1. Ethnography, analogy, and theory

The late Pleistocene (from very approximately 150 kya) was a crucial and controversial period in human social life, for it was a time in which the pace of technological change and the extent of regional differentiation increased markedly. The control of form became astonishingly precise;Footnote ¹ the range of worked materials increased, with bone, horn, and ivory becoming widely used; composite tools became a routine part of human toolkits, and material symbols became a regular aspect of human material culture. These changes are often summarized in the claim that at this stage human life became “behaviorally modern.” It was once thought that these changes were very abrupt, probably signaling some upward step in intrinsic cognitive capacities. It is now known that these aspects of material culture had more ancient and dispersed origins, and few now attribute the changes to a genetically based cognitive upgrade.Footnote ² Instead, the main candidate explanation is that these late Pleistocene changes were the result of enhanced social learning and hence more reliably cumulative culture, a view developed in different forms in Henrich (Reference Henrich2016), Heyes (Reference Heyes2018), Tennie et al. (Reference Tennie, Premo, Braun and McPherron2017), and Sterelny (Reference Sterelny2012). How is such a view developed and tested? In interpreting the Pleistocene, archaeologists have routinely appealed to ethnographic information. For example, Frank Marlowe explicitly sees the foraging life of the Hadza as a model of the Pleistocene foragers who foraged in the same area and in a similar physical and biological landscape (Marlowe Reference Marlowe2010). Much more controversially, Francesco d’Errico and his co-authors have recently argued that the material culture of the Border Cave in South Africa is so strikingly similar to very recent San artefacts as to show strong cultural continuity, even though the Border cave deposits and the San are separated by over 40 k years (d’Errico et al. Reference d’Errico, Backwell, Villa, Degano, Lucejko, Bamford and Higham2012).Footnote ³ The aim of this paper is to defend the relevance of ethnographic information in reconstructing late Pleistocene social worlds, and, in particular, reconstructing the organization of social learning in those worlds. That is particularly important in this case. For the hypothesis that the late Pleistocene saw a marked change in the reliability of cultural accumulation has so far rested heavily on formal models of cultural evolution (see especially Henrich Reference Henrich2004; Powell et al. Reference Powell, Shennan and Thomas2009; Premo and Kuhn Reference Premo and Kuhn2010).

In sum, this paper has three aims. (1) It defends the relevance of ethnographic information to archaeological interpretation through a detailed case study illustrating a more general method. (2) It links that case study to the Pleistocene innovation spike, supporting the claim that this spike depended on enhanced social learning. (3) The paper then details both the scope and the limits of the approach illustrated by the case study.

Despite its prevalence, the use of ethnographic data in archaeological reconstruction remains controversial. It has typically been taken to depend on analogy, that is, on the idea that past human cultures are relevantly similar to ethnographically known ones, and hence that we can legitimately extrapolate from (more or less) contemporary cases to more ancient ones. In turn, this is often seen as “presentist,” implicitly denying the possibility that the social and economic lives of ancient humans were qualitatively different from those in the ethnographic record. In response, Alison Wylie, and more recently Adrian Currie, have argued that these similarity-based inferences can be sound and should be assessed on a case-by-case basis (Wylie Reference Wylie1982, Reference Wylie1985; Currie Reference Currie2016). Both Wylie and Currie suggest that analogical inference is warranted to the extent that the observable similarity between an ethnographic and an archaeological case is evidence for a causal similarity. To take a specific instance discussed by Currie, it is argued that the Pleistocene cave art of glacial Western Europe is so strikingly similar to shamanistic imagery that we can infer that they are the result of a shared neuro-cognitive mechanism. As the two art traditions have a common causal base, we can use what we know about shamanistic symbolism to interpret that Pleistocene art. However, if we can use ethnographic data to construct and test claims about causal mechanisms, similarity between past and present is no longer essential. We can use ethnographic case studies to build, refine, and test a general model of forager social organization and the environmental factors that explain variation in that social organization. These environmental factors can be outside the contemporary range.

