The Minds of Gods: a Comparative Study of Supernatural Agency

Cognition 129 (2013) 163–179

Contents lists available at SciVerse ScienceDirect

Cognition

journal homepage: www.elsevier.com/locate/COGNIT

The minds of gods: A comparative study of supernatural agency

Benjamin Grant Purzycki

Centre for Human Evolution, Cognition, and Culture, University of British Columbia, 1871 West Mall, Vancouver, British Columbia V6T 1Z2, Canada article info abstract

Article history: The present work is the first study to systematically compare the minds of gods by exam- Received 11 October 2012 ining some of the intuitive processes that guide how people reason about them. By exam- Revised 21 June 2013 ining the Christian god and the spirit-masters of the Tyva Republic, it first confirms that the Accepted 24 June 2013 consensus view of the Christian god’s mind is one of omniscience with acute concern for interpersonal social behavior (i.e., moral behaviors) and that Tyvan spirit-masters are not as readily attributed with knowledge or concern of moral information. Then, it reports Keywords: evidence of a moralization bias of gods’ minds; American Christians who believe that God Religion is omniscient rate God as more knowledgeable of moral behaviors than nonmoral informa- Minds of gods Supernatural Punishment Hypothesis tion. Additionally, Tyvans who do not readily report pro- or antisocial behavior among the Omniscience things that spirit-masters care about will nevertheless rate spirit-masters’ knowledge and Cultural consensus analysis concern of moral information higher than nonmoral information. However, this knowledge Theological correctness is distributed spatially; the farther away from spirits’ place of governance a moral behavior takes place, the less they know and care about it. Finally, the wider the breadth of knowledge Tyvans attribute to spirit-masters, the more they attribute moral concern for behaviors that transpire beyond their jurisdiction. These results further demonstrate that there is a significant gulf between expressed beliefs and intuitive religious cognition and provides evidence for a moralization bias of gods’ minds. Ó 2013 Elsevier B.V. All rights reserved.

1. Introduction the world are especially knowledgeable and concerned with human behavior (Boyer, 2001; Johnson, 2005; Pet- For a long time, researchers of religion have argued that tazzoni, 1955), but the speciﬁc domains of concern people religion functions to maintain and promote cooperation explicitly attribute to gods appear to vary from place to between individuals (Durkheim, 2001[1915]; Malinowski, place. As far as we know, these domains are limited to 1932; Rappaport, 1979, 1999). Contemporary experimen- interpersonal social behavior, ritual, environmental re- tal evidence suggests that religious priming can boost source management, and etiquette (Purzycki, 2012; Pur- cooperative behavior and can minimize defection on social zycki & Sosis, 2011). However, while religion appears to rules (Bering, McLeod, & Shackelford, 2005; Johnson, 2005; positively mediate interpersonal social behavior and curb Johnson & Bering, 2006; Piazza, Bering, & Ingram, 2011; antisocial behavior, we currently have very little in the Randolph-Seng & Nielsen, 2007; Shariff & Norenzayan, way of understanding the cognitive foundations for these 2007). Behavioral and cross-cultural studies also suggest effects. If religion heightens levels of prosociality, do gods that participation in religious rituals reliably conveys com- without explicit concern for interpersonal social behaviors mitment to the group and contributes to overcoming prob- (e.g., theft, generosity, murder, honesty, etc.) still prime lems inherent to human cooperation (Soler, 2012; Sosis & moral systems and domains? Bressler, 2003; Sosis & Rufﬂe, 2003, 2004; Sosis, Kress, & The cognitive science of religion literature regularly Boster, 2007). Many have also observed that gods around claims that the ability to mentalize is the essence of religious cognition; our commitment to supernatural beings requires the ability to make sense of other minds (Atran, E-mail address: [email protected]

