Writings‎ > ‎

In Mysterious Ways: On petitionary prayer and subtle forms of supernatural causation

Maarten Boudry & Johan De Smedt



The psychology of prayer and supernatural causation has received surprisingly little attention from empirical researchers. This paper discusses implicit belief patterns about the causal mechanisms by which God effects changes in the world. We offer a psychological account of belief in supernatural causation based on the existing empirical literature on petitionary prayer, incorporating mechanisms of psychological self-correction and rationalization, confirmation bias, and folk physics. We propose that religious believers ‘prefer’ modes of divine action that are subtle and indistinguishable from the natural course of events: given that the causal structure of our world is partly inscrutable, beliefs in subtle and unascertainable modes of supernatural causation will be compelling and cognitively appealing because they are more susceptible to occasional confirmation and less vulnerable to repeated disconfirmation. In other words, believers who request supernatural interventions that are subtle and indistinguishable from the natural course of events will have a better chance of finding themselves in a situation in which they can attribute the events in question to God answering their prayers. We argue that such individual psychological factors play a role in the cultural transmission of prayer practices as well, leading to culturally widespread beliefs in subtle forms of supernatural causation.

Keywords: petitionary prayer; supernatural causation; divine intervention; cognitive science of religion; psychological self-correction; theological incorrectness; cultural transmission; epidemiology of religious representations



God moves in a mysterious way

His wonders to perform

(William Cowper)


God is silent, now if we can only get Man to shut up

(Woody Allen)



Petitionary prayer is one of the most widespread expressions of religious behaviour, but surprisingly little is known with regard to its psychology.[1] Religious believers all over the world have attempted to engage in interactions with gods, spirits, witches, dead ancestors and other supernatural beings (Zaleski and Zaleski 2005), asking them to intervene on their behalf and bring safety, good fortune (or bad fortune for others), cure from illness, and many other goods. The psychological dimension of this interaction with supernatural beings is a relatively unexplored domain. According to believers, how does God go about answering prayers? Under which circumstances may one expect supernatural beings to intervene in the natural world? In the gospel of Mark, we read that ‘What things soever you desire, when you pray, believe that you receive them, and you shall have them’ (Mark 11:24). However, even the devout will admit that these are rather high hopes, and often the book of Psalms is more on the mark: ‘Why do you stand afar off, O Lord? Why do you hide yourself in times of trouble?’ (Psalm 10:1) In the Christian tradition, the problem of God’s silence in times of trouble has been pondered by countless theologians and ordinary believers, and as with the classical problem of theodicy, many ingenious rationalizations have been put forward (e.g. Murray 1993; Howard-Snyder and Moser 2002; Swinburne 2004; but see Schellenberg 2006).

In this paper we discuss a related but equally pertinent question: what implicit beliefs do people entertain about the mechanisms through which supernatural causation is effected in the natural world? If people see God as an agent capable of bringing about things in the natural world, of which He is not a part, what implicit assumptions do they make about his modus operandi[2]? More technically, how do people make sