R1.1. Things that we did not say

We are not arguing that:

• The presence of Big Gods is the only magic bullet that caused societies to scale up; there are no other elements beside “Big Gods” in the “religious packages” that contributed to the process of escalating social complexity (Dutton & Madison; Krueger; Watts, Bulbulia, Gray, & Atkinson [Watts et al.]).

• Prosocial religions are a necessary condition for large-scale cooperation; therefore, without prosocial religions, large-scale cooperation is impossible (Beit-Hallahmi; Demetriou, Makris, & Pnevmatikos [Demetriou et al.]; Galen; Seewald, Hechler, & Kessler [Seewald et al.]).

• There are only two types of gods: Big Gods and completely amoral and indifferent gods, without intermediate cases (Boyer & Baumard, Brazil & Farias, Watts et al.; see Roes for methodological clarifications).

• Only monotheistic religions have Big Gods, or any kind of supernatural agents that could induce greater prosociality; polytheisms are incompatible with Big Gods (Costello; Dutton & Madison).

• Big Gods always precede social complexity (Watts et al.).

• Other than world religions, there are no other mechanisms that build social solidarity on a large scale, such as ethnic bonds, cultural norms and traditions grounded in moral emotions, and secular institutions and ideologies (Beit-Hallahmi, Galen, Seewald et al.).

• Religion occupies a special domain in human cognition (Boyer & Baumard)

• Prosocial religions encourage indiscriminate and universal love, or that the prosociality in world religions is desirable or morally good for everyone (Galen, Krueger, Roes; see McKay & Whitehouse for discussion).

• The prosociality that world religions promote do not cause intergroup intolerance, conflict, violence, and within-group inequality, injustice, and exploitation (Galen; Krueger; Hobson & Inzlicht). Or that prosocial religions cannot galvanize cooperation unless they completely eliminate within-group conflict or exploitation (Seewald et al., Soler & Lenfesty).

Assuming that these misunderstanding arose from a lack of clarity in our target article, we expand upon these points below.

R1.2. Prosocial religions are part of a much larger network of causes

Our argument is grounded in the idea that cultural evolutionary pressures selected for a broad and diverse suite of mechanisms that facilitated large-scale cooperation. As we explain in the target article, we hypothesize that complexes of culturally transmitted traits – beliefs, values, practices, traditions that (1) sustain within-group solidarity and (2) promote success in competition with other social groups – are culturally selected to the extent that they allow groups to survive and outcompete other groups. We hypothesize that any cultural traits, regardless of whether they are rooted in religious cognition or not, that directly or indirectly promote in-group solidarity in increasingly expanding and competing groups, are targets of cultural selection, meaning they are more likely to persist through time and space. Therefore, we take the view that prosocial religions are an important cause in a network of causes, but they are not a necessary, perhaps not even a sufficient cause, of large-scale cooperation (Atran & Henrich Reference Atran and Henrich2010; Norenzayan Reference Norenzayan2015). The framework outlined here is therefore compatible with other causal pathways that also promote large-scale cooperation, such as third-party monitoring and punishment institutions (Fehr & Fischbacher Reference Fehr and Fischbacher2004b; Henrich Reference Henrich2006; Henrich & Henrich Reference Henrich, Henrich, Ensminger and Henrich2014), the rule of law and effective policing (Hermann et al. 2008; Norenzayan & Gervais Reference Norenzayan and Gervais2015; Norris & Inglehart Reference Norris and Inglehart2004), modes of production, markets, and exchange (Henrich et al. Reference Henrich, Ensimger, McElreath, Barr, Barrett, Bolyanatz, Cardenas, Gurven, Gwako, Henrich, Lesorogol, Marlowe, Tracer and Ziker2010a), moral emotions harnessed and amplified by cultural traditions (Frank Reference Frank1988; Henrich et al. Reference Henrich, Boyd and Richerson2012; Keltner et al. Reference Keltner, Kogan, Piff and Saturn2014), social safety nets (Hruschka et al. Reference Hruschka, Efferson, Jiang, Falletta-Cowden, Sigurdsson, McNamara, Sands, Munira, Slingerland and Henrich2014; Norris & Inglehart Reference Norris and Inglehart2004), ethnic solidarity (Henrich & Henrich Reference Henrich and Henrich2007, Ch. 9), and so on. In fact, an interesting cluster of open questions are precisely how the religious elements we hypothesize interact with these other mechanisms, and when and how religious mechanisms are replaced with alternatives, as appears to have happened in some societies. As should be clear from the above citations, several of us have worked on these other mechanisms (as have many others we cited in the target article).

Although prosocial religious packages may compete with these alternative routes in particular cases (e.g., religious and secular sources of social safety nets), there is no a priori theoretical reason why these two routes are mutually incompatible. We are puzzled, therefore, why several commentators seem to think that they inevitably are. Krueger, for example, points out that there are sources of public morality other than prosocial religions, which of course we agree with entirely and have integrated into our framework (see sect. 1, 2.4, 7.2, and 7.3 in the target article). Similarly, Beit-Hallahmi, Galen, and Seewald et al. point out that secular ideas and institutions can also increase prosociality; that in priming experiments, reminders of secular notions such as “jury” and “judge” also increase prosociality. In fact, that was precisely the point of one of the earliest priming studies we conducted (Shariff & Norenzayan Reference Shariff and Norenzayan2007) that these commentators refer to. As McCauley points out, there are important questions about the psychological dynamics involved, but this observation can be accommodated within several theoretical frameworks, including the one presented in this target article.

Given this reasoning, readers might wonder, then: Why focus on the causal role of prosocial religions rather than other factors in the scaling up of human societies? There are two important reasons. For one, recall that one of the primary goals of our analysis is to build an account of how world religions spread and stabilized; this necessarily involves a close look at the specific role of prosocial religions in this process. Two, treatments of large-scale cooperation in economics, psychology, geography, and evolutionary biology have considered a broad range of factors, but religion is rarely ever discussed. For example, The Company of Strangers (Seabright Reference Seabright2004), a delightful book that is about large-scale cooperation, considers many possible contributing causes but not the potential role of world religions. In Guns, Germs & Steel (Diamond Reference Diamond1997a), religious beliefs and practices are rarely mentioned; the rise of priesthood classes is discussed briefly as a consequence (not a contributing cause) of population explosions and settled agriculture, which opened the door for “kleptocracy.” Conversely, many widely read books on religion tend to either fail to mention any connection with large-scale prosociality or cooperation (parochial or otherwise) or forcefully argue against such effects (Dawkins Reference Dawkins2006; Harris Reference Harris2005), though typically without engaging with the empirical literature.

R1.3. There is a multidimensional spectrum of increasingly potent supernatural punishment and interventionism, not presence versus absence of Big Gods

To our regret, our use of the term Big Gods, as a catchy and compact rhetorical heuristic, has done real damage to the comprehensibility of our argument, resulting in a series of crucial misconceptions. This is unfortunate, because the use of the term Big Gods in this target article is meant more as a rhetorical device or theoretical placeholder rather than as an exclusive term; moreover, one of the main goals of this target article was precisely to move beyond this term and present the theoretical framework in its entirety, elaborate and fill in important gaps, and in the process respond to emerging critiques. Hence, the title of our article did not mention Big Gods but rather referred to the cultural evolution of prosocial religions.

We therefore see Big Gods as the extreme end of one of the key mechanisms in our framework: the intensity of supernatural punishment and interventionism culturally prevalent in particular place and time. This conceptualization allows for (1) a spectrum of intermediate cases where gods can have some interventionist capacities and some moral concern without fully being classified “Big Gods,” and still play an important causal role in the gradual escalation of large-scale cooperation and (2) other supernatural punishment beliefs, emotions, and ensuing norms that may or may not be tied to gods, let alone to only the Big Gods, and let alone to only the great monotheisms of the Abrahamic religions. These include, among other things, moralizing gods with a more circumscribed scope of interventionism, but also notions of imminent justice and karmic beliefs (see Banerjee; White, Sousa, & Prochownik [White et al.]).

This issue is of particular concern when scholars rely on ethnographic or historical data. Roes is quite correct in pointing out that, in databases such as the Ethnographic Atlas or Standard Cross Cultural Sample (SCCS), the variable concerning high gods is often coded as a sharply binary variable (present or absent), whereas the issue of whether or not a given supernatural being counts as a high god, or a moralizing high god, is often subject to interpretation and admits to gradations and qualifications (e.g., Purzycki Reference Purzycki2013; Purzycki & McNamara Reference Purzycki, McNamara, Cruz and Nichols2016). Binary coding of this variable may be a convenient methodological tool, but it should not influence our theorizing about the cultural distribution of these beliefs in the historical and ethnographic record. Therefore, our framework does not hinge on testing for the presence versus absence of Big Gods in the ethnographic record, as Dutton & Madison imply. This also addresses Krueger's worry about how Big Gods originate in the first place. The important details about origin questions are invariably hard to pin down and require extensive, patient work. But in principle, it is not a mystery how belief in Big Gods come about: They arise from modifications of preexisting beliefs and practices that over historical time become targets of cultural evolutionary selection pressures. We think Watts et al., in their study of Austronesian societies before European contact, present such a case (see sect. R2.2).

In our view, the best methodological response to this concern about historical data is to adopt an expert-centered approach to coding variables, and below we describe just such an effort in which our team is engaged. That is, when attempting to convert “thick,” qualitative historical or ethnographic data into quantifiable data – a necessary step for large-scale quantitative analyses – the task of making this conversion should be entrusted to the historical or ethnographic experts, who are best able to weigh the multiple and subtle factors that would argue for one coding rather than another (Slingerland & Sullivan, Reference Slingerland and Sullivanin press). Moreover, it is important to capture the entire range of supernatural monitoring and punishment beliefs, including intermediate-sized or non-agentic supernatural enforcers of the social orders (see religiondatabase.org; Table R1).

Table R1. Some relevant entries and variables from the Database of Religious History (DRH; http://religiondatabase.org/)

Both Dutton & Madison and Costello focus on monotheism as the key problem in our theoretical framework. We had hoped that the use of the term “Big Gods” consistently in its plural form would have prevented this misconception. The great monotheisms are, of course, quintessential examples of prosocial religions (and raise interesting questions regarding their potential distinctiveness). Nevertheless, they are a subclass of prosocial religions, and some are arguably not true monotheisms in practice (e.g., Catholic saint worship). As we discuss in section R2.3, we fully agree with White et al. that polytheistic prosocial religions, particularly the karmic religions, are as much part of our explanatory framework as the monotheistic ones. Admittedly, we said more about the Abrahamic religions than the karmic religions, for the main reason that there is a dearth of behavioral science research on the latter, a gap our ethnographic team is beginning to rectify. We applaud White et al.'s efforts to gather actual data from practitioners of these traditions.

R1.4. Prosocial religions comprise a dynamic suite of converging traits

Building and expanding on the previous point, the intensity of supernatural punishment is a central component of the process, but certainly not the only component, that drove the coevolutionary process that is the central thrust of our argument. The Victorian title that we use above (para. 4) should leave no doubt that our framework includes an entire suite of mechanisms that likely contributed to the escalating social complexity of human societies in the last 12,000 years. Watson-Jones & Legare provide a cogent argument for the importance of rituals in building social solidarity. We agree there are a variety of ways by which rituals play a critical role, and research in this area promises to be important and complementary to the framework we described in the target article. One interesting question important for the functioning of prosocial religions is whether some rituals are able to galvanize solidarity beyond face-to-face groups, in imagined moral communities of strangers (e.g., Purzycki & Arakchaa Reference Purzycki and Arakchaa2013).

This is why we encourage doing away with fruitless semantic debates about the term “religion.” Definitions in research, to the extent that they are useful, arise from the gradual maturing of particular theories and are then operationalized for testing against the body of evidence. Definitions rarely ever precede theories, and in developing definitions from theories, pretheoretical intuitions are dropped (not imposed). Hence, contra Beit-Hallahmi, we think the scientific study of religion can proceed by first developing good theories to explain specific phenomena and then operationalizing precise variables in ways suitable for empirical inquiry. This could be done without worrying about the exact conceptual boundaries of the broad term “religion,” that is, with necessary and sufficient features. This approach also allows researchers to account for the important differences between religious traditions, as Beit-Hallahmi is calling for. As many cognitive scientists of religion before us have suggested, the only reason to use the folk term “religion” at all is for it to serve as a convenient pointer to a common set of phenomena about which we hypothesize.

R1.5. Parochial prosociality

Aside from the term “Big Gods,” which caused so much misunderstanding, the other “culprit” in our target article turned out to be the term “prosociality.” Galen, Roes, Krueger, and Soler & Lenfesty express various reservations regarding this term. And although our argument centers on parochial prosociality, McKay & Whitehouse ask whether there are prospects for world religions transcending it.

In the present framework, escalating intergroup competition is a potent driver of cultural evolution. This has important implications for the type of prosociality we would expect to proliferate. As an evolutionary strategy, parochial prosociality outcompetes both indiscriminate prosociality and a self-interested strategy in a wide range of conditions. As Bowles (Reference Bowles2008) puts it, intergroup conflict is altruism's midwife. Applying this insight to prosocial religions, we think that the boundaries of the parochial altruism that this midwifery mediates stretch outward toward coreligionists, but typically not beyond (with some important and interesting exceptions, discussed at the end of this section). Religious cooperation and religious conflict are not incompatible – they are the two sides of the same coin.

It appears that several commentators were unwilling to label a set of behaviors “prosocial” unless they refer to indiscriminate prosociality toward anyone. This led to double confusion in the commentary by Galen, when he wrote: “Conflated terms such as ‘coreligionists including strangers’ and ‘anonymous individual from the community’ are oxymoronic from the standpoint of distinguishing a complete stranger – possibly an out-group member – from someone who shares some group affiliation with the participant” (para. 3). Of course, the question we are asking, precisely, is how some religious communities came to be world religions, where coreligionists are no longer confined to a face-to-face community of regular interactants. Instead, they expanded into an imagined moral community comprising strangers – that is, people one has never met or heard of – who, moreover, often transcend ethnic, linguistic, and geographic boundaries, such as in Islam, the Mormon Church, Mahayana Buddhism, Bah'ai, and so forth. We need an explanation for why, for example, Mexican Mormons would be willing to cooperate with, or even sacrifice for, Nigerian Mormons on the other side of the globe. Yet, such expanded cooperation is nevertheless bounded, “in-groupish,” and tethered to tribal instincts. With real or imagined intergroup conflict and rivalries, it often turns toxic toward those who fall outside of the moral boundaries of the group, fueling indifference, prejudice, and violence toward religious out-groups and toward nonbelievers (for reviews, see Atran & Ginges 2012; Gervais & Norenzayan Reference Gervais, Norenzayan, Clarke, Powell and Savulescu2013; Norenzayan Reference Norenzayan2013).

We speculate that another related source of miscommunication is the attribution of inherent virtue to “prosociality” or “cooperation,” a move that we did not make. Prosociality and cooperation in and of themselves are neither good nor bad. The same cooperative tendencies can be directed toward building roads or feeding the homeless, or, at the other extreme, toward the organized plundering of another group or waging war. The moral valence of such acts often turns out to be in the eye of the beholder. This is also true in the case of some bacteria, which cooperate and coordinate in “quorum sensing” to mount infectious attacks on their hosts (e.g., Winans & Bassler Reference Winans and Bassler2008) – good for the bacteria, bad for the host organism (an insight that is now being harnessed to create second-generation antibiotics). To use a more extreme example closer to our species, consider suicide attacks. The attacker sacrifices his own life and is seen as a virtuous martyr by his own community and, simultaneously, as a violent mass murderer by the group receiving the brunt of the attack (Atran Reference Atran and Roberts2011).

Does this mean that religious prosociality can never transcend the boundaries of the religious ingroup, as McKay & Whitehouse ask? We agree it can, under some conditions that may turn out to be extremely interesting and deserving of deeper examination. One hypothesis is that some prosocial religions may encourage extended prosociality when the targets of prosociality are also potential converts, which enhances the community's cultural survival and expansion at the expense of rival groups that do not practice such extended prosociality. For example, Stark (Reference Stark1996) argues that one of the key reasons Christianity spread so rapidly in the early Roman Empire, aside from high fertility rates, was that Christian altruism toward pagans led to steady conversions, particularly in times of mass epidemics and accompanying population decline. Therefore, the “spillover effects” that McKay & Whitehouse's commentary explores are not confined to the lab but exist in the real world as well.

The effects of ritual participation on prosocial behavior might illustrate such spillover effects. Hobson & Inzlicht discuss recent data from their own work showing that rituals increase out-group hostility. This is an interesting finding, and not incompatible with the hypothesis that the prosocial effects of ritual are parochial. However, McKay & Whitehouse cite several papers showing that rituals, or aspects of rituals such as synchrony, increase prosociality even toward out-groups. We point out two additional studies that show that the effects of some specifically religious ritual participation may indeed extend beyond in-group boundaries. One remarkable field study by Clingingsmith et al. (Reference Clingingsmith, Khwaja and Kremer2009) compared a group of Pakistani Muslims who participated in the Hajj, the annual Muslim pilgrimage to Mecca, to a matched group of participants who were on a waiting list but did not win the lottery to go to the Hajj (hence, the study's design, by utilizing random assignment, largely eliminates selection explanations). They found greater intergroup tolerance among Hajj participants toward Christians and other out-groups. In another study discussed in our target article, Xygalatas et al. (Reference Xygalatas, Mitkidis, Fischer, Reddish, Skewes, Geertz, Roepstorff and Bulbulia2013) found that participation in the annual Hindu Kavadi ritual increased identification with the broader Mauritian culture by Hindu Mauritians. We think this issue is far from settled, and there likely are important, unidentified mechanisms and boundary conditions that would help us answer this question not just regarding ritual but also regarding other religious elements.

R2.1. Prosocial religions predating or falling outside of the Abrahamic traditions

In critiquing our use of cross-cultural evidence, Watts et al. are correct in observing that the Ethnographic Atlas (EA) and Standard Cross Cultural Sample (SCCS) “high god” cultures are overwhelmingly either Abrahamic religions or plausibly influenced by Abrahamic religions. This is, however, precisely why we provide alternative, non-Abrahamic evidence, such as that from early Mesopotamia, Egypt, and China. Although the lack of solid written or archeological evidence complicates efforts to ascertain whether or not moralistic gods played a role in the rise of the first large-scale agricultural societies approximately 9,000 years ago, several historical trends pointed out in our target article are worth reiterating. After we lay these out, we return to discuss concerns regarding Watts et al.'s own treatment of the independence problem.

Monumental religious architecture – a key feature of prosocial religions according to our framework – appears at least coterminously with the rise of large-scale civilizations in both the Old and New Worlds, and – in the case of sites such as Çatalhöyük and particularly Göbekli Tepe – arguably precedes them. Boyer & Baumard maintain that “temples or massive offerings to the gods do not show that the populace was committed to these religious symbols, but more prosaically that coercive authorities could rely on high taxes and large amounts of forced labor” (para. 4). However, this raises the question: Why did archaic, socially complex societies devote a significant proportion of their wealth to apparently nonutilitarian structures? If monumental architecture were not serving the social functions we hypothesize, it is difficult to see how cultures that instead spent their wealth on increased agricultural efficiency, irrigation, defensive walls, superior weapons, or better roads, all else being equal, would not outcompete cultures that buried a significant proportion of their wealth in the ground with dead people.

Moreover, as we document in the target article, once the earliest written records appear, which help determine more precisely the content of ancient religious beliefs, we find powerful gods concerned with public prosociality playing a central role in ancient societies as diverse as ancient China, Egypt, and Mesopotamia. We completely agree with Watts et al. that “future work should (1) systematically catalogue religious variation both within and outside the Abrahamic traditions” and “(3) identify at what points in the evolution of big societies each kind of supernatural agent plays a functional role” (para. 5). To this end, we have been devoting considerable resources to developing the Database of Religious History (DRH), intended to become a publicly accessible, comprehensive database of historical cultures around the globe, with an initial focus on religious variables but plans to expand as well to political, economic, technological, military, and ecological data. The DRH is only in the very early, pilot stages of data collection as we hone its usability and functionality, but even the relatively limited data we have collected to date demonstrate that, as coded by independent experts, morally concerned gods with the power to punish and reward are certainly present outside the Abrahamic traditions and before any hypothesized “Axial Age” (see Table R1).