2002, pp. 59–63; Barrett, 2004; Bering, 2011; Guthrie, focuses on the content (i.e., representational domains) of 1980, 1995). However, remarkably little has been done to gods’ minds. empirically investigate the cross-cultural similarities and Boyer (2001, p. 144) observes that people are far more variation in representations of gods’ minds and how this interested in the minds of their gods rather than other fea- relates to religious prosociality. As such, many questions tures and posits that people are more likely to endow remain unanswered. For example, do people have a predis- supernatural agents with access to domains of ‘‘socially position towards associating interpersonal social behaviors strategic information’’ (Boyer, 2000, 2002). He defines this such as theft and generosity, deceit and honesty, cruelty as ‘‘the subset of all the information...that activates the and kindness with supernatural agents? If present, does mental systems that regulate social interaction’’ (2001, p. this predisposition operate even in contexts where super- 152). Bering (2011) defines ‘‘strategic social information’’ natural agents are not explicitly concerned about human as ‘‘any information that, once exposed, could influence morality? Do people around the world draw upon different one’s reproductive success’’ (p. 174). Having access to this cognitive systems to think about the knowledge and con- reputational information is often held to be the primary cerns of their gods? The present study initiates the formal distinction between gods and other counterintuitive examination of these questions by systematically investi- agents (Atran, 2002; Barrett, 2008; Boyer, 2001, 2002); gating how people from two different cultures—American gods who care about such information are more salient Christian students and Buddhist-animist Tyvans from to our minds than other agents without such abilities southern Siberia—make sense of their gods’ minds. (e.g., cartoons, mythological creatures). These theorists of- The following sections review current thinking and evi- ten explain retention of and commitment to supernatural dence on how and why it is that humans commit to gods. agent concepts by appealing to their cognitive salience, Following this is a discussion of the important distinction but very little work has been done to get a reliable account between ‘‘theologically correct’’ beliefs and how intuitive of the variation in attributed domains of concern to gods. cognitive processes often betray culturally specific postu- However, one thread of research suggests that the links be- lates. I then review some of the theological discussions of tween supernatural agents, their minds, and socially stra- the Abrahamic god’s mind followed by a brief description tegic knowledge function to mediate social behavior and of religion in the Tyva Republic. Drawing from this frame- thus contribute to the evolution of cooperation. work, I present hypotheses and their tests regarding how people ought to reason about their supernatural agents 1.2. Gods’ minds and conclude by illustrating future directions in the study of gods’ minds. 1.2.1. Supernatural punishment and cooperation Some evolutionary theories culminate into what is known as the Supernatural Punishment Hypothesis (Sch- 1.1. Representing supernatural minds loss & Murray, 2011). One strand of this hypothesis (Ber- ing & Johnson, 2005; Johnson, 2005) suggests that a Recently, the question of why humans are religious has psychological commitment to punishing deities is a part become reinvigorated by the cognitive and evolutionary of our evolved repertoire of cognitive systems. Another sciences of religion. While cognitive approaches attempt (Norenzayan & Shariff, 2008; Shariff & Norenzayan, to unravel the proximate mechanisms at work in religious 2007) posits that because of the prosocial effects of pun- cognition, evolutionary approaches seek to determine ishing supernatural agents, such concepts are more preva- whether or not religion promotes survival and lent in large populations. Together, this research shows reproduction. that supernatural agents often function to inhibit self- Many argue that the essence of religious cognition is the interested behavior or free-riding, and this strengthens representation of supernatural minds, an ability stemming cooperation which significantly contributes to individual from our normal, everyday theory of mind systems (see well-being. Baron-Cohen, 1995; Premack & Woodruff, 1978). In his People are more likely to engage in prosocial behavior foundational work, Guthrie (1980, 1995) argues that or suppress antisocial behavior when primed with super- anthropomorphism—granting human-like characteristics natural agent concepts (Bering & Johnson, 2005; Bering to non-human entities—is a central feature of religion. et al., 2005; Johnson, 2005; Johnson & Bering, 2006; Piazza, Anthropomorphism includes human-like forms, and attrib- Bering, & Ingram, 2011; Randolph-Seng & Nielsen, 2007; uting comprehension of symbolic thought and human Rossano, 2007; Shariff & Norenzayan, 2007). We also see motivations to non-humans. Barrett (2004) and Keil (Bar- similar effects with the priming of secular justice concepts rett & Keil, 1996) suggest that humans are equipped with (Shariff & Norenzayan, 2007) as well as mere indicators of a ‘‘hyperactive agency detection device’’ (HADD) that rap- agency such as a picture of two eyes (Haley & Fessler, idly attributes mental states to events and objects lacking 2005; cf. Nettle et al., 2013). Importantly, explicit models agency. This contributes to our religious inclinations inso- of whether God is primarily forgiving or punishing affect far as we frequently appeal to gods’ mental states to ex- behavior in economic games (Shariff & Norenzayan, plain a host of phenomena. Recently, Norenzayan, 2011). While the vast majority of these studies have been Gervais, and Trzesniewski (2012) found that those who conducted among Westerners in the Abrahamic tradition, suffer from deficiencies in theory of mind are less likely Xygalatas (2013) found that among Hindu Mauritians, reli- to believe in God. While these works address the innate gious location primes also boost cooperation. Together, cognitive machinery behind mentalizing gods, other work these works suggest that god and other religious concepts B.G. Purzycki / Cognition 129 (2013) 163–179 165 engage theory of mind and regulatory systems that signif- their explicitly held concerns. If there is such a bias, it icantly alter social behavior. If this is the case, there should would suggest that gods play a role in mediating interper- be a cognitive bias towards moralizing the content of our sonal social behavior regardless of explicit cultural beliefs. gods’ minds. However, not all gods care about human virtue or interpersonal social behaviors like cooperation, gen- 1.2.3. Gods and moral concern erosity, cheating, and so forth. In other words, not all gods Anthropologists and other social scientists have demon- care about ‘‘morality’’. strated that people in highly specialized state-level societies primarily believe in explicitly moralistic high gods (Johnson, 2005; Lahti, 2009; Roes & Raymond, 2003; San- 1.2.2. Moral models, systems, and interpersonal social derson, 2008; Stark, 2001; Swanson, 1960; Wallace, behavior 1966). Because of the heightened anonymity and unac- Of course, what is explicitly ‘‘moral’’ for one population countability fostered by population density, the argument is not necessarily moral for another (Shweder, Mahapatra, goes, moralistic deities function to curb antisocial behavior & Miller, 1987; Shweder, Much, Mahapatra, & Park, 1997), and promote cooperative behavior. These gods tend to just as what might be ‘‘socially strategic’’ might vary cross- know everything in order to see antisocial behaviors other- culturally. Researchers continue to struggle with how to wise not witnessed by other humans. As such, they func- properly define morality (see Haidt, 2008; Rossano, 2008 tion as powerful appeals to influence others’ behavior. for reviews). Some consider morality to be prescriptive However, while many traditional societies appear to have and generalizable models of welfare and rights (Turiel, moralistic (Boehm, 2008) and omniscient deities (Pettazz- 1983, p. 3; Smetana, 2006, p. 121). Rossano (2008), for in- oni, 1955) as well, just how widespread such beliefs are/ stance, defines morality as the ‘‘cultivation and application were within such societies remains unknown. of virtue’’, thus emphasizing the connection between mod- We do know, however, that not all supernatural agents els of what it means to be virtuous and how they are en- are explicitly concerned with interpersonal social behav- acted. Haidt (2008), however, defines moral systems as ior; in many traditions, gods appear to be primarily con- ‘‘interlocking sets of values, practices, institutions, and cerned with ritual behavior, resource management, and/ evolved psychological mechanisms that work together to or nonmoral taboos and etiquette (see Atran et al., 2002; suppress or regulate selfishness and make social life possi- Lansing, 2007; Lansing & Kremer, 1993; see Purzycki, ble’’ (70). 2011, 2012, pp. 232–252; Purzycki & Sosis, 2011 for further As mentioned earlier, supernatural agents seem to be discussion) or appear to be altogether unconcerned with concerned with: (a) interpersonal social behavior (i.e., humans. Yet, Boehm (2008) analyzed 18 ethnographic ac- how we behave toward one another with a cost or benefit counts of foraging societies and found that while there is to others), (b) resource management (i.e., how we behave significant variation in the particular behaviors supernatu- toward resources that others might be able to use), (c) non- ral agents sanction, each of these societies have concepts of moral etiquette (i.e., how we behave in ways that identify the supernatural punishment of immoral behavior. There us as a socially appropriate person), and (d) ritual behavior are many other ethnographic accounts of small-scale soci- (i.e., how we behave towards the gods). These can all be eties with moralistic supernatural agents. For example, construed as moral or ‘‘socially strategic’’, depending on according to some accounts, the ancestor-spirits of the one’s semantic flexibility.1 Ju/hoansi get angry and kill people when they are ‘‘quarrel- The present work uses ‘‘moral’’ merely as shorthand for some and unpleasant to other people’’ (Lee, 2003, p. 129; a conceptual domain of interpersonal social behavior with cf. Marshall, 1962). The Nuer’s god ‘‘sees and hears all that an immediate cost or benefit to other people (e.g., theft, happens and he can be angry and can love’’ (Evans-Prit- murder, harm, generosity, kindness, assistance, etc.) and chard, 1956, p. 7). So, if we assume that such accounts ‘‘moralistic deities’’ as gods who care about such behaviors. are accurate and representative of these populations, even It does not rely on any particular definition of morality, in small-scale societies, moral concern and/or omniscience however, and I use it in this fashion to avoid cumbersome can be central features of gods’ minds. redundancies. Rather, the present concerns focus on How intuitive, then, are the omniscience and moral whether or not there are cognitive biases towards endow- concern we attribute to supernatural agents? It appears a ing supernatural agents with such domains of socially stra- priori that most people around the world assume and/or tegic knowledge even when such knowledge is not among believe that gods know more than humans, even if they are not thought of as being absolutely all-knowing. And 1 Even though ‘‘culture’’ may play an important role in the formation of while it might be effortless for Westerners to claim that moral models and reasoning (Sachdeva, Singh, & Medin, 2011), presumably all gods are ‘‘moralistic’’, people explicitly endow their all populations have various conceptual domains approximating to bene- gods with different concerns. So, in order to examine the ficial and harmful interpersonal social behaviors, regardless of whether or question of the intuitiveness of omniscience and moralism, not they are appealed to in the domain of religion (Brown, 1991, p. 139). Generalizable norms, interpersonal moral behaviors, information about we must turn to the distinction between culturally ‘‘cor- rituals, nonmoral taboos, how to behave in specific contexts, and other rect’’ beliefs and intuitive cognitive systems. It may be forms of etiquette may exist everywhere even though their content may the case that a moralization bias of gods’ minds exists in vary superficially across populations. These domains can overlap and/or all populations, but reflective cultural models may override nest within one another. So, in one context, representational models of or harness this bias in particular contexts. Much like when what it means to be a good or bad person may entirely overlap with models of what a god cares about (e.g., in the Abrahamic tradition), whereas in we say that the sun rises even though we know it does not, other contexts, they may not (e.g., in the Tyvan case). culturally correct models of the world’s entities are often 166 B.G. Purzycki / Cognition 129 (2013) 163–179 inconsistent with our intuitions about those entities’ prop- ‘‘God is everywhere’’, people will also readily say that ‘‘God erties (Barrett, 1999, p. 325). came down from heaven’’ thereby limiting his form to the physical by virtue of cognitive systems devoted to folk- physics. Such conceptions are, therefore, ‘‘theologically 1.3. Theological and cultural ‘‘correctness’’ incorrect’’ or inconsistent insofar as they tap into ‘‘online’’, intuitive faculties designed for rapid, everyday computa- Sperber (1997) distinguishes between what he calls tions. These computations produce outputs with content ‘‘reflective’’ and ‘‘nonreflective’’ beliefs (see too McCauley, that is inconsistent with more abstract beliefs.2 2011; Slone, 2004). Reflective beliefs are those generated Does this distinction relate to how we make sense of by way of conscious weight assignment and distribution, our gods’ minds? The idea that a god knows anything is debate, and deliberation (Barrett, 2004, p. 7). On the other indicative of a nonreflective, intuitive process; we intui- hand, we generate nonreflective or intuitive beliefs with tively assume gods have minds by virtue of evolved theory little to no conscious manipulation: ‘‘All beliefs that are of mind systems. However, by comparison, the idea that the output of perceptual processes are intuitive in a stan- ‘‘God knows everything’’ is a reflective, theologically cor- dard psychological sense, and so are all beliefs that are rect religious belief about the Abrahamic god. One can con- the output of spontaneous and unconscious inferential template its implications, bring it forth to consciousness processes taking intuitive beliefs as premises’’ (Sperber, when asked, critically examine and question its veracity, 1997, p. 78). Barrett (2004) discusses how this distinction and so forth. However, theological correctness is a matter characterizes the logical and factual inconsistencies people of how people’s conscious, reflective beliefs correspond frequently entertain about their deities. In other words, na- to the authoritative and/or shared models held by a partic- tive intuitions often get in the way of what we are ‘‘sup- ular tradition (Barrett, 1999); active computational pro- posed’’ to say. cesses produce different results from the stored doctrinal Regarding reflective beliefs, Barrett (1998) and Keil ideas we retain. If a tradition does not have a written (Barrett & Keil, 1996) reason that ‘‘If subjects were asked source for their ‘‘abstract theology’’, laypeople often appeal directly what they believed about God, responses would to the authority of religious leaders and popular cultural tend to fit into an abstract theology’’ (223). In other words, models of tradition. Theological incorrectness, then, is a upon reflection, people will explicitly state semantically matter of how cognitive biases systematically betray the stored religious postulates that are consistent with theo- content of cultural models. In the case of gods’ minds, logical teachings. However, people often recall narratives one such bias may be knowledge of and concern for so- that are inconsistent with explicit, theologically correct be- cially strategic behavior. liefs. For example, while a narrative in one experiment makes no indication that there is a perceptual conflict for God when hearing birds singing and the roar of a jet en- 1.4. The Abrahamic god’s mind gine, one participant recalls the story as: ‘‘God was listen- ing to two birds singing in a tall tree next to an airport. People often say that the Abrahamic god knows every- When a large jet landed, God listened to it because he thing. There are numerous references in the Bible (e.g., could no longer hear the birds. Then he listened to the Matthew 10:30–31; Luke 12:4–7; Psalms 147:5) and the birds again’’ (Barrett & Keil, 1996, p. 239). Even though par- Quran (Surat Al-‘An‘a¯m 6:59; Al-Nur 24:35) to God’s omni- ticipants say that God perceives all, they nevertheless vio- science. There are also many classical discussions about late this when relying on intuitive processes about normal gods’ knowledge of everything. For example, Homer’s humans’ perceptual abilities. Likewise, while some say that Odyssey (7th century BCE) has many tales of gods hearing the prayers of Odysseus. Much later, Saint Augustine (Book 2 Biases resulting in theological inconsistency quite likely come from 5 Section 9) critically examined Cicero’s position, arguing different cognitive resources. While we may have cognitive modules that that ‘‘to confess that God exists, and at the same time to rapidly compute specific domains of information (see Anderson, 2007; deny that He has foreknowledge of future things, is the Fodor, 1983), we also have clusters of information stored as cultural or computational ‘‘schemas’’ (see D’Andrade, 1995, pp. 122–149; Strauss & most manifest folly’’ (Dods, 1888, p. 190). Also, the logical Quinn, 1997). In the case of the former, as already discussed, we may think consistencies of God’s omniscience and humans’ free-will of God has having a physical body with perceptual limitations by virtue of have also been a topic of serious debate among theologians our default inferences about agents. In the case of the latter, we might have and philosophers (Abbruzzese, 1997; Grim, 1983; Hughes, a bias towards thinking of the Abrahamic god as having a beard, for 1995, pp. 64–107; Kapitan, 1991; Kretzmann, 1966). instance, but this is likely because of the regularity of our exposure to such images. Individuals might think this, but do not believe in it and Moreover, within the Abrahamic holy books, there are understand that they might not be supposed to believe in it. Such an numerous references to God’s punitive stance with promises inconsistency stems from our schematic prototypes of God. Readers may of both reward and punishment (Hebrews 12:6; 1 John 5:16– refer to Purzycki (2012, pp. 40–87) for some of my thoughts on the 17; Surat Al-Ma¯’idah 5:38; Surat An-Nisa¯’ 4:40; Thessalo- relationship between intuitive or modular inferences and explicit ‘‘cultural’’ assumptions. For the present work, I assume ‘‘culture’’ is, at the very least, nians 1:5–9; Psalm 73). Again, this stance is not limited to explicit schematic knowledge whereas ‘‘intuitive’’ refers to deeper faculties, the deities of the world religions (Pettazzoni, 1955) and its but the results and any interpretations are not affected by such assump- ubiquity suggests something fundamental about the way tions. As such, an important unanswered but immediately relevant the religious mind works; supernatural agent concepts may question to the present work is if supernatural agents trigger evolved engage evolved social monitoring modules (see Cosmides & moral systems, merely tap into conceptual schematic domains of morality, a combination of both, or neither (and perhaps these are all different ways Tooby, 1989, 1992; Purzycki et al., 2012; Sugiyama, Tooby, of saying the same thing). I discuss this briefly in Section 4. & Cosmides, 2002) and/or Haidt’s (2008) aforementioned B.G. Purzycki / Cognition 129 (2013) 163–179 167