A focus on using ethnographic information to build a causal model has two advantages. First, historically recorded forager cultures are impressively varied. In documented times, foragers have negotiated an immense range of physical and biological environments: from the tropics to the high latitudes; from rain forest to tundra, to sea edges to desert; living largely in peace to permanent hostility to neighbors, and much in-between. This variation in the lifeways of ethnographic foragers (celebrated in Kelly [Reference Kelly2013]) is a crucial resource for theory-building. It enables us to explore both the sensitivity and the insensitivity of forager lifeways to variation in the environment. This paper will argue that the forager education system is relatively insensitive to variations in the physical and biological environment. In a companion paper, I develop a parallel argument about a decidedly variable aspect of forager lives: their very differing levels of investment in subsistence technology (Sterelny Reference Sterelny2021). That paper shows that high investment subsistence toolkits (often found in high latitude forager communities) are responses to resource portfolios that have restricted diversity, but contain high value targets whose locations are relatively predictable. So, this theory-applying strategy does not depend on identifying relatively invariant features of forager lifeways, even though in this paper the focal example is relatively invariant.

Second, it expands the relevance of ethnographic information. Many ancient environments were very different from recent ones. In places, that is true even of the Early and Mid Holocene. Suppose, for example, we aimed to reconstruct the lifeways of the Australian foragers of the late Pleistocene and earlier Holocene. We could not rely on similarities across time of technology and subsistence economy of the kind d’Errico and his colleagues claim to see between the recent San and the Border caves foragers. Nor could we rely (as does Marlowe in his discussion of the Hadza) on the enduring stability of the physical environment. Physically and biologically, Pleistocene and early Holocene Australia did not resemble contemporary conditions. For example, before the last glacial cycle took hold, large parts of inland Australia, whilst still being arid by the standards of other continents, were significantly less arid than now. Fifty thousand years ago, Lake Eyre and Lake Frome (now huge salt pans, very occasionally flooded) were not just full; they constituted a single huge body of inland water. As recently as 32 kya, Lake Frome was indeed a lake, with a volume 40 times that of recent ephemeral floods. In contrast, at the peak of the last ice age, the interior was not just colder but drier, with a weaker monsoon. Ethnographic information would be irrelevant to Australian Pleistocene or early Holocene foraging if its value depended on establishing similarities between present and past.

So, we need to build a general model and use that model in connection with Pleistocene environmental data to reconstruct the lifeways of past forager societies. Our best guiding theory is human behavioral ecology, broadly understood to include evolutionary game theory and life history theory. Crucially, this theoretical approach makes minimal, relatively uncontroversial assumptions about human action (Laland and Brown Reference Laland and Brown2002). Its grounding assumption is that in environments that agents have long experienced, and when faced with decisions they and/or their social partners have faced many times in the past, their choices approximate those that would maximize their expected inclusive fitness.Footnote ⁴ This broad-church version of human behavioral ecology rests on the view that in well-practiced environments, humans are intelligent and flexible enough to learn the choices that lead to the outcomes they welcome, and that the outcomes they welcome enhance their fitness.Footnote ⁵

That said, behavioral ecology does have significant limitations over and above the requirement that environments are stable enough for agents to zero in on suitable responses. Agents must have access to relevant information and be autonomous enough to act on their choices. On the view developed here, we can use causal models based on behavioral ecology to interpret patterns of behavior when (i) the activities in question are central to fitness; (ii) they take place in environments in which agents have time and opportunity to identify their relevant affordances; and (iii) agents have sufficient autonomy to choose options that advance their own interests. I return to these constraints in the final section to test whether they undermine the main arguments of the paper and to consider the scope and the limits of this strategy for using ethnography to inform archaeology. For now, I aim to show the power of this approach through an example: the use of information about social learning amongst ethnographically known foragers to estimate the social learning environment of the late Pleistocene. I shall suggest that this supports the idea that the late Pleistocene acceleration of both innovation and differentiation reflects a social environment better adapted for efficient social learning.

