference Huber2005) and Odling-Smee and Laland (Reference Odling-Smee and Laland2011), among others, agree that part of the information that permits social learning can be stored in the environment.

2.4. Group-Level Explanations of Human Culture

A fourth type of work in the cultural evolution literature treats culture as a phenomenon that also occurs at the group level. There are a variety of slightly different types of group-level culture, but the upshot of all of these is that it is possible for groups to possess group-level cultural traits that have an evolutionary history in their own right.

All group selection models, cultural or genetic, involve an interaction between two evolutionary forces: a selective force acting on group-level traits and a selective force acting on individual-level traits; most commonly in group selection models the group-level trait is simply some aggregate of the traits of the individuals making up the group (e.g., the proportion of altruists making up a group). The forces acting on the group-level traits (e.g., the fitness effects of there being a high proportion of altruists in the group) then interact with the forces acting on the individual-level traits (e.g., the fitness effects of being a nonaltruist within that group) to determine the fate of the whole group (Sober and Wilson Reference Sober and Wilson1998). The fitness effects of these forces can then be felt at the level of the individual (in the group [and individual] selection process called “Multi-level Selection 1,” or MLS1) or at the level of the group (in the group [and individual] selection process called “Multi-level Selection 2,” or MLS2). This means that the consequence of positive group selection is either an increase in the number of individuals making up the group (MLS1) or (in MLS2) that the group itself splits and produces more groups (Okasha Reference Okasha2006). Such processes are generally referred to as “cultural group selection” when the group-level traits are composed of traits that are socially transmitted (rather than genetically heritable) at the individual level. Cultural group selection can take similar forms with models that are both effectively MLS1 (Henrich and Boyd Reference Henrich and Boyd1998) and MLS2 (Soltis, Boyd, and Richerson Reference Soltis, Boyd and Richerson1995; Bowles Reference Bowles2006). However, either way, the group-level cultural trait is taken by these models to have an effect and a cultural evolutionary history in its own right, which interacts with and even against the effect of the individual-level traits in the same group.

To make this more obvious, consider multilevel selection explanations that are not simply aggregative but in which there are coherent, functional, group-level behaviors or traits of whole groups and organizations that are composed of potentially highly various contributions from their members. One obvious example is from Wilson’s (Reference Wilson2002) work on the (cultural) evolution of religion, which sees religion as a cultural group-level adaptation for constituting individuals into an organization that can drive adaptive collective action and promote group or individual fitness. Many religious groups, businesses, and other organizational activities are made up of interacting human parts: consider a religious ritual in which different members perform different behaviors possibly using different artifacts or constructed environments. In these sorts of cases the functional effects of these group-level traits do not come just from the particular contribution of any individual-level trait but from their interaction in a group-level whole.

Another plausible “group-level” cultural phenomenon is found in work on cultural phylogenies. Cultural phylogenetics examines large-scale traditions in individual-level cultural traits. Traditions are “bundles” of lineages of similar, usually related, individual-level cultural traits within a group; as such they constitute group-level cultural phenomena. Traditions can vary over time and between groups because of the nature of the lineages composing them; cultural phylogenetics seeks to determine the relationships between similar traditions in different groups, in particular whether they are formed by phylogenetic branching or assimilation, by examining patterns of differences and similarities between those traditions (Jordan and Shennan Reference Jordan and Shennan2003; Matthews et al. Reference Matthews, Tehrani, Jordan and Collard2011). The nature of such traditions can be used to determine relationships between the groups that use them and have also been applied to nonhuman animal cultural traditions (Lycett, Collard, and McGrew Reference Lycett, Collard and McGrew2009).

Work on cultural group selection and phylogenetics, therefore, also seems to use another type of culture concept, a concept of group culture, which, while composed of individual-level culture, nevertheless warrants its own set of cultural evolutionary models and explanations. Cultural evolutionists do not often explicitly call these various group-level cultural phenomena “group-level culture,” although they come pretty close, for example, describing group-level “cultures” that have such evolutionary effects (Bowles and Gintis Reference Bowles and Gintis2011, 16, 33) or “cultural variation … at the level of groups” in group-level traits such as “solidarity” (Richerson and Boyd Reference Huber2005, 14) or “cultural information” in the form of genuine group-level properties (Chudek and Henrich Reference Chudek and Henrich2014) or even group-level cultural transmission (Laland et al. Reference Laland, Odling-Smee and Feldman2000, 138).

