Religion is not an evolutionary adaptation per se, but a recurring cultural by-product of the complex evolutionary landscape that sets cognitive, emotional, and material conditions for ordinary human interactions. Religion exploits only ordinary cognitive processes to passionately display costly devotion to counterintuitive worlds governed by supernatural agents. The conceptual foundations of religion are intuitively given by task-specific panhuman cognitive domains, including folkmechanics, folkbiology, and folkpsychology. Core religious beliefs minimally violate ordinary notions about how the world is, with all of its inescapable problems, thus enabling people to imagine minimally impossible supernatural worlds that solve existential problems, including death and deception. Here the focus is on folkpsychology and agency. A key feature of the supernatural agent concepts common to all religions is the triggering of an “Innate Releasing Mechanism,” or “agency detector,” whose proper (naturally selected) domain encompasses animate objects relevant to hominid survival – such as predators, protectors, and prey – but which actually extends to moving dots on computer screens, voices in wind, and faces on clouds. Folkpsychology also crucially involves metarepresentation, which makes deception possible and threatens any social order. However, these same metacognitive capacities provide the hope and promise of open-ended solutions through representations of counterfactual supernatural worlds that cannot be logically or empirically verified or falsified. Because religious beliefs cannot be deductively or inductively validated, validation occurs only by ritually addressing the very emotions motivating religion. Cross-cultural experimental evidence encourages these claims.

The target article proposes that “counterintuitive beliefs in supernatural agents” are shaped by cognitive factors and survive because they foster empathic concern and counteract existential dread. I argue that they are shaped by motivational forces similar to those that shape our beliefs about other people; that empathic concern is rewarded in a more elementary fashion; and that a major function of these supernatural beliefs may be to provide a more flexible alternative to autonomous willpower in controlling not only dread but also many other unwelcome urges.

In sketching a preliminary scientific theory of religion, Atran & Norenzayan (A&N) generally agree with cognitive scientists of religion in the factors that coalesce to form religion. At times they misrepresent, however, the notion of “counterintuitive” concepts as they apply to religious concepts, confusing counterintuitive with counterfactual, category mistakes, and logical contradiction.

Atran & Norenzayan's (A&N's) target article effectively combines the insights of evolutionary biology and interdisciplinary cognitive science, neither of which alone yields sufficient explanatory power to help us fully understand the complexities of supernatural belief. Although the authors' ideas echo those of other researchers, they are perhaps the most squarely grounded in neo-Darwinian terms to date. Nevertheless, A&N overlook the possibility that the tendency to infer supernatural agents' communicative intent behind natural events served an ancestrally adaptive function.

From a religious studies perspective, Atran & Norenzayan (A&N) succeed in arguing for the influence of evolved cognitive functions in religious phenomena. To develop their argument further, four suggestions are offered: (1) Look beyond the ordinary to the extraordinary; (2) culture matters more than ever; (3) theists need not despair, atheists ought not celebrate; and (4) dreaming is a primal wellspring of religion.

Atran & Norenzayan (A&N) correctly claim that religion reduces emotions related to existential concerns. Our response adds to their argument by focusing on religious differences in the importance of emotion, and on other emotions that may be involved in religion. We believe that the important differences among religions make it difficult to have one theory to account for all religions.

According to embodied cognition theory, our physical embodiment influences how we conceptualize entities, whether natural or supernatural. In serving central explanatory roles, supernatural entities (e.g., God) are represented implicitly as having unordinary properties that nevertheless do not violate our sensorimotor interactions with the physical world. We conjecture that other supernatural entities are similarly represented in explanatory contexts.

In the case of religion, explanations based on emotion should be privileged over those based on “cold” cognition. The origins of religious beliefs are as critical to understanding religion as are the group phenomena which sustain them. In addition, religion's relationship to the growth of knowledge is neglected by the target authors. The balance between the costs and benefits of religion will vary depending upon the phase of an individual society's cultural evolution.

