To be published in Behavioral and Brain Sciences (in press)
© Cambridge University Press 2003
Word Count: Text = 12,594 , References = 2,265, Total = 16, 042
RELIGION’S EVOLUTIONARY LANDSCAPE:
COUNTERINTUITION, COMMITMENT, COMPASSION, COMMUNION
CNRS – Institut Jean Nicod
Department of Psychology
2136 West Mall
Vancouver, BC V6T 1Z4
Agency, Death anxiety, Evolution, Folkpsychology, Maya, Memory, Metarepresentation, Morality, Religion, Supernatural
Religion is not an evolutionary adaptation, but a recurring by-product of the complex evolutionary landscape that sets cognitive, emotional and material conditions for ordinary human interactions. The conceptual foundations of religion are intuitively given by task-specific panhuman cognitive domains, including folkmechanics, folkbiology, folkpsychology. Core religious beliefs minimally violate ordinary intuitions about the world, with its inescapable problems. This enables people to imagine minimally impossible supernatural worlds that solve existential problems, including death and deception. 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.
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, 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, 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.
In every society, there is:
(1) widespread counterfactual and counterintuitive belief in supernatural agents (gods, ghosts, goblins, etc.);
(2) hard-to-fake public expressions of costly material commitments to supernatural agents, that is, offering and sacrifice (offerings of goods, property, time, life);
(3) mastering by supernatural agents of people’s existential anxieties (death, deception, disease, catastrophe, pain, loneliness, injustice, want, loss); and
(4) ritualized, rhythmic sensory coordination of (1), (2) and (3), that is, communion (congregation, intimate fellowship, etc.).
In all societies there is an evolutionary canalization and convergence of (1), (2), (3) and (4) that tends towards what we shall refer to as “religion”; that is, passionate communal displays of costly commitments to counterintuitive worlds governed by supernatural agents. Although these facets of religion emerge in all known cultures and animate the majority of individual human beings in the world, there are considerable individual and cultural differences in the degree of religious commitment. The question as to the origin and nature of these intriguing and important differences we leave open.
This theoretical framework drives our program of research. The framework is the subject of a recent book (Atran 2002a). Here, a more comprehensive set of experimental results and observations are introduced to support integration within an evolutionary perspective that envisions religion as a converging by-product of several cognitive and emotional mechanisms that evolved for mundane adaptive tasks (for somewhat similar, independently researched, views of religion as an emergent by-product of numerous domain-specific psychological mechanisms, see Kirkpatrick 1999, Boyer 2001).
The current experiments suggestively support this long-term research program. We hope the findings will stimulate further tests and refinements to assess the empirical viability of this framework. The aim of this paper is to foster scientific dialogue between cultural anthropology, cognitive, developmental and social psychology, and evolutionary biology regarding a set of phenomena vital to most human life and all societies. The present article mainly concerns the first (1) and third (3) criteria of religion. This introductory section first presents in general terms the overall intellectual framework that interrelates all four criteria, discusses some obvious objections to these generalizations, and offers some caveats.
The criterion (1) of belief in the supernatural rules out commitment theories of religion as adequate, however insightful. Such theories underplay or disregard cognitive structure and its causal role. Commitment theories attempt to explain the apparent altruism and emotional sacrifice of immediate self interest accompanying religion in terms of long-term benefits to the individual (Alexander 1987, Irons 1996, Nesse 1999) or group (Boehm 1999, D.S.Wilson 2002) - benefits that supposedly contribute to genetic fitness or cultural survival. They do not account for the cognitive peculiarity of the culturally universal belief in beings who are imperceptible in principle, and who change the world via causes that are materially and logically inscrutable in principle. They cannot distinguish Marxism from Monotheism, secular ideologies from religious belief (Atran 2002a).
The criterion (2) of costly commitment rules out cognitive theories of religion as adequate, however insightful. Cognitive theories attempt to explain religious belief and practice as cultural manipulations of ordinary psychological processes of categorization, reasoning and remembering (Lawson & McCauley 1990, Atran & Sperber 1991, Boyer 1994, Barrett 2000, Andersen 2000, Pyysiäinen & Anttonen 2002). They do not account for the emotional involvement that lead people to sacrifice to others what is dear to themselves, including labor, limb and life. Such theories are often short on motive and unable to distinguish Mickey Mouse from Moses, cartoon fantasy from religious belief (Atran 1998:602; cf. Boyer 2000, Norenzayan & Atran 2003). They fail to tell us why, in general, the greater the sacrifice – as in Abraham offering up his beloved son - the more others trust in one’s religious commitment (Kierkegaard 1955).
This article extends the idea (first suggested by Sperber 1975) that religious thought and behavior can be explained as mediated by ordinary mental mechanisms, which can be scientifically studied regardless of whether religions are true or not in a metaphysical sense. In this “mentalist” tradition, the focus so far has been on cognition and culture; that is, on how religious ideas are mentally constructed, transmitted across minds and acquired developmentally. To be sure, there have been recent attempts by cognitive scientists studying religion to consider the role of emotion, and growing realization that religion cannot have a purely cognitive explanation that fails to take into account the social dilemmas motivating religious beliefs and practices (Whitehouse 1999, McCauley & Lawson 2002, Pyysiännen 2001). But there is still little analytic or empirical integration of (1) and (3).