2.5. Studies of the Origin of Culture

Finally, a small but significant amount of cultural evolutionary science is aimed at understanding the origin of culture (Boyd and Richerson Reference Boyd and Richerson1996; Richerson and Boyd Reference Richerson, Boyd, Heyes and Huber2000). This work, in particular, is focused on understanding why capacities for social learning, especially cumulative social learning in humans, might have evolved. There are two ways this work could be understood as describing the origin of human “culture.” One way would be the origin of the mass of socially learned cultural traits and types of cultural evolutionary processes that humans have; it could also mean the origin of the human capacity for culture. However, it does seem to mean both: when Boyd and Richerson (Reference Boyd and Richerson1996), for example, ask “why culture is common but cultural evolution is rare” they are referring at least in part to the learned traits and the process. When they and other cultural evolutionists describe “culture” as an adaptation or as a trait that has evolved by natural selection (e.g., Richerson and Boyd Reference Huber2005, 99–147; Mesoudi Reference Mesoudi2011, 190–91), they must mean the capacity for culture: while single cultural traits can be adaptations in the sense that some cultural evolutionary processes are adaptive, “culture” the adaptation is not any one individual-level cultural trait or even the mass of them, since neither of these things was the target of the natural selection process on which these scientists are focusing in this case. Instead, the target of this natural selection process was the mechanisms for social learning and related capacities that permit the acquisition of these cultural traits: for example, when Richerson and Boyd (Reference Huber2005) discuss why culture may or may not be an adaptation, their focus is on the capacity for social learning. The cumulatively socially learned and culturally evolved mass of information and the processes that produced it explain why the capacities evolved by natural selection, but they are not what evolved in this sense.

2.6. Why All These Different Uses of the Culture Concept?

So far I have given some reason to think that there are multiple uses of the term “culture” in the cultural evolutionary sciences. The question is what in fact explains the different ways these terms are being used. My contention is that the best way to understand this case is that there are in fact multiple culture concepts in the cultural evolution literature. There are, however, some alternative possibilities. One is that the different uses indicate the presence of different scientific projects and that as a matter of fact what I have been treating as a relatively unitary field (cultural evolution) is much more disparate than supposed. For example, culture might be used to refer to both culture in animals and culture in humans because there are two fields (animal culture and human culture fields) that use different culture concepts. This, however, does not seem to be the case. The different uses above occur within the work of particular researchers and overlap between them: for example, Laland, an important researcher in niche construction is an equally important researcher of animal culture and has discussed at length human gene culture coevolution and group-level culture, sometimes within the same papers (Laland et al. Reference Laland, Odling-Smee and Feldman2000; Laland and Hoppitt Reference Laland and Hoppitt2003). Furthermore, the most plausible divisions in the cultural evolutionary sciences do not map onto the different uses above. For example, dual inheritance theorists and the epidemiology of representation researchers might plausibly differ on whether to understand individual-level culture in terms of representations, rather than behaviors, but in practice both groups use both versions of individual-level culture as described earlier. Their differences also would not explain the distinct appeals to group culture, individual culture, and niche-constructed environments described above, for example, since EoR researchers rarely deal with group or niche construction models. Instead, the different uses of the culture concept occur in different types of models and explanations, which deal with different, if related, phenomena.

However, the possibility that I want to consider in detail is that the different uses of “culture” and the phenomena they describe can in fact be gathered under some sufficiently broad single definition of culture. There are a variety of proposed culture concepts in the philosophical and cultural evolution literature that might be able to do the work. In the next section, I argue that, while we cannot completely rule out any such definition, at the very least, none of the current attempted definitions of culture can be successfully applied to all the uses of “culture” described above. This leaves the remaining possibility: that there are in fact distinct, if related, culture concepts.