‘‘moral systems’’ precisely because religion evolved to over- As commonly found throughout Inner Asia, due to the come challenges to human cooperation. Manichean-influenced Uighurs, the Tyvan shamanic Purzycki et al. (2012) tested this prediction in a re- worldview is a tripartite model (Upper, Middle, and Lower sponse-time task. We found that Christians who believe Worlds; see Eliade, 1964; Khomushku, 2008; Purzycki, that God knows everything answered questions about 2012, pp. 214–230). Some supernatural agents in this re- God’s knowledge of negative moral behavior (e.g., Does gion’s traditional pantheon appear to be moralistic, but God know that John cheats on his taxes?) significantly reliable data on this is lacking. For example, the lord of quicker than positive moral behaviors. Moreover, positive the Lower World, Erlik-Khan, has been reported as moralis- moral behaviors (e.g., Does God know that Ann gives to tic and punitive (Khomushku, 2008; Van Deusen, 2004). the homeless?) were answered significantly quicker than However, there are many other cases where he is explicitly nonmoral questions (e.g., Does God know that Richard’s reported as neither (Eliade, 1964, p. 173; Harvilahti, 2000; cat is hungry?). We found identical results for a trial asking Heissig, 1980, p. 54), or even benevolent and helpful the same questions about a fictitious, omniscient govern- (Chadwick, 1936; Halemba, 2006, p. 46; Pegg, 2001, p. ment, but there were no such differences for a trial with 102). an omniscient, but non-interfering alien species. This Spirit-masters are of our world; the Middle world of the raised the question of whether or not agents’ explicit con- shamanic tradition. Cher eezi come in both anthropomor- cerns mediate responses to questions about knowledge; phic and zoomorphic forms and these forms correspond God cares about cheating, giving to those who need it, to resource type; anthropomorphic spirits are lords of lar- and other ‘‘moral’’ behaviors and perhaps this alters the ger regions whereas zoomorphic spirits govern particular accessibility of moral domains. discrete resources (Purzycki, in press). Tyvans regularly One way to address this would be to test whether or not make offerings to spirit-masters at designated ritual cairns a similar propensity to intuitively moralize supernatural (ovaa) and at the resource over which they lord. Cairns are agents’ knowledge is present in a context where moral typically found on territorial borders and have a close asso- information is not salient among the concerns attributed ciation with place among Inner Asians (Humphrey & Onon, to spirits. If a population’s gods explicitly care about things 1996; Kristensen, 2004; Vreeland, 1953). Purzycki and other than wickedness and kindness, but members of that Arakchaa (2013) found that participation in cairn rituals society exhibit systematic inclinations towards moralizing conveys trustworthiness to other Tyvans and this effect is their gods’ minds, then this would support the prediction likely to be found elsewhere with this tradition. that humans intuitively attribute supernatural agents with In general, however, spirit-masters are not acutely con- socially strategic information even if that information is in a cerned with interpersonal moral behavior or human vir- domain that is not the culturally correct domain of gods’ tues like trustworthiness; they are pleased by rituals and attributed concerns. If this were the case, it would suggest angered primarily by people ruining, soiling, or over- that supernatural agent concepts—regardless of their ex- exploiting their resources. Importantly, there is very little plicit concerns—correspond to underlying social monitor- overlap between Tyvans’ explicit, freelisted models of good ing systems. The present work tests this hypothesis in and bad people and what angers or pleases cher eezi; Ty- the Tyva Republic. vans rarely list common things in these domains (Purzycki, 2011, 2012, pp. 232–252). In fact, in ethnographic interviews, Tyvans often explicitly denied that spirit-masters 1.5. The Tyva Republic have concern for interpersonal social behavior (Purzycki, 2010).4 Moreover, it should be noted that Tyvans seldom The Tyva Republic (also known as ‘‘Tuva’’) lies just consider cher eezi as an actively punishing agent; while north of western Mongolia in southern Siberia. About half bad things might happen if one fails to stop and perform rit- of Tyva’s population (ca. 243,000) lives in rural villages uals at designated cairns, one’s luck is negatively affected and/or engage in transhumant pastoralism of various live- rather than the spirit angrily imposing its wrath upon non- stock. In the northeast, people herd reindeer whereas those participants (Purzycki, 2010, in press). In other words, unlike living in the forested mountains and steppe regions pri- the Abrahamic god, they are not widely held as judges or marily herd sheep, goats, and cattle. punitive agents. As such, this context represents an ideal Religion in Tyva is a complex, syncretic system that place to test predictions about latent cognitive processes incorporates elements of animism, totemism, shamanism, responsible for moral reasoning and their association with and Buddhism. There is considerable regional variation in gods’ minds. If such a bias is present, it would suggest that belief and practice as these traditions are rooted in place, religion need not be explicitly about prosociality or antiso- landscape, and local ecology. One of the older threads cial behavior to be an effective means of mediating interper- within this rich tapestry is that of the local totemic spir- sonal social behavior. it-master or cher eezi (lit. ‘‘owner’’ or ‘‘master’’ of the place, pl. cher eeleri3). Cher eezi are the supernatural lords of resources (e.g., lakes, rivers, trees, etc.) and regions (e.g., kin- 4 This does not mean that Tyvans would say ‘‘no’’ if they were asked if a based territories, regions that are off-limits to human person who participates in rituals is a ‘‘better person’’ than one who does exploitation, and political districts) (Purzycki, 2010, 2011, not or that they would deny that someone who pollutes a river is ‘‘bad’’. The in press). fluidity of ‘‘moral’’ and ‘‘socially strategic’’ can make just about anything a ‘‘moral’’ issue and it is obvious that there are moral dimensions (in Haidt and Rossano’s senses) to behavioral prescriptions. These are questions 3 For the sake of clarity, I use cher eezi as a group noun. about the degree to which conceptual domains emically share content. 168 B.G. Purzycki / Cognition 129 (2013) 163–179

1.6. Predictions 2. Culturally correct models of gods’ minds

1.6.1. Consensus models 2.1. The Abrahamic god’s mind As discussed above, our intuitive cognitive processes often interfere and run counter to explicit, reflective, and The theologically correct version of the Abrahamic god’s theologically correct beliefs. In order to generate what knowledge consists of everything whereas his concerns would approximate to the culturally or theologically cor- lean toward moral behavior. Cultural consensus models rect models of gods’ minds, I use cultural consensus analy- should indicate that God knows all manner of things—mor- sis (CCA; Batchelder & Romney, 1988; Borgatti & Halgin, al and nonmoral alike—whereas his concerns should be 2011; Romney, Weller, & Batchelder, 1986; Weller, 2007) limited to interpersonal social behaviors. to find where agreement lies within each of the two samples. CCA finds pools of agreement within samples and 2.1.1. Methods identifies individual deviations from these pools. So, a Participants were given a survey (Appendix A) that con- small subsample of people might have more agreement sisted of a number of components in the following order: in a particular domain (i.e., higher consensus). Addition- (a) basic demographic questions, (b) questions from a ally, CCA generates the ‘‘correct’’ responses to questions modified religiosity scale (Nicholas, 2004; Nicholas & Durr- based on these pools of agreement. As such, it has been heim, 1995; Rohrbaugh & Jessor, 1975), (c) one long-re- invaluable for field researchers (e.g., Boster & Johnson, sponse question asking participants to describe what God 1989; Caulkins, 2004; de Munck, Dudley, & Cardinale, is in a few statements, (d) one free-list question asking 2002; Miller, Kaneko, Bartram, Marks, & Brewer, 2004). participants to list up to 10 things about what God is, (e) As the ‘‘theologically correct’’ characterization of God’s a 50-question survey about God’s knowledge and 50 sub- mind is omniscient and moralistic, a consensus analysis questions about whether or not God cared about such should demonstrate that while God knows everything, he information, and (f) another long-response question only cares about morality. And, there should be little vari- requesting that participants summarize what they believe ation in the responses and therefore a high degree of con- God knows and cares about. sensus. However, the same questions posed to Tyvans The 50-question survey’s instructions read: ‘‘Please an- about spirit-masters should yield very little agreement as swer the following questions to the best of your ability. If morality is not among their readily-expressed concerns. you are an atheist, please answer how you think people who believe in God would answer’’. The randomized questions were designed to account for a number of domains 1.6.2. Moralization bias drawn from elsewhere (Purzycki et al., 2012). Ten ques- If moralizing gods’ minds is intuitive, people should en- tions were designed to be strange and illogical (e.g., Does dow gods with more knowledge and concern of moral God know how tall he is?; Does God know to create a information than nonmoral information. In the case of round circle?) in order to distract participants from any the Abrahamic god, responses should be consistent with streamlined responses. These responses were therefore theologically correct models of God’s mind if there is no not included in the following analyses. such bias. Alternatively, if there is a bias, then there should Twenty questions contained hypothetical moral con- be significantly more affirmative responses to God’s tent. Half of these questions consisted of behaviors that knowledge of moral information than to nonmoral infor- take place on the University of Connecticut campus and mation. In the case of the Tyva Republic, where local spir- the other half take place in various places around the it-masters are not as concerned about interpersonal world. Of these moral items, 5 were good behaviors (e.g., morality, if a moralizing bias exists, Tyvans should say that Does God know if I was honest to my friends in St. Peters- spirit-masters know moral behavior better than they know burg?; ...if I were generous to someone here at UConn?) non-moral information. Alternatively, as moral informa- and 5 were bad (e.g., ...if I quarreled with someone when tion is not among the primary concerns of spirit-masters, I was in Brazil?; ...if I hurt another person here at UConn?). there should be no differences between knowledge and Twenty questions were about mundane behaviors of no concern of moral and nonmoral information ascribed to obvious consequence. Half were conceivably knowable in these spirits. my presence (e.g., ...know that my hair is brown?; ...know what color my shirt is?) whereas the other half consists of properties knowable beyond me (e.g., ... what is growing in my yard at home?; ...what is under my bed?). Partici- 1.6.3. Omniscience, space, and moral concern pants’ answers were on a scale of 2 to À2: ‘‘Yes’’ (2), ‘‘Prob- As local spirits associated with place, cher eezi’s knowl- ably’’ (1), ‘‘I don’t know’’ (0), ‘‘I doubt it’’ (À1), and ‘‘No’’ edge base is likely to be grounded in its territory or re- (À2). Again, each item was accompanied with a sub-ques- source of mastery. If this is the case, Tyvans should use tion of whether or not God cares about such information. proximity of information to spirit-masters’ location to mediate what information spirits know. Moreover, wider breadth of knowledge explicitly attributed to spirits should 2.1.2. Participants predict higher concern for morality if there is indeed a po- Participants were recruited from an introductory sitive relationship between omniscience and concern for anthropology course at the University of Connecticut and morality. given course credit for participation. In order to isolate B.G. Purzycki / Cognition 129 (2013) 163–179 169 believers in an omniscient god, only those participants diverged from the interval model. As per the conventional who answered (a) that they believed that God is all-know- criterion of a 3-to-1 ratio of the first and second eigen- ing and all-powerful and (b) that they are sure that God values, these data show strong evidence of a single culture really exists and that He is active in their lives in prelimin- of consensus using the two approaches (Table 1; Table C3). ary questions were used in this analysis (N = 88; 41 wo- In this analysis, God is knowledgeable of all information men, one did not report sex; age M = 19.45, SD = 0.96; equally. In summary, by consensus, God knows all of the four did not report age). Sixty reported that they were information contained in the questions, but only cares Catholic and 14 reported that they were Protestant. Among about the moral information. the fill-in responses for those who circled ‘‘Other’’, 9 reported that they were Christian, two reported that they 2.2. Tyvan cher eezi’s minds were Lutheran, one reported Greek Orthodox affiliation, one reported being a ‘‘born again Christian’’, and one re- Using consensus analysis, we can determine the cultur- ported being a ‘‘non-denominational Christian’’. While ally correct responses to similar questions posed to Tyvans one did not answer the question, all others reported being about cher eezi. As such questions are inconsistent with raised in Christian households. Four items from the religi- reflective beliefs, there should be considerable variation osity scale had high intercorrelations with each other and in Tyvans’ responses. the scores from the first Varimax rotated factor were used as an index of religiosity (eigenvalue = 2.86; 71.53% of the 2.2.1. Participants variance was explained; a = 0.86). Participants (N = 88; 52 women; age M = 38.79, SD = 13.41; three did not report age) were interviewed in 2.1.3. Results the capital city of Kyzyl between the months of March Using UCINET 6 (Borgatti, Everett, & Freeman, 2002), re- and August of 2010. The reported religious affiliations sponses were analyzed with cultural consensus analysis. were: 56 Buddhist, 2 Shamanist, 26 Buddhist and Shaman- One CCA method treats the data as though there are dis- ist, two Christians, one answered ‘‘Other’’ with no elabora- crete, correct answers to the questions (àladiscrete multi- tion, and one did not answer the question. Only ethnic ple choice) whereas the other treats the data as interval Tyvans who could speak and read Tyvan well were inter- data. These data can be construed as both; if we draw viewed. All participants were non-specialists (i.e., not sha- the ‘‘correct’’ responses from some Biblical sources and mans or lamas). Survey questions were translated into other authorities, the theologically correct response should Tyvan, back-translated to English, and edited for consis- be a resounding ‘‘yes’’ for all responses and not nuanced. tency. With the help of assistants, all data were collected However, the scale itself was more nuanced in order to de- using written surveys administered on the streets, and in tect more subtle phenomena in other analyses (see below). various schools, clinics, and other institutions throughout As such, I present both analytical models (see Appendix C Kyzyl. Surveys took around 20–30 min to complete. for answer keys). Of God’s attributed concerns (Table 1), the interval 2.2.2. Methods model maintained the assumption of one pool of agree- Surveys consisted of a number of tasks. One was a free- ment as indicated by the ratio between the eigenvalue of list task designed to elicit (a) qualities of a good Tyvan per- the first and second factor of the subject-by-subject corre- son, (b) qualities of a bad Tyvan person, (c) what pleases lation matrix. The discrete model was borderline; if the and (d) angers spirit-masters (Purzycki, 2011, 2012, n.d.). largest eigenvalue is at least three times that of the next These data were compiled and analyzed in order to deter- largest, then there is—according to convention—evidence mine if there was significant overlap between models of of more agreement (Bernard, 2011, p. 375; Borgatti & Hal- good/bad Tyvan people and if these corresponded to mod- gin, 2011; Handwerker, 2001, p. 195).5 The discrete model els of spirit-masters’ concerns (Purzycki, 2012, pp. 232– shows that the correct answer to all moral items as ‘‘2’’ (yes) 252). Another component of the surveys was designed to and the correct answer to nonmoral items as ‘‘À2’’ (no) elicit attributed breadth of knowledge to spirit-masters (Table C2). In summary, consensus models suggest that (see below). God only cares about the moral information. In order to test and potentially control for effects of For the questions concerning God’s knowledge, there master type (human vs. animal) on attributions of omni- was a good fit with the consensus model even though thir- science and moralization, there were three variants of the teen individuals diverged from the discrete model and four survey (Appendix B): the introduction of one variant