2. The forager education system

The forager education system is both distinctive and relatively invariant, despite the variation in other aspects of forager life. A broad range of ethnographic sources converge on the following picture: (i) a forager’s education develops over three main phases. As an infant and young toddler, a forager learns primarily from his/her parents (mostly the mother). From childhood to early adolescence (from approximately 4 yo to 12 yo), the young forager spends most of the day in a mixed age/mixed gender play-group. In this group, learning is a mix of collaboration, individual exploration, practice, peer-to-peer horizontal transmission, and with some information flowing through participation in adult activities. From early adolescence, gender segregation becomes more marked, and social learning becomes predominantly intergenerational, with a variable mix of oblique and vertical transmission. (ii) Forager children are free-range, with plenty of time to play and explore. In most cases, the playgroup operates with remarkably little adult supervision. Forager children do, indeed, forage, but their motivation is as much recreational and social as economic. Their targets and hence their profits reflect this. (iii) Forager children acquire a basic forager competence remarkably early. At about 12, they are not yet experts, as they are still to acquire only their most demanding skills (large game hunting, complex craft techniques, esoteric knowledge). (iv) Explicit teaching is important to a forager’s education, but it is not the dominant form of social learning. It is mostly restricted to norm acquisition, esoteric knowledge, and the most challenging subsistence skills. (v) Notwithstanding the importance of the playgroup, children are not excluded from adult activities. Given the intimacy and compactness of mobile forager campsites, there is a lot of public information available, and children have the opportunity to look, to eavesdrop, to learn by helping, and to benefit from occasional advice. In the rest of this section, I will expand on this skeletal summary.

3. Optimized social learning?

One’s overwhelming impression of the forager education system is of its great efficiency. It is impressively fit for purpose. First: the three-phase organization of the forager learning is very cost effective for adults. There is no escape from the burden of babies and infants, falling mainly on the mother. But through middle childhood (from about four to about twelve), children are largely educated through the playgroup. They learn through a combination of informational trickle-down from older children, collaborative problem solving, practice, and trial and error exploration, somewhat leavened by observation, eavesdropping, and the provision of equipment by the adult world. The playgroup is self-sustaining, as older children fade into early adolescence, infants toddle in. Consequently, foragers largely escape the opportunity costs of having to educate their children (as distinct from their infants and their adolescents). As children emerge into adolescence, explicit teaching becomes more important. However, except perhaps from some cases of the transmission of esoteric knowledge, this teaching is rarely sustained or high cost. Chabu men do indeed explain navigation, point out and identify tracks, and demonstrate the correct grip on a spear. But these are brief episodes in the course of a hunt. Likewise, storytelling is inexpensive by design, confined to nights (and seasons) where there is little else to do. Young foragers need food, shelter, and clothes, but they are cheap to educate; perhaps the greatest cost is the provision of equipment with which to practice.

Second, the forager education system seems to have an extraordinarily high success rate. While there is clearly variation in skill level, there seems to be no ethnographic reports of forager children who just do not get it, who never become passably competent at the skills essential to their lives, given their gender and community. Unless shamans count, there are no reports of amiable incompetents who have to be fed and protected by others, though of course infanticide weeds out obviously doomed babies (Hrdy Reference Hrdy1999). The system achieves this near-universal success despite the fact that forager children have a lot to learn. As Peter Richerson and Rob Boyd regularly point out, foragers need rich and detailed knowledge of their local patch (Richerson and Boyd Reference Richerson and Boyd2005; Boyd Reference Boyd2018). Motivation is part of the explanation of near universal competence. Forager children are highly motivated to acquire forager skills (they may be less motivated to internalize forager norms, hence the heavier footprint of teaching and negative feedback). Allowing children to learn at their own time and pace, sometimes in collaboration with friends, enhances the desire to learn. However, it may also be significant that the skills in question rarely require brutal drudgery. Men enjoy hunting, perhaps even when it is not the most efficient means of subsistence (Smith et al. Reference Smith, Bird and Bird2003). Likewise, women forage together enjoyably and socially (Lowe Reference Lowe2002). Many activities that derive from foraging are contemporary sports and recreations. That is not true of the subsistence farming of grains and other starches. No-one goes hoeing recreationally, let alone turns it into an Olympic sport. Subsistence farming often requires many hours of drudgery in the sun, and where they can, farmers often force others to do this work. The more regimented and disciplined character of farming childhoods may be linked to the unappealing character of farming labor. Forager children are motivated to learn, in part because they are learning skills they will enjoy exercising as adults, not just skills that are required of them as adults.

Third, the forager education system is resilient. Despite the small size of forager camps, there is ample redundancy, especially for the more basic skills. Social information is available through many modes. These include imitation and demonstration (the Dene, for example, apparently demonstrate tool-making carefully but without explanation [Lew-Levy et al. Reference Lew-Levy, Rachel Reckin, Cristóbal-Azkarate and Ellis-Davies2017, 376]); emulation (the Gidra apparently provide their children with their first bow-arrow set, but expect them to use this as a template in producing replacements [Lew-Levy et al. Reference Lew-Levy, Rachel Reckin, Cristóbal-Azkarate and Ellis-Davies2017, 382]); provisioning with real and toy equipment; advice; and explicit instruction. Once children are part of the playgroup, social information flows from many sources and with lots of collaborative symmetrical learning (Lew-Levy et al. Reference Lew-Levy, Adam Boyette, Hewlett and Lamb2020). Moreover, for the most part, learning is not time pressured. At least for the routine skills of daily life, the physical intimacy of forager camps, and the tolerance by adults of children tagging along in adult activities, ensures that there are many opportunities to look and experiment. The acquisition of core forager competences does not depend on the high-fidelity flow of information from model to novice in any specific learning episode. There are many second chances.

In sum, the three-phase organization of social learning is one in which children have a lot of autonomy to shape their own days, learn at their own pace by their own efforts, but with social support from adults and peers. They then become gradually integrated into adult pursuits in their teens, with more explicit teaching. That organization results in children becoming competent adults at minimal educational cost to their parental generation. Can we project that organization back to late Pleistocene foragers? I suggest below that the forager pattern was largely in place in the late Pleistocene, and only in the late Pleistocene, for it depends both on the efficiency of forager foraging and on socio-cultural adaptations that were probably in place only then. If so, this strengthens the case for the social learning account of the late Pleistocene innovation acceleration, for that acceleration would then be concurrent with the emergence of communities optimized for such learning.

4. Forager education in the late Pleistocene

The forager education system has important prerequisites that collectively scaffold the adaptive organization of forager education. These include:

1. (i) elaborated kinship systems. These provide mothers with material and social support that ease the burdens of babies and infants, and they help integrate children into their social world. While mothers are the primary carers of the very young, forager children have regular and close physical contact with their fathers and other relatives (Hewlett et al. Reference Hewlett, Hudson, Boyette, Fouts, Lavi and Friesem2019).

2. (ii) The playgroup and free-range childhood depend on the generally peaceable nature of forager social worlds. Forager mothers very rarely must worry about infanticideFootnote ⁶ or other forms of child-directed violence. This is far from universally true of primate mothers (Hrdy Reference Hrdy1999, Reference Hrdy2009).

3. (iii) The three-phase organization of learning depends on the open and fluid character of forager residence. This makes it possible for mothers (and other kin) to support adult daughters in different camps and gives foragers entering adolescence access to a much wider circle of potential models. Movement between residential groups pools information: adolescents hear about potential models outside their immediate circle and sometimes have a chance to learn from those models. Moreover, while basic competence probably depends only on the skills and information present in a single residential group, the fact that groups are networked, with information flowing between them, makes the preservation of information and skill more resilient (Shennan Reference Shennan2008; Premo and Kuhn Reference Premo and Kuhn2010) and supports fine-tuning expertise (Henrich Reference Henrich2004).

4. (iv) Time to play and explore depend on the prosperity of forager subsistence economies (Sahlins Reference Sahlins, Lee and DeVore1968). Most adult foragers do not have to work long and hard to gather just enough for themselves and their dependents. Richard Gould reports that even Western Desert Australian foragers, living as they do in a profoundly inhospitable environment, rarely have to work more than four hours a day (Gould Reference Gould1980, 78–89). If foragers were living on the very edge of necessity, there would be much stronger pressure to integrate children into the subsistence economy (as in farming and herding cultures).