Religions invoke supernatural agents (Tylor 1958, Horton 1967) to deal with (3) emotionally eruptive existential anxieties (Malinowski 1961), such as death and deception (Becker 1973, Feuerbach 1972, Freud 1990).  All religions, it appears, have “awe-inspiring, extraordinary manifestations of reality” (Lowie1924:xvi). They generally have malevolent and predatory deities as well as more benevolent and protective ones. Supernatural agent concepts trigger our naturally-selected agency-detection system, which is trip-wired to respond to fragmentary information, inciting perception of figures lurking in shadows, and emotions of dread or awe (Guthrie 1993; cf. Hume 1956). Granted, nondeistic “theologies”, such as Buddhism and Taoism, doctrinally eschew personifying the supernatural or animating nature with supernatural causes. Nevertheless, common folk who espouse these faiths routinely entertain belief in an array of gods and spirits that behave counterintuitively in ways that are inscrutable to factual or logical reasoning. Even Buddhist monks ritually ward off malevolent deities by invoking benevolent ones, and conceive altered states of nature as awesome.
Conceptions of the supernatural invariably involve the interruption or violation of universal cognitive principles that govern ordinary human perception and understanding of the everyday world. Consequently, religious beliefs and experiences cannnot be reliably validated (or disconfirmed as false) through consistent logical deduction or consistent empirical induction. Validation occurs only by (4) collectively satisfying the emotions that motivate religion in the first place. Through a “collective effervescence” (Durkheim 1995), communal rituals rhythmically coordinate emotional validation of, and commitment to, moral truths in worlds governed by supernatural agents. Rituals involve sequential, socially interactive movement and gesture and formulaic utterances that synchronize affective states among group members in displays of cooperative commitment. Through the sensory pageantry of movement, sound, smell, touch and sight, religious rituals affectively coordinate actors’ minds and bodies into convergent expressions of public sentiment (Turner 1969) – a sort of N- person bonding that communicates moral consensus as sacred, transcending all reason and doubt (Rappaport 1999). Sensory pageantry also assures the persistence and transmission of the religious beliefs and practices it infuses.
These four conditions do not constitute the necessary and sufficient features of “religion.” Rather, they comprise a stipulative (working) framework that delimits a causally interconnected set of pancultural phenomena that comprises the object of our study. One may choose to call phenomena that fall under this set of conditions “religion” or not; however, for our purposes their joint satisfaction is what we mean by the term. Nevertheless, we offer this working framework as an adequate conceptualization that roughly corresponds to what most scholars consider religion. This framework is concerned with the pancultural foundations of religion; accordingly, our conceptualization is broad in scope. Surely, religions are manifested in culturally diverse ways, and are shaped by local cultural contexts. Elsewhere, scholars have examined how the distinctive paths that religions take shape psychological tendencies (e.g., Weber 1946; Shweder, et al., 1997). Our framework is not incompatible with these approaches. Indeed, it offers candidates for the psychological building blocks of religion, which then are culturally exploited in distinct but converging paths.
More critical are the many ethnographic reports interpreting that some people or some societies make no hard and fast distinction (1) between the natural and supernatural, or (2) between costly sacrifice and the social redistribution of material or social rewards; or that (3) religions are as anxiety-activating as anxiety-assuaging, or (4) sometimes devoid of emotional ritual. In addition, (5) there is considerable psychological and sociological evidence for the health and wellbeing benefits of religion, which suggests that religion may be adaptive and not simply a by-product of evolutionary adaptations for other things. We address each of these objections next.
1.1. The Natural versus the Supernatural
We base our argument regarding the cognitive basis of religion on a growing number of converging cross-cultural experiments on “domain-specific cognition” emanating from developmental psychology, cognitive psychology and anthropology. Such experiments indicate that virtually all (non brain-damaged) human minds are endowed with core cognitive faculties for understanding the everyday world of readily perceptible substances and events (for overviews, see Hirschfeld & Gelman 1994, Sperber et al. 1995, Pinker 1997). The core faculties are activated by stimuli that fall into a few intuitive knowledge domains, including: folkmechanics (object boundaries and movements), folkbiology (biological species configurations and relationships), and folkpsychology (interactive agents and goal-directed behavior). Sometimes operation of the structural principles that govern the ordinary and “automatic” cognitive construction of these core domains are pointedly interrupted or violated, as in poetry and religion. In these instances, counterintuitions result that form the basis for construction of special sorts of counterfactual worlds, including the supernatural, for example, a world that includes self-propelled, perceiving or thinking mineral substances (e.g., Maya sastun, crystal ball, Arab tilsam [talisman]) or beings that can pass through solid objects (angels, ghosts, ancestral spirits) (cf. Atran & Sperber 1991, Boyer 1994).
These core faculties generate many of the universal cognitions that allow cross-cultural communication and make anthropology possible at all. For example, even neonates assume a naturally occurring rigid body cannot occupy the same space as another (unlike shadows), or follow discontinuous trajectories when moving through space (unlike fires), or change direction under its own self-propelling initiative (unlike animals), or causally effect the behavior of another object without physical contact (unlike people) (Spelke et al. 1995). When experimental conditions simulate violation of these universal assumptions, as in a magic trick, neonates show marked surprise (longer gaze, intense thumb sucking, etc.). Children initially expect shadows to behave like ordinary objects, and even adults remain uncertain as to how shadows move. This uncertainty often evokes the supernatural.
All known societies appear to partition local biodiversity into mutually exclusive species-like groupings (Darwin 1859, Diamond 1966, Atran 1990, Berlin 1992), and to initially identify nonhuman organisms according to these groupings rather than as individuals (unlike the immediate local identification of individual human faces and behaviors, Atran 1998; cf. Hirschfeld 1996). Individualized pets and taxonomic anomalies, such as monsters, become socially relevant and evocative because they are purposely divorced from the default state of “automatic” human cognition about the limited varieties of the readily perceptible world, that is, “intuitive ontology” (Atran 1989, Boyer 1997; cf. Sperber 1975). This commonsense ontology is arguably generated by task-specific “habits of mind” that evolved selectively to deal with ancestrally recurrent “habits of the world” that were especially relevant to hominid (and in some cases, pre-hominid) survival: inanimate substances, organic species, persons.