3.1. Culture as Information

I want to begin with what I think is the most intuitively obvious way to define culture, in terms of information. One property shared by all of the various entities or phenomena that anyone wants to count as “culture” is that they carry, transmit, or help to transmit information. The definition of culture in terms of information is extremely popular in the cultural evolution literature. For example, Richerson and Boyd describe culture as “information capable of affecting individuals’ behavior that they acquire from other members of their species through teaching, imitation, and other forms of social learning” (Reference Huber2005, 5) that can also be present in artifacts or the environment; Mesoudi et al. (Reference Mesoudi, Whiten and Laland2006, 331) largely follow Richerson and Boyd’s definition. However, the notion of information used in these definitions is essentially disposable: these authors often say that by “information capable of affecting behavior” they mean, essentially, mental states of various kinds (Richerson and Boyd Reference Huber2005, 5; Mesoudi Reference Mesoudi2011, 4). Consequently, I will include these authors in the next section, “Culture as Phenotypic Traits.” By far the best remaining candidate for a genuinely information-based account of culture is Ramsey (Reference Ramsey2013). Ramsey defines culture (partly following Mesoudi Reference Mesoudi2011) as “information transmitted between individuals or groups, where this information flows through and brings about the reproduction of, and a lasting change in, [a] behavioral trait” (Ramsey Reference Ramsey2013, 466). So while culture is clearly necessarily either a kind of information or an information-bearing state, activity, or property, not just any such information or state is culture; Ramsey claims the appropriate limitation is that cultural information travels through a behavioral channel (466).

Ramsey’s purpose in giving a definition of culture is not necessarily to unify the various cultural phenomena described in the cultural evolutionary literature, but he does seem to be aware of the different phenomena involved: for example, he takes seriously the idea that there is cultural information present in niche-constructed environments and artifacts (Reference Ramsey2013, 466), and his appeal to groups in the definition suggests he takes seriously the idea that one could have group-level cultural traits. So could his definition actually do this work? Unfortunately, used for this purpose, his definition has a variety of problems. The first problem is that the definition is not going to successfully handle group-level culture. While individual-level culture plausibly “flows through a behavioral trait” in that the performance of an individual-level cultural behavior is required for social learning to occur and for that behavior to be reproduced, the same is not true of group-level behavior. While there is some reasonable sense to be made of the notion of a group-level behavior (e.g., a group ritual), that behavior is reproduced by the individual-level social learning of its component individual-level behaviors and not necessarily by observation (or anything else) of the group-level behavior as such. Hence, the group-level “behavior” that is reproduced is not necessarily the behavior through which cultural information flows.

The second problem for Ramsey is that he defines culture in terms of information itself, rather than in terms of information carriers, such as behaviors, mental representations, artifacts, or environments. This does raise an important issue: as Sperber and Claidiere (Reference Sperber and Claidiere2008) note, information as such has no causal power; instead, it is the carriers and processors of information that generate behavior and have the other effects attributed to culture. This means cultural information as such cannot cause behavior, changes in the environment, and so on, as most cultural evolutionary explanations suggest it can. Ramsey’s main reason for this move is that he wants to maintain a distinction between cultural and genetic causes of traits in which cultural traits are those for which cultural information (vs. genetic information) is the primary difference maker in their origin; if culture is to be the cause of phenotypic traits, it cannot also be those traits themselves. The problem with making this distinction, as Lewens (Reference Lewens2015) argues, is that Ramsey offers us no clear way to make sense of the notion of genetic versus cultural information, except via the notion of a cultural channel, but there is similarly no clear way to demarcate a cultural channel from a genetic one. Lewens also suggests that cultural traits can be caused by culture in the sense that they are caused by the cultural traits of other people (via social learning), so the phenotypic definition of culture can be preserved even in the face of this requirement.

The third problem with Ramsey’s definition for capturing the evolutionary culture concept is that it is just too broad. While Ramsey takes seriously the need to account for multiple types of cultural transmission and information, he includes as culture any information that “flows through a behavioral channel” so long as it reproduces the behavior in the appropriate way. Ramsey’s reason for this looseness is that he deliberately wants to include epigenetically transmitted behaviors in “culture” where the epigenetic change leading to the behavior in the offspring is caused by the same behavior in the parent. This means that even if his definition could account for most cultural phenomena described in section 2, it would not be a definition of culture as used by the cultural evolutionary scientists. The reason is that these scientists are focused on understanding a cluster of phenomena that involve the spread of informational states via social learning as such, rather than merely via behavior. Ramsey chooses to avoid definitions of culture in terms of social learning because the concept of social learning is not very well defined (as discussed earlier, social learning and individual learning form a rather fuzzy continuum), but in so doing he makes his definition something rather different from what is currently being used in the main scientific programs studying cultural evolution. While this may be his intent—this may be as much a prescriptive as a descriptive proposal—it does mean that this definition will not do the work needed to account for the uses of culture in the cultural evolutionary literature.