5. (v) Some forms of teaching are independent of language. But explicit teaching obviously depends on the prior establishment of language and other tools of human communication.

It is difficult to date the origin of these scaffolds, but they are ubiquitous across known forager cultures, and that strongly suggests that these scaffolds were available to the African foragers that emerged from Africa and settled the world, beginning sometime around 100 kya–80 kya. It is conceivable that these scaffolds were convergently assembled after the out-of-Africa migration. But the great variety of forager physical, biological, and demographic environments suggests inheritance from the African homeland rather than parallel cultural evolution. We can assume then, with reasonable confidence, that these background enabling conditions were in place for late Pleistocene foragers (say, those living 100 kya–12 kya). But they were probably not in place much before that. Subsisting in difficult environments (full desert, very high latitudes, and, a little counter-intuitively, dense rainforests) depends on very high levels of foraging technique and on extensive networks of social support, especially in regions that are both challenging and unpredictably variable. These networks are made possible by the open and fluid character of forager residential practices. While there are always uncertainties about negative evidence, humans do not seem to have established in such habitats before the late Pleistocene (Gamble Reference Gamble2013), suggesting that the essential props of the forager education pattern were not fully in place until then.

These considerations suggest that late Pleistocene forager education was probably broadly similar to the ethnographic template. Important elements of the forager pattern are consequences of the basic demography and economy of forager life: life in a physically compact, emotionally intimate camp of somewhere between 30 and 50 individuals. As Kelly has shown, it is no accident that forager bands are roughly that size. It is a consequence of energetic and ecological constraints that would have acted in the Pleistocene too (Kelly Reference Kelly2013). In groups of that size, playgroups must be mixed in age and gender; there are not enough children for narrower sorting.Footnote ⁷ Likewise, the ready availability of public information is a consequence of the physical compactness and intimacy of forager camps. The same physical intimacy implies the availability of many models, and many modes, through which information can flow from the more informed to the less informed. Pleistocene children would have similar opportunities to observe and eavesdrop. Pleistocene adults were probably as tolerant of children handling their tools and raw materials, as tolerant of onlooking and eavesdropping, as willing to supply tools for practice, play and use, as than those known from ethnography. For it is in adults’ interest that children self-educate. Self-education reduces the costs of children to their parents, especially if learning-by-doing enables children to partially supply their own needs. Likewise, we can expect Pleistocene forager children to be as highly motivated to learn adult skills as those known from ethnography and for the same reasons. It is also probable that Pleistocene children’s time budgets were relaxed in similar ways, again for reasons that parallel those of contemporary foragers. Forager life is time rich in most seasons, for mobile forager group size is constrained by the resource envelop of the least productive season. Moreover, many forager cultures manage risk by more-or-less compulsory sharing (Lewis et al. Reference Lewis, Vinicius, Strods, Mace and Migliano2014). This dis-incentivizes hard work. A forager and his/her family will get little fitness benefit by working harder than their peers, for they are providing a public rather than a private good. The consequence is a relaxed adult time budget for much of the year.

In brief, most of the factors that explain the stability of the ethnographically-described forager education system were also factors in the late Pleistocene. There are two exceptions, and those exceptions imply that Pleistocene forager education was somewhat more expensive to the adult generation. The foragers described in the reports on which I have drawn were all to some degree connected to the market economy. Their toolkit incorporated materials sourced through that economy. Martu men hunted with rifles from vehicles. That is an extreme case, but the Chabu, while making their own spears, tipped them with metal, which they did not forge. Pleistocene foragers had to make all their own equipment and source the necessary raw materials (or perhaps only almost all: there might have been some trade). So, the artisan skill load was higher, and acquiring these skills probably required more explicit instruction. Second, Pleistocene habitats were probably more dangerous. The Pleistocene predator guild was significantly larger than that of the late Holocene, and some of its larger members were probably less wary of humans. The survivors have had 10,000+ more years to learn to avoid us. Could the playgroup have been quite so free-range and independent of adult supervision and protection in Pleistocene forests and woodlands? I suspect not. These considerations suggest a Pleistocene education system quite similar to that attested from ethnography, but with a larger footprint of adult teaching in the acquisition of craft skills and somewhat greater adult vigilance over the mixed age and gender playgroup. However, as noted above, forager childhoods even at 200 kya may have been very different. Ethnography thus supports the idea that the late Pleistocene acceleration of technical innovation depended on enhanced social learning.

5. Caveats, scope, and limitations

These inferences are only as good as broad-church behavioral ecology, the framework on which they depend. While that framework has many strengths, it has real limitations too, and it is time to consider whether my conclusions about the late Pleistocene are undermined by those limits. To begin with, behavioral ecology is essentially an equilibrium theory: if the environment is changing rapidly or agents are migrating into novel environments, adaptive lags and other misfires are likely to be common and important. However, while climate change in both the late Pleistocene and Holocene has been challenging, forager economic organization and risk management are very resilient, and the forager education pattern seems to be fairly insensitive to environmental difference.

Second, behavioral ecology is an individualist, bottom-up approach to social theorizing: patterns in social life are aggregated outcomes of individual decisions. This approach has limits when applied to communities structured by social hierarchy and by institutions (Lewens Reference Lewens2015). But in forager communities, (adult) individual choice is not seriously constrained by power difference and institutions of inequality (though of course agents must consider others’ responses to their choices). So, it is well-suited to explaining forager social life. Third, even if individual agents are making optimal decisions, optimization is always optimization under constraint. Choices are constrained by available technology and by the social environment, and one reason why behavioral ecology is such a powerful tool is that its use reveals puzzles and paradoxes. We seem to see far from optimal patterns of action in familiar environments. These anomalies reveal unexpected and subtle constraints on choice, and this is a most important virtue of the behavioral ecology framework (Gould Reference Gould1980).Footnote ⁸ One such subtle constraint is the availability of information. The informational environment must be transparent enough for the experienced agent to be able to estimate, with reasonable reliability, the costs and benefits of particular choices. This is not always available, even in familiar environments. Arguably, the male gerontocracy of Central Australia depends on information asymmetries in that culture (Keen Reference Keen2006; Sterelny Reference Sterelny2015). However, forager adults probably have a good appreciation of the requirements of competence, of how difficult skills are to acquire, of the extent to which particular children have mastered the forager curriculum, and of the interactions of other adults in the band with those children. These are compact, intimate communities (Boehm Reference Boehm1999). It is unlikely that informational constraints limited adaptive choice in social learning and teaching, either in ethnographically known forager cultures or the final Pleistocene.

Finally, in identifying a social practice as adaptive, it is always necessary to ask: adaptive for whom? The fitness interests of parents and their children overlap, but they are not identical. In general, parents optimize their fitness by spreading their resources across their children. However, each child would prefer a disproportionate share of that resource envelope. Even in relatively egalitarian communities, a system of the distribution of resources that is adaptive for some is not necessarily adaptive for all. But the forager pattern of education seems genuinely win-win, a good education for the children at a low price to the parents. One reason is that information is not like food. Interpreting a track for one boy does not diminish the capacity to show it to another. Indeed, demonstration, explanation, and advice are efficient; they need not be one-on-one. Parents are not forced into difficult resource allocation decisions. However, while it is true that information shared is not information lost, an agent sharing information loses a relative fitness advantage over others. For subsistence skills, that potential advantage does not seem to lead to information hoarding. Such hoarding might not be a practical option, as it would be hard to exploit information while keeping it private, and there are social rewards for informational generosity, with prestige accruing to those who share their expertise (Henrich and Gil-White Reference Henrich and Gil-White2001). Moreover, there will often be multiple social sources of that information, as well as the prospect of relearning from the material world itself. Finally, mobile foragers manage risk by relying on aid from others to protect against bad luck and trouble. It may be imprudent to hoard information that decreases the foraging efficiency of those on whom you may depend. In short, the forager system of educating subsistence skills is adaptive for the members of forager communities, and for that reason, late Pleistocene foragers probably organized subsistence information flow in a similar way.Footnote ⁹