What testable evidence there is indicates that, sometime after age three and except for severe autistics, most any person understands that most any other person can entertain perceptions, beliefs and desires different from one’s own, and that these different mental states differentially cause people’s behaviors
(Wimmer & Perner 1983, Avis & Harris 1991, Baron-Cohen 1995, Knight et al. 2001). Granted, there is experimental evidence for cultural variations in causal attribution of social behavior to personality traits versus social situations (Choi et al. 1999), and anecdotal interpretations of cultural behaviors as indicating an inability to distinguish between true and false beliefs, or reality from desire (cf. Lévy-Bruhl 1966, Lillard 1998). But contrary to the anecdotal evidence, experimental evidence suggests that children growing up in very different cultures, soon develop similar understanding of core aspects of human behavior as a function of beliefs and desires (Avis & Harris, 1991; Flavell, et al 1983). Furthermore, there is no generally accepted body of evidence indicating that our simian cousins can simultaneously keep in mind the thoughts of others, or, equivalently, entertain multiple possible and different worlds from which to select an appropriate course of action (Premack & Woodruff 1978, Hauser 2000; cf. Hare et al. 2001 for intriguing experiments suggesting rudimentary perspective taking in chimps). Without the ability to entertain multiple possible worlds, belief in the supernatural is inconceivable.
Within the emerging work on domain specificity there are controversies and doubts, as in any young and dynamic science. But the findings sketched above are widely replicated. Admittedly, there are alternative approaches to understanding cognition, such as connectionism, artificial intelligence, or phenomenology. Using any of these other approaches to model religion would no doubt present a different picture than the one we offer. We leave it to others to work the alternatives.
1.2. Costly Sacrifice versus Redistribution
One evolutionary problem with religion is explaining how and why biologically unrelated individuals come to sacrifice their own immediate material interests to form genetically incoherent relationships under an imagined permanent and immaterial authority. Altruism occurs when an organism’s behavior diminishes its own fitness and enhances the fitness of some other organism or organisms. Fitness is a measure of an organism’s reproductive success. The sacrifice of an organism for its relatives – a mother for her children, a brother for his siblings, an ant for its colony, a bee for its hive – lowers an organism’s individual fitness (also called “classical” or “Darwinian” fitness) because it compromises the individual’s ability to bear and raise offspring. Nevertheless, such kin altruism may also enhance the individual’s “inclusive fitness” by allowing surviving relatives to pass on many of the individual’s genes to future generations (Hamilton 1964). But what motivates the sort of nonkin cooperation characteristic of human religious commitment?
Unlike other primate groups, hominid groups grew to sizes (Dunbar 1996) that could not function exclusively on the basis of kin-selection (commitment falls off precipitously as genetic distance increases between individuals) or direct reciprocity (ability to directly monitor trustworthiness in reciprocation decreases rapidly as the number of transactions multiply). Larger groups of individuals out compete smaller groups in love and war (Axelrod 1984). A plausible hypothesis, then, is that the mechanisms for successful promotion of indirect reciprocity – including both religious and nonreligious behaviors - were naturally-selected in response to the environmental problem-context of spiraling social rivalry among fellow conspecifics, or “runaway social competition” (Alexander 1989). As “fictive kin” (Nesse 1999), members of religious groups perform and profit from many tasks that they could not do alone, one by one, or only with family. Thus, “Among the Hebrews and Phoenicians… the worshipper is called brother (that is, kinsman or sister of the god)” (Robertson Smith 1972:44n2). “Brotherhood” is also the common term applied today among the Christian faithful and to the fraternity (ikhwan) of Islam.
Indirect reciprocity occurs when individual X knows that individual Y cooperates with others, and this knowledge favors X cooperating with Y. Consider a population whose individuals have the option to cooperate or not. Suppose individual X randomly meets individual Y. If Y has a reputation for cooperation, and if X cooperates with Y, then X’s reputation likely increases. If X does not cooperate with Y, then X’s reputation likely decreases (see Nowak & Sigmund 1998 for various simulations). The basic idea is to help those who are known to help others. Reputation for religious belief is almost always reckoned as sincere social commitment, and such reputation is invariably linked to costly and hard-to-fake expressions of material sacrifice or concern that goes beyond any apparent self-interest.
Although calculations of economic or political utility often influence religious practices (Stark 2000), to conclude that’s all there is to religious commitment and sacrifice is unwarranted. In religious offerings, there is usually a nonrecuperable cost involved both in the selection of the item offered and in the ceremony itself. Thus, for the Nuer of Sudan, substituting a highly-valued item (cow) by one less-valued (fowl or vegetable) is allowable only to a point, after which “a religious accounting might reveal that the spirits and ghosts were expecting a long overdue proper sacrifice, because accounts were out of balance so to speak” (Evans-Pritchard 1940:26). Religious sacrifice usually costs something for the persons on whose behalf the offering is made. That is why “sacrifice of wild animals which can be regarded as the free gift of nature is rarely allowable or efficient” (Robertson Smith 1894:466). In many cases, the first or best products of one’s livelihood goes to the gods, as with the first fruits of the Hebrews or the most perfect maize kernels of the Maya. Most, if not all, societies specify obligatory circumstances under which religious sacrifice must be performed, regardless of economic considerations. Reviewing the anthropological literature, Raymond Firth (1963:16) surmises: “In all such cases the regular religious need to establish communication with god or with the spirit world… would seem to be pressing and primary. ‘Afford it or not’.”