Table 1 Fits to interval and multiple choice consensus models for God’s concern and knowledge.

Concern Knowledge Discrete Interval Discrete Interval No. of negative competencies 13 1 13 4 Largest eigenvalue 29.64 58.34 52.69 24.41 2nd Largest eigenvalue 12.01 3.15 11.04 3.60 Ratio of largest to next 2.47a 18.53 4.77 6.78

a Violates assumption of 3-to-1 ratio required to determine the presence of a single culture. 170 B.G. Purzycki / Cognition 129 (2013) 163–179

Table 2 Fits to interval and multiple choice consensus models for cher eezi’s concern and knowledge.

Concern Knowledge Discrete Interval Discrete Interval No. of negative competencies 50 11 36 6 Largest eigenvalue 19.19 31.24 17.42 33.66 2nd Largest eigenvalue 12.90 6.73 13.79 7.85 Ratio of largest to next 1.49a 4.64 1.26a 4.29

a Violates assumption of 3-to-1 ratio required to determine the presence of a single culture. See Table C5–C6 for answer keys.

(n = 37) read as follows: ‘‘The cher eezi of Bayan-Kol is a moral items and none of the nonmoral items (Table C5) beautiful woman on a horse. Imagine that we are at her whereas the interval model suggests that they probably ovaa. Please answer the following questions as though know the local moral items, while distant moral behaviors we were there. Please circle your answers below’’. The sec- range from ‘‘I don’t know’’ to ‘‘Probably not’’. However, ow- ond treatment’s introduction was similar, but used a large ing to the considerable variation in responses, there is an bull spirit-master and Ubsa-Khol Lake (n = 23). The third extreme lack of fit. As such, by eliminating the divergent variant’s (n = 28) introduction used Adargan Arzhaan (in individuals, we can examine the subgroups with different the Övür district) that has a small marmot as its master...’’ response patterns. These examples were taken from previous interviews (Pur- Table 3 details the consensus analyses for the primary zycki, 2010).6 cultures’ responses for concern and knowledge items. For The survey treatments consisted of 39 questions: 10 the concern items, the remaining 38 were notably consis- moral items done in proximity to the ovaa (e.g., Does the tent in their responses (no one correlated negatively with cher eezi of this place know if I stole from another person the model). Discrete responses were consistently ‘‘I don’t here?), 10 moral items conducted at a significant distance know’’ while the responses in the model for interval data from the ovaa (e.g., ...if I lied to someone when I am at vary (Table C6). home in America?), 10 non-moral items knowable at the The primary pattern of agreement for the knowledge ovaa (e.g., ...that my eyes are blue?), 9 non-moral items items (Table C7) also show a strong lack of fit for the dis- only knowable at a distance from the ovaa (e.g., ...that crete model, but only two individuals diverged from the the car parked at my house is orange?). Of the moral items, model. According to the discrete responses, spirit-masters half were positive and half were negative social behaviors. only know the moral items that take place at the spirits’ Participants’ answers were on a scale of 2 to À2: ‘‘Of place of mastery, only some of the moral items that take course’’ (2), ‘‘Probably’’ (1), ‘‘I don’t know’’ (0), ‘‘I doubt place elsewhere, and none of the nonmoral items (see Sec- it’’ (À1), and ‘‘No’’ (À2). As was the case for the UConn sam- tion 3.2.4). ple, each item was accompanied with a sub-question of Assessing the responses from 50 individuals who di- whether or not spirits cared about such information. These verged from the primary model of the concern items, we items were answered after the free-lists (Purzycki, 2011, once again find a strong lack of fit in the discrete model 2012, pp. 232–252, n.d.), which were not revisable upon with 23 individuals diverging from the consensus (Table 4). completion. This secondary group shows that spirit-masters care only about moral behaviors occurring at their place of mastery but do not care about moral behaviors taking place else- 2.2.3. Results where or any of the nonmoral information (Table C8). Consensus analyses demonstrate that there is a consid- In summary, there was a high degree of consensus erable amount of variation in responses with the Tyvan among Christian students at the University of Connecticut. sample (Tables C4–C8). The interval cultural consensus Consensus models for God are theologically consistent models of spirit-masters’ knowledge and concern indicate insofar as God knows everything, and cares only about a single culture with only six and eleven divergent individ- the moral information. By contrast, Tyvans show consider- uals respectively. The discrete multiple-choice models, ably more variation in their responses. This further sug- however, indicate a strong lack of fit (Table 2). gests that these questions are indeed inconsistent with The discrete consensus models of spirit-masters’ con- any widespread understanding of spirit-masters’ minds. I cerns show that the culturally correct response as consis- now turn to a direct cross-cultural comparison that further tently ‘‘I don’t know’’, but 50 diverged from this pool. The emphasizes the relative lack of concern and knowledge for interval model shows higher positive responses for spirit- moral information attributed to spirit-masters. masters’ concern of local moral behaviors than distant moral behaviors with consistently lower scores for all nonmoral items (Table C4). 3. Cross-cultural comparison The knowledge items show a similar pattern with 36 individuals diverging from the consensus model. The dis- 3.1. Knowledge and concerns crete model suggests that spirit-masters know all of the If Tyvans popularly hold that spirit-masters know and 6 The data reported here include data reported elsewhere (Purzycki, care about morality, then there should be no significant 2011) and with the addition of the second treatment. differences between Tyvan and American ratings. B.G. Purzycki / Cognition 129 (2013) 163–179 171

Table 3 Fits to interval and multiple choice consensus models for primary cultures’ concern and knowledge attributions.

Concern (n = 38) Knowledge (n = 52) Discrete Interval Discrete Interval No. of negative competencies 0 5 2 3 Largest eigenvalue 17.06 10.87 13.80 22.56 2nd Largest eigenvalue 3.91 3.11 11.52 4.51 Ratio of largest to next 4.36 3.50 1.20a 5.22

a Violates assumption of 3-to-1 ratio required to determine the presence of a single culture. See Tables C6–C7 for answer keys.

Table 4 nonmoral information (Table 6). In other words, for these Fits to interval and multiple choice consensus models for secondary items, the only thing that God and spirit-masters share is culture’s concern attributions. their lack of concern for nonmoral information. Discrete Interval No. of negative competencies 23 6 Largest eigenvalue 11.12 20.43 3.2. Biases toward moralizing gods’ minds 2nd Largest eigenvalue 8.48 3.75 Ratio of largest to next 1.31a 5.44 3.2.1. God’s mind a Violates assumption of 3-to-1 ratio required to determine the pres- These participants claimed that God knows everything ence of a single culture. See Table C8 for answer key. and is all-powerful. If they maintain their explicit convic- tions of God’s omniscience to answer questions, only one factor should load in a factor analysis; reflective, theologi- However, there are significant differences between these cally correct thinking should yield one factor—the ‘‘God’s populations in their mean attributions of knowledge knowledge’’ factor. However, as found in previous experi- (V = 0.62, F(13,162) = 20.58, p<0.001) and concern mental research (Purzycki et al., 2012), if moral domains (V = 0.63, F(13,161) = 21.06, p<0.001) to their spirits are more accessible when thinking about God’s attributed (Table 5, Figs. 1–3). The Christian god had significantly knowledge, then the analysis should detect this by virtue higher (p<0.001) ratings than cher eezi for all variable cat- of degree of conviction. Participants’ mean scores of each egories other than concern for both proximate and distant category were used for analyses.

Table 5 Basic statistics of question type ratings.