We agree with Watts et al. on the challenges of dealing with the potential historical relationships between societies in the SCCS and the Ethnographic Atlas, and we welcome new methods of inquiry to complement existing analyses. Nevertheless, we caution against rushing to embrace analytical techniques imported from genetic evolution – used to reconstruct species phylogenies – to cultural evolution. Cultural evolution is in some crucial respects unlike genetic evolution (see discussion in sect. R3.1). Species, for example, are not subject to intergroup competition that creates massive and directed horizontal transmission of only some traits. Therefore, we think the first step should be to benchmark phylogenetic techniques to cultural history using known historical cases. To our knowledge, and to the knowledge of two experts on cultural phylogenies we consulted, Mark Collard and Quentin Atkinson (personal communication), that has yet to be done.

R2.2. Compatibility of the present theoretical framework with several cultural evolutionary scenarios

Huebner & Sarkissian, Demetriou et al., and Watts et al. present versions of the claim that the wide-ranging evidence we reviewed in the target article is compatible with several possible historical-cultural pathways to large-scale cooperation. We agree. In section 3.2.3, paragraph 4, of the target article, we say that our framework “allows for multiple causal pathways, including the possibility that in some societies prosocial religions played a minor or no role, or that their role emerged late in the process.” We also cautioned against the fallacy of assuming there is a single, overarching scenario that is true everywhere and at all times:

It might be that part of the confusion arises from a lack of familiarity with modern cultural evolutionary theory, or falsely attributing unilinear evolutionism of the early twentieth century to our framework, or both.

Huebner & Sarkissian raise some important issues about how some societies might get large-scale cooperation off the ground through nonreligious mechanisms, such as mundane monitoring by other individuals or the ritual-enhanced internalization of values. Focusing on China, Sarkissian (Reference Sarkissian2015) has previously noted that Chinese religious worldviews posited a relatively impersonalized cosmic order in comparison with, for example, the Abrahamic faiths. Although Shang Di or tian (“Heaven”) of the Shang or Western Zhou was a clearly full-blooded, anthropomorphic Big God, capable of being angered and sending down “mandates” or orders, it should be noted that this Big God did not communicate directly with humans but only indirectly through omens or signs of pleasure (e.g., battle success or good harvests) or displeasure (e.g., natural disasters, military defeats, and popular uprisings). Huebner & Sarkissian further observe that the focus of early Chinese religions tended to be upon individual self-monitoring, bolstered by social monitoring and reliance on such mechanisms as shame. In addition, from at least the time of the state of Qin in the fourth century BCE (the state that eventually went on to defeat the other “Warring States” and found the Qin Dynasty in 221 BCE), the early Chinese developed extremely elaborate and far-reaching legal and bureaucratic institutions that may, Sarkissian (Reference Sarkissian2015) has argued, have provided an essentially “secular” alternative to religious-based cooperation.

We point out that the early Chinese high god tian continued to be the object of supernatural reverence throughout imperial Chinese history, receiving offerings and sending down rewards and punishments. In addition, the ability of individuals to self-monitor and internalize norms requires a panoply of rituals and costly displays that are at the center of our proposed cultural complex. Finally, even in the highly bureaucratic and legalistic governments of the Qin and later dynasties, it is not at all clear that supernatural agency is entirely left behind. In important state rituals and sacrificial cults, supernatural endorsement still plays a major role, and belief in the existence and power of supernatural agents remains an important component of individual belief at all levels of society. Though clearly the developmental sequences vary, the current evidence on early China is consistent with the influence of intergroup competition on “scaling up,” and for the role of supernatural beliefs and rituals in that process.

In their study of Austronesian societies discussed in their commentary, Watts et al. (see Watts et al. Reference Watts, Greenhill, Atkinson, Currie, Bulbulia and Gray2015) found that “broad supernatural punishment,” but not “high gods” per se, played a role in social complexity. This is not only compatible with our framework, but it is also supportive of it. Because there has been limited social complexity in Austronesia (relative to, in particular, Eurasia), there is no reason to expect widespread prevalence of Big Gods. As we noted, these gods represent one corner of a multidimensional spectrum, and certainly no relationship would be expected between the two variables, and none was found. In a cultural region where small-scale societies predominate, one would instead expect a gradual ramping up of moral concern and punishment of norm-violations tied to smaller gods and to other cultural beliefs and practices. And indeed that is what Watts et al. found.

R2.3. The cultural evolution of karma

The issue of impersonal supernatural order brings us to two commentaries, by White et al. and Banerjee. White et al. take us to task for not exploring in sufficient depth the role of karmic religions in human cultural history despite the fact that we do point out the importance of karmic religions and cite some recent ethnographic literature on their psychological impact. We also acknowledge that beliefs in impersonal supernatural orders might do much of the same work as anthropomorphic deities. Our research teams are currently pursuing empirical studies of karmic beliefs in several field sites.

However, we agree with White et al.'s caution to researchers to avoid the fallacy of attributing theological correctness (Barrett Reference Barrett2004) to adherents of karmic – or any other – religion. Although the official theology of some forms of Buddhism, for example, might involve a completely impersonal karmic order, once we leave the scholar's or monk's study, anthropomorphic beings often reappear with a vengeance These include not only punishing and rewarding deities, but also fearsome demons portrayed in popular religious art as the personifications of the karmic order, gripping the wheel of samsara in their teeth and talons (e.g., Lopez Reference Lopez2005). In fact, most Buddhist traditions as actually practiced focus on the veneration of particular deities, who are typically portrayed in anthropomorphic form, and Buddha treated as a Big God (e.g., Purzycki Reference Purzycki2013). Some of the forms of Buddhism that are most widespread in East Asia center on personal devotion to Buddhas capable of rewarding one with rebirth in an eternal Pure Land (Payne & Tanaka Reference Payne and Tanaka2004). As with the case of tian in early China – or, indeed, any religious tradition – it is probably the case that within the same tradition we can find varying degrees of personification of the cosmic order.

Banerjee (see also Banerjee & Bloom Reference Banerjee and Bloom2014; Reference Banerjee and Bloom2015) opens an important line of inquiry by suggesting that teleology, or the intuition that objects and events occur for a purpose (Kelemen Reference Kelemen2004), is a likely cognitive foundation for karmic beliefs that then become targets of cultural evolution. White et al. point to theories that suggest that intuitions of fairness and immanent justice are the basis for this type of thinking (see Baumard & Chevallier Reference Baumard and Chevallier2012). We agree that these ideas of retribution and reward are prominent around the world, but much evidence shows that “fairness” is highly variable across diverse societies (e.g., Henrich et al. Reference Henrich, Ensimger, McElreath, Barr, Barrett, Bolyanatz, Cardenas, Gurven, Gwako, Henrich, Lesorogol, Marlowe, Tracer and Ziker2010a). The egalitarian ideal of fairness referred to in this literature has been tested in the West primarily – and in the industrialized world exclusively (see Furnham Reference Furnham2003).

Regardless of what the underlying cognitive processes may be, explicit karma-like beliefs seem to be widespread (see Obeyesekere Reference Obeyesekere2002) – especially if we broaden our outlook to include related beliefs like witchcraft and mana (seen across Oceania). We echo Banerjee's and White et al.'s emphasis on this topic. We think a good research direction for the future is to investigate how these beliefs are related to social complexity, social inequality, moral psychology, cooperation, and conflict.