In sum, religious sacrifice generally runs counter to calculations of immediate utility, such that future promises are not discounted in favor of present rewards. In some cases, sacrifice is extreme. Although such cases tend to be rare, they are often held by society as religiously ideal: for example, sacrificing one’s own life or nearest kin. Researchers sometimes take such cases as prima facie evidence of “true” (nonkin) social altruism (Rappaport 1999, Kuper 1996), or group selection, wherein individual fitness decreases so that overall group fitness can increase (relative to the overall fitness of other, competing groups) (Sober & D.S. Wilson 1998, D.S. Wilson 2002). But this may be an illusion.
A telling example is contemporary suicide terrorism (Atran 2003a). Consider the “Oath to Jihad” taken by recruits to Harkat al-Ansar, a Pakistani-based ally of Al-Qaida., which affirms that by their sacrifice they would help secure the future of their “family” of fictive kin: “Each [martyr] has a special place – among them are brothers, just as there are sons and those even more dear.” In the case of religiously-inspired suicide terrorism, these sentiments are purposely manipulated by organizational leaders, recruiters and trainers to the advantage of the manipulating elites rather than the individual (much as the fast food or soft drink industries manipulate innate desires for naturally scarce commodities like fatty foods and sugar to ends that reduce personal fitness but benefit the manipulating institution). No “group selection” is involved, only cognitive and emotional manipulation of some individuals by others.
1.3. Relieving versus Provoking Anxieties
Often the naturally eruptive anxieties that bring on the supernatural are artificially (purposely) excited then assuaged (Durkheim 1995). It might seem, then, that the problem of religion’s ability to neutralize suffering is akin to the wag about the salesman who throws dirt on the rug in order to demonstrate the vacuum cleaner’s ability to remove it. Consider initiation rituals that involve “rites of terror” (Whitehouse 1996), as among Native American Cheyenne and Arapaho (Lowie 1924), Walbiri (Meggitt 1965) and other aboriginals of the Central Australian Desert (Spencer & Gillen 1904), Mountain Ok Baktaman (Barth 1975) and Ilahita Arapesh of Highland Papua New Guinea (Tuzin 1982), or Candombolé Nagô sects of African-Brazilian Bahia (Carneiro 1940, Omari 1994). These arouse existential anxieties by culturally mimicking and manipulating seemingly capricious and uncontrollable situations that naturally provoke them: terror and risk of death from unidentifiable sources, the menace of infirmity and starvation through physical ordeal and deprivation, the injustice of whimsical oppression, sudden isolation and loneliness. Often initiates temporarily manifest behaviors and cognitions associated with persons clinically diagnosed as suffering abuse, stress or trauma, including: re-experiencing the events (nightmares, intrusive memories, flashbacks), avoidance (amnesia of the event, refusal to talk about or think about it) and hyperarousal (startle response, fitful sleep, poor concentration) (cf. Newport & Nemeroff 2000).
Still, there are important differences between such initiations and stress syndromes (e.g. posttraumatic stress disorder). Stress sufferers who permanently lose memory and undergo reduced immune response often suffer from chronic stress and lack of effective social support (Khansari et al. 1990, Dhabar & McEwen 1999). By contrast, even the most severe and emotionally aversive religious initiations end in positive exhibitions of social acceptance:
Boys and girls are made to recognize members of The People [Navajo] and are introduced to full participation in ceremonial life.... The first boy is led out beside the fire. The figure in the white mask makes a mark on each shoulder with sacred cornmeal.... Then, using a different falsetto cry, the black-masked figure lightly strikes… other places on the body, and the one who uses the reeds varies the time interval between touching the boy and uttering his cry, so its unexpectedness causes the boy to start convulsively…. Then the one who wore the black mask places it over the face of each child in turn…. All the children are told to look up and always remember the Holy People. The reversal of the masks is a very intelligent psychological act, for it allows the child to see that the dread figure is actually someone he knows, or at least a human being, and thus the ritual is robbed of some of its terror.... The ceremony closes with the admonition to each child not to betray to uninitiates what he has seen. (Kluckholn & Leighton 1974:207-208; cf. Turnbull 1962:225)
Through the stress that these exaggerated sensual displays induce, rites of passage furnish emotionally costly and memorable - but ultimately satisfying - commitments to the group and its supernatural agents.
In brief, these life rehearsals incite the very emotions and existential anxieties that motivate religious beliefs and quests for deliverance. Then, by assuaging and resolving the ensuing distress, successful completion of the ritual performance authenticates the religious thoughts and actions. This confirms the efficacy of religious belief and ritual performance in fusing cosmos to culture by overcoming the dreads and uncertainties of both spontaneously occurring natural events and the manipulated happenings of the social world.
1.4. Emotional Ritual
Although there is wide variation in the degree of sensory pageantry associated with religious rituals (Whitehouse 1999, McCauley & Lawson 2002), religious rituals habitually – perhaps invariably - include displays of social hierarchy and submission typical of primates and other social mammals (outstretched limbs baring throat and chest or genitals, genuflection, bowing, prostration, etc.). Even priests and kings must convincingly show sincere submission to higher supernatural authority lest their own authority be doubted (Aristotle 1958, Burkert 1996; cf. Watanabee & Smuts 1999).
Most often, religious rituals involve repeated, generally voluntary, and usually reversible states of emotional communion in the context of formulaic social ceremonies. Here, supernatural agents, through their surrogates and instruments, manifest themselves in people’s affections. The ceremonies repetitively occur to make highly improbable, and therefore socially unmistakable, displays of mutual commitment. Within the congregation’s coordinated bodily rhythms (chanting, swaying, tracking, etc.), in conjunction with submission displays, individuals show that they feel themselves identifying with, and giving over part of their being to, the intensely felt existential yearnings of others. This demonstration, in turn, conveys the intention or promise of self-sacrifice by and towards others (charity, care, defense, support, etc.), without any specific person or situation necessarily in mind.