God Cher Eezi Total M SD Mdn M SD Mdn M SD Mdn Knowledge items Moral Here Bad 1.81 0.38 2.00 1.20 0.88 1.40 1.51 0.74 2.00 Moral Here Good 1.79 0.41 2.00 1.00 0.85 1.00 1.40 0.77 1.90 Moral Here Total 1.80 0.39 2.00 1.10 0.80 1.16 1.45 0.72 1.84 Moral Elsewhere Bad 1.77 0.43 2.00 0.23 1.16 0.00 1.00 1.17 1.40 Moral Elsewhere Good 1.78 0.41 2.00 0.33 1.18 0.00 1.06 1.14 1.80 Moral Elsewhere Total 1.78 0.41 2.00 0.28 1.13 0.00 1.03 1.13 1.40 Moral Total 1.79 0.40 2.00 0.69 0.82 0.65 1.24 0.85 1.53 Nonmoral Here 1.31 0.90 1.80 À0.44 1.33 À0.17 0.44 1.43 0.78 Nonmoral Elsewhere 1.30 0.92 1.80 À0.42 1.23 À0.60 0.44 1.38 0.60 Nonmoral Total 1.31 0.90 1.80 À0.43 1.25 À0.55 0.44 1.39 0.78 Here Total 1.56 0.59 1.85 0.37 0.83 0.11 0.96 0.93 1.00 Elsewhere Total 1.54 0.61 1.88 À0.07 1.07 À0.08 0.74 1.18 0.98 Total 1.55 0.60 1.84 0.14 0.91 0.00 0.85 1.04 1.00 Concern items Moral Here Bad 1.62 0.68 2.00 0.91 1.02 1.20 1.26 0.93 1.60 Moral Here Good 1.50 0.73 2.00 0.74 1.00 1.00 1.12 0.95 1.20 Moral Here Total 1.56 0.69 2.00 0.83 0.95 0.00 1.20 0.91 1.50 Moral Elsewhere Bad 1.60 0.69 2.00 0.12 1.12 0.00 0.86 1.18 1.00 Moral Elsewhere Good 1.56 0.70 2.00 0.23 1.10 0.00 0.89 1.13 1.00 Moral Elsewhere Total 1.58 0.69 1.90 0.19 1.07 0.00 0.88 1.13 1.10 Moral Total 1.57 0.68 1.85 0.51 0.87 0.55 1.04 0.94 1.18 Nonmoral Here À0.44 1.18 À0.40 À0.42 1.28 0.00 À0.43 1.23 À0.16 Nonmoral Elsewhere À0.47 1.18 À0.50 À0.37 1.19 0.00 À0.42 1.18 À0.25 Nonmoral Total À0.46 1.17 À0.50 À0.39 1.21 À0.05 À0.42 1.19 À0.31 Here Total 0.56 0.72 0.48 0.24 0.87 0.08 0.40 0.81 1.00 Elsewhere Total 0.55 0.72 0.38 À0.09 1.05 0.00 0.23 0.95 0.15 Total 0.56 0.71 0.44 0.07 0.91 0.00 0.31 0.85 0.13 172 B.G. Purzycki / Cognition 129 (2013) 163–179

Fig. 1. Conﬁdence intervals (a = 0.05) of mean ratings for item categories of gods’ knowledge.

Overall, God was reported to be significantly more (M = À0.46, SD = 1.17, eigenvalue = 25.00%, variance ex- knowledgeable than concerned about all items, plained = 40.83, a = 0.99).7 F(1,174) = 99.41, p<0.001, x = 0.60. God was attributed Sex (F(1,79) = 2.07, p = 0.15), age (F(1,79) = 0.002, with greater knowledge of than concern for moral items, p = 0.96), and religiosity (F(1,79) = 1.09, p = 0.30) showed Welch’s F(1,140.02) = 6.93, p = 0.01, adj. x = 0.18. God no significant effects on the average amount of God’s con- knew more than cared about good moral items (Welch’s cern of the moral items. Moreover, sex (F(1,79) = 1.19, F(1,139.04) = 8.88, p = 0.003, adj. x = 0.21) as well as bad p = 0.28), age (F(1,79) = 2.13, p = 0.15), and religiosity moral items, Welch’s F(1,140.88) = 4.91, p = 0.03, adj. (F(1,79) = 1.09, p = 0.30) had no significant effects on the x = 0.15. However, God did not know good or bad moral average amount of God’s concern of nonmoral items. information differently (F(1,174) = 0.01, p = 0.91), nor did An ANOVA demonstrates that this sample attributed he have any differential concern for good and bad items, God’s concerns as significantly more moral than nonmoral, F(1,174) = 0.58, p = 0.46. According to this sample, then, F(1,174) = 201.19, p<0.001. These results are wholly God knows moral information greater than he cares about unsurprising and entirely consistent with the results from it and shows no significant differences between His knowl- the consensus analyses. However, God’s knowledge is edge of and concern for good or bad behaviors. inconsistent with participants’ stated beliefs that God knows everything and the results of the consensus analyses. 3.2.2. God’s concerns Regarding the items of what God cares about, a factor analysis solving for two factors demonstrates that there 3.2.3. God’s knowledge 8 are two discrete components (Fig. 3, left): the moral com- With some overlap of two items on an initial analysis, a ponent (M = 1.59, SD = 0.69, eigenvalue = 5.25;% variance Varimax rotated principle components analysis of God’s explained = 39.65; a = 0.98) and the nonmoral component knowledge of this information (without the two overlapping items) yields two factors (Fig. 3, right): a moral factor

7 One item loaded awkwardly (UCONGC2) likely because it ran into the page following the corresponding knowledge item. As such, this item was 8 The items were whether or not God knew if I was honest to my friends deleted from all further analyses of God’s attributed concerns. Dropping in St. Petersburg (ELSEWRG5) and if I told a lie to someone at UConn this item raised the alpha of the moral items by 0.01 and their grand mean (UCONB1). Because these two items loaded highly on both factors for by 0.03. unknown reasons, I eliminated them from all further analyses. B.G. Purzycki / Cognition 129 (2013) 163–179 173

Fig. 2. Conﬁdence intervals (a = 0.05) of mean ratings for item categories of gods’ concerns.

Fig. 3. Rotated factor loadings plot of God’s concern (left) and knowledge (right) items.

(M = 1.79, SD = 0.40, eigenvalue = 5.25%; variance ex- An ANOVA demonstrates that mean response to ques- plained = 65.79, a = 0.98) and a nonmoral factor (M = 1.31, tions about God’s knowledge of moral items was signifi- SD = 0.90, eigenvalue = 25.00;% variance explained = 13.82, cantly higher than mean response to nonmoral questions, a = 0.99). The very presence of two factors is a case of theo- F(1,174) = 21.45, p<0.001. Moreover, a paired t-test of logical inconsistency insofar as God’s access to all knowable the standard deviations of each category reveals that there information should be collectively represented as one factor; was more variance among the responses to the nonmoral as participants initially claimed that God is omniscient, questions (M = 0.66, SD = 0.38) than moral questions there should not be any significant differences across cate- (M = 0.22, SD = 0.29, t(87) = 9.51, Bonferroni’s adjusted gories. Neither sex nor age had significant effects on mean p<0.001); there is more consistency in responses to God’s response to moral or nonmoral items (all p P 0.42). knowledge of moral items than with nonmoral items. 174 B.G. Purzycki / Cognition 129 (2013) 163–179

Table 6 not constitute a single factor. Moreover, unlike the Abraha- Multivariate general linear model F-ratios for knowledge and concern mic god, these items clustered into three domains (Fig. 4, items. Table 7): (a) moral behaviors conducted at the ovaa Dependent variable Knowledge Concern (a = 0.94, M = 1.09, SD = 0.81), (b) moral behaviors con- df = 1(174) df = 1(173) ducted away from the ovaa (a = 0.95, M = 0.27, SD = 1.13), F-ratio F-ratio and (c) all non-moral items (a = 0.99; M = À0.47, Moral Here Bad 36.43 30.46 SD = 1.22). Tyvans clearly make distinctions between do- Moral Here Good 61.15 33.59 mains of knowledge that cher eezi have in two dimensions: Moral Here Total 54.19 34.81 Moral Elsewhere Bad 136.78 111.10 proximity of act to the ovaa and whether or not the act is Moral Elsewhere Good 117.74 94.81 moral. The same proximate moral (a = 0.95; M = 0.85, Moral Elsewhere Total 136.36 109.37 SD = 0.93), distant moral (a = 0.94; M = 0.17, SD = 1.03), Moral Total 126.14 83.61 and nonmoral (a = 0.99; M = À0.44, SD = 1.18) factors Nonmoral Here 105.98 0.00* Nonmoral Elsewhere 109.71 0.17* loaded for the concern items. Sex showed no significant ef- Nonmoral Total 111.84 0.04* fects for the responses to each of these factors’ mean re- Here Total 119.8 8.38 sponses (p>.05) and age only showed significant effects Elsewhere Total 149.57 25.10 for nonmoral knowledge, F(36,85) = 1.78, p = 0.04. Total 146.33 17.47 The fan of plots represents the distance from the spirit- Independent variable (IV) = supernatural agent. master’s location; the cluster of items in the lower right * p > 0.05 (Nonparametric tests yielded the same significant differences sector are moral behaviors that take place at the spirit- for all tests). master’s ritual cairn, behaviors that take place in Ak Dovu- rak and Kyzyl—distant cities in Tyva—are plotted between 3.2.4. Spirit-masters’ minds behaviors that take place in America, Moscow, and St. Recall from Section 1.4 that there were three variations Petersburg. Clearly, Tyvans are reasoning about spirit-mas- in treatments of spirit-masters’ forms (woman, marmot, ters’ knowledge of moral behavior geographically. and bull). ANOVAs demonstrate that cher eezi type had If there is a general tendency to attribute moral concern no significant effects on mean ratings of proximate moral and knowledge to supernatural agents, then moral items knowledge items (F(2,85) = 0.24, p = 0.79), concern should rate significantly higher than the non-moral items. (F(2,85) = 0.34, p = 0.72), on distant moral knowledge Overall, spirit-masters are more knowledgeable (Welch’s (F(2,85) = 1.39, p = 0.26) or concern (F(2,85) = 1.24, F(1,150.74) = 49.53, p<0.001, adj. x = 0.46) and concerned p = 0.30), or on distant nonmoral knowledge (Welch’s F(1,158.28) = 32.46, p<0.001, adj. x = 0.39) (F(2,85) = 1.51, p = 0.23) or concern (F(2,85) = 0.50, about moral information than nonmoral information. p = 0.61). So, even though type of resource is correlated If proximity to location plays a significant role in how with type of spirit-master (Purzycki, in press), the actual Tyvans conceive of cher eezi’s access to information, this form of spirit does not have an effect on Tyvan conceptions should also affect responses insofar as distant moral acts will of moralization or breadth of knowledge. As such, all re- be both less known and less cared about than proximate acts. sults reported below are from the three pooled treatments. Overall, the proximity of information to the spirit-masters’ A factor analysis using Varimax rotation solving for location (F(1,348) = 10.66, p = 0.001) and moral content three factors demonstrates that the knowledge items do (F(1,348) = 85.56, p < 0.001) had significant effects on

Fig. 4. Factor loadings plot of cher eezi’s knowledge items. B.G. Purzycki / Cognition 129 (2013) 163–179 175

Table 7 Summary of rotated factor analysis for spirit-masters’ knowledge and concern items.

