R2. CGS is a framework

Commentators Gintis and Mace & Silva observe that CGS is a framework rather than an explanation. Barclay & Krupp seem to be expressing the same idea when they describe CGS as a plausible force while arguing that its action needs to be demonstrated in actual cases.

We agree. We see evolutionary theory as a framework, an accounting system, for keeping track of heritable variation through time. A CGS framework is based on the empirical data strongly suggesting that heritable cultural variation often exists between groups at scales larger than groups linked by kinship or a history of reciprocation (target article, sect. 4). Thus, such groups will often respond to some form of selection and many case studies suggest that they do (sects. 5 and 6). The CGS framework has been used to develop explanations for concrete phenomena, for example, the expansion of the Nuer at the expense of the Dinka in the Southern Sudan in the 19th century. Kelly (Reference Kelly1985) developed a CGS explanation for the Nuer expansion based on the classic ethnography of Evans-Pritchard (Reference Evans-Pritchard1940). Kelly argued that the bride-price system of the Nuer caused them to count descent more deeply than the Dinka, which made it possible for them to mobilize larger groups of fighters than the Dinka. Hence the Nuer more frequently won fights than the Dinka and incorporated defeated Dinka into their tribes. It is only space constraint and the patience of readers that prevented us from offering many more, and more detailed, examples. The empirical argument in our article is that keeping accounts of group-level cultural variation, and on the impact of cultures selecting genes, is frequently necessary to explain events in human evolution.

Turchin & Currie also use the concept of a framework to describe CGS. They and colleagues are building a historical database in which they expect that the CGS framework will frequently need to be employed to explain features of human history. Other historically minded scholars have adopted the cultural evolutionary framework exactly because it offers a disciplined method for investigating and comparing quite concrete cases (e.g., Steinmo Reference Steinmo2010). It is in the details of such cases that researchers can establish (or not) that an observed cultural difference between groups is actually affecting the outcome of their competition in a manner consistent with an explanation rooted in CGS, as Barclay & Krupp correctly insist must be done.

R3. Measuring cultural variation

Three commentaries include remarks on the difficulty of accurately measuring cultural variation: Mace & Silva, Fuentes & Kissel, and Ross & Atkinson. All three critique our use of the F _ST statistic in particular (also see related discussions in sect. R8.) This is an important point. The most ambitious attempt to catalog the world's living cultures was the descriptive work of classical ethnographers. This led to the development of an important database, the Human Relations Area Files (http://hraf.yale.edu/), that has supported a great deal of quantitative comparative work. The descriptive work on cultural evolution and gene-culture coevolution has been conducted by historians, archaeologists, and geneticists. This is likewise being compiled into databases (target article, sect. 3.2).

Unfortunately, information on variation within societies is very scanty compared to estimates of between-society variation. The individual-level data that we used in section 4.1 to estimate F _ST between various cultural units provide a large number of responses collected from large samples, but we cannot be certain about the extent to which the traits represent culturally heritable variation as opposed to variation influenced by direct environmental effects or genetic variation. We did choose questions (see the appendix at the end of the target article) that seem intuitively to be cultural and related to norms and institutions. It is also undoubtedly the case that competition between the political units known as “countries” is seldom the key level at which selection takes place. Cultural evolutionists have a big task before them to get better estimates of the heritable components of within- and between-group variations.

Despite these limitations, we maintain that our cultural F _ST estimates provide readily available evidence that between-group variation is sufficient for selection to act at the group level (Rogers Reference Rogers1990). Two specific critiques of cultural F _ST require further discussion, however: First, the applicability of the traits measured to CGS, and second, whether variation along geographic clines, not nominal groups, affect the scope of CGS.

Mace & Silva and Fuentes & Kissel worry about whether our cultural F _ST measurements assess variation relevant to cooperation. We note that, (1) many of the cultural traits from which we derived cultural F _ST estimates do reflect cooperative altruism, as is made clear in the appendix. For example, we include questions about willingness to sacrifice personal resources for a public good, specifically in the domain of climate change, government involvement in health care, and taxation by the state. Further, (2) we note that there is no reason why we would need to limit cultural F _ST calculations to beliefs or behaviors related to altruism or cooperation. Every cultural or genetic trait can have its own evolutionary trajectory, and this trajectory can be decomposed using the Price equation (Price Reference Price1970) into change due to individual-level and group-level effects. Moreover, altruistic traits are arguably the least interesting traits with which one could study the relative effects of individual-level and group-level selection quantitatively. By definition, the regression of individual-level fitness on individual-level allele frequency is negative for any altruistic trait (a result from Price Reference Price1972). The adaptive evolution of an altruistic trait therefore must be due to group-level selection (although not necessarily CGS). Investigation of intergroup partitioning of variance in beliefs related to abortion is interesting in that such beliefs relate very directly to fitness outcomes and could, in principle, be studied quantitatively.

Further, broader cultural differences between groups, or measurements across a range of cultural traits, may approximate the relevant suite of differences resulting in a cultural expansion due to group selection. In this view, our measurements of F _ST are trying to estimate the theoretically important heritable variation that selection might be acting on. Some of that variation is, of course, generated by a complex variety of processes. For example, the military expansion of the Nuer at the expense of the Dinka (Kelly Reference Kelly1985) is presumably not just due to differences in political organization. Differences in other cultural traits, such as subsistence strategy and bride wealth, are likely also important. Likewise, modern political groups differ on a large range of possibly integrated issues that may cause one group to increase and the other to dwindle. The multidimensional quality of cultural expansions often requires us to try to identify and quantify the heritable variation that selection may be acting on. Similar levels of complexity are seen with genetic variation. Genetic F _ST measurements must also concede to complexity. Genetic quantitative characters are influenced by multiple contextual and contingent environmental variables within the processes of, for example, gene regulation, phenotypic plasticity, development, and epigenetics more generally. See our discussion of the problems biologists have linking genes to phenotypes in section R14, which is quite parallel to the difficulties in linking questionnaire responses to behavior.

Ross & Atkinson raise the issues of nominal groups and spatial variation. Our measure of cultural F _ST is a direct expression of the partitioning of variation in questionnaire responses across the classification of individuals into discrete categories or groups. We have computed cultural F _ST values between ethnolinguistic, political, and religious groups (appendix), adding to other work computing cultural F _ST between a larger sample of nations and four small-scale East-African societies (Bell et al. Reference Bell, Richerson and McElreath2009, Tables S1 and S2). These categories are in some sense ad hoc, in that we could classify these same individuals into any number of other categories/groups and produce a different F _ST value. However, the data we used were obtained from opinion surveys which sampled units that the survey formulators considered to be meaningful social units. We make no ontological assumption that empirical group boundaries are discrete, or that any one particular group classification is true. We simply illustrate that, if we assume a given classification of individuals into groups, we can calculate the scope for group selection to act on the focal cultural trait as a function of this classification. It is true that many human groups have formal criteria for membership and are often symbolically marked by language, dialect, and other differences. Such groups often compete (target article, sect. 5). In section 2.1 we discussed how symbolic marking can create boundaries to the flow of culture between groups. These boundaries are never completely impermeable. The project of understanding just how cultural diffusion is regulated from an evolutionary point of view is in its infancy (Jordan Reference Jordan2015), although a considerable literature exists on how symbolic and formal boundaries affect the diffusion of innovations.