Collective religious ritual always seems to involve ancestrally primitive communicative forms that Tinbergen calls “ritualized social releasers” (1951:191-192). Social releasers exhibit sense-evident properties, “either of shape, or colour, or special movements, or sound, or scents,” which readily elicit a well-timed and well-oriented cooperative response in a conspecific: for mating, parenting, fighting, defense, food gathering, and the like. But humans appear to be the only animals that spontaneously engage in creative, rhythmic bodily coordination to enhance cooperation. Unlike, say, avian mating calls or flight formations, human music or body dance, which are omnipresent in worship, can be arbitrarily and creatively recomposed.
A key feature of the creativity of
human worship is use of music in social ritual. Even the Taliban, who
prohibited nearly all public displays of sensory stimulation, promoted a
cappella religious chants. Nearer to home, in a survey of persons who reported
a religious experience (
Much of the intimate connection between music and religion remains a puzzle. One possible account sees music as an invitation to interpersonal relationships, creating emotional bonds between people, through the “attunement” of somatic states – much as the rocking and cooing behavior of mother and infant attunes the parental bond (Stern 1985). This is especially apparent in “call-response” format, as in Yoruba dances and Hebrew services. Moreover, in religious contexts, music is frequently experienced as authorless, like the sacred texts that often accompany it. The pre-tonal religious music of small-scale societies usually has its mythic beginnings in the origins of the world, which invites audiences to share in a sense of timeless intimacy. For the Catholic Church, Gregorian chants were taught to men by birds sent from heaven. Even Bach, Mozart and Beethoven were but vehicles of The Divine’s call to communion.
1.5. “Mind-Blind” Functionalism: Sociobiology, Group Selection, Memetics
Finally, our account opposes other evolutionary approaches to religion and culture, including much sociobiology (Harris 1974, E.O. Wilson 1978), group-selection theory (Sober & Wilson 1998, Boehm 1999) and memetics (Dawkins 1976, Dennett 1997). These alternatives are “mindblind” to the cognitive constraints on religious beliefs and practices, viewing religion and culture as bundles of functionally-integrated, fitness-bearing traits: for example, packages of environment-induced rituals (the material infrastructure underlying ideational superstructure), machinelike patternings of collective norms (worldviews) or partnerships of invasive and authorless ideas (memeplexes).
Proponents of these alternatives do not deny that minds have causally “proximate” roles in generating religious behaviors - as they may in generating economic behaviors – or that cognition may form part of some “ultimate” explanation of religion. Nevertheless, a common claim is that a meaningful causal account of such behaviors requires initial focus on measurable relationships between putative fitness-motivating factors in religious behaviors and ostensible fitness consequences (Dennett 1995:358-359, Sober & D.S. Wilson 1998:182,193; cf. Lumsden & E.O.Wilson 1981): for example, between individuals needing protein in animal-poor environments and ritual human sacrifice (E.O. Wilson 1978, Harris 1974), between ideas endeavoring to propagate themselves and proselytizing for altruism (Lynch 1996, Blackmore 1999), or between groups competing for survival and Judaism’s alleged cultural and genetic separatism (MacDonald 1998, D.S. Wilson 2002). These arguments are presented through selective use of anecdotal evidence, rather than being reliably tested and demonstrated.
sociobiological claims to the canons of “scientific materialism,” the causal
account that is supposed to produce religious practices (e.g., Aztec
cannabilistic sacrifice) from their ostensible material functions (e.g.,
compensating for lack of large game as sources of protein in
It is also notoriously difficult to establish measurable criteria by which whole cultures/societies or worldviews/memeplexes can have fitness consequences. Functional accounts are often synthetic abstractions: for example, a lone anthropologist’s normative digest of some culture that in reality has no clear boundaries and no systematically identifiable structural functions. Indeed, most reported “norms” are too semantically open-ended to have specific contents, such as the Ten Commandments: even members of the same church congregation fail to provide interpretations of the Ten Commandments that other congregation members consistently recognize as being interpretations of the Ten Commandments (Atran 2001). There are no “replicating” or even definite or definable cultural units for natural selection and vertical (transgenerational) or horizontal (contemporaneous) transmission (e.g., memes can be anything from a gender marker to partial tune, cell phone, cooking recipe, political philosophy, etc.). These facts render implausible all attempts to explain religions (or cultures with a religious element) as discrete or integrated functional systems (for reviews and analyses of specific arguments, see Atran 2001, 2002a, 2003b, 2003c).
All human societies pay a price for religion’s material, emotional, and cognitive commitments to unintuitive, factually impossible worlds. Functional evolutionary (“adaptationist”) arguments for religion often try to offset its clear functional disadvantages with greater functional benefits. There are many different and contrary explanations for why religion exists in terms of beneficial functions served. These include functions of social (bolstering group solidarity, group competition), economic (sustaining public goods, surplus production), political (mass opiate, rebellion’s stimulant), intellectual (e.g., explain mysteries, encourage credulity), health and well being (increase life expectancy, accept death), and emotional (terrorizing, allaying anxiety) utility. Many of these functions have obtained in one cultural context or another; yet all also have been true of cultural phenomena besides religion.
Such descriptions of religion are not wrong; however, none of these accounts provides explanatory insight into cognitive selection factors responsible for the ease of acquisition of religious concepts by children, or for the facility with which religious practices and beliefs are transmitted across individuals. They have little to say about which beliefs and practices – all things being equal – are most apt to survive within a culture, most likely to recur in different cultures, and most disposed to cultural variation and elaboration. None predicts the cognitive peculiarities of religion, such as:
Why do agent concepts predominate in religion?