Knowledge Items Concern Items Moral Moral Item Nonmoral Prox. Distant Item Nonmoral Prox. Distant Eigenvalues 20.08 6.65 3.34 Eigenvalues 19.18 7.41 2.91 % Variance 51.49 17.06 8.55 % Variance 49.17 19.00 7.47 a 0.99 0.94 0.95 a 0.99 0.95 0.94

response. These two factors also had significant interaction ef- happens to it, 45 (51.1%) answered that they know only fects on response, F(1,348) = 12.16, p = 0.001. what transpires in their area, five (5.7%) answered that More specifically, there were significant effects for they know all that takes place in Tyva, and 25 (28.4%) an- proximity of moral act on masters’ knowledge (Welch’s swered that they know everything that happens on earth. F(1,157.27) = 30.93, p < 0.001, adj. x = 0.38) and concern Five individuals answered with multiple responses and as (F(1,174) = 17.76, p<0.001, x = 0.29). In other words, these scale items range in the breadth of knowledge cher where behaviors take place plays an important role on eezi are attributed, these responses were converted to how Tyvans reason about spirit masters’ minds; cher eezi means. Because participants overwhelmingly answered are reported to know and care more about moral acts con- the question in at least two ways, I dichotomized the re- ducted at their places of mastery than those taken place sponses (0 = only knows what happens to it, only know distantly from such places. things that happen in their domain; 1 = knows everything There were no significant differences between their that happens in Tyva, everything that happens on earth). knowledge of good and bad behaviors conducted proxi- If there is a relationship between omniscience and con- mately (F(1,174) = 2.25, p = 0.14) or distantly, cern for morality, then attributed breadth of knowledge F(1,174) = 0.33, p = 0.57. There were also nonsignificant should predict knowledge of and concern for moral behav- differences between their concern for proximate iors. While there were no significant effects for knowledge- (F(1,174) = 0.004, p = 0.95) or for distant (F(1,174) = 0.08, breadth on knowledge (F(1,76) = 0.00, p = 0.99) or concern p = 0.78) good or bad behaviors. (F(1,76) = 0.02, p = 0.90) of proximate moral items, attrib- Moreover, while cher eezi were reported as more knowl- uted breadth of knowledge does in fact predict knowledge edgeable of local moral behaviors than concerned with (F(1,76) = 9.05, p = 0.004, x = 0.31) and concern them (F(1,174) = 4.12, p = 0.04, x = 0.13), there were no (F(1,76) = 9.08, p = 0.004, x = 0.31) for distant moral differences between their knowledge and concern for dis- behaviors. This is consistent with previous findings of the tant moral behaviors, F(1,174) = 0.33, p = 0.57. positive relationship between omniscience and moral con- In summary, participants gravitate towards moralizing cern across populations, but these results are the first of spirit-masters’ knowledge and concerns, but this moraliza- their kind from a single population. tion is bound by location of the behavior in proximity to spirits’ domain of mastery. Even though moral concerns 3.4. The moralization bias of spirit-masters’ minds are not the culturally correct concerns of cher eezi, these results support the view that socially strategic information is As reported above, spirit-masters have place-bound more accessible when reasoning about supernatural knowledge of moral information. A stronger test of the agents. moralization bias hypothesis would be to analyze only data for individuals who did not list any interpersonal so- 3.3. Omniscience and moral concern cial behaviors among the things which anger spirit-masters (see Purzycki, 2010, 2012, pp. 232–252, n.d.). If If there is a relationship between omniscience and mor- Tyvans who do not freelist moral behaviors among the alization, attributed breadth of knowledge should predict things which concern spirit-masters but rate their knowl- knowledge and concern of morality. As addressed in edge and concern of moral behavior higher than nonmoral numerous cross-cultural studies (Johnson, 2005; Lahti, information, this would strongly support the presence of 9 2009; Sanderson, 2008; Stark, 2001; Swanson, 1960), there such a bias. is a strong positive relationship between omniscience and Among the responses for those who did not list a single moral concern. To test for such a relationship among Ty- moral item among the things which please or anger spirit- vans, one question asked early in the survey inquired about masters (n = 33; Mage = 40.68; SDage = 15.56; two did not the breadth of spirit-masters’ knowledge. The possible re- report age; 13 males, one did not report sex; 13 reporting sponses were: (4) She/It knows everything that happens in the world, (3) She/It only knows everything that hap- 9 Only 10 participants listed a moral item in the free-list task eliciting pens in Tyva, (2) She/It only knows everything that hap- behaviors which please spirit-masters. Also, the only significant effects for pens in her area, and (1) She/It only knows what happens listing moral items for the ‘‘anger’’ freelists were found on cher eezi’s knowledge of good moral behaviors which take place nearby to her. (F(1,63) = 6.21, p=0.02) and overall knowledge of moral behaviors (nearby Out of 88 individuals, seven (8%) did not answer, three (F(1,63) = 4.44, p=0.04) which take place nearby their place of mastery. No (3.4%) answered that the spirit-master only knows what other domains showed effects from having listed a moral item in this task. 176 B.G. Purzycki / Cognition 129 (2013) 163–179 being both Buddhist and shamanist; 20 characterized data regarding what gods know and care about. Cultural themselves as just Buddhist), we find a similar pattern to anthropologists and other researchers of religion have a the overall sample. Once again, proximity of information lot to gain from the collection of such data in order to assess (F(1,128) = 4. 64, p = 0.03) and moral content the major assumptions which have been guiding research (F(1,128) = 35.11, p < 0.001) had significant effects on re- for the past century, as do the people whose traditions we sponse and also showed significant interaction effects, may document who want an accounting of their beliefs F(1,128) = 4.84, p = 0.03. In summary, even though all of for future generations. the participants in these analyses did not list a single inter- Perhaps it is the case that domains of interpersonal so- personal behavior among the things which please or anger cial behavior are nested within other specific domains (e.g., spirit-masters, moral content had a significant effect on ritual behavior, resource maintenance, etiquette etc.). responses. Making sense of gods’ minds may prime these closely re- lated clusters of cultural knowledge even though some domains are not among the culturally correct concerns 4. Discussion ascribed to supernatural agents (see Huebner, Lee, & Haus- er, 2010 on the relationship between convention and The present work demonstrates that: (1) while partici- morality). Indeed, when children believe that a ghost is pants claim that God knows everything, they show a par- watching them, they cheat less in games even though no ticular sensitivity towards moralizing the content of explicit moral concern is attributed to the ghost (Piazza God’s knowledge, (2) Tyvans likewise show biases towards et al., 2011). While many Tyvans feel no obligation to stop moralizing spirit-masters’ knowledge and concerns even and pay their respects, it may be an underlying assumption though this is not among their collectively held concerns, that people should stop that brings moral systems to bear (3) Tyvans use geography when considering spirit-masters’ on responses to these questions. knowledge and concern of moral information, and (4) attri- It may very well be the case that humans have cognitive butions of omniscience to spirit-masters predicts attrib- algorithms that tap into moral domains upon detection of uted concern for distant moral behavior. The results agents both supernatural (Bering & Johnson 2005; Bering suggest that there is an underlying moral component to et al., 2005; Johnson, 2005; Johnson & Bering, 2006; Piazza, representing gods even though they may not explicitly Bering, & Ingram, 2011; Randolph-Seng, & Nielsen, 2007; care about such ‘‘moral’’ behaviors. In other words, despite Rossano, 2007; Shariff & Norenzayan, 2007) and natural the world’s religious diversity and cultural models, inter- alike (Bateson, Nettle, & Roberts, 2006; Ernest-Jones, personal social behavior is an essential constant in reli- Nettle, & Bateson, 2011; Gray, Young, & Waytz, 2012; gious cognition. Gods may explicitly not care about Haley & Fessler, 2005). In other words, perhaps mere morality, but moral domains nevertheless lie beneath the agency-detection automatically prepares us for social surface of explicit, reflective representations. As such, reli- interaction. This may explain why there were no signifi- gious systems around the world may indeed be essentially cant differences in moralization between the zoomorphic about interpersonal social regulation and monitoring and anthropomorphic spirit-masters. Again, however, we regardless of whether or not moral concern is explicitly (Purzycki et al., 2012) found that a non-interfering alien attributed to gods. More concretely, nonmoral gods may species did not elicit different response times between nevertheless promote human morality. These results raise moral and nonmoral questions. many compelling questions. If the concern items were asked about actual people, the How ‘‘cultural’’ is the moralization of gods’ minds? In moral questions would likely be rated higher; actual hu- the case if Tyva, as morality is not a central or salient fea- mans certainly care about being assaulted or treated with ture of spirit-masters’ concerns (Purzycki, 2010, 2011, kindness. And, people care about a lot of things, so another 2012), such concerns do not appear to be ‘‘cultural’’ insofar remaining question is whether or not personal concerns as they are not as much a part of the widespread repertoire can also drive explicit attributions of gods’ concerns. In of explicitly attributed concerns as ritual and resource the case of the present study, moral content renders re- maintenance are. In this case, moralization is intuitive. Per- sponses inconsistent with explicit, culturally correct postu- haps Tyvans use analogical reasoning in their post hoc mor- lates. In one sitting, students claim both that God knows alizations of cher eezi’s minds insofar as such question everything, but knows moral information better than non- might be priming the concerns other agents in the Tyvan moral information. In Tyva, people might not readily freel- pantheon; moral domains may be a latent association ist or discuss moral behaviors among the things which among Tyvan conceptions of spirit-masters by virtue of spirit-masters care about, but they appear to both know their association with other supernatural agents (e.g., Bud- and care more about the latter relative to nonmoral infor- dha). A stronger test of this sort would be to conduct the mation when asked in a survey limited to such questions. same study in a tradition without moralistic deities or con- Additionally, evidence suggests that omniscience can be sistent contact with traditions having such deities to deter- quite intuitive; even young children grant supernatural mine whether or not there are similar effects. It remains agents with access to more knowledge than human agents doubtful, however, that such a society exists. Unfortu- (Barrett, Richert, & Driesenga, 2001; Knight, et al., 2004; nately, outside of my own ethnographic work in Tyva (Pur- Wigger, Paxson, & Ryan, 2013). This suggests that the cog- zycki, 2010, 2012) and data collected in Fiji (Shaver, 2012, nitive foundations of omniscience may be more intuitive pp. 115–123), there is presently no available systematically than not and ‘‘culture’’ merely builds on and exaggerates collected, quantifiable, and therefore directly comparable the intuitive stance of omniscience. But, is there a B.G. Purzycki / Cognition 129 (2013) 163–179 177 moralization bias of supernatural agents in children? contexts (Purzycki, 2010, 2011, 2012). Only a close Developmental psychologists are in a prime position to an- accounting of the relationships between religious cogni- swer these questions. tion, ritual, and context will allow us to answer such ques- Why is there a close association between omniscience tions with confidence. and concern for moral behavior? As found in cross-cultural studies which demonstrate a positive relationship between Acknowledgements attributed breadth of knowledge and concern for interpersonal social behavior (Johnson, 2005; Sanderson, 2008; Many, many thanks to Candace Alcorta, Tayana Arak- Swanson, 1960), the more spirit-masters know, the more chaa, Justin Barrett, Adam Cohen, Brian Donahoe, the ECC they care about moral behaviors. However, there was var- gang at UConn, W. Penn Handwerker, Jessica McCutcheon, iation in attributed breadth of knowledge in the case of Ty- Nadia Oorzhak, the Saryglars, Rich Sosis, Valentina vans and perhaps this variation exists in most traditions Süzükei, the students and staff at the Tyvan Institute for (see Boehm, 2008; Pettazzoni, 1955). Still, the question re- Humanitarian Research, the Tyvan National Orchestra, mains as to how such conceptual packages become both and the many people on opposite sides of the globe who more widespread and part of the explicit repertoire gods’ participated in this study. Thanks, too, to Rita McNamara, minds. The Abrahamic god is more consistently considered Matt Rossano, John Shaver, Aiyana Willard, and to the omniscient, so the question of what mechanism is respon- anonymous reviewers of previous drafts of this paper for sible for such a shift remains unknown. As already dis- their helpful, critical comments. This work was financially cussed, some researchers suggest that under particular supported by Oxford University’s Cognition, Religion and constraints, moralistic and omniscient gods are culturally Theology Project, the University of Connecticut’s Depart- selected because they maximize prosocial behavior (Shariff ment of Anthropology, and the SSHRC-funded Cultural & Norenzayan, 2007). As societies become more complex, Evolution of Religion Research Consortium (CERC). individual behavior more easily hidden from others, concepts of omniscient moralistic high gods may become not only more easily promulgated, but also more salient in Appendix A. Supplementary data individual minds. Cultural evolutionary approaches might focus on the source of the information such as authorities, Supplementary data associated with this article can be peers, and parents (Boyd & Richerson, 1985; Richerson & found, in the online version, at http://dx.doi.org/10.1016/ Boyd, 2005). However, the present research suggests that j.cognition.2013.06.010. the intuitions for moral concern are already there—they simply may not become explicit, reflective, and widespread until there are socioecological pressures to make References them as such. In other words, their explicitness may be Abbruzzese, J. E. (1997). The coherence of omniscience: A defense. more effective as a social mediating device under particu- International Journal for Philosophy of Religion, 41(1), 25–34. lar conditions. The case might be the same for omniscience Anderson, J. R. (2007). How can the human mind occur in the physical (Peoples & Marlowe, 2012; Roes & Raymond, 2003; Wal- universe? New York: Oxford University Press. lace, 1966). Atran, S. (2002). In gods we trust: The evolutionary landscape of religion. Oxford: Oxford University Press. Finally, the ultimate question of whether or not various Atran, S., Medin, D., Ross, N., Lynch, E., Vapnarsky, V., Ucan Ek’, E., et al. explicit conceptions of gods’ minds have particular behav- (2002). Folkecology, cultural epidemiology, and the spirit of the ioral and fitness effects requires further investigation. Is commons: A garden experiment in the Maya lowlands, 1991–2001. Cultural Anthropology, 43(3), 421–450. there variation in levels of prosociality for populations Baron-Cohen, S. (1995). Mindblindness: An essay on autism and theory of who explicitly moralize their deities? Recall that evidence mind. Cambridge: The MIT Press. demonstrates that when God is conceived as being forgiv- Barrett, J. L. (1998). Cognitive constraints on Hindu concepts of the divine. Journal for the Scientific Study of Religion, 37, 608–619. ing rather than punishing, people will actually defect on so- Barrett, J. L. (1999). Theological correctness: Cognitive constraint and the cial obligations more often (Shariff & Norenzayan, 2011). study of religion. Method and Theory in the Study of Religion, 11, This suggests that cultural models of gods do play an 325–339. Barrett, J. L. (2004). Why would anyone believe in God? New York: AltaMira important role in interpersonal social behavior. Cross-cul- Press. turally, it may be the case that punishing, moralistic gods Barrett, J. L. (2008). Why Santa Claus is not a God. Journal of Cognition and make lasting, widespread cooperation more sustainable Culture, 8(1–2), 149–161. Barrett, J. L., & Keil, F. C. (1996). Conceptualizing a nonnatural entity: than they would if they were not moralistic. Again, how- Anthropomorphism in God concepts. Cognitive Psychology, 31, ever, explicit models of gods’ concerns may correspond 219–247. to local pressures; they mediate sociality differently Barrett, J. L., Richert, R. A., & Driesenga, A. (2001). God’s beliefs versus according to local conditions (Purzycki & Sosis, 2011). In mother’s: The development of nonhuman agent concepts. Child Development, 72, 50–65. the case of Inner Asia, ritual cairns mark territories and Batchelder, W. H., & Romney, A. K. (1988). Test theory without an answer thus prime supernatural agency. If it is the case that ritual key. Psychometrika, 53(1), 71–92. functions to minimize antisocial behavior through the Bateson, M., Nettle, D., & Roberts, G. (2006). Cues of being watched enhance cooperation in a real-world setting. Biology Letters, 2(3), priming of supernatural agency, it suggests that there are 412–414. important long-term, complex relationships between Bering, J. (2011). The belief instinct: The psychology of souls, destiny, and the socioecological pressures and human cognition that play meaning of life. New York: W.W. Norton & Company. Bering, J. M., & Johnson, D. D. P. (2005). ‘‘O Lord...you perceive my an important role in how prosocial priming becomes thoughts from afar’’: Recursiveness and the evolution of supernatural differentially expressed and distributed across diverse agency. Journal of Cognition and Culture, 5, 118–142. 178 B.G. Purzycki / Cognition 129 (2013) 163–179