Ross & Atkinson's finding that geographic variation explains more variation in folktales than currently observed “group boundaries,” shows that defining the group structure of a human population is complicated, but we cannot conclude from this that it is a worthless endeavor. Clinal geographic structure is a kind of group structure, as is nominal group identification. A cultural F _ST for a given trait could be calculated using nominal group membership (religion, political party, etc.), or using classification based on geographic structure to any arbitrary level of geographic resolution (nation, state, county, city, postal code, area code, city block, etc.), or even on classification based on the interaction of geographic structure and nominal group membership. The scope of group selection to operate on the given cultural trait could be analyzed in light of each of these classifications. Of course, to understand the system under consideration requires more than estimating the scope for selection. We have to move from the scope for selection to the concrete norms, institutions, and payoffs to variant behaviors to understand what is evolving and why. Folktales may encode norms and be relevant to some CGS process acting on the groups in which they become common, or they could merely spread because individuals find some of them sufficiently entertaining. See also Barclay & Krupp's comments on the need to understand how a cultural trait affects the success of individuals and/or groups in particular cases.

R4. Little unites the three forms of CGS

The study of cultural evolution has long been bedeviled by the complexity of the selective processes operating on cultural variation. Natural selection is important; the cultural traits of individuals who survive and reproduce are more likely to be passed on than those of individuals who do not thrive. But the capacity of individuals to decide (consciously or unconsciously) what cultural variants to learn and what to pass on is also important. Individuals also influence the innovation of new cultural variants. Moreover, collective decision-making occurs in many societies using institutions like tribal councils and parliaments. All of these processes act as “forces” influencing the direction of cultural evolution. Cavalli-Sforza and Feldman (Reference Cavalli-Sforza and Feldman1981) called these latter types of forces “cultural selection,” and Boyd and Richerson (Reference Boyd and Richerson1985) talked about “decision-making forces” arising from individuals choosing how to behave given the choices offered by their culture and the opportunity to innovate. Morin is quite right that one form of CGS we describe, the differential growth, proliferation, and extinction of cultural groups, is a form of natural selection. The other two, differential imitation and differential migration, are selective decision-making processes, not natural selection. The main thing all three have in common is that they are driven by variation among groups. We are not wedded to any particular terminology in this matter. If people feel more comfortable seeing these as three distinctive processes rather than ones with enough family resemblance to merit the CGS term, we have no objection. At any rate, we hope that the similarities and differences are spelled out well enough in section 2.

Morin's more substantive claims are debatable. He argues that natural section on cultural variation is more likely to favor altruism (traits costly at the individual level) than selective imitation and selective migration because the latter are based on individual decisions. There are two problems with this argument. First, imitation and migration decisions are typically made in the face of considerable causal opacity. The decision maker, even if rational and self-interested, cannot understand all of the causal connections between an institution and the benefits it confers. The imitator and especially the migrant make their decisions on the basis of results, not causal pathways. If a measure of altruism is essential to make an institution work, at least some imitators or migrants may acquire the habit of altruism by imitation or teaching. It is good to recall that if a group selection process is favoring altruism, it will be because, averaged over all the altruists in the meta-population, altruistic types do better than non-altruists. In currently highly successful populations, altruists may be relatively numerous, even if not as successful as some of the rarer non-altruists in the successful population. Unless immigrants or imitators can compute the advantages of altruism versus defection rather accurately, they are likely to often imitate the more numerous altruists. Second, recall the tribal social instincts hypothesis (sect. 2.2). Humans are plausibly well-adapted genetically to living in societies in which a measure of altruism is necessary to produce the benefits of cooperation. In public goods game experiments, cooperation gets established if a sufficient number of “leaders” pay costs to discourage selfish behavior, using whatever tools the experimenters give them, such as punishment or communication (e.g., Baum et al. Reference Baum, Paciotti, Richerson, Lubell and McElreath2012). A minority of people are rationally selfish and a plurality are conditional cooperators, yet in the laboratory with enough tools to work with, altruists can establish near perfect cooperation. This result is common in Western samples, but worldwide behavior is more variable (Herrmann et al. Reference Herrmann, Thöni and Gächter2008). This finding is consistent with theoretical analyses which show that selection for altruism tends to be frequency dependent (Boyd & Richerson Reference Boyd and Richerson1992b). For example, if the costs to altruistic punishers for punishing a given defector are low relative to the cost imposed on the defectors, the frequency of altruistic punishers at equilibrium can be small in proportion. At least in the lab and in theory, a sufficient minority of altruists is necessary to generate cooperation in sizeable groups of non-relatives. There may well be a genetic basis for this altruism driven by culture-led gene-cultural coevolution (sect. 2.2).

Morin's remark that acquisition of behaviors by diffusion will not necessarily lead to demographic consequences is technically true but misleading. Borrowers can acquire a neutral cultural variant that is statistically but not causally related to the success of the donor society. Normally, borrowers expect, or at least hope, that the cultural trait they have copied will have a similar positive effect on the functioning of their society. Once a variant is transferred, it will often have similar effects in the host society as it did in the donor society. However, we should not underestimate the problems of mismatches with other institutions, with other aspects of culture, and with the environment. The decidedly mixed results of efforts to transfer modern Western institutions to the former European colonies after WWII testify to the problem (Gibson et al. Reference Gibson, Andersson, Ostrom and Shivakumar2005). We do not see the relevance of any alliance between the donor and borrowing society. Competing organizations often borrow institutional and military innovations from rivals (see sect. 5.2).

R5. Institutions are important

Several commentaries underscore the importance of institutions. Institutions only operate when a substantial number of people obey their dictates; hence they are inherently group-level traits (sect. 4.3). At the same time, the schedule of rewards offered for conformity to the institution damps down individual-level variation in behavior in the groups that are operating the institution. The punishment of non-conformists prevents them from benefitting from the rewards the institution offers. Institutions vary from group to group. CGS may well act on other sorts of cultural variation, but if institutions exist they are prime candidates to evolve by such processes.

Read argues that CGS does not take into account that cultural idea systems are constituted at an organizational rather than an individual level. Perhaps we were not sufficiently clear, but, in the main, what institutions do is create organizations. Kinship systems as idea systems are used by many peoples to create kin-based organizations such as clans. It is the concrete organizations that actually compete for resources. Ideas may compete directly to be represented in human brains, as in the cross-cultural borrowing we discuss. Presumably, the attractiveness of the institutions of other societies is mostly related to potential adopters' estimates of how much an institution would improve the functioning of their own society. The Netsilik seal-hunting partnership cultural idea system is an institution in our terms, and the Netsilik use it over and over again to create concrete partnerships. Mace & Silva underscore the important point that kinship norms, political systems, and religion are generally group-level traits.

Houdek & Novakova introduce the interesting idea of frozen plasticity. Their idea is that as societies become more institutionally complex, interactions between them make it increasingly difficult to change dysfunctional institutions. Eventually, large-scale events like the loss of wars or major natural disasters lead to a punctuated restructuring of social institutions during which time innovations are made, often prosocial ones. Turchin (Reference Turchin2006) makes a similar point. We did not intend to imply that the CGS framework is committed to steady and irreversible progress toward better institutions, and if we seem to have done so, we apologize. The stylized facts suggest a trend in the Holocene toward ever more complex societies supporting an ever larger human population, but a more detailed view reveals this global trend to be frequently interrupted by local crises, retrenchment, collapse, and the social extinction of whole societies. A mature theory will have to explain both the trend line and the substantial deviations around it. We might imagine that in crises involving societies with more or less equally complex institutions, the societies with the least frozen and least dysfunctional institutions innovate their way out of the crisis, while their marginally more dysfunctional competitors perish. In this way, on the multi-millennial time scale, the capacity of societies to manage institutional complexity has gradually improved, whereas on the millennial time scale much more complex historical processes dominate the dynamics (e.g., Davies Reference Davies2011).

R6. The relationship between psychological processes and cultural evolution

Several commentaries discuss the relationship between psychological processes and CGS, either as alternatives or complements. In our review of the CGS framework (sect. 2), we noted the many ways in which the population-level processes of CGS depend upon individual-level psychological processes. In our empirical discussion, we focused on the early-developing cognitive structures that support the human social learning system (sect. 3.1) and the role of norm acquisition in making institutions possible (sect. 4.2).

Henrich & Boyd's commentary is very close to our view on the relationship between psychological and CGS processes. The important point is that that norms operate at the individual level by reinforcing, co-opting, and modifying gene-based predispositions. For example, Nisbett and Cohen (Reference Nisbett and Cohen1996) compared levels of the hormones cortisol and testosterone in Southern and Northern undergraduates (from the U.S.) when exposed to an insult. Southern participants disproportionately subscribed to a culture of honor norm, which on other behavioral measures indicated that Southerners react much more aggressively to mild insults than Northerners. The hormone measures show how emotional centers in the brain stem can be recruited by a culturally transmitted norm (Panksepp & Biven Reference Panksepp and Biven2012). Nisbett and Cohen review the historical data related to the evolution of the Southern Culture of Honor and the child-rearing practices that cause it to develop.

When considering gene-based predispositions, it must be remembered that by the end of infancy humans have developed a sophisticated capacity for social learning (sect. 3.1). The genetic and cultural contributions to adult cognitive structures will therefore co-develop from early childhood as well as coevolve in evolutionary time. We as yet do not know in much mechanistic detail how cognition develops or how it evolved. Features like incest avoidance, kin recognition, and basic emotions are probably widely shared with other species, making animal models relevant to the discussion. Historic information extracted from the genome, together with improvements in the conventional paleoanthropological record, may one day tell us much more about how human brains and cognition evolved (Richerson et al. Reference Richerson, Boyd and Henrich2010). Cross-cultural developmental studies, which Henrich & Boyd and their co-workers are pioneering, will help tease apart genetic and cultural contributions to development. We think we know enough to be confident that the outline of the relationship between genes and culture in development that Henrich & Boyd outline is correct but our current understanding resolves few details.

We are sympathetic to Tooby & Cosmides' argument that much is built on gene-based social psychological innovations that evolved during the Pleistocene. We briefly reviewed the “tribal social instincts” hypothesis in section 2.2 and discuss it in relation to the sufficiency of individual-level processes in sections R7 here and 2.2 in the target article. Tooby & Cosmides argue that individuals with neural adaptations evolved in Pleistocene forager societies can self-organize into large cooperative structures without the aid of cultural evolutionary processes. They suggest that ancestral foraging societies were small enough that individual-level social skills are sufficient to explain their social structure, and that these skills likewise suffice to explain the much larger scale and more complex societies of today.

Tooby & Cosmides question the realism of experimental games that find altruistic or other prosocial behavior. All experiments involve simplifications. It is not clear that the experimental game studies they cite in support of their argument are any more realistic. See Zefferman's (Reference Zefferman2014a) critique of a simulation study by Delton et al. (Reference Delton, Krasnow, Cosmides and Tooby2011) in this regard (see also Delton & Krasnow Reference Delton and Krasnow2014; Zefferman Reference Zefferman2014b). The most ambitious attempts to relate experimental games to the real world are the cross-cultural studies of Henrich et al. (Reference Henrich, Boyd, Bowles, Camerer, Fehr and Gintis2004; Reference Henrich, Ensminger, McElreath, Barr, Barrett, Bolyanatz, Cardenas, Gurven, Gwako, Henrich, Lesorogol, Marlowe, Tracer and Ziker2010a) (see sects. 6.2, 6.3, and 4.3 in the target article). These studies were conducted mostly in institutionally simple hunting and gathering, herding, and horticultural societies by ethnographers of the people studied. This design was adopted so as to be able to infer from the extensive ethnographic knowledge whether or not game behavior matched behavior in the society concerned. They summarize their main findings as follows:

In section 6.2 we also briefly discussed the social psychologist Daniel Batson's (Reference Batson2011) experimental program designed to test whether psychological altruism was necessary to explain participants' behavior in circumstances where they could volunteer to aid someone in need. Batson's experiments used deception to create scenarios that were as realistic as possible, and he found evidence that some participants behaved altruistically. Tooby & Cosmides' discussions of List (Reference List2007) and Sheskin et al. (Reference Sheskin, Wynn and Bloom2014) are incomplete. In List's taking treatments, dictator giving was substantially reduced relative to the standard dictator game, but the amount of taking was quite modest. In a treatment where dictators and their partners had to work to earn their money, very few dictators chose to take at all. List emphasizes that prosocial morality plays a big role in his results. In Sheskin et al.'s study, young children, but not older ones, chose to pay a cost to reduce the welfare of others relative to their own. Whether this change with age was a product of an innate developmental process or socialization, or both, is unanswered by this study.

We agree that ancestral forager societies were small compared to the giant societies in which many humans lived during the last 5,000 years, but we question the suggestion that they were so small that individual-level social skills can explain their social structure. Human forager societies are already large compared to chimpanzee and bonobo communities. Forager societies have culturally variable norms and institutions regulating family life and interfamily cooperation. As Hill et al. (Reference Hill, Wood, Baggio, Hurtado and Boyd2014) show, the band scale units of forager tribes are fluid, and cooperation with many distantly related or unrelated people is common. In ethnographic cases, and by at least the Upper Paleolithic, trade networks linked local groups on a subcontinental scale. It is true that these institutions can be scaled up considerably with a more productive economy than hunting and gathering. The Turkana, in our example in sect. 5.1, use a subsistence economy based on livestock raising and trade to support a society numbering around one million with institutions hardly different from those of most forager societies. It is the largest society known to us that remains without formal leadership institutions, although people vary greatly in prestige. Most societies larger than a few thousand people evolve formal institutions of leadership and governance. We argue that culture-led gene-culture coevolution had extensively remodeled our social psychology some time before 50,000 years ago. We agree with Tooby & Cosmides that major new genetic adaptations supporting the formation of giant societies are unlikely to have occurred during the last 5,000 years (although minor adjustments may have). It seems likely that large-scale societies arose through a process of institutional innovation, which we understand in some detail in particular cases (e.g., Currie et al. Reference Currie, Greenhill, Gray, Hasegawa and Mace2010). In section 6, we pointed to evidence that language, social control, and religion and ritual all show evidence of “design” features consistent with the operation of CGS.