R2.4. Back to “Axial” and “pre-Axial” religions

We commend Boyer & Baumard for raising the WEIRD challenge! Cultural and historical diversity is the very focus of the comparative approach that we have used. Our empirical evidence includes ethnographic research in small-scale societies, cross-cultural comparisons more broadly, and historical evidence from a variety of cultures both within and outside of modern WEIRD contexts. Having done this work, we diverge from Boyer & Baumard's interpretation of the historical evidence.

Boyer & Baumard dismiss the importance of the Mesopotamia, Egypt, Mesoamerica, and Early China data as instances of “a common but misleading confusion between the religions of large-scale archaic societies … and ‘Axial Age’ religions with moralizing and spiritual doctrines that appeared only in a small subset of these societies” (para. 2). In our opinion, the confusion actually lies in (1) a failure to recognize the continuities between these early religious traditions and the later “Axial Age” religions, (2) a lack of recognition of the actual diversity of the posited Axial Age religions, and (3) the relevance of moralistic religious traditions to our central hypothesis.

The idea of an Axial Age was developed against the background of the metaphysical view of the originator of the term, Karl Jaspers (Jaspers Reference Jaspers and Bullock1953), who saw it as the product of a maturing world Spirit (Weltgeist). Once this metaphysical notion is put aside, the concept has little theoretical value or empirical support. Empirically, the counterexamples for a supposedly conceptually unified “Age” are many. For example, during the Warring States period (fifth through third centuries BCE) Chinese religious systems such as Confucianism or Daoism – cited by Baumard et al. (Reference Baumard, Hyafil, Morris and Boyer2015) as paradigmatic Axial Age religions – in fact lacked major supposed hallmarks of the Axial Age, such as an otherworldly orientation or focus on the afterlife. Confucius, for example, famously dismissed a question about the fate of human beings after death (Slingerland Reference Slingerland2003, p. 115). On the flip side, as described in the target article, there is also ample historical evidence that several pre-Axial Age religions were supportive of public morality. New Kingdom Egypt, which well preceded the Axial Age, had many of the characteristics of an Axial Age religion (Assmann Reference Assmann2003). Our particular interpretations of the historical record aside, independent historians identify various “Axial Age” qualities as being present in religious traditions from pre-Axial Age ancient Egypt, China, the non-Christian Mediterranean, and the ancient Near East (see Table 1).

Finally, we acknowledge the distinction between a concept of “spiritualized” morality in something like a Kantian sense (Kant 1785/1964) – moral behavior freely chosen and driven solely by intrinsic motivation – and moral behavior resulting from other motives. We agree with Boyer & Baumard that, for example, a Confucian scholar enjoined to follow the Way for its intrinsic moral superiority, in the face of social disapproval, poverty, even death, looks somewhat different from a second millennium BCE Zhou Dynasty king anxious about following moral and ritual structures lest he lose the “Mandate of Heaven,” and with it his wealth and his head. However, it is unclear why Boyer & Baumard think that our hypothesis applies to the former only. Moreover, unless one is committed to the belief that this shift from prudential morality to “true” moral commitment is the result of a mysterious Weltgeist, the most plausible way to understand the change – as well as the one most in accord with the actual historical evidence – is as the result of a gradual process occurring at different times and in different ways around the world.

The fundamental flaw in the Axial Age hypothesis is its failure to recognize the gradual nature of cultural change. This is why it is unable to account for religions possessing “Axial” qualities that arise outside of the proposed time span, such as Western Zhou and Babylon (1,000 years before), or Islam (800 years after), or elements of the supposed Axial Age monolith appearing independently or sporadically in the historical record. Atran amplifies these points and provides additional evidence of pre-Axial moral religions, including evidence for their role in regulating long-distance trade. The gradual nature of this coevolutionary process that intertwines religious elements with social complexity is also illustrated by Watts et al.'s study in Austronesian societies.

R3.1. Memetics

Blackmore's commentary focuses on memetics as an important explanatory framework in the study of religion. In our view, all of the useful insights from memetics have long been a part of the broader field of cultural evolution (Boyd & Richerson Reference Boyd and Richerson1985; Cavalli-Sforza & Feldman Reference Cavalli-Sforza and Feldman1981). To see this, realize that memetics builds on the notion of replicators (Dawkins Reference Dawkins1976). Cultural replicators, termed memes, are discrete, gene-like entities that can evolve and spread to the degree that they can replicate with high fidelity and fecundity. Focus is given to the meme's eye view and on the ability of different memes to proliferate and colonize other minds (Blackmore Reference Blackmore1999). This view can plausibly describe some limited cases of cultural evolution. However, much of cultural evolution involves (1) continuous (nondiscrete) traits transmitted with substantial errors or systematic transformations (no replication), (2) a constellation of powerful psychological “attractors” that shape traits (Sperber Reference Sperber1996), (3) social learning abilities that combine information gleaned from multiple other people, and (4) the influences of individual experience or trial-and-error learning (Claidiere & Sperber Reference Claidiere and Sperber2007; Henrich & Boyd Reference Henrich and Boyd2002; Henrich et al. Reference Henrich, Boyd and Richerson2008).

Suppose, for example, that a learner observes (with error) and averages the arrow lengths of the three best hunters in his village in deciding how long to make his arrows. He might observe lengths of 15, 16, and 20 inches, and then make his arrow 17.5 inches. Where is the replication or the replicator? None of the models' arrows is 17.5 inches long, so there is actually no replication at all. Yet, a simple cultural evolutionary model, rooted empirically in what is known about human social learning (Morgan & Laland Reference Morgan and Laland2012; Morgan et al. Reference Morgan, Rendell, Ehn, Hoppitt and Laland2012), can show that if many individuals are doing this (and making many errors as well as adjustments using trial-and-error learning) over generations, the average arrow length will home in on the optimal arrow length. Hence, modern cultural evolutionary theory allows us to model and study a much broader diversity of psychological processes, including those involving empirically well-established learning mechanisms and cognitive biases. Of course, within this framework, memes can be modeled and studied, and it is even possible to show when and why memes can sometimes provide a suitable approximation of otherwise more complex phenomena (Henrich & Boyd Reference Henrich and Boyd2002).

R3.2. Cultural group selection

Another element of our cultural evolutionary framework that sowed confusion among some commentators was our reliance on cultural group selection as one important component of the explanatory framework (Beit-Hallahmi, Blackmore, Krueger, Viciana et al.). We argue that intergroup competition may have shaped cultural evolution, favoring the spread of particular sets of supernatural beliefs and ritual practices. This use of “cultural group selection” often gives rise to two misunderstandings. One inappropriately applies critiques relevant for genetic group selection to cultural group selection (Henrich Reference Henrich2004). The other confuses debates about the importance of intergroup competition with a rather technical debate about which mathematical and conceptual accounting system is preferable when building evolutionary models (Fletcher & Doebeli Reference Fletcher and Doebeli2009; McElreath & Boyd Reference McElreath and Boyd2007).