Bering, J. M., McLeod, K. A., & Shackelford, T. K. (2005). Reasoning about Hughes, G. (1995). The nature of God: An introduction to the philosophy of dead agents reveals possible adaptive trends. Human Nature, 16, religion. New York: Routledge. 360–381. Humphrey, C., & Onon, U. (1996). Shamans and elders: Experience, Bernard, H. R. (2011). Research methods in anthropology: Qualitative and knowledge and power among the Daur Mongols. Oxford: Clarendon quantitative approaches (5th ed.). Lanham: Alta Mira Press. Press. Boehm, C. (2008). A biocultural evolutionary exploration of supernatural Johnson, D. D. P. (2005). God’s punishment and public goods: A test of the sanctioning. In J. Bulbulia, R. Sosis, E. Harris, R. Genet, C. Genet, & K. supernatural punishment hypothesis in 186 world cultures. Human Wyman (Eds.), Evolution of religion: Studies, theories, and critiques Nature, 16, 410–446. (pp. 143–152). Santa Margarita: Collins Foundation Press. Johnson, D. D. P., & Bering, J. (2006). Hand of God, mind of man: Borgatti, S. P., Everett, M. G., & Freeman, L. C. (2002). Ucinet for windows: Punishment and cognition in the evolution of cooperation. Software for social network analysis. Harvard: Analytic Technologies. Evolutionary Psychology, 4, 219–233. Borgatti, S. P., & Halgin, D. S. (2011). Consensus analysis. In D. B. Kapitan, T. (1991). Agency and omniscience. Religious Studies, 27(1), Kronenfeld, G. Bennardo, & V. C. de Munck (Eds.), A companion to 105–120. cognitive anthropology (pp. 171–190). Malden: Blackwell Publishing. Khomushku, O. M. (2008). Traditional religious beliefs of the peoples of Boster, J. S., & Johnson, J. C. (1989). Form or function: A comparison of Sayano-Altai and problems of religious syncretism. Anthropology and expert and novice judgments of similarity among fish. American Archeology of Eurasia, 46(3), 51–63. Anthropologist, 91(4), 866–889. Knight, N., Sousa, P., Barrett, J. L., & Atran, S. (2004). Children’s attributions Boyd, R., & Richerson, P. J. (1985). Culture and the evolutionary process. of beliefs to humans and God: Cross-cultural evidence. Cognitive Chicago: University Of Chicago Press. Science, 28, 117–126. Boyer, P. (2000). Functional origins of religious concepts: Ontological and Kretzmann, N. (1966). Omniscience and immutability. The Journal of strategic selection in evolved minds. The Journal of the Royal Philosophy, 63(14), 409–421. Anthropological Institute, 6, 195–214. Kristensen, B. M. (2004). The living landscape of knowledge. An analysis of Boyer, P. (2001). Religion explained: The evolutionary origins of religious ahamanism among the Duha Tuvinians of Northern Mongolia. Institut thought. New York: Basic Books. for Antropologi, Københavns Universitet. Specialeafhandling til Boyer, P. (2002). Why do gods and spirits matter at all? In I. Pyysiäinen & Kandidateksamen, Specialerække nr. 317. (Accessed (pp. 68–92). New York: Continnuum. 21.09.12). Brown, D. E. (1991). Human universals. Boston: McGraw Hill. Lahti, D. C. (2009). The correlated history of social organization, morality, Caulkins, D. D. (2004). Identifying culture as a threshold of shared and religion. In E. Voland & W. Schiefenhövel (Eds.), The evolution of knowledge: A consensus analysis method. International Journal of religious mind and behavior (pp. 67–88). New York: Springer. Cross Cultural Management, 4(3), 317–333. Lansing, J. S. (2007). Priests and programmers: Technologies of power in the Chadwick, N. K. (1936). Shamanism among the Tatars of Central Asia. The engineered landscape of Bali. Princeton: Princeton University Press. Journal of the Royal Anthropological Institute of Great Britain and Lansing, J. S., & Kremer, J. N. (1993). Emergent properties of Balinese Ireland, 66, 75–112. water temple networks: Coadaptation on a rugged fitness landscape. Cosmides, L., & Tooby, J. (1989). Evolutionary psychology and the American Anthropologist, New Series, 95(1), 97–114. generation of culture, Part II. Case study: A computational theory of Lee, R. B. (2003). The Dobe Ju/’Hoansi. Belmont: Wadsworth Publishing. social exchange. Ethology and Sociobiology, 10, 51–97. Malinowski, B. (1932). Argonauts of the Western Pacific: An account of Cosmides, L., & Tooby, J. (1992). Cognitive adaptations for social exchange. native enterprise and adventure in the archipelagoes of Melanesian New In J. H. Barkow, L. Cosmides, & J. Tooby (Eds.), The adapted mind: Guinea. New York: E.P. Dutton and Co.. Evolutionary psychology and the generation of culture (pp. 163–228). Marshall, L. (1962). !Kung Bushman religious beliefs. Africa: Journal of the New York: Oxford University Press. International African Institute, 32(3), 221–252. D’Andrade, R. (1995). The development of cognitive anthropology. McCauley, R. (2011). Why religion is natural and science is not. Oxford: Cambridge: Cambridge University Press. Oxford University Press. de Munck, V. C., Dudley, N., & Cardinale, J. (2002). Cultural models of Miller, M. L., Kaneko, J., Bartram, P., Marks, J., & Brewer, D. D. (2004). gender in Sri Lanka and the United States. Ethnology, 41(3), 225–261. Cultural consensus analysis and environmental anthropology: Durkheim, É. (2001[1915]). The elementary forms of religious life. New Yellowfin Tuna fishery management in Hawaii. Cross-Cultural York: Oxford University Press. Research, 38(3), 289–314. Eliade, M. (1964). Shamanism: Archaic techniques of ecstasy. New York: Nettle, D., Harper, Z., Kidson, A., Stone, R., Penton-Voak, I. S., & Bateson, M. Arkana. (2013). The watching eyes effect in the Dictator Game: It’s not how Ernest-Jones, M., Nettle, D., & Bateson, M. (2011). Effects of eye images on much you give, it’s being seen to give something. Evolution and everyday cooperative behavior: A field experiment. Evolution and Human Behavior, 34(1), 35–40. Human Behavior, 32(3), 172–178. Nicholas, L. J. (2004). The association between religiosity, sexual fantasy, Evans-Pritchard, E. E. (1956). Nuer religion. New York: Oxford University participation in sexual acts, sexual enjoyment, exposure, and reaction Press. to sexual materials among Black South Africans. Journal of Sex and Fodor, J. (1983). The modularity of mind: An essay on faculty psychology. Marital Therapy, 30, 37–42. Cambridge: The MIT Press. Nicholas, L. J., & Durrheim, K. (1995). Religiosity, Aids, and sexuality Gray, K., Young, L., & Waytz, A. (2012). Mind perception is the essence of knowledge, attitudes, beliefs, and practices of Black South-African morality. Psychological Inquiry, 23(2), 101–124. first-year university students. Psychological Reports, 77, Grim, P. (1983). Some neglected problems of omniscience. American 1328–1330. Philosophical Quarterly, 20(3), 265–276. Norenzayan, A., Gervais, W. M., & Trzesniewski, K. H. (2012). Mentalizing Guthrie, S. E. (1980). A cognitive theory of religion. Current Anthropology, deficits constrain belief in a personal god. PLoS One, 7(5), e36880. 21(2), 181–203. Norenzayan, A., & Shariff, A. F. (2008). The origin and evolution of Guthrie, S. E. (1995). Faces in the clouds: A new theory of religion. New religious prosociality. Science, 322, 58–62. York: Oxford University Press. Pegg, C. (2001). Mongolian music, dance, and oral narrative. Seattle: Haidt, J. (2008). Morality. Perspectives on Psychological Science, 3(1), University of Washington Press. 65–72. Peoples, H. C., & Marlowe, F. W. (2012). Subsistence and the evolution of Halemba, A. (2006). The Telengits of Southern Siberia: Landscape, religion religion. Human Nature, 23(3), 253–269. and knowledge in motion. New York: Routledge. Pettazzoni, R. (1955). On the attributes of God. Numen, 2, 1–27. Haley, K. J., & Fessler, D. M. T. (2005). Nobody’s watching? Subtle cues Piazza, J., Bering, J. M., & Ingram, G. (2011). ‘‘Princess Alice is watching affect generosity in an anonymous economic game. Evolution and you’’: Children’s belief in an invisible person inhibits cheating. Journal Human Behavior, 26(3), 245–256. of Experimental Child Psychology, 109, 311–320. Handwerker, W. P. (2001). Quick ethnography. New York: Alta Mira Press. Premack, D., & Woodruff, G. (1978). Does the chimpanzee have a theory of Harvilahti, Lauri. (2000). Altai oral epic. Oral Tradition, 15(2), 215–229. mind? Behavioral and Brain Sciences, 1(4), 515–526. Heissig, W. (1980). The religions of Mongolia. London: Routledge and Purzycki, B. G. (2010). Spirit masters, ritual cairns, and the adaptive Kegan Paul. religious system in Tyva. Sibirica, 9(2), 21–47. Huebner, B., Lee, J. J., & Hauser, M. D. (2010). The moral–conventional Purzycki, B. G. (2011). Tyvan cher eezi and the socioecological constraints distinction in mature moral competence. Journal of Cognition and of what supernatural agents know. Religion, Brain and Behavior, 1(1), Culture, 10(1–2), 1–26. 31–45. B.G. Purzycki / Cognition 129 (2013) 163–179 179