3.2. Culture as Phenotypic Traits

Views of culture as phenotypic states are fairly common in the cultural evolution literature. There are some differences, however, in which phenotypic states are taken to be culture. The most common view is that culture is socially transmitted mental representations. The individuals most explicitly holding this view are those evolutionary psychologists and cognitive anthropologists who advocate an EoR view of culture (e.g., Tooby and Cosmides Reference Tooby, Cosmides, Barkow, Cosmides and Tooby1992; Boyer Reference Boyer1999; Atran Reference Atran2001; Sperber and Claidiere Reference Sperber and Claidiere2008); memeticists such as Dawkins (Reference Dawkins and Blackmore1999); some of the dual inheritance theorists such as Durham (Reference Durham1991, 3–5), who defines culture as “shared ideational phenomena”; and some evolutionary behavioral economists, such as Bowles and Gintis (Reference Bowles and Gintis2011, 13) who define culture in terms of preferences and beliefs. We can add to these many of the “culture as information” theorists such as Richerson and Boyd (Reference Huber2005) and Mesoudi et al. (Reference Mesoudi, Whiten and Laland2006) who, as shown previously, take cultural information “in brains” to really refer to socially learned mental representations or similar states.

However, it is also possible to take culture to be socially learned behavior. Behavior here can be read as overt behavior—actual movements—patterns of behavior, or behavioral dispositions. A few individuals hold an at least partly behavioral view in the literature: for example, Jablonka and Lamb, define behavioral culture as “a system of socially transmitted patterns of behavior, preferences and products of animal activities that characterize a group of social animals” (Reference Jablonka and Lamb2005, 160). Finally, one can take culture to be (at least sometimes) socially learned or transmitted artifacts, as proposed in part by Dawkins (Reference Dawkins and Blackmore1999), Richerson and Boyd (Reference Huber2005, 63), and Sterelny (Reference Sterelny2012).

The question, then, as with the previous definition of culture, is whether these kinds of definitions can capture the range of uses of the culture concept described in section 2. Let us first consider how well they do at making sense of the way the concept of culture is used in individual-level work. As I discussed earlier, in practice, real individual-level cultural evolutionary models use a variety of these different phenotypes as their focus in different cases. So in determining how well such phenotype-based definitions of culture account for scientific practice, we need to explain why these differences arise. One possibility is that one type of phenotypic definition is right, and the others are either wrong or operationalizations of that one; another possibility is that it genuinely varies from case to case which sort of phenotype is the “right” focus of the model; another possibility is that they are simply equivalent, and the variations are just a matter of preference. In favor of this last view, many cultural lineages involve social learning in which behaviors or artifacts are “copied” to form mental states, which produce behaviors, which produce artifacts, which are “copied” to form a new generation of mental states, and so on; each component is necessarily part of the cultural lineage, and each gets “reproduced” to some extent.

So why think that one type of cultural trait in these sequences is what is really reproduced and the others are not? One argument that has arisen in the memetics literature is that the mental representations view of culture permits a natural parallel between cultural and genetic evolution: there is a “replicator” and an “interactor” in both cases. In the genetic case it is the genes and the phenotype, and in the cultural case the mental representation and the resultant behaviors or artifacts (Hull Reference Hull and Plotkin1988). The claim is that mental representations are then what “really” evolves and is reproduced across generations (Dawkins Reference Dawkins and Blackmore1999). However, it is not clear why this distinction is important; even in some biological evolutionary processes (e.g., codon bias or meiotic drive), the so-called replicators are also the interactors.

Another way that arguments have been run for preferring the mental representations view over the behavior or artifact view is that most human social learning processes reproduce one or the other “properly” at the expense of the other. The argument is usually that the behaviors and artifacts produced in cultural lineages tend to be error prone relative to what is intended by their producers. Individuals reproducing these behaviors or artifacts often correct for these errors: they reconstruct what is intended (the mental representation of the producer), rather than what is actually produced (Sperber Reference Sperber and Aunger2000). However, this will not decide the issue either. While behaviors may be error prone, mental representations are invisible: the only way to “copy” mental representations is via observing the behavior they produce, and hence these mental representations may also be error filled. Furthermore, the idea that copying accuracy determines what is really forming cultural lineages is mistaken: cultural transmission can occur even in the face of imperfect replication (Henrich, Boyd, and Richerson Reference Henrich, Boyd and Richerson2008).