Why do supernatural-agent concepts are culturally universal?
Why are some supernatural agent concepts inherently better candidates for cultural selection than others?
Why is it necessary, and how it is possible, to validate belief in supernatural agent concepts that are logically and factually inscrutable?
How is it possible to prevent people from deciding that the existing moral order is simply wrong or arbitrary and from defecting from the social consensus through denial, dismissal or deception?
Our argument does not entail that religious beliefs and practices cannot perform social functions, or that the successful performance of such functions does not contribute to the survival and spread of religious traditions. Indeed, there is substantial evidence that religious beliefs and practices often alleviate potentially dysfunctional stress and anxiety (Ben-Amos 1994, Worthington et al. 1996) and maintain social cohesion in the face of real or perceived conflict (Allport 1956, Pyszczynski et al. 1999). It does imply that social functions are not evolutionarily responsible for the cognitive structure and cultural recurrence of religion. This article addresses these and related issues with cross-cultural experiments and observations.
2. The Supernatural Agent: Hair-Triggered Folkpsychology
Religions invariably center on supernatural agent concepts, such as gods, goblins, angels, ancestor spirits, jinns. In this section, we concentrate on the concept of agency, a central player in what cognitive and developmental psychologists refer to as “folkpsychology” and the “theory of mind.” agency, we speculate, evolved hair-triggered in humans to respond “automatically” under conditions of uncertainty to potential threats (and opportunities) by intelligent predators (and protectors). From this perspective, agency is a sort of "Innate Releasing Mechanism" (Tinbergen 1951) whose proper evolutionary domain encompasses animate objects but which inadvertently extends to moving dots on computer screens, voices in the wind, faces in the clouds, and virtually any complex design or uncertain circumstance of unknown origin This insight into the supernatural as the by-product of a hair-triggered agency detector was first elaborated by Guthrie (Guthrie 1993; cf. Hume 1957). We further ground it in the emerging theory of folkpsychology.
A number of experiments show that children and adults spontaneously interpret the contingent movements of dots and geometrical forms on a screen as interacting agents who have distinct goals and internal motivations for reaching those goals (Heider & Simmel 1944, Premack & Premack 1995, Bloom & Veres 1999, Csibra et al. 1999). Such a biologically-prepared, or “modular,” processing program would provide a rapid and economical reaction to a wide – but not unlimited – range of stimuli that would have been statistically associated with the presence of agents in ancestral environments. Mistakes, or “false positives,” would usually carry little cost, whereas a true response could provide the margin of survival (Seligman 1971, Geary & Huffman 2002).
Our brains, it seems, are trip-wired to spot lurkers (and to seek
protectors) where conditions of uncertainty prevail (when startled, at night,
in unfamiliar environments, during sudden catastrophe, in the face of solitude,
illness, or prospects of death, etc.). Plausibly, the most dangerous and
deceptive predator for the genus Homo since the Late Pleistocene has
been Homo itself, which may have engaged in a spiraling behavioral and
cognitive arms race of individual and group conflicts (Alexander 1989). Given
the constant menace of enemies within and without, concealment, deception and
the ability to generate and recognize false beliefs in others would favor
survival. In potentially dangerous or uncertain circumstances, it would be best
to anticipate and fear the worst of all likely possibilities: presence of a
deviously intelligent predator. How else could humans have managed to
constitute and survive such deadly competitive groups as the Iatmul
All the Nāga tribes are, on occasion, head-hunters, and shrink from no treachery in securing these ghastly trophies. Any head counts, be it that of a man, woman, or child, and entitles the man who takes it to wear certain ornaments according to the custom of the tribe or village. Most heads are taken... not in a fair fight, but by methods most treacherous. As common a method as any was for a man to lurk about the water Ghāt of a hostile village, and kill the first woman or child who came to draw water…. Every tribe, almost every village is at war with its neighbour, and no Nāga of these parts dare leave the territory of his tribe without the probability that his life will be the penalty. (Crooke 1907:41-43)
Throughout the world, societies cast
their enemies as physically or mentally warped supernatural beings, or at least
in league with the supernatural. Originally, nāga “applied to
dreaded mountain tribes, and [was] subsequently used to designate monsters generally”
(Werner 1961:284). The dragons of ancient
From an evolutionary perspective, it’s better to be safe than
sorry regarding the detection of agency under conditions of uncertainty. This
cognitive proclivity would favor emergence of malevolent deities in all
cultures, just as the countervailing Darwinian propensity to attach to
protective caregivers would favor the apparition of benevolent deities. Thus, for
the Carajá Indians of Central Brazil, intimidating or unsure regions of the
local ecology are religiously avoided: “The earth and underworld are inhabited
by supernaturals…. There are two kinds. Many are amiable and beautiful beings
who have friendly relations with humans…. The others are ugly and dangerous
monsters who cannot be placated. Their woods are avoided and nobody fishes in
their pools (Lipkind 1940:249).” Nearly identical descriptions of supernaturals
can be found in ethnographic reports throughout the
In addition, humans conceptually create information to mimic and manipulate conditions in ancestral environments that originally produced and triggered our evolved cognitive and emotional dispositions (Sperber 1996). Humans habitually “fool” their own innate releasing programs, as when people become sexually aroused by make-up (which artificially highlights sexually appealing characteristics), fabricated perfumes or undulating lines drawn on paper or dots arranged on a computer screen, that is, pornographic pictures. Indeed, much of human culture – for better or worse - can be arguably attributed to focused stimulations and manipulations of our species’ innate proclivities.