Putting all this together, the limits of the behavioral ecology framework do not undermine the conclusion of section 4: the Holocene template is likely to fit the late Pleistocene as well. More generally, this framework provides a worked model for the use of ethnographic information in interpreting Holocene and Pleistocene sites. In the introduction, I suggested that we can use behavioral ecology to interpret the past when the activities in focus are central to fitness, when the agents have learned enough to navigate their environments competently, and when they have enough autonomy to act in their own interests. These conditions are satisfied in this case. However, this framework is limited in important ways. For one thing, these models are somewhat coarse-grained. Signaling theory, for example, can tell us that material symbols functioning within a community will vary in systematic ways from those functioning between communities (Kuhn and Stiner Reference Kuhn, Stiner, Mellars, Boyle, Bar-Yosef and Stringer2007a, Reference Kuhn and Stiner2007b). But it cannot tell us what those signals meant to and for those agents. Given the arbitrariness of the relationship between sign and target, and hence the many degrees of freedom in establishing sign systems, we have every reason to be skeptical about the many attempts to interpret the meaning of the material symbols of cultures like Catalhoyuk (Whitehouse and Hodder Reference Whitehouse, Hodder and Hodder2010), or Neolithic megaliths like StonehengeFootnote ¹⁰ or Gobekli Tepe. It is not just that we do not have to hand a theory that allows us to interpret arbitrary signs; we have theoretical reason to suspect that there is no such theory. Consequently, I am very skeptical of one prominent use of ethnographic information, its use in the interpretation of rock art (VanPool Reference VanPool2009; Bednarik Reference Bednarik2011; Berrocal Reference Berrocal2011). As Adrian Currie pointed out, in some cases the inference is based on the supposition of cultural continuity between the makers of rock art and their ethnographically described descendants. But even if the historical connection is real, we lack evidence for the stability of interpretative traditions over millennial time spans (Hiscock Reference Hiscock2020). When shamanistic iconography is used to interpret the art of glacial Europe, the causal model is even more fragile, since it depends on the assumption that a fine-grained neural architecture is common to the Pleistocene inhabitants of glacial Europe and some near-contemporary forager populations. That assumption seems undermined by accumulating evidence of neural plasticity (discussed in an archaeological context in d’Errico and Colage [Reference d’Errico and Colage2018]).

However, this framework is not limited to forager subsistence and its skills. For example, Robert Kelly, building on the work of Bruce Winterhalder, has developed a model of the factors that condition interactions between communities (Kelly Reference Kelly2013). The core idea links the risk of unpredictable variation in conditions to the spatial extent of that variation. When variation-driving factors are localized, good and bad times for one community tend not to be experienced by neighboring communities. In those circumstances, the rewards for intercommunal cooperation are high, predicting the existence of material exchange networks and other ways of maintaining cooperative connections. When variance is determined by regional factors, bad times here are bad times next door, and opportunities for mutual benefit are limited. Moreover, when variance is high, bad times are times of genuine stress, perhaps crisis. This is a scenario predicting high degrees of territoriality and high risks of conflict. As Kelly shows, this general model is well-supported in the ethnographic record, and he uses it to interpret the archaeological record of Wyoming (Kelly Reference Kelly2013; Pelton et al. Reference Kelly, Pelton, Robinson, Lavi and Friesem2019). There, fluctuations in both population size and climate exposed those communities to both cooperation-favoring and territoriality-favoring conditions, arguably reflected in fluctuations in the flow of exotic materials like obsidian.

I have argued that the use of ethnography to interpret the archaeological record can accommodate deep differences between present and past lives and that the relevance of ethnographic information is enhanced by the great variation documented in that record. I have used a specific example developed in some detail to support a cultural learning explanation of late Pleistocene social change. But in one fundamental respect Wylie and Currie are right. There is no one-size-fits-all assessment of inferences from ethnography to archaeology. Each case must be assessed on its specific merits. The strength of such an inference depends on the fit between the specific case and the general causal model and the degree to which that model is supported.