Tooby & Cosmides and Krasnow & Delton seem to suggest that the study of cultural evolution should be postponed until research in evolutionary psychology has revealed whether or not the patterns observed can be entirely explained by evolved cognitive mechanisms. But that is not how science works. Darwin's explanation of evolution based on heritable variation was valuable even though his theory of inheritance was wrong. Even today our understanding of the genetic inheritance system is incomplete. Krasnow & Delton contrast the “detailed, unique, and novel predictions” of Evolutionary PsychologyFootnote ¹ with the “retrodictions” of CGS of facts we already know, such as why humans have languages. Privileging prediction over retrodiction in in this way leads to some odd conclusions. Would they disparage Darwin's work on the Origin of Species because he was merely retrodicting facts he already knew – that species existed and are generally well adapted to their environments? Science often makes good progress when some components of the phenomenon being studied have to be treated like black boxes. Such investigations motivate and guide the science that aims to open up the boxes.

We are as enthusiastic about the adaptationist approach to human behavior as Krasnow & Delton. Boyd and Richerson's (Reference Boyd and Richerson1985) book is basically an extended evolutionary analysis of how human culture functions as an adaptation. To do this analysis properly, you have to take into account the population-level properties of culture, just as you have to take the population-level properties of genes into account to understand organic evolution. The theoretical work outlined in sect. 2 suggests that one peculiarity of the cultural inheritance system is that it can support considerable variation at the level of groups, and accordingly that it can support the evolution of group functions if there is competition between groups. The target article is mostly an empirical sketch of the evidence available on these axiomatic points. Krasnow & Delton do not impugn this evidence. Their objections depend upon an argument based on G. C. Williams' (1966) gene-centered view of evolution, a view that he subsequently modified (Williams Reference Williams1992) to include the idea of multi-level selection, based on the work of Price (Reference Price1970; Reference Price1972). Maynard Smith and Szathmáry (Reference Maynard Smith and Szathmáry1995) attributed the major transitions of evolution (including human origins) to rare innovations that led to the operation of selection at higher levels. Regarding parsimony, recall Einstein's dictum “Everything should be made as simple as possible, but no simpler.” Trying to understand human evolution without taking into account the population-level properties of culture is to use too simple a framework.

The field of Evolutionary Psychology, pioneered by Tooby and Cosmides (Reference Tooby and Cosmides1989; Reference Tooby, Cosmides, Barkow, Cosmides and Tooby1992), has received much criticism for the narrowness of its program (e.g., Bolhuis et al. Reference Bolhuis, Brown, Richardson and Laland2011), but their enterprise fits well into a larger evolutionary science of human behavior along the lines of Henrich & Boyd's commentary. Barrett's (Reference Barrett2015) recent book, and work such as Gervais et al. (Reference Gervais, Willard, Norenzayan and Henrich2011), show how naturally psychological and population based approaches fit together. In this regard Buss' (2014) textbook Evolutionary Psychology, cited by Krasnow & Delton, is interesting in that it does not contain a discussion of the human social learning system that has been well described by developmental and comparative psychologists (sect. 3.1). This is an odd omission, given that the human social learning system is perhaps the best-described human cognitive module known to have evolved considerably during the Pleistocene. There is also no discussion of institutions or norms in Buss (Reference Buss2014).

We appreciate Suárez & Koenig's raising the evidence on the selectivity of even young children's social learning. The broader theory of cultural evolution (e.g., Boyd & Richerson Reference Boyd and Richerson1985; Cavalli-Sforza & Feldman Reference Cavalli-Sforza and Feldman1981) developed mathematical models to investigate the effect on social learning selectivity based on genetic predispositions and social learning selectivity based on earlier acquired culture. These early models of bias were derived from the literature on persuasion and the acquisition of innovations by adults. The more recent studies of social learning in children have filled important gaps. The idea that children are evaluating potential role models for expertise, truthfulness, conformity to norms, and affiliation with the child's group are important planks in the Tribal Social Instincts hypothesis, showing that children do favor prosocial and in-group models. This, we suggest, reflects an evolutionary history of living in norm- and institution-bound societies, which create a tendency for norms and institutions to culturally evolve in prosocial directions. We are also interested in the variation in children's social learning biases. In section R4 we noted that young adults vary substantially in their strategies in the Public Goods Game. When these young adults were younger, would they have had different social learning biases that led them to acquire a different set of moral norms?

This is consistent with the idea that selection of prosocial tendencies under CGS will be frequency dependent. The rational component of children's imitation may tend to smooth out intergroup differences. However, it is important to remember the observation of Harris (2012) that in many matters children have so little knowledge that they must accept the testimony of adults. In the case of complex artifacts and institutions, even adult experts do not seem to have deep causal knowledge of how these cultural constructs function (Sterelny Reference Sterelny2012). Such causal opacity will remove much of the influence of rational choices, and causal clarity will make such choices very powerful. In laboratory experiments with young adults, payoff-biased social learning strategies are commonly used if sufficient information is available, and more conformity is used when information is less definitive (McElreath et al. Reference McElreath, Bell, Efferson, Lubell, Richerson and Waring2008).

Vigil & Kruger's suggestion that honest communication of pain is an important part of social life is plausible. It parallels the argument we made for the case of language in section 6.1. More generally, it is common for people to signal their emotional states, positive as well as negative. Signaling systems in which self-interested deception is possible can evolve only if there is a certain level of trust between senders and receivers of signals, especially if the signals are inexpensive. Expressions of pain often trigger helping behavior in others. This is a reliable reaction. It has led to the development of the strategy of terrorist bombers to set off a second bomb as people rush in to aid the victims of the first. Images of the pain suffered by a particular group are used to recruit charitable donations and volunteers to aid them in natural disasters and wars. Vigil & Kruger provide evidence that people who identify with an in-group can be powerfully affected by images of harm done to in-group members. At the same time, some people exaggerate their suffering to attract aid they do not deserve, and such behavior is disparaged and policed much like any other lie.

R7. Individual-level processes are important

Several commentaries remind us that cultural evolution includes many phenomena that are distantly, if at all, connected to group-level processes. Most cultural evolutionary analysis asks how individual-level decisions affect the evolution of culture where the agents making the decisions are assumed to be boundedly rational individuals pursuing their inclusive fitness interest (e.g., Boyd & Richerson Reference Boyd and Richerson1985). In this regard we endorse many of the remarks in the commentaries of Amir, Jordan, & Rand (Amir et al.); Singh, Glowacki, & Wrangham (Singh et al.); Tooby & Cosmides; Houdek & Novakova; and Nesse. However, when it comes to human social life, we are convinced that individual-level processes are often heavily influenced by the products of CGS processes.

Amir et al. present an excellent model of individual decision-making that we think is largely compatible with CGS. In their social heuristics hypothesis (SHH), Amir et al. argue that individuals internalize the strategies that are typically advantageous in daily life. We argue that institutions substantially determine which strategies are typically advantageous. Institutions are group-level traits that evolve due to CGS (sect. 4.3). Amir et al. give the example of how voting institutions might guide individual behavior. Importantly, voting institutions, and social choice institutions more generally, vary widely across groups, even across modern democratic states. While individual behavior might be explained by SHH, given a particular set of voting institutions, voting institutions themselves are group-level traits because an individual cannot unilaterally decide to use different voting rules than the rest of her group. Since voting institutions can only change at the group level, they are subject to CGS. If, as Amir et al. suggest, individuals can migrate to join groups with voting institutions more to their liking, this would be a mechanism by which CGS could operate (sect. 2.1). Additionally, Amir et al.'s idea that institutions homogenize individual behavior is precisely what potentiates CGS processes by suppressing individual-level variation and creating group-level variation. Because sanctions and rewards can stabilize institutions that have highly variable functionality, institutional innovators cannot easily move from suboptimal to optimal institutions (see sect. 4.3). In summary, we see SHH as a theory to explain human behavior, given variation in norms, institutions, and innate predispositions, whereas CGS and gene-culture coevolution aim to explain norms, institutions, and innate predispositions themselves. An important future project might be formally integrating ideas from SHH and CGS to more tightly link gene-culture coevolutionary models with the findings of cross-cultural experimental psychology and economics.