To the first confusion, many commentators may have been intellectually weaned on skepticism toward the relevance of group selection for genetic evolution (e.g., see Blackmore). One important source of this skepticism is the tendency of migration among groups to deplete the genetic variation between groups through mixing. Because variation fuels the engine of selection, as the variation among groups declines, so too will the strength of selection between groups. Meanwhile, mixing between groups sustains within-group variation, and hence the within-group components of selection often come to dominate (Hamilton Reference Hamilton and Fox1975; Wilson & Wilson Reference Wilson and Wilson2007).

However, with regard to cultural evolution, a quarter century of detailed theoretical work has revealed that cultural evolution is much less susceptible to these effects because of the nature of cultural learning and the speed of cultural evolution (Boyd & Richerson 1990; Reference Boyd and Richerson2002; Guzman et al. Reference Guzman, Rodriguez-Sickert and Rowthorn2007). There are actually many reasons for this crucial distinction, but one key set of reasons involves the tendency of rapid, error-prone, and biased cultural evolution to give rise to many different self-reinforcing stable equilibria (effectively, different social norms). In these equilibria, immigrants from other groups adopt local norms purely based on within-group selection pressures – so they or their children culturally learn how to behave “properly” in the new cultural environment. Unlike in genetic evolution, where genes cannot readily change, between-group differences can be sustained culturally, and thus the power of the group selection component of the overall evolutionary process is preserved. Empirical work now confirms that, in fact, between-group cultural variation is much greater than between-group genetic variation (Bell et al. Reference Bell, Richerson and McElreath2009; Richerson et al., Reference Richerson, Hillis, Newson, Baldini, Frost, Bell, Demps, Mathew, Newton, Narr, Ross, Smaldino, Waring and Zeffermanin press). Hence, one must resist the temptation to simply take lessons (or preconceived notions) from genetic evolution and apply them haphazardly to cultural evolution without systematically considering the differences in the two inheritance systems (Beit-Hallahmi, Blackmore, Krueger, Viciana et al.). Of course, as with inclusive fitness formulations (Nowak et al. Reference Nowak, Tarnita and Wilson2010), important and legitimate controversies persist, so we point readers to two recent target articles in this journal and their respective commentaries on cultural group selection (Richerson et al. Reference Richerson, Hillis, Newson, Baldini, Frost, Bell, Demps, Mathew, Newton, Narr, Ross, Smaldino, Waring and Zeffermanin press; Smaldino Reference Smaldino2014.

A lack of appreciation for the important difference between genetic and cultural evolution also leads to the common, yet confused, belief that cultural group selection models require within-group homogeneity. Though this is true of many simple models, it is not a general property. For example, many cultural evolutionary models give rise to mixed equilibria, with considerable stable within-group heterogeneity, that generate higher group-level payoffs than other equilibria (e.g., Henrich & Boyd Reference Henrich and Boyd2008). This can happen in a situation in which different individuals culturally acquire different skills and can then combine these skills to create higher-order group-specific traits (e.g., specialization and trade). Notably, in these situations, what emerge are truly group-level traits (the mixed equilibria) that are not reducible to individual-level traits (Miller & Page Reference Miller and Page2007; Smaldino Reference Smaldino2014). This logic can apply, for example, to rituals: Some participants may know only one part of a ritual, whereas another subgroup knows only another part of that ritual – so the group is culturally heterogeneous in terms of its ritual knowledge internally, but relatively homogenous compared with other groups that have different practices. But, working together, all of the participants can jointly produce emotionally more powerful rituals that strongly bind groups together, which can in turn produce measurable somatic fitness-effects (for further discussion, see Purzycki et al. Reference Purzycki, Haque, Sosis, Watts and Turner2014).

To the second issue, there is apparent confusion about the question of whether intergroup competition might be important in shaping cultural evolution, with an argument among theorists about which accounting systems or conceptual interpretation is the best way to handle this intergroup competition (Lehmann et al. Reference Lehmann, Keller, West and Roze2007). Some argue that researchers should always narrowly restrict themselves to purely inclusive fitness approaches and interpretations, whereas others argue for the importance of a variety of approaches and take seriously the value of multilevel selection models, at least for some types of problems. That is, one can typically model intergroup competition using any of at least three different conceptual and mathematical frameworks: (1) inclusive, (2) individual, or (3) multilevel accounting. The debate here is not about whether intergroup competition matters; it is about how to think about intergroup competition. Relying exclusively on purely individual-level accounting or inclusive fitness accounting can make one miss the fact that intergroup competition is driving important outcomes.

The failure to recognize the centrality of intergroup competition is well illustrated by Viciana et al.'s commentary and simulation model. Viciana et al. develop a simulation model in which the evolution of fairness (the sharing of resources) is entirely driven by competition among groups, where groups compete for membership – that is, Viciana et al. developed a classic cultural group selection model (see Boyd & Richerson Reference Boyd and Richerson2009) without realizing it! Here is their setup: The simulation begins with a large diversity of groups, and these groups are characterized by group-level (not individual-level) traits that prescribe entry costs and some degree of sharing (one possibility is no sharing at all). Groups then compete for membership as individuals make calculations about which groups will give them the highest payoff individually. Analyzing this setup within a multilevel selection framework immediately reveals that the only driving force is intergroup competition: Behaviorally (phenotypically) there is zero variation within groups in individuals' willingness to share or to pay entry costs. Instead, all of the behavioral variation in paying access costs and in sharing is between groups. Migrants into new groups must immediately adopt the traits of the group they migrate into. This means that the selective forces within groups are not doing anything, and all of the evolutionary change is driven by different group-level traits. Viewed from the point of view of cultural evolutionary theory, this is very similar to models in which social norms emerge as self-reinforcing stable equilibria (creating group-level traits), and then groups at different stable equilibria (with different norms) compete (Boyd & Richerson 1990; Reference Boyd and Richerson2002; Reference Boyd and Richerson2009; Henrich Reference Henrich2004). We commend Viciana et al. for developing an interesting model of how fairness norms can evolve through intergroup competition (though a bit more simulation work is needed).

Nevertheless, it is not wrong to conceptualize this model using either individual fitness accounting (averaging across groups) or inclusive fitness. But, in this case, those conceptualizations conceal the fundamental dynamics driving the evolution of fairness in Viciana et al.'s model – the competition to attract and retain migrants. There is certainly a sense in which thinking of this as “partner choice” is relevant, mostly in order to compare it with other more quintessential partner choice models (e.g., Hruschka & Henrich Reference Hruschka and Henrich2006). But, individuals in Viciana et al.'s models are actually not evaluating and selecting specific other individuals as potential partners – so a “partner choice” interpretation seems a bit strained. In their simulation, individuals are actually picking groups according to which group gives them the best payoffs, and different groups have different institutional traits. Eventually, groups not selected by anyone go extinct forever. If one turns off the intergroup competition in their model, fairness no longer evolves at all, ever. Hence, intergroup competition gradually selects the particular combination of access costs and sharing rules that best harness and exploit the fitness-maximizing psychology (or “receiver psychology” as in Soler & Lenfesty) of the simulation's agents.