However, if we do take the view that one “culture as phenotype” view is correct and the others are not, then we have to account for why we see different uses of behavior, artifacts, and mental representations in different models (see the examples given in sec. 2.1). One possible reason is that the different views of culture are simply operationalizations of a more specific culture concept: cultural information cannot be observed directly, whereas the particular phenotypic traits that carry it can. As one might expect if this were true, the notion of “culture” used depends to some extent on the type of explanation being used: for example, cultural evolutionary scientists are more likely to use models describing artifacts as opposed to behavior when the behaviors are inaccessible in themselves apart from the artifacts (Henrich Reference Henrich2004). However, some of these cultural evolutionary models and explanations are abstract “proofs of concept”: as such, the appropriate forms of testing for these models do not involve observations of either behaviors or mental states but instead things like mathematical simulations of populations, and so on; consequently, there is no need to operationalize (e.g., Claidiere and Sperber Reference Claidiere and Sperber2007; Henrich and Boyd Reference Henrich and Boyd2008). Nevertheless, it is behaviors, rather than mental representations, that are still what is described on the models.

Most of the previous cases have been based on dividing up the same lineages into their behavior, mental representation, or artifact parts and then deciding that one or the other of these is the “real” culture in all cases. However, Jablonka and Lamb’s (Reference Jablonka and Lamb2005) discussion of “symbolic culture” suggests another problem for this way of thinking. Consider, for example, a lineage of cakes baked by individuals on the basis of the same recipe. In practice, many disparate types of things may form parts of such a cake lineage: actual cakes, recipe-describing linguistic behavior, artifacts (recipe cards or books), mental states (mental representations of recipes), and cake-baking behaviors. While there is always a mind-behavior-mind or mind-behavior-artifact-mind pattern in these lineages, the behaviors and artifacts in each such sequence may be different. For example, the sequence might go from the mental representation of a recipe (in one person) to cake-baking behavior to cake to the mental representation of recipe (in the next person) to speaking of recipe to mental representation of recipe (in the third person) to writing behavior to recipe card to mental representation (in the fourth person), and so on. So why (and how) are such different components making up these lineages? It is worth noting that, when the behaviors and artifacts in these lineages are not actually the cake or cake-making behaviors themselves, they are symbolic/linguistic items meant to induce the same mental content as observing the cake-making behavior would, that is, to reproduce the mental representation of the recipe in the absence of the cake. One natural interpretation of these cases is that in some chunks of the lineage what is being reproduced is a cake (an artifact) or the cake-making behavior; in others it is the mental representation of the recipe. Jablonka and Lamb describe the former as “behavioral culture” and the latter as “symbolic culture.” Any definition of culture in terms of only one particular sort of phenotype may automatically exclude these natural ways of thinking about such cases. A well-formulated culture concept should perhaps not be so specific.

Of course, the discussion so far in this section simply relates to how well phenotypic definitions of culture capture the individual-level culture concepts used in cultural evolutionary models. How about culture as behavior and culture as mental representations and their capacity to deal with cases of group-level culture? Culture as mental representations is going to have trouble here. Group-level traits, of the sort described earlier, are certainly caused by or composed of individuals’ mental representations and behaviors, but they are not simply those representations; there is also not a group-level mind. It is plausible, alternatively, that there are genuinely group-level behaviors or artifacts; however, group-level traits are not “socially learned” or, at least, not in the same way as individual-level traits are and the above definitions suggest. In fact, Jablonka and Lamb (Reference Jablonka and Lamb2005) get around this problem by defining behavioral culture entirely in terms of patterns of behavior characterizing groups and avoid the individual- versus group-level distinction altogether.

Similarly, the culture as mental representations concept falls short in dealing with cases in which culture is a transmitted environment. If culture is mental representations, then it cannot possibly exist in nonmental environments; transmitted environments are caused by behaviors, rather than being those behaviors. “Culture as artifacts” could possibly work here as long as artifacts are read to include transmitted environments, but of course it will have problems capturing the other uses of “culture.” It is probably best, therefore, to take these “culture as phenotype” views to be, at best, attempts to formalize a definition of individual-level culture and not applicable to any of the other uses of the culture concept.