Both Singh et al., and Houdek & Novakova make claims about how endogenous processes of cultural change will affect the emergence of group-beneficial institutions. Singh et al. argue that endogenous processes of cultural change can lead to norms that are group beneficial because the interest of powerful leaders align with the interest of the group. They therefore argue that CGS is unnecessary for explaining group-beneficial institutions. In contrast, Houdek & Novakova point to the several pathways by which the action of self-interested actors can wreck the success of a group. We agree that punishment can sustain antisocial behavior as readily as it can sustain prosocial behavior and know that history is rife with examples of coercive, extractive institutions that were not optimal for the group, but benefited the people imposing these rules (e.g., Turchin & Nefedov Reference Turchin and Nefedov2009). And we argue that this is exactly why the processes of CGS are needed to explain the unusual level of prosocial outcomes that we do see in human societies, such as property rights, redistributive mechanisms, safety nets for those struck by misfortunes, military defense, and public capital facilities.

We think the diversity of opinion on this issue reflects the reality that both self-interest and cooperation do occur. Self-interested actors maneuver rules to their advantage, as evidenced by the corruption that plagues many institutions. Furthermore, individuals with more bargaining power in a situation – people with wealth, influence, coercive authority, weapons, private militias, family ties, social connections, and so forth – will try to extract more for themselves. Such a situation often leads to rules and outcomes that are not beneficial for the group. In fact, it is seldom (if ever) the case that rules and outcomes that are optimal for the group are achieved.

If self-interested leaders often create rules that also happen to benefit the group, as Singh et al. argue, shouldn't group-beneficial outcomes be widespread in animals? Hierarchies and power differences are extremely common in social animals. Yet, the presence of powerful individuals with coercive capacity has not created animal societies with complex rules for solving public goods problems, or enforcing property rights, or any of the myriad other forms of cooperation commonplace in human societies. When cooperative outcomes are observed in animal societies, they are typically among kin. Singh et al. do not consider the full strategy space of rules that are available to self-interested leaders. They bring up the example of Cheyenne leaders, who implemented a rule regarding thieving of horses: “Now we shall make a new rule. There shall be no more borrowing of horses without asking.… [If] the taker tries to keep them, we will give him a whipping.” They conclude that the leaders impose such a rule because preventing theft and maintaining social order benefits the leaders. But wouldn't a rule in which the leader takes this stolen horse for himself benefit the leader even more? What about a rule requiring such thieves to hand over their wives to the leader? Or a rule that says that the punishment only applies to people the leader does not like? Let us assume for a moment that leaders do always attempt to establish rules in their own interest. Nevertheless, they will occasionally make rules favoring group interests either because of an alignment of elite and non-elite preferences, or because of errors of judgment or exogenous changes. No matter the intent of the original rule-makers, CGS processes will act to favor such rules and disfavor rules that are actually selfish. We do not want to say that the interests of leaders and followers never align without evolved institutional incentives, just that this mechanism alone can only account for the limited amounts cooperation found in societies with dominance hierarchies.

Tooby & Cosmides speculate that modern institutions can be far better explained by neural adaptations in dense persisting social networks of intelligent cultural agents than by CGS. Once again, one cannot discuss individual behavior without considering the possibility that humans have adapted by gene-culture coevolution to live in societies shaped by CGS. As we discussed in section R4, it takes a largish minority of altruistic punishers to establish cooperation in laboratory public goods games, consistent with CGS models (Boyd et al. 2003). In societies with poor institutions, experimental participants can't establish cooperation using the punishment tool (Herrmann et al. Reference Herrmann, Thöni and Gächter2008). Leaders are leaders either by virtue of coercive power or by grants of prestige from their followers, or both. Prestige is a strong motivator of behavior (Henrich & Gil-White Reference Henrich and Gil-White2001), and even the powerful may act in the group interest to gain prestige. Also, selfish political operators can't advocate for new institutions based on selfish advantage unless they are dictators. To the extent that they need the consent or at least the acquiescence of others, they will at least need to convince followers that the innovation is in the group's interest even if the group is only a narrow ruling class. If the new institution turns out to be only in the leaders' interests, their prestige and perhaps authority will often by diminished by unpopularity, resistance, and revolt (Insko et al. Reference Insko, Gilmore, Drenan, Lipsitz, Moehle and Thibaut1983). Thus, endogenous processes can aid in the creation of group-beneficial institutions, but for this to work on any scale, the prosocial instincts would have had to be favored by some process like CGS acting via gene-culture coevolution in the past.

Nesse brings up the very important concept of social selection. Certainly many non-human social phenomena are shaped by social selection. Indeed, it seems likely that in any social species social selection plays an important role. For example, the mutual grooming seen in many social mammals must be the product of social selection disfavoring individuals who were unacceptably rough or ticklish. The very ubiquity of social selection argues against it being capable, on its own, of generating societies of the size and complexity of human societies. Runaway sexual selection is a common phenomenon. If a runaway form of partner choice social selection were possible, shouldn't we see many non-human societies with extensive cooperation between non-relatives? As we argued in sections 2, 4.3, and 6.2 of the target article, culturally transmitted institutions act to greatly amplify social section. They are rules that deliver rewards and sanctions to individuals, they vary substantially from society to society, and they influence the outcome of inter-organizational competition. We think it is CGS on institutions that put humans on a runaway path to semi-domestication. Pleistocene culture-driven gene-culture coevolution would have favored the evolution of guilt, pity, commitment, empathy, and generosity, much as Darwin (Reference Darwin1874) imagined in the epigraph from The Descent of Man at the start of our target article. As we elaborate in sections R10 here and 4.3 in the target article, even the ubiquitous simple partnerships, like marriage and trade, are heavily institutionalized even in simple societies.