These manipulations can activate and play upon several different cognitive and emotional faculties at once. Thus, masks employ stimuli that trigger our innate, hyperactive facial-recognition schema. Masks also employ stimuli that activate, amplify and confound emotions by highlighting, exaggerating or combining certain facial expressions. Moreover, like two-dimensional drawings of the Nekker cube for which there is no stable three-dimensional interpretation, masks can produce feelings of unresolved anxiety or “uncanniness.” In many religious ceremonies, for example, as a mask rotates away (e.g., clockwise) from an onlooker, who now gazes on the mask’s hollow back, the onlooker perceives a three-dimensional face emerging in the other direction (counterclockwise) from inside the back of the mask (cf. Dawkins 1998). Such manipulations can serve cultural ends far removed from the ancestral adaptive tasks that originally gave rise to those cognitive and emotional faculties triggered, although manipulations for religion often centrally involve the collective engagement of existential desires (e.g., wanting security) and anxieties (e.g., fearing death).
Recently, numbers of devout American Catholics eyed the image of
Mother Theresa in a cinnamon bun sold at a shop in
In sum, supernatural agents are readily conjured up because natural selection has trip-wired cognitive schema for agency detection in the face of uncertainty. Uncertainty is omnipresent; so, too, the hair-triggering of an agency-detection mechanism that readily promotes supernatural interpretation and is susceptible to various forms of cultural manipulation. Cultural manipulation of this modular mechanism and priming facilitate and direct the process. Because the phenomena created readily activate intuitively given modular processes, they are more likely to survive transmission from mind to mind under a wide range of different environments and learning conditions than entities and information that are harder to process (Atran 1998, 2001). As a result, they are more likely to become enduring aspects of human cultures, such as belief in the supernatural.
3. Counterintuitive Worlds
In this section we unpack the idea of the supernatural as a counterintuitive world that is not merely counterfactual in the sense of physically implausible or nonexistent. Rather, the supernatural literally lacks truth conditions. A counterintuitive thought or statement can take the surface form of a proposition (e.g., “Omnipotence [i.e., God] is insubstantial”) but the structure of human semantics is such that no specific meaning can be given to the expression and no specific inferences generated from it (or, equivalently, any and all meanings and inferences can be attached to the expression). The meanings and inferences associated with the subject (omnipotence = physical power) of a counterintuitive expression contradict those associated with the predicate (insubstantial = lack of physical substance), as in the expressions “the bachelor is married” or “the deceased is alive.”
All the world’s cultures have religious myths that are attention-arresting because they are counterintuitive. Still, people in all cultures also recognize that such beliefs are counterintuitive, whether or not they are religious believers (Atran 1996). In our society, for example, Catholics and non-Catholics alike are unquestionably aware of the difference between Christ’s body and ordinary wafers, or between Christ’s blood and ordinary wine. Likewise, Native American Cowlitz are well aware of the difference between the deity Coyote and everyday coyotes, or between Old Man Wild Cherry Bark and ordinary wild cherry bark (Jacobs 1934:126-133).
Religious beliefs are counterintuitive because they violate what studies in cognitive anthropology and developmental psychology indicate are universal expectations about the world’s everyday structure, including such basic categories of “intuitive ontology” (i.e., the ordinary ontology of the everyday world that is built into the language learner’s semantic system) as person, animal, plant and substance (Atran 1989). They are generally inconsistent with fact-based knowledge, though not randomly. Beliefs about invisible creatures who transform themselves at will or who perceive events that are distant in time or space flatly contradict factual assumptions about physical, biological and psychological phenomena (Atran & Sperber 1991). Consequently, these beliefs more likely will be retained and transmitted in a population than random departures from common sense, and thus become part of the group’s culture. Insofar as category violations shake basic notions of ontology they are attention-arresting, hence memorable. But only if the resultant impossible worlds remain bridged to the everyday world can information be stored, evoked and transmitted.
As a result, religious concepts need little in the way of overt cultural representation or instruction to be learned and transmitted. A few fragmentary narrative descriptions or episodes suffice to mobilize an enormously rich network of implicit background beliefs (Boyer 1994). For instance, if God is explicitly described as being jealous and able to move mountains, He is therefore implicitly known to have other emotions, such as anger and joy, and other powers, such as the ability to see and touch mountains or to lift and sight most anything smaller than a mountain, such as a person, pot, pig or pea.
Invocation of supernatural agents implicates two cognitive aspects of religious belief:
(i) activation of naturally-selected conceptual modules, and
(ii) failed assignment to universal categories of ordinary ontology.
Conceptual modules are activated by stimuli that fall into a few intuitive knowledge domains, including: folkmechanics (object boundaries and movements), folkbiology (species configurations and relationships), and folkpsychology (interactive and goal-directed behavior). Ordinary ontological categories are generated by further, more specific activation of conceptual modules. Among the universal categories of ordinary ontology are: person, animal, plant, substance.
To give an example, sudden movement of an object stirred by the wind may trigger the agent-detection system that operates over the domain of folkpsychology, and a ghost invoked to interpret this possibly purposeful event. In normal circumstances, a sudden movement of wind might activate cognitive processing for agents, but would soon deactivate upon further analysis ("it's only the wind"). But in the case of (bodiless) supernatural agents, the object-boundary detectors that operate over the domain of folkmechanics, and which are required to identify the agent, cannot be activated. The same cognitive conditions operate when supernatural beings and events, like ghosts or gods, are evoked in religious ceremonies, whether or not there is any actual triggering event (e.g., a sudden movement of unknown origin or other uncertain happening). In such cases, assignment to the person or animal category cannot be completed because ghosts and gods have counterintuitive properties (e.g., movements and emotions without physical bodies). This results in a potentially endless, open-textured evocation of possible meanings and inferences to interpret the event; however, the process can be provisionally stopped, and the semantic content somewhat specified, in a given context (e.g., a Sunday sermon that fixes interpretation of a Biblical passage on some particular community event in the preceding week).