This excellent outline of evolutionary hypotheses is compromised by severe reductionism. Other writings succeed in granting theism ontological significance without compromising rigor. The discussion of counterintuitiveness neglects coherence in memory. Bearing in mind our severely limited working memory capacity, susceptibility to religious mythologies may comprise an adaptive heuristic approach to summarizing the contingencies of the most far-reaching of life's problems.

In this commentary I attempt to extend the argument made by Atran and Norenzayan in two ways. First, I distinguish between the causes and the consequences of religious belief and speculate on the positive and negative consequences of religion. Second, I raise some questions about individual differences in religiosity and suggest that the origins of nonbelief are worth investigating.

In arguing for religion as a side effect of everyday cognition, Atran & Norenzayan (A&N) provide useful analyses of the strengths of the “naturalness-of-religion” position over others; however, experimental shortcomings limit the contributions of their empirical work. A relevant addendum involves considering research on children's orientation to teleological explanations of natural phenomena, which suggests that relatively rich cognitive proclivities might underlie religious thought.

Atran & Norenzyan (A&N) claim that an appreciation of the evolved inferential machinery underlying supernatural beliefs can greatly aid us in understanding regularities in culturally shared conceptions of religion. I explore how their model provides insight into why culturally shared tales of horror (e.g., horror movies) often combine religious and predatory content.

Atran & Norenzayan (A&N) are correct that religion is an evolutionary by-product, not an adaptation, but they do not go far enough. Once supernatural beliefs are enabled by processes they describe, numerous social-cognitive mechanisms related to attachment, social exchange, coalitional psychology, status and dominance, and kinship are crucial for explaining the specific forms religion takes and individual and cultural differences therein.

Atran & Norenzayan (A&N) rightly criticize cognitive theories for failure to explain sacrifice and commitment. But their attempt to reconcile cognitivism with commitment theory is unconvincing. Why should imaginary entities be effective in punishing moral defectors? Heavy costs are entailed in enforcing community-wide social contracts, and behavioural ecology is needed to explain how and why evolving humans could afford these costs.

Terror management theory and research can rectify shortcomings in Atran & Norenzayan's (A&N's) analysis of religion. (1) Religious and secular worldviews are much more similar than the target article supposes; (2) a propensity for embracing supernatural beliefs is likely to have conferred an adaptive advantage over the course of evolution; and (3) the claim that supernatural agent beliefs serve a terror management function independent of worldview bolstering is not empirically supported.

Atran & Norenzayan (A&N) have proposed a study of religion based in the cognitive sciences. Their final conclusions, however, incorporate functionalist definitions. Further, key features by which they characterize religion are not instantiated by some historical evidence. Nevertheless, the foci of their arguments are central to any study of religion and should provoke further research and experimentation along the lines suggested.

The authors attempt to explain the ubiquity and persistence of human religion by invoking innate, domain-specific cognitive furniture, while dismissing the potential of other approaches, such as memetics, to produce “mindful” understandings of religion. This commentary challenges the explanatory adequacy of cognitive nativism, suggesting that memetics has as much claim to utility and “mindfulness” as innate mental modules do.

Some forms of religion may in some cases alleviate existential anxieties and help maintain morality; yet religion can also persist without serving any such functions. Atran & Norenzayan (A&N) are unclear about the importance of these functions for a theory of the recurrence of religious beliefs and behaviors.

Atran & Norenzayan (A&N) fail to address several problems with commitment theory as it relates to non-kin altruism in religious contexts. They (1) provide little support for the contention that religious sacrifices function as signals, (2) do not distinguish between religious specialists and lay believers, and (3) conflate definitions of cooperation and sacrifice.

Atran & Norenzayan (A&N) have not adequately supported the epistemic component of their proposal, namely, that God does not exist. A weaker, more probable hypothesis, not requiring that component – that the benefits of religious belief outweigh those of disbelief, even though we do not know whether or not God exists – is available. I counsel them to use Ockham's razor, eliminate their negative epistemic thesis, and accept the weaker hypothesis.