Lamba describes a field study of variation in cooperation (measured by an experimental public goods game) in a sample of 14 villages within an ethnic group in India (Lamba & Mace Reference Lamba and Mace2011). Some of this variation could be explained statistically by population size, age, and sex, which Lamba attributes causally to individual-level effects. As far as we can see, it could also involve some component of micro-cultural variation, for example, norms of cooperation might differ by gender. However, all experimental game studies we know of find ample individual-level variation, including the studies of Henrich et al. (Reference Henrich, Boyd, Bowles, Camerer, Fehr and Gintis2004; Reference Henrich, McElreath, Barr, Ensminger, Barrett, Bolyanatz, Cardenas, Gurven, Gwako, Henrich, Lesorogol, Marlowe, Tracer and Ziker2006; Reference Henrich, Ensminger, McElreath, Barr, Barrett, Bolyanatz, Cardenas, Gurven, Gwako, Henrich, Lesorogol, Marlowe, Tracer and Ziker2010a). Henrich et al. (Reference Henrich, Boyd, McElreath, Gurven, Richerson, Ensminger, Alvard, Barr, Barrett, Bolyanatz, Camerer, Cardenas, Fehr, Gintis, Gil-White, Gwako, Henrich, Hill, Lesorogol, Patton, Marlowe, Tracer and Ziker2012a) report that the variation between communities, estimated for some but not all of their samples of societies, is approximately the same as the variation that Lamba and Mace (Reference Lamba and Mace2011) reported and is rather smaller than the between-cultures variation in the Henrich et al. studies. We do not doubt that as the scale of measurement of variation shrinks, individual-level effects will become increasingly important, all the more so to the extent that cultural variation is concentrated at a larger scale. If people in a regional-scale society all operate under the same institutions, perhaps the main sources of individual variation in behavior are environmental and genetic. Lamba (Reference Lamba2014) also reports finding only modest effects of cultural transmission as measured by changes in individuals' behavior between two rounds of a one-shot public goods game. In the second round, some individuals tended to conform to the mean contribution in the first round and some tended to adopt the payoff-maximizing contribution, but there was a strong mode at no change. This finding is in line with laboratory studies of social learning (e.g., McElreath et al. Reference McElreath, Bell, Efferson, Lubell, Richerson and Waring2008). A major limitation of such studies is that one brief social learning experience cannot be expected to have a large impact on habits and norms that have been acquired over a lifetime. The experiments reported in section 4.2 of the target article show that children acquire norms quite readily.

R8. On what units does CGS operate?

Some authors raise important issues concerning the units and levels of selection. As we remarked in section 1, the CGS framework can be applied to many different concrete entities; it includes the concept of multi-level selection. It is applicable to easily copied fragments of culture (which Whiten & Erdal call “meme-complexes”); to structural elements within a population, such as business organizations (discussed in sect. 5.2); to ethnolinguistic groups, corporate kin groups, clubs, nations, religions; and even to multi-ethnic empires (Turchin Reference Turchin2009).

Whiten & Erdal suggest that the relevant groups for CGS should be groups of cultural replicators (meme-complexes), rather than groups of people. While meme-complexes, like other cultural traits, can vary between groups or increase and decrease in frequency as part of a CGS process, in the CGS framework, they are not the groups themselves. However, if meme-complexes are self-reinforcing, they may become particularly stable within groups once adopted, similar to Smaldino's (Reference Smaldino2014) concept of a “group level trait.” Groups may, as a result, become strongly influenced by particular meme-complexes, and cultural group selection may operate on meme-complexes with particular force.

When applying the CGS framework to a particular case, we think it is as important to analyze the ecology of competition and cooperation when defining groups as it is to analyze the patterns of cultural variation. CGS is likely to be operating where there is a coincidence of a pattern of cultural variation and a pattern of conflict. Lamba seems to suggest that ethnolinguistic groups have a privileged position in the CGS framework. That is not our view. She finds considerable variation in cooperative behavior within the ethnic group she studied, partly related to demographic and ecological processes. This is not surprising once one takes into account the relevant scales of collective action for the traits she examines. For example, Lamba describes an experiment where villagers are asked to divide a measure of salt between themselves and other villagers. If salt division is a collective action problem typically encountered at the scale of villages, then villages would be the relevant group in the CGS framework. The ethnolinguistic group would not be the relevant group unless the problem of salt division could be solved by collective action at that larger scale. Similarly, one of the best studied possible examples of CGS is warfare between clans within ethnic groups in Highland New Guinea (Soltis et al. Reference Soltis, Boyd and Richerson1995). Mace & Silva and Acedo-Carmona & Gomila also describe field situations in which the locus of cooperation and conflict is at smaller scales than what a Western anthropologist might define as a cultural group. In section 5.1 we described the cooperation and competition among the Turkana which are organized at the level of a large ethnic group. In section 5.2 we discussed business organizations that are typically subnational in scale, although large modern firms are often multinational in their operations and multi-ethnic in their staff. Clearly, human societies are both complex and diverse. Doing full justice to this diversity is beyond the scope of our target article, but we would not want to be misunderstood as thinking that the CGS approach is committed to a particular claim about the kinds of groups that compete.

Acedo-Carmona & Gomila raise the important issue of the persistence of medium-scale diversity within state-level societies. They argue that CGS predicts regional cultural homogenization in the long run, and that such homogenization does not occur in their cases in Oaxaca and Northern Ghana. This is not quite right because, while it is true that, in a stable environment, a simple CGS model predicts that competing groups will have similar group success at equilibrium, social evolution is a complex design problem (Boyd & Richerson Reference Boyd, Richerson, Nitecki and Nitecki1992a). There are likely to be many equally successful but rather different ways of organizing societies. Furthermore, medium-sized organizations often evolve better institutions from the bottom up than can be created by top-down organizations like nation-states (Ostrom Reference Ostrom2005). CGS for cooperative institutions at the scale of neighborhoods, villages, churches, business organizations, and the like probably often favors institutions that protect the local autonomy of medium-sized organizations which are successful at solving a particular problem. For example, around Lake Titicaca, Peruvian lakeside villagers actively defend fishing territories that are completely illegal under Peruvian law. Yet they succeed in defending these rights against trespass by neighboring villagers, interference from state officials, and use by state-sanctioned private companies (Levieil & Orlove Reference Levieil and Orlove1990). Multiple equilibria in complex systems and relative efficiency of medium-scale organization will therefore counter the homogenizing tendency of CGS and preserve medium-scale diversity within bigger units like states or ethnolinguistic groups.

Santana, Patel, Chang, & Weisberg (Santana et al.) make a similar argument focused on the phenomenon of religious syncretism. They point out that selective conquest, migration, or imitation seldom results in total replacement of pre-existing culture. More often there is a new hybrid culture, a process sometimes called “ethnogenesis.” Among other things, ethnogenesis offers the possibility of cultural recombination, discovering favorable new combinations of cultural traits not present in either of the contributing lineages. As Santana et al. correctly observe, there must be evidence that traits acquired or retained during ethnogenesis actually had group functional effects that drove acquisition or retention if we are to attribute their evolution to CGS. We pointed to literatures in section 5 that document how competing organizations use techniques such as head-hunting and espionage in highly organized ways to acquire the successful techniques of rivals. (For the borrowing of military technology, see Horowitz [Reference Horowitz2010].) Still, Santana et al. are surely correct that many cultural traits spread between groups or are retained in ethnogenesis for their individual-level functions, not to mention the frequent diffusion of harmful practices such as tobacco smoking and the consumption of fast food. The quantitative study they suggest to examine the levels of organization that are affected by a sample of diffusing innovations would be very interesting.

Santana et al.'s case of the Catholic veneer adopted by local Mexican communities is an interesting example related to those described by Acedo-Carmona & Gomila. As Foster's (Reference Foster1960) classic account of America's Spanish heritage documented, the pattern of highly heterodox local beliefs in America replicates a pattern found in Spain, despite energetic and largely successful attempts by colonial authorities to prevent the spread of unorthodox religious ideas from Spain to the New World. Again, local differentiation of beliefs could be related to multi-level CGS favoring one set of institutions at the village scale correlated with local heterodox religious beliefs and another associated with Catholic orthodoxy among national elites. As we discuss in section 6.3, religions seem to function to provide symbolic group boundaries, moral justifications for institutions, and rituals that express group solidarity. The “veneer” inter-level relationship between elite Catholic orthodoxy and village-level heterodoxy is particularly well studied in the case of the New Mexican Pueblos. The Pueblo peoples resisted Spanish domination violently, culminating in a spectacularly, if temporarily, successful revolt in 1680. The Spanish returned 12 years later much chastened. “Veneer” Catholicism symbolized the Puebloans' acquiescence to Spanish overlordship in exchange for a Spanish tolerance of substantial Pueblo autonomy, an arrangement that persisted under Mexican and U.S. rule (Simmons Reference Simmons and Ortiz1979).