Several commentaries focused solely on within-group dynamics. Soler & Lenfesty, for example, seem to think that if priests or elites figure out how to use religious beliefs to scale up cooperation, then intergroup competition is ruled out or rendered unnecessary. Of course, cultural group selection models include both within-group dynamics (elites are out for personal gains within their group) and between-group dynamics (more cooperative groups win wars or attract more migrants). Intergroup competition will, indeed, favor the cultural evolution of better and better cultural technologies for exploiting “receiver psychology” and use elites and priests as command and control mechanisms. Elites may and often do operate to push things in directions that narrowly benefit themselves. However, if the elites push things too far as they highjack various cultural technologies, and if intergroup competition is intense, they will, in the long run, get crushed by surrounding groups with better institutional forms that more effectively restrained those elites, and galvanize cooperation for success in intergroup competition. Turchin (Reference Turchin2005; Reference Turchin2009; Reference Turchin2011), in discussing and modeling these kinds of dynamic historical processes, has shown how we can observe the push and pull of both forces.

Much ethnographic work in the South Pacific illustrates how cultural beliefs actually inhibit the ability of elites to overly exploit receiver psychology for their own benefits. In Fiji, anyone can cultivate mana (a kind of supernatural power or effectiveness), but improper behavior can turn mana against careless practitioners. Having mana gives a person greater social standing and is used to explain the potency of both prosocial healing in traditional medicine and antisocial damage from sorcery. Mana abused for antisocial ends poses a danger of turning against the abuser, producing illness or misfortune. In the case of Fijian villages, one mitigates the danger of mana turning against the practitioner by carefully observing traditional village values and norms (Katz Reference Katz1999). Sau, a similar power or potency that specifically legitimates Fijian chiefs' hereditary authority, depends on a chief's observation of his obligations to the vanua (land and people). If a chief fails in these duties, he may lose sau and can be ousted from power, thus providing feedback between leader and follower that does not merely capitalize on receiver psychology (McNamara Reference McNamara2014). Overall, detailed ethnographic studies do not support a simple “exploitation” or “kleptocracy” story for religion.

R4.1. Good versus mean gods

Johnson & Cohen note that societies where order and prosperity are high may favor the spread of beliefs in benevolent supernatural agents in addition to the authoritarian gods implied in our discussion of supernatural punishment. Research has long recognized that there are psychological benefits associated with the sunnier, more benevolent side of religions. WEIRD people with more benevolent God beliefs have higher self-esteem and coping (Francis et al. Reference Francis, Gibson and Robbins2001; Ironson et al. Reference Ironson, Stuetzle, Ironson, Balbin, Kremer, George and Fletcher2011). Correlational and experimental research indicates that belief in Heaven brings with it notable well-being benefits compared with belief in Hell (Shariff & Aknin Reference Shariff and Aknin2014). And, along with their colleagues, Johnson & Cohen have begun to empirically show that priming a benevolent, rather than authoritarian, God encourages forgiveness of a transgressor and the willingness to aid out-groups, at least for some religious groups (Johnson et al. Reference Johnson, Li, Cohen and Okun2013).

Johnson & Cohen's commentary raises an important question about the societal conditions under which different balances of supernatural benevolence and punishment would emerge. In a cultural evolutionary framework such as ours, we expect that supernatural agents should tend toward punishment and authoritarianism when existential insecurity is rampant, secular institutions are weak, and real or perceived intergroup threats are high. On the other hand, when societies have reduced existential insecurity, secular alternatives to prosocial religions are firmly in place and keep the cooperative wheels turning, and intergroup competition is largely a matter of attracting converts, the balance would tilt toward supernatural benevolence. This reasoning produces specific hypotheses that can be tested against the historical and ethnographic records. For example, we would expect that over time, increases in existential security and the development of reliable rule of law in a society should be followed by a greater shift toward ideas of supernatural benevolence and less concern with supernatural punishment.

This brings us to explanations of religions as carriers of mechanisms for uncertainty reduction, favored by Dutton & Madison and Brazil & Farias. Our framework is compatible with this perspective. However, we also point out that some supernatural beliefs in many societies increase uncertainty and fear. In many cases, witchcraft beliefs promote anxieties and paranoia that others are out there, working to do one harm, often out of jealousy. What kinds of cultural evolutionary pressures explain why such belief systems are so stable, and why do they sometimes change?

Prosocial religions offer means to reduce existential uncertainties that are endemic in human life, particularly in places and times of rampant existential threats such as poverty, short life spans, and natural disasters. Some of us have done research that shows effects consistent with this claim (Atran & Norenzayan Reference Atran and Norenzayan2004; Laurin et al. Reference Laurin, Shariff, Henrich and Kay2012b; Norenzayan & Hansen Reference Norenzayan and Hansen2006; Norenzayan et al. Reference Norenzayan, Dar-Nimrod, Hansen and Proulx2009). However, there are interesting and unexplored questions in this area ripe for research. Are ideas of supernatural punishment and benevolence equally good at soothing existential anxieties? How are supernatural agency and moral concern related (Purzycki Reference Purzycki2013)? Moreover, both historically and cross-culturally, strong rule of law and existential security go hand in hand and jointly contribute to what is called secularization (Norris & Inglehart Reference Norris and Inglehart2004). Yet, why are these conditions of secularization producing a different kind of existential threat – loss of meaning in life, with palpable consequences, such as elevated rates of depression and suicide in more secularized societies (Oishi & Diener Reference Oishi and Diener2014)?

R4.2. Gods versus Govs

McCauley raises interesting questions regarding our claim of the interchangeability of God and Gov (to borrow the catchy term by Huebner & Sarkissian, who raise similar questions). In past work we found that, although both religious priming and secular priming had comparable effects on prosocial behavior, religious primes more reliably influenced believers, whereas the secular primes influenced everyone (Shariff & Norenzayan Reference Shariff and Norenzayan2007). We have also extended the theoretical implications of this idea to show that secular and supernatural sources of prosociality are at least partly interchangeable: Across nations, the rule of law (Norenzayan & Gervais Reference Norenzayan and Gervais2015) and experimentally induced secular primes (Gervais & Norenzayan 2012) reduce believers' distrust of atheists, presumably because they undermine the folk intuition that religion is necessary for morality. Kay et al. (Reference Kay, Whitson, Gaucher and Galinsky2009) have found similar effects of interchangeability of God and Gov when it comes to threats to personal control (also see Laurin et al. Reference Laurin, Shariff, Henrich and Kay2012b).

