Norenzayan et al. gather an array of evidence illustrating that elements of Big Gods (e.g., supernatural monitoring) increase group cohesion and cooperation. However, the hypothesis as stated, frequently conflates concepts encompassing parochial prosociality – group cohesion in the context of intergroup competition – with extended or universal prosociality involved in cooperation among strangers. It is beyond question that religiosity increases the former. However, the authors’ hypothesis that religion directs morality beyond the boundaries of the shared ethnic or religious in-group, to include strangers and known or presumed out-group members, is not supported. This distinction is crucial because the authors are presenting Big Gods as not only enabling successful intergroup competition but also as enabling societies to “scale up.”

As Norenzayan et al. concede, Big Gods subsume mechanisms that are not solely prosocial but rather contain multiple competing forces. In some places, the authors correctly qualify this pattern as referring to prosociality directed toward group members and occurring alongside intolerance or prejudice against out-group members. However, in other instances, the authors refer simply to prosociality, invoking concepts like empathy and cooperation as causal mechanisms. In section 4, they suggest studies of charitable giving and volunteering illustrate the prosocial effects of religious engagement. However, it is unclear whether this represents extended prosociality because of the uncontrolled status of the recipient (i.e., often the religious group itself; Galen Reference Galen2012). Religious belief is less predictive of charity or volunteering outside the group (Galen et al. Reference Galen, Sharp and McNulty2015; McKitrick et al. Reference McKitrick, Landres, Ottoni-Wilhelm and Hayat2013).

This conflation of concepts is also featured in the studies in section 4.2 purportedly demonstrating that semantic priming of religion increases “cooperation with strangers.” Most of the constituent studies did not specify the relationship between partners (i.e., fellow group member vs. stranger), which is necessary to distinguish parochialism from extended prosociality. For example, in McNamara et al. (Reference McNamara, Norenzayan and Henrich2016), the beneficiary of generosity was described as an outsider from another (Fijian) island, but a coreligionist. Similarly, other studies included designs in which participants knew the group identity of their partners (Hadnes & Schumacher Reference Hadnes and Schumacher2012) or included results illustrative of selective, not universal, prosociality (Pichon & Saroglou Reference Pichon and Saroglou2009). In order to label these effects as relevant to a pluralistic social context, it is necessary to demonstrate that religious priming activates prosociality regardless of the targets’ group membership (e.g., not only “coreligionists”). Any group favoritism promoted by religiosity in small societies is irrelevant to large-scale societies in which anonymous strangers cannot be presumed coreligionists. For the same reason, any “deep trust and commitment … characteristic of global religious communities” (sect. 5.2) cannot be extrapolated to pluralistic large-scale societies. Out-group-inclusive trust is not associated with religiosity (Welch et al. Reference Welch, Sikkink and Loveland2007) but can appear so because trusting “most people” connotes in-group members to those in more religious countries, but out-group members or strangers to those in less religious countries (Delhey et al. Reference Delhey, Newton and Welzel2011). The use of terms such as “stranger” and “anonymous” (sect. 3) to refer to individuals known to be from a given island or from within the community is oxymoronic from the standpoint of distinguishing a complete stranger – possibly an out-group member – from someone who shares some group affiliation with the participant.

Another problem with the Big Gods theory, as Norenzayan et al. partly concede, is that phenomena attributed to religion are by-products of more generalized, secular mechanisms. For example, supernatural monitoring is a subset of a broader social monitoring function. Equivalent effects are elicited by priming social scrutiny or self-awareness (Gervais & Norenzayan Reference Gervais and Norenzayan2012a). Other contextual primes shown to promote honesty include mirrors and bright lights, which activate intuitions such as “what would others think of me?” (Chiou & Cheng Reference Chiou and Cheng2013; Diener & Wallbom Reference Diener and Wallbom1976). Supernatural concepts such as “God is watching” or “avoiding the evil eye” are thus variations of social monitoring intuitions projected as stemming from external agents, rather than uniquely religious in character.

Similarly, the authors often state that prosocial effects (e.g., in sect. 4) are attributable to “religious commitment.” However, naturalistic as well as experimental studies indicate that prosociality is promoted by secular factors such as general group involvement, rather than by uniquely prosocial effects of religious beliefs (Galen et al. Reference Galen, Sharp and McNulty2015; Thomson Reference Thomson2015). Many of the studies in the meta-analysis found varied effects depending on the specific primed content such as “religion” versus “God” (only the latter associated with out-group prosociality; Preston & Ritter Reference Preston and Ritter2013). Hence, any prosocial priming effects are not the result of “religious belief” but of certain versions of religious as well as secular content exhibiting positive or reward-related semantic associations (Harrell Reference Harrell2012; Pichon et al. Reference Pichon, Boccato and Saroglou2007).

In sum, Norenzayan et al. concede throughout their impressive body of work that religious influences are: (1) not necessary for prosociality; (2) intertwined with non-prosocial influences; (3) context dependent; and (4) reliably linked to in-group cohesion rather than extended prosociality. In numerous places, the language used to describe religious group solidarity is properly qualified as referring only to within-group benefits. But elsewhere, phrases are used such as “large-scale cooperation” and “benefitting others” without the crucial qualifier “within the group.” What may seem to be a picayune terminological issue becomes more serious when extrapolated to a generalized conclusion that religious concepts have prosocial effects. In modern pluralistic societies consisting of individuals from mixed religious and ethnic backgrounds, group cohesion is not tantamount to extended prosociality, and indeed often opposes it. As stated by the authors, sacred non-negotiable beliefs exacerbate the “dark side” of intergroup intolerance by sanctifying and moralizing it (sect. 5.3, para. 3).

Therefore, group cohesion should not even warrant the term prosociality for the same reason that selective nepotism does not. It is one thing for religiosity to connote concepts such as “God is watching and wants you to be nice to fellow group members,” but this is not equivalent to more abstract moral enhancement such as “treat all others the way you want to be treated” or simply “be nice to others.” In many cases (e.g., interactions with a coreligionist), the resulting actions could be identical. However, if the interaction is not with a presumed group member, the two concepts will predict different forms of behavior.