Regarding Mace & Silva's Northern Ireland case study (Silva & Mace Reference Silva and Mace2014), we appreciate their raising the issue of parochial altruism. Some plausible CGS models do predict the evolution of a combination of in-group cooperation and out-group hate, but in section 4.2 we dissociated ourselves from the view that this relationship is a necessary product of CGS in general. Empirically, no law-like association between in-group cooperation and out-group dislike has been found (Brewer Reference Brewer2007). This is not hard to understand. Peaceful relations between groups make possible gains from trade while warring with neighbors is generally negative sum, costly to both sides. Only in very asymmetric conflicts or surprise attacks are aggressors reasonably certain to come out ahead and in the case of surprise attacks, victory is usually temporary. Large organizations, like states and empires, often offer mediation or legal services to settle disputes between the smaller organizations they contain. When such services are efficient and just, the constituent organizations are typically happy with the results (North et al. Reference North, Wallis and Weingast2009). In simple societies, the lack of mediation institutions can lead a society toward extinction, at least in a competitive environment (Knauft Reference Knauft1985). Even in cases where enmities are raw and deep-seated, as in Northern Ireland, mediation by the United Kingdom and Ireland, assisted by the United States, eventually damped down violent conflict there.

De Dreu & Balliet's point that intergroup interactions are mixed motive rather than zero sum is likewise appreciated. Groups that trade and make peace may also be competitors. Thus, firms in an industry may come together to form a trade group that lobbies, advertises, and does research to benefit the industry as a whole, while still competing vigorously on price, quality, and so forth. In our example of the 19th century dye industry in section 5.2, it was cooperation between competing German firms to lobby states to create research professorships in chemistry that gave German firms collectively a decisive advantage over their British rivals. Societies find it expedient to ally with others to fight a threatening common enemy even if they continue to compete vigorously in other spheres. Any organization that does not have a robust set of institutions to make possible trade and alliance is liable to be outcompeted by ones that do. The Comanche dominated the Southern Plains of North America militarily for a couple of centuries in part because they made alliances and trade relations with other Indian tribes and with Europeans (Hämäläinen Reference Hämäläinen2008). A tight coupling between in-group love and out-group hate would make alliance-forming difficult, something that CGS is likely to disfavor, all else equal. Models also exist which show cooperation arising in a game to survive in harsh unstable environments, even in the absence of conflict between groups (Smaldino et al. Reference Smaldino, Schank and McElreath2013). A failure to cooperate may lead to group extinction due to a failure to extract sufficient resources to maintain a viable population, even in the absence of competition. We do not classify this as a form of CGS because it does not involve between-group variation and inter-group competition, but that does not mean it is unimportant in favoring group-level cooperation. This process would also tend to produce in-group love but not out-group hate.

Herzog proposes that pet-keeping can best be explained by CGS. He presents no obvious group fitness benefits of pet-keeping. The best evidence provided suggests only that, as the author states, pet-keeping is like fashion, and it spreads through cultural transmission. It is influenced by parental preferences, prestige, and conformity. Pet-keeping is apparently not a cultural trait that is appreciably influenced by CGS processes. The fact that people readily form attachments to pets illustrates how flexible the human attachment system is, but the capacity to form attachments to social groups is a much more consequential example for CGS processes.

R9. Variations in the strength of CGS

Commentators Gelfand, Roos, Nau, Harrington, Mu, & Jackson's (Gelfand et al.'s) point that the strength of CGS processes will vary as a function of the threats a society faces makes sense. Some of the changes in the commitment to groups are short term in nature, for example, the compulsory mobilization of a society fighting an existential war. The Mu et al. (Reference Mu, Kitayama, Han and Gelfand2015) study shows that priming subjects to think of an out-group threat immediately increases cooperation. These are the sorts of psychological predispositions that might result from a long evolutionary history of group living (sect. 2.2). Boehm's (Reference Boehm1984) study of Montenegrin feuding described institutions for suspending feuds when it was necessary to raise a tribal army to fend off Muslim invaders. Montenegrins' willingness to suspend feuds was no doubt aided by the clear danger to the whole society posed by potential Muslim conquest. Other changes are longer term, as evidenced by institutional differences across American states. Some scholars have argued that these differences have a historical character derived from patterns of colonial immigration from different parts of Britain (Fischer Reference Fischer1989; Nisbett & Cohen Reference Nisbett and Cohen1996). Immigrants may also settle in regions and situations to which their norms and institutions are adapted, making the correlations in the Mu et al. study hard to interpret causally, as Gelfand et al. recognize. The role of possible genetic adaptations to living in tight versus loose societies is interesting, but the arrow of causation also has to be established. Human genetic diversity is strongly correlated with distance from Africa, implying that drift was a strong factor in our evolution (Pickrell & Reich Reference Pickrell and Reich2014). Institutions might coevolve with gene frequency differences that were a product of chance factors. Ancient DNA can be used to characterize the genotypes of people before and during a putative culture-caused selective sweep and hence rule out the possibility that a pre-existing genetic difference influenced cultural evolution (Itan et al. Reference Itan, Powell, Beaumont, Burger and Thomas2009).

R10. Ancient preadaptations

It goes without saying that human culture and CGS processes rest on ancient roots. Even something as novel as our capacity to form large-scale societies is likely to be a mere modification of traits present in our ancestors. Chisholm, Coall, & Atkinson (Chisholm et al.) offer mother–infant attachment as an example. This is a bond that is practically required by mammals and is thus phylogenetically ancient. It is one that has been remodeled for many different purposes in the course of social evolution, just as Chisholm et al. argue. See also Panksepp and Biven (Reference Panksepp and Biven2012, Chapters 8–10). Attachment to groups, for example, looks as if it depends upon some of the same subcortical roots as individual–individual attachments (Haslam Reference Haslam2001).

Burkart & van Schaik argue that cooperative breeding provides a much more recent precursor of why humans rather than some other great ape evolved into the hyper-cooperative niche. The comparative primatology of cooperative breeding is certainly striking, as their experiments show. Both generalized cooperation and social learning appear to be enhanced in cooperative breeders. Cooperative breeding is certainly a candidate to have favored more cooperation in early hominins and to have amplified the advantages of social learning by exposing selective learners to a greater number of potentially useful cultural variants. Larger social groups could also have favored improved communication. Our only worry is our original one in the target article (sect. 1) that the paleoanthropological record is almost silent on the evolution of social organization. Stiner and Kuhn (Reference Stiner, Kuhn, Hublin and Richards2009) argue from diet and artifact types that a sexual division of labor did not evolve until the Upper Paleolithic, suggesting that large-brained hominins, closely related to modern humans, had very different social organization.