Does this mean that strong governments predictably spell the end of religions? And how irreversible is this process? On the one hand, there is strong evidence of secularization in response to strong rule of law and government effectiveness, along with elevated existential security (Norris & Inglehart Reference Norris and Inglehart2004; Solt et al. Reference Solt, Habel and Grant2011). On the other hand, this process may not occur everywhere, it may be more fragile than previously thought, and it could be undone quickly. McCauley notes Sibley and Bulbulia's (Reference Sibley and Bulbulia2012) study, which found a spike of religiosity among New Zealanders directly affected by the Christchurch earthquake, in a country that is otherwise steadily secularizing (see also Bentzen Reference Bentzen2013, for extensive evidence of greater religiosity in places prone to natural disasters). As McCauley puts it, “Reversals of the shrinking religiosity characteristic of secularized populations may be no more than a natural disaster away” (para. 8). As the world confronts the possibility of prolonged natural disasters tied to climate change, this finding by Sibley and Bulbulia is of especial importance. We think these are excellent questions that demand more attention.

R4.3. Emotion, motivation, cognition

Ejova proposes a number of intriguing hypotheses regarding how extravagant religious displays may act in part by producing feelings of awe, which in turn may increase prosocial tendencies. The potential role of awe, as well as the other moral emotions such as compassion, empathy, guilt, pride, anger, and shame in prosocial religions, are an unexplored area that deserves greater attention. Another important area that is ripe for research is the precise mechanisms of how and when religious beliefs and behaviors are implicated in self-control. Hobson & Inzlicht express skepticism at this link, but we need not choose between two claims, that either religious beliefs and practices enhance self-control, or they don't. Laurin et al. (Reference Laurin, Kay and Fitzsimons2012a), for example, find a complex pattern, showing that thoughts of a controlling God increase temptation resistance but decrease goal pursuit. Clearly, we want to see more research to recognize and tease apart these complexities. Following up on McCullough & Willoughby's (Reference McCullough and Willoughby2009) seminal review and analysis of the religion–self-control link, Reynolds & Baumeister provide a thoughtful blueprint and a set of fruitful hypotheses that can guide future research.

Similarly, Lindeman & Svedholm-Häkkinen caution readers against over-interpreting the findings regarding cognitive processes that underlie religious beliefs. We agree, insofar as any new literature that is in its infancy should be treated as suggestive and provisional. We offer two points to add to Lindeman & Svedholm-Häkkinen's analysis. One, we note that religious beliefs are complex and multidetermined. Cognitive processes are one key ingredient, but they are one of several causal pathways that also include motivational and cultural learning mechanisms (Norenzayan Reference Norenzayan2013). Therefore, we should not expect overwhelming effect sizes from cognitive variables alone on religious beliefs. Two, we note that despite the relatively small literature, a diversity of methods, including neuroimaging, individual difference analyses, and experimental approaches converge in pointing to the same cognitive mechanisms such as mentalizing, mind-body dualism, and teleological intuitions (for a review, see Willard & Norenzayan Reference Willard and Norenzayan2013).

R4.4. CREDs and developmental psychology

Several commentaries (Boyer & Baumard; Corriveau & Chen; Huebner & Sarkissian; Viciana et al.; Watson-Jones & Legare) attended to our discussion of Credibility Enhancing Displays (CREDs). Here we address some of the key issues.

The application of culture-gene coevolutionary thinking to cultural transmission provides a systematic research program for studying cultural learning in both children and adults. Unlike most work in psychology, the approach begins by using evolutionary theory, often rooted in formal mathematical models, to derive predictions about the when, where, and from whom of cultural learning (Boyd & Richerson Reference Boyd and Richerson1985; Henrich & Gil-White Reference Henrich and Gil-White2001; Laland Reference Laland2004; Nakahashi et al. Reference Nakahashi, Wakano and Henrich2012). For example, relevant to the comments by Corriveau & Chen, this body of theoretical work predicted many of the later empirical findings of developmental psychologists on the effects of cultural learning cues such as prestige and attention (Chudek et al. Reference Chudek, Heller, Birch and Henrich2012), dialect (Kinzler et al. Reference Kinzler, Corriveau and Harris2011), competence/success (Chudek et al. Reference Chudek, Brosseau, Birch, Henrich, Banaji and Gelman2013), conformity (Corriveau et al. Reference Corriveau, Fusaro and Harris2009), and sex (Shutts et al. Reference Shutts, Kinzler and DeJesus2013). This approach has allowed a vast body of work on cultural learning, from both developmental and social psychology, as well as economics, biology, and anthropology, to be systematically integrated within the same framework (Chudek et al. Reference Chudek, Brosseau, Birch, Henrich, Banaji and Gelman2013; Hoppitt & Laland Reference Hoppitt and Laland2013; Mesoudi Reference Mesoudi2009).Footnote ²

Two challenges with regard to CREDs were raised by the commentators: one that urged us to integrate more developmental psychology and address particular experimental findings (Corriveau & Chen) and a second that made sweeping (and surprising) historical claims about the lack of CREDs in non-WEIRD societies. With regard to Corriveau & Chen, it is worth noting that the original paper on CREDs (Henrich Reference Henrich2009) deploys much supporting laboratory data from developmental psychologists (Harris et al. Reference Harris, Pasquini, Duke, Asscher and Pons2006; Harris & Koenig Reference Harris and Koenig2006). As with the more recent studies that Corriveau & Chen highlight (e.g., Corriveau et al. Reference Corriveau, Chen and Harris2015), this work shows that children are not automatically inclined to believe in things like germs and angels. Such findings run contrary to the proposals of some by-product theories that downplay cultural learning influences except for content biases (e.g., Barrett Reference Barrett2004). Instead, this developmental evidence suggests that children's inclinations to believe in invisible agents or entities depend heavily on the combination of the testimony they hear and the actions associated with that testimony.

What is striking about the studies discussed by Corriveau & Chen is that children exposed to religious CREDs, either through church attendance or parochial school, not only believed in characters that seemed like those from religious stories they had likely heard, but also believed in other fantastical or magical characters. We think this is interesting, but it is not evidence against CREDs. If anything, it suggests that CREDs are capable of shaping belief inferences in a broader range of domains.Footnote ³ The CREDs hypothesis points developmental psychologists toward: (1) the cost of beliefs and actions of children's models, as well as relevance of testimony (testimony is only one element of cultural learning), and (2) the need to study children naturalistically, outside of the laboratory, in daily life.

Boyer & Baumard's comments about CREDs run contrary to much work in anthropology (Sosis et al. Reference Sosis, Kress and Boster2007; Whitehouse Reference Whitehouse1995), archaeology, and history (Atran & Henrich Reference Atran and Henrich2010). As we noted in the target article, detailed fieldwork in New Guinea (Tuzin Reference Tuzin1976; Reference Tuzin2001), for example, reveals the painful initiation rites are used by the Ilahita Arapesh in support of the expressed beliefs in the gods of the Tambaran, a cult they explicitly copied from the highly successful and aggressively expanding Abelam. The Ilahita Arapesh believed that by performing the rituals they could persuade the gods to bestow blessing on their community. They eventually became the largest and most successful community in the region. Similarly, in the highlands of New Guinea, decades of painstaking ethnography by Wiessner reveals how competition among groups favored the spread of complexes of supernatural beliefs, particular ideologies, and rituals that promoted solidarity (Wiessner Reference Wiessner2002; Wiessner & Tumu Reference Wiessner and Tumu1998).