1. We make no conceptual distinction between “culture” and “society” or “mind” and “brain.”

2. This framework is also informed by the first author's (Atran’s) ethnographic sojourns among Lowland Maya (Mesoamerica), Druze mountaineers (Middle East), Pashtun nomads (Central Asia), Tamil Hindu farmers (South India), and Ladakhi Buddhist tanshumants (Himalaya), and by the second author's (Norenzayan’s) familiarity with the religious civil wars of Lebanon (1975–1991).

3. Evolutionarily, at least some basic emotions preceded conceptual reasoning: surprise, fear, anger, disgust, joy, sadness (Darwin 1872/1965; Ekman 1992). These may have further evolved to incite reason to make inferences about situations relevant to survival decisions. Existential anxieties are by-products of evolved emotions, such as fear and the will to stay alive, and of evolved cognitive capacities, such as episodic memory and the ability to track the self and others over time. For example, because humans are able to metarepresent their own selves and mentally travel in time (Wheeler et al. 1997), they cannot avoid overwhelming inductive evidence predicting their own death and that of persons to whom they are emotionally tied, such as relatives, friends, and leaders. Emotions compel such inductions and make them salient and terrifying. This is “The Tragedy of Cognition.” All religions propose a supernatural resolution in some minimally counterfactual afterlife.

4. Although the Buddha and the buddhas are not regarded as gods, Buddhists clearly conceive of them as “counter-intuitive agents” (Pyysiännen 2003). The Chinese Buddhist Pantheon in-cludes the 18 Lohan, or supernatural guardian angels known for their great wisdom, courage, and supernatural power, and the four Si-Ta-Tien-Wang, or Guardian Kings of the four directions (akin to the Maya Chaak). In Sri Lanka, the Sinhalese relics of the Buddha have miraculous powers. In India, China, Japan, Thailand, and Vietnam, there are magic mountains and forests associated with the Buddha; and the literature and folklore of every Buddhist tradition recount amazing events surrounding the Buddha and the buddhas.

5. Experiments with adults in the United States (Barrett & Keil 1996) and India (Barrett 1998) further illustrate the gap between theological doctrine and actual psychological processing of religious concepts. When asked to describe their deities, subjects in both cultures produced abstract and consensual theological descriptions of gods as being able to do anything and anticipate and react to everything at once; always knowing the right thing to do; and being able to dispense entirely with perceptual information and calculation. When asked to respond to narratives about these same gods, the same subjects described the deities as being in only one place at a time, puzzling over alternative courses of action, and looking for evidence in order to decide what to do (e.g., to first save Johnny, who's praying for help because his foot is stuck in a river in the United States, and the water is rapidly rising; or to first save little Mary, whom He has seen fall on railroad tracks in Australia where a train is fast approaching).

6. One distinction between fantasy and religion is knowledge of its source. People know or assume that public fictions (novels, movies, cartoons, etc.) were created by specific people who had particular intentions for doing so. Religious believers assume that utterances or texts connected with religious doctrines are authorless, timeless, and true. Consequently, they don't apply ordinary criteria of relevance to religious communications to figure out the speaker's true intentions or check on whether God is lying or lacking information (Sperber & Wilson 1986).

7. As Dan Sperber (1996) asked in an open communication to the Evolution and Human Behavior Society: “Is fitness a matter of having descendants with a recognizable ideology? Of population size? Of variations in size (expansion)? Of duration? Of some weighted combination of size and duration? What of social systems that expand rapidly at the expense of heritability (empires)?” Without answers to such questions (and none seem forthcoming) the idea of societal-level fitness is hopelessly vague.

8. For each natural domain, there is a proper domain and (possibly empty) actual domain (Sperber 1994). A proper domain is information that is the cognitive module's naturally selected function to process. The actual domain of a module is any information in the organism's environment that satisfies the module's input conditions whether or not the information is functionally relevant to ancestral task demands – that is, whether or not it also belongs to its proper domain. For example, cloud formations and unexpected noises from inanimate sources (e.g., a sudden, howling gush of wind) readily trigger inferences to agency among people everywhere. Although clouds and wind occurred in ancestral environments, they had no functional role in recurrent task problems with animate beings. Similarly, moving dots on a screen do not belong to agency's proper domain because they could not have been involved with ancestral task demands. Like clouds and wind, moving dots on computer screens belong to its actual domain. A parallel example is food-catching behavior in frogs. When a flying insect moves across the frog's field of vision, bug-detector cells are activated in the frog's brain. Once activated, these cells in turn massively fire others in a chain reaction that usually results in the frog shooting out its tongue to catch the insect. The bug-detector is primed to respond to any small dark object that suddenly enters the visual field (Lettvin et al. 1961). If flying insects belong to the proper domain of frog's food-catching module, then small wads of black paper dangling on a string belong to the actual domain.

9. Psychoanalytic (Erikson 1963; Freud 1913/1990) and attachment (Bowlby 1969; Kirkpatrick 1998) theories hold that primary deities are surrogate parents who assuage existential anxieties. But ethnographic reports indicate that malevolent and predatory deities are as culturally widespread, historically ancient, and socially supreme as benevolent deities. Examples include cannibalistic spirits of small-scale Amazonian, sub-Saharan African, and Australian aboriginal societies, as well as bloodthirsty deities of larger-scale civilizations that practiced human sacrifice, such as Moloch of the Ancient Middle East, the death goddess Kali of tribal Hindus, and the Maya thunder god Chaak. Psychological findings on false-belief tasks (see below) further indicate that beliefs about people are not the basis of beliefs about God because the developmental trajectories of these two belief sets diverge from the outset.