Human cooperative breeding is heavily institutionalized in ethnographically known small-scale societies by rules about kinship and marriage. Mathew et al. (Reference Mathew, Boyd, Van Veelen, Richerson and Christiansen2013) suggest that the human capacity to cooperate in many different activities is based on our ability to create institutions to manage diverse forms of cooperation (see sect. 4.3 of the target article). Thus, an alternative hypothesis is that simple institutions preceded cooperative breeding in humans and made it possible. Perhaps families were one of the earliest social institutions. On present evidence both hypotheses seem equally plausible, not to mention any number of more complex coevolutionary scenarios one can imagine in which genes and culture played alternating leading and lagging roles.

R11. Pleistocene adaptations generated by CGS

Commentators Gintis, Fuentes & Kissel, and Whiten & Erdal observe that the target article includes little mention of CGS and culture-driven gene-culture coevolution in the Pleistocene. They outline several examples of adaptations that may very well have been built by these processes far earlier than the examples we discuss. We agree that these are good candidates to be explained by CGS and culture-driven gene-culture coevolution along the lines of the tribal social instincts hypothesis (sect. 2.2). However, in this article we emphasized empirical evidence that is as close to unassailable as it is possible to get. Inevitably, this resulted in an emphasis on more recent cases. Humans seem to have been quite rare before about 50,000 years ago (Li & Durbin Reference Li and Durbin2011). Hence, the paleoanthropological data record is scanty and hard to interpret. Language, institutions, and other traits that are central to the operation of CGS leave poor proxies in that record. The temporal resolution of the record is poor, so it is hard to be sure whether cultural evolution is the leading or lagging variable in coevolutionary events. Ongoing improvements in the paleoanthropological record will help make hypotheses about deep time events more testable, and there is great promise for extracting more genetic history from living autosomal DNA, supplemented by recovery of subfossil DNA (Richerson et al. Reference Richerson, Boyd and Henrich2010). Gintis, Fuentes & Kissel, and Whiten & Erdal raise questions about the most important events in our evolutionary history. We look forward to advances that will help us better understand them.

R12. Relationship of CGS to the expanded evolutionary synthesis

We thank Fuentes & Kissel for raising the issue of the relationship between CGS and the extended evolutionary synthesis (EES). We are quite willing to include CGS as part of the EES. Cultural evolutionists often cite and are cited by other EES authors. For example, we cited Laland et al. (Reference Laland, Odling-Smee and Myles2010; Reference Laland, Sterelny, Odling-Smee, Hoppitt and Uller2011) in our target article, and both of these papers cite numerous publications by other cultural evolutionists. If our tribal social instincts hypothesis is correct, the evolution of the cultural inheritance system under CGS had major impacts on the genetic component of our social psychology, not to mention many other aspects of our phenotype. Two of the three CGS mechanisms we discuss are based on selective decisions, not natural selection, and as we have remarked here in section R4, individual and collective decisions regarding institutional innovation and adoption are potentially biased in a prosocial direction due to the ancient remodeling of our social psychology by CGS. In section 6.2, we discussed the genetic evolutionary effect that punishment on deviants would have if indeed there is (or was) genetic variation in tendencies to deviate from the strictures of norms and institutions. Barclay & Krupp defend a decidedly non-EES picture in their remark that proponents of CGS are confusing proximate and ultimate causation. Exactly who is confused is the central issue in the debate over EES. For those who accept the evidence for culture-led gene-culture coevolution, it is the opponents of the CGS who are confused (Laland et al. Reference Laland, Sterelny, Odling-Smee, Hoppitt and Uller2011).

R13. Selection of extended behavior patterns

Simon makes a useful point in her commentary that selection on patterns of behavior has to consider the effects on fitness that accrue over time and across individuals. Individuals may pay a cost for being prosocial at one time but be rewarded later because they had achieved a reputation for generosity. Carriers of a cultural variant or genotype encouraging generosity may incur a net lifetime cost not paid by the ungenerous. In the case of such an altruistic group-selected trait, generosity may increase because even though it is counter-selected within every group where it is present, the success of groups with many generous types relative to those with few can result in the generous trait increasing in frequency in the meta-population. Thus, Simon is not quite correct to say that “cooperation … was selected because it commonly enhanced the individual's long-term fitness” in the sense that there is, by definition, always a within-group disadvantage to altruistic generosity.

Much of evolutionary theory is devoted to understanding how selection trades off costs and benefits of extended patterns of behavior. Life history theory studies how the effect of sacrifice of reproduction at one stage of the life history can or cannot be favored by increased reproduction at another stage. Evolution in temporally varying environments will tend to favor types that have lower mean fitness in good times but that resist having very low fitness in bad times. We did not consider these important issues in our target article in favor of issues more directly relevant to the evolution of cooperation. As Equation 1 in section 4.1 depicts, in the context of multi-level selection, the ratio of group benefits to individual costs that can be favored by group selection depends upon the ratio of within-group to between-group heritable variation for the trait concerned.

R14. The evidence is incomplete

Commentators Mace & Silva, Barclay & Krupp, and Turchin & Currie remark that the evidence supporting CGS is inadequate to support any strong claims for the utility of the framework. Mace & Silva point out that the case for CGS to the exclusion of other explanations is hard to prove. We feel that there is no need to prove the case to the exclusion of all other processes that influence or may influence human behavior. Given that human societies are complex and diverse, we do not expect a single model derived from the CGS framework, or any other framework, to provide a complete and universal explanation. In many cases, competing hypotheses or models are not mutually exclusive.

Barclay & Krupp say that we have established that CGS processes are plausible, but that we have not shown that a particular group trait was caused by cultural transmission and directly caused differences in group fitness. We did point to evidence that norms are transmitted (see sects. 3 and 4.2) and that the outcomes of competition are often explained by institutional and other cultural differences between groups (sect. 5). Institutions are group-level traits by their nature (see sect. R5 here). We have pointed to a small selection of concrete empirical cases in our article and additional ones in this response, but Barclay & Krupp are right that in no particular case can we put all the pieces of a complete CGS explanation together. This is a serious problem in the science of evolution generally (Brandon Reference Brandon1994). Biologists know a lot about genetic transmission, mainly from laboratory studies, and a lot about the evolution of phenotypic characters in the wild, but in most cases, connecting genes to phenotypic characters has proven difficult. Many good studies of the transmission of norms have been done, and the role of institutions in inter-organizational competition has also been well-studied. But it is hard to conceive of a study that would allow us to understand how a complex institution is built from the ground up by the social learning of norms. The complexity of cultural inheritance and of cultural evolution precludes easy progress on many important questions. Evolutionary biology is still a thriving scientific field 155 years after the Origin of Species; we do not think the study of cultural evolution will be a perfected science a century from now.

In this regard, we enthusiastically endorse the path forward painted by Turchin & Currie. The gold standard in empirical inference has become fitting competing models to data using techniques such as Markov Chain Monte Carlo simulations and testing the goodness of fit of these models to data using Bayesian and information theoretic measures (e.g., Gerbault et al. Reference Gerbault, Allaby, Boivin, Rudzinski, Grimaldi, Pires, Vigueira, Dobney, Gremillion, Barton, Arroyo-Kalin, Purugganan, Rubio de Casas, Bollongino, Burger, Fuller, Bradley, Balding, Richerson, Gilbert, Larsen and Thomas2014). Advances in genomics have furnished huge amounts of historical data on the course of evolution. Human languages, living and dead, have long been mined for human historical information, and to some extant non-linguistic data have been used as well (see sect. 3.2). As Turchin & Currie suggest, archaeological and historical data can surely be assembled into databases that can support model-fitting and model-comparison approaches.