Purzycki, B. G. (2012). Finding minds in the natural world: Dynamics of the Shweder, R. A., Mahapatra, M., & Miller, J. G. (1987). Culture and moral religious system in the Tyva Republic. Doctoral dissertation. University development. In J. Kagan & S. Lamb (Eds.), The emergence of morality in of Connecticut. young children (pp. 1–83). Chicago: University of Chicago Press. Purzycki, B. G. (in press). Toward a cognitive ecology of religious Shweder, R. A., Much, N. C., Mahapatra, M., & Park, L. (1997). The ‘‘big concepts: An example from the Tyva Republic. Journal for the three’’ of morality (autonomy, community, divinity) and the ‘‘big Cognitive Science of Religion, 1. three’’ explanations of suffering. In A. M. Brandt (Ed.), Morality and Purzycki, B. G., & Arakchaa, T. (2013). Ritual behavior and trust in the Tyva health (pp. 119–169). New York: Routledge. Republic. Current Anthropology, 54(3), 381–388. Slone, D. J. (2004). Theological incorrectness: Why religious people believe Purzycki, B. G., & Sosis, R. (2011). In U. Frey (Ed.). Our gods: Variation in what they shouldn’t. New York: Oxford University Press. supernatural minds. In Essential building blocks of human nature (pp. Smetana, J. G. (2006). Social-cognitive domain theory: Consistencies and 77–93). New York: Springer-Verlag. variations in children’s moral and social judgments. In M. Killen & J. Purzycki, B. G., Finkel, D. N., Shaver, J., Wales, N., Cohen, A. B., & Sosis, R. (2012). Smetana (Eds.), Handbook of moral development (pp. 119–153). What does God know? Supernatural agents’ access to socially strategic Mahwah: Lawrence Erlbaum Associates. and nonstrategic information. Cognitive Science, 36(5), 846–869. Soler, M. (2012). Costly signaling, ritual and cooperation: Evidence from Purzycki, B. G. (n.d.). The spirit-master religious complex of the Tyva Candomblé, an Afro-Brazilian religion. Evolution and Human Behavior, Republic. 33(4), 346–356. Randolph-Seng, B., & Nielsen, M. E. (2007). Honesty: One effect of primed Sosis, R., & Bressler, E. R. (2003). Cooperation and commune longevity: A religious representations. International Journal for the Psychology of test of the costly signaling theory of religion. Cross-Cultural Research, Religion, 17(4), 303–315. 37(2), 211–239. Rappaport, R. A. (1979). Ecology, meaning, and religion. Berkeley: North Sosis, R., Kress, H. C., & Boster, J. S. (2007). Scars for war: Evaluating Atlantic Books. alternative signaling explanations for cross cultural variance in ritual Rappaport, R. A. (1999). Ritual and religion in the making of humanity. costs. Evolution and Human Behavior, 28, 234–247. Cambridge: Cambridge University Press. Sosis, R., & Ruffle, B. (2003). Religious ritual and cooperation: Testing for a Richerson, P. J., & Boyd, R. (2005). Not by genes alone: How culture relationship on Israeli religious and secular kibbutzim. Current transformed human evolution. Chicago: University Of Chicago Press. Anthropology, 44, 713–722. Roes, F. L., & Raymond, M. (2003). Belief in moralizing gods. Evolution and Sosis, R., & Ruffle, B. (2004). Ideology, religion, and the evolution of Human Behavior, 24(2), 126–135. cooperation: Field tests on Israeli kibbutzim. Research in Economic Rohrbaugh, J., & Jessor, R. (1975). Religiosity in youth: A personal control Anthropology, 23, 89–117. against deviant behavior. Journal of Personality, 43(1), 136–155. Sperber, D. (1997). Intuitive and reflective beliefs. Mind and Language, Romney, A. K., Weller, S. C., & Batchelder, W. H. (1986). Cultural as 12(1), 67–83. consensus: A theory of culture and informant accuracy. American Stark, R. (2001). Gods, rituals, and the moral order. Journal for the Scientific Anthropologist, 88(2), 313–338. Study of Religion, 40(4), 619–636. Rossano, M. J. (2007). Supernaturalizing social life: Religion and the Strauss, C., & Quinn, N. (1997). A cognitive theory of cultural meaning. evolution of human cooperation. Human Nature, 18, 272–294. Cambridge: Cambridge University Press. Rossano, M. J. (2008). The moral faculty: Does religion promote ‘‘moral Sugiyama, L. S., Tooby, J., & Cosmides, L. (2002). Cross-cultural evidence of expertise’’? International Journal for the Psychology of Religion, 18(3), cognitive adaptations for social exchange among the Shiwiar of 169–194. Ecuadorian Amazonia. Proceedings of the National Academy of Sciences Sachdeva, S., Singh, P., & Medin, D. (2011). Culture and the quest for of the United States of America, 99(17), 11537–11542. universal principles in moral reasoning. International Journal of Swanson, G. E. (1960). The birth of the gods: The origin of primitive beliefs. Psychology, 46(3), 161–176. Ann Arbor: University of Michigan Press. Sanderson, S. K. (2008). Religious attachment theory and the biosocial Turiel, E. (1983). The development of social knowledge: Morality and evolution of the major world religions. In J. Bulbulia, R. Sosis, E. Harris, convention. Cambridge: Cambridge University Press. R. Genet, C. Genet, & K. Wyman (Eds.), The evolution of religion: Studies, Van Deusen, K. (2004). singing story, healing drum: Shamans and theories, and critiques (pp. 67–72). Santa Margarita: Collins storytellers of Turkic Siberia. Seattle: University of Washington Press. Foundation Press. Vreeland, H. H. (1953). Mongol community and kinship structure. New Schloss, J. P., & Murray, M. J. (2011). Evolutionary accounts of belief in Haven: Human Relations Area Files. supernatural punishment: A critical review. Religion, Brain and Wallace, A. F. C. (1966). Religion: An anthropological view. New York: Behavior, 1(1), 46–99. McGraw-Hill. Shariff, A. F., & Norenzayan, A. (2007). God is watching you: Priming god Weller, S. C. (2007). Cultural consensus theory: Applications and concepts increases prosocial behavior in an anonymous economic frequently asked questions. Field Methods, 19(4), 339–368. game. Psychological Science, 18, 803–809. Wigger, J. B., Paxson, K., & Ryan, L. (2013). What do invisible friends Shariff, A. F., & Norenzayan, A. (2011). Mean gods make good people: know? Imaginary companions, God, and theory of mind. International Different views of God predict cheating behavior. The International Journal for the Psychology of Religion, 23(1), 2–14. Journal for the Psychology of Religion, 21, 85–96. Xygalatas, D. (2013). Effects of religious setting on cooperative behavior: Shaver, J. (2012). The behavioral ecology of Fijian religion. Doctoral A case study from Mauritius. Religion, Brain & Behavior, 3(2), 91–102. dissertation. University of Connecticut.