10. Another example from ethology offers a parallel. Many bird species have nests parasitized by other species. Thus, the cuckoo deposits eggs in passerine nests, tricking the foster parents into incubating and feeding the cuckoo's young. Nestling European cuckoos often dwarf their host parents (Hamilton & Orians 1965): “The young cuckoo, with its huge gape and loud begging call, has evidently evolved in exaggerated form the stimuli which elicit the feeding response of parent passerine birds… . This, like lipstick in the courtship of mankind, demonstrates successful exploitation by means of a ‘super-stimulus’” (Lack 1968). Late nestling cuckoos have evolved perceptible signals to manipulate the passerine nervous system by initiating and then arresting or interrupting normal processing. In this way, cuckoos are able to subvert and co-opt the passerine's modularized survival mechanisms.

11. Aristotle (1963) was the first to point out in his Categories that such counterintuitive expressions cannot even be judged false because no set of truth conditions could ever be definitely associated with them. He gave the example of “two-footed knowledge.” According to him, “two-footed” could be sensibly (truly or falsely) applied to all animals but not to any sort of knowledge. This is because knowledge falls under the ontological category of nonsubstantial things, whereas being two-footed falls under the altogether distinct ontological category of substantial things. Trying to put together things from different ontological categories produces a “category mistake.” For Aristotle, the world that could be properly described in ordinary Greek was the world that is (nomologically). This led him to conflate the world's ontological structure (what philosophy and science consider to be the ultimate “stuff” composing the world) with the semantic structure of language (the constraints that govern the ordinary relations between words and thoughts). Subsequent philosophers have reinterpreted the notion of a category mistake as a logical or semantic “type confusion” (Pap 1963; Sommers 1963). Cognitive and developmental psychologists have experimentally shown that children across cultures do not violate such categorical constraints on language learning when attempting to learn the meaning of words (Keil 1979; Walker 1992).

12. Science, like religion, uses metarepresentation in cosmology building: for example, in analogies where a familiar domain (e.g., solar systems, computers, genetic transmission) is used to model some initially less familiar system (e.g., atoms, mind/brains, ideational transmission). In fact, science and religion may use the same analogies; however, there is a difference in these uses. Science aims to reduce the analogy to factual description, where the terms of the analogy are finally specified, with no loose ends remaining and nothing left in the dark: Atoms are scientifically like solar systems if and only both can be ultimately derived from the same set of natural laws. Whereas science seeks to kill the metaphor, religion strives to keep it poetic and endlessly open to further evocation. In religion, these ideas are never fully assimilated with factual and commonsensical beliefs, like a metaphor that metarepresents the earth as a mother but not quite, or an angel as a winged youth but not quite.

13. According to Boyer (1994; 1997; 2000), bodiless supernaturals are counterintuitive because they think and act but lack physical substance. The matter is not so simple. First, experiments with infants and adults indicate that ordinary intuitions about causal agents do not require knowledge or perception of material substance, only the expectation (perhaps never actually realized) that there ultimately is a physical source of intentional action (Csibra et al. 1999). Ontological violations block such expectations being realized even in principle (e.g., invisible agents versus heard but unseen beings). They countermand rules for eventual processing, not actual perception. Second, not all mental states are equally bound to ordinary intuitions about bodies. Recent studies indicate that children from 5 years on up more readily attribute epistemic mental states (see, think, know) to beings in the afterlife than psychobiological mental states (hunger, thirst, sleepiness) (Bering & Bjorklund 2002). Ordinary distinctions between mind and body (e.g., dreaming) thus seem to provide at least some intuitive support for extraordinary beings with disembodied minds (Hobbes 1651/1901).

14. Barrett and Nyhof (2001, p. 79) list as common items: “a being that can see or hear things that are not too far away”; “a species that will die if it doesn't get enough nourishment or if it is severely damaged”; “an object that is easy to see under normal lighting conditions.” Such items fall so far below ordinary expectations that communication should carry some new or salient information that Barrett and Nyhof (2001, pp. 82–83) report: “common items were remembered so poorly relative to other items… . In some instances of retelling these items, participants tried to make the common property sound exciting or unusual.” In other words, some subjects tried to meet minimum conditions of relevance (Sperber & Wilson 1986). For the most part, common items failed these minimum standards for successful communication.

15. Highest degradation was observed in the mostly MCI and all INT conditions, conforming to an inverse quadratic function, F(3, 89) = 4.49, p .05. Memory degraded least in the Mostly INT condition, and increased as the proportion of MCI beliefs increased, resulting in a linear trend, F(2, 65) = 3.53, p = .06.

16. Only additional evidence could show whether children “continue” to think of God in the same way after they become aware of false beliefs (as Barrett et al. 2001 intimate), or (as seems more likely) come to have different reasons for thinking that God would not be deceived.

17. To deal with deficits in counterfactual thinking, St. Paul's Church in Alabama (Trenton Diocese) has a special program for autistics: “The church requires that children who receive Holy Communion be able to recognize the difference between ordinary bread and the Eucharist… . The St. Paul's program was designed to teach the difference” (Rev. Sam Sirianni, cited in Raboteau 2000).