Ordinary ontological categories always involve more specific processing over the folkmechanics domain (nonliving objects and events).
- Only substance involves further processing that is exclusive to folkmechanics.
- plant involves additional processing over the folkbiological domain (every organism is assigned to one and only one folk species).
- animal involves supplemental processing over the domains of folkbiology (every animal is assigned uniquely to a folk species) as well as folkpsychology (animal behavior is scrutinized as indicating predator or prey, and possibly friend or foe).
- person involves more specific processing over the folkpsychological domain (human behavior is scrutinized as indicating friend or foe, and possibly predator or prey) and the folkbiological domain (essentialized group assignments, like race and ethnicity).
The relationship between conceptual modules and ontological categories is represented as a matrix in Table 1. Changing the intuitive relationship expressed in any cell generates what Boyer (2000) calls a “minimal counterintuition” (cf. Barrett 2000). For example, switching the cell ( - folkpsychology, substance) to ( + folkpsychology, substance) yields a thinking talisman, whereas switching ( + folkpsychology, person) to ( - folkpsychology, person) yields an unthinking zombie.
These are general, but not exclusive, conditions on supernatural beings and events. Intervening perceptual, contextual or psycho-thematic factors, however, can change the odds. Thus, certain natural substances - mountains, seas, clouds, sun, moon, planets – are associated with perceptions of great size or distance, and with conceptions of grandeur and continuous or recurring duration. They are, as Freud surmised, psychologically privileged objects for focusing the thoughts and emotions evoked by existential anxieties like death and eternity. Imaginary or actual violation of fundamental social norms also readily lends itself to religious interpretation (e.g., ritual incest, fratricide, status reversal).
Finally, supernatural agent concepts tend to be emotionally powerful because they trigger evolutionary survival templates. This also makes them attention-arresting and memorable. For example, an all-knowing bloodthirsty deity is a better candidate for cultural survival than a do-nothing deity, however omniscient. In the next part, we address some of the cognitive processes that contribute to the cultural survival of supernatural beliefs.
4. Cultural Survival: A Memory Experiment
Many factors are important in determining the extent to which ideas achieve a cultural level of distribution. Some are ecological, including the rate of prior exposure to an idea in a population, physical as well as social facilitators and barriers to communication and imitation, and institutional structures that reinforce or suppress an idea. Other factors are psychological, including the cognitive and emotional ease with which an idea can be accommodated, represented and remembered, the intrinsic interest that it evokes in people so that it is processed and rehearsed, and motivation and facility to communicate the idea to others.
One complex of psychological factors concerns the apparent sensitivity to religious ideas in young children. Studies of American and European children indicate that most children through grade 1 (age 6-7) think that God is present everywhere, can hear prayers and see everything, and is near when one feels troubled or happy. This lends credence to the Jesuits’ mantra of “Give me a child till the age of seven and I’ll give you a Believer for life.” Sentiments about God’s pervasiveness in life seem to degrade with age unless institutionally supported, and God’s presence and guidance become associated more with danger and difficulties (Thun 1963, Goldman 1964, Tamminen 1994).
Of all cognitive factors, however, mnemonic power may be the single most important one at any age (Sperber 1996). In oral traditions that characterize most of human cultures throughout history, an idea that is not memorable cannot be transmitted and cannot achieve cultural success (Rubin 1995). Moreover, even if two ideas pass a minimal test of memorability, a more memorable idea has a transmission advantage over a less memorable one (all else being equal). This advantage, even if small at the start, accumulates from generation to generation of transmission leading to massive differences in cultural success at the end.
One of the earliest accounts of the effects of memorability on
transmission of natural and nonnatural concepts was Bartlett’s (1932) study of how British university students remembered,
and then transmitted a culturally unfamiliar story (a Native American folk
tale). Over successive retellings of the story, some culturally unfamiliar
items or events were dropped. Other unfamiliar items were distorted, being
replaced by more familiar items (e.g., a canoe replaced by a rowboat).
Recent studies, however, suggest that under some conditions counterintuitive beliefs are better recalled relative to intuitive beliefs (Boyer & Ramble 2001). Barrett & Nyhof (2001) asked people to remember and retell Native American folk tales containing natural as well as nonnatural events or objects. Content analysis showed that participants remembered 92% of minimally counterintuitive items, but only 71% of intuitive items.
Although suggestive, these studies leave several issues unresolved. For one: why don’t minimally counterintuitive concepts occupy most of the narrative structure of religions, folktales and myths? Even casual perusal of culturally successful materials, like the Bible, Hindu Veda or Maya Popul Vuh, suggests that counterintuitive concepts and occurrences are a minority. The Bible is a succession of mundane events - walking, eating, sleeping, dreaming, copulating, dying, marrying, fighting, suffering storms and drought - interspersed with a few counterintuitive occurrences, such as miracles and appearances of supernatural agents like God, angels, and ghosts.
An answer to this puzzle may lie in examining memorability for an
entire set of beliefs taken as a single unit of transmission, rather than
individual beliefs. Accordingly, we conducted a study to examine the
memorability of intuitive (INT) and minimally counterintuitive (MCI) beliefs
and belief sets over a period of a week (for examples, see Table 2). One group of 44
We replicated this finding with a different set of ideas, where a
sharper distinction was made between counterintuitive ideas and ideas that are
intuitive but bizarre, and between degrees of counterintuitiveness. 107
Intuitive ideas enjoyed highest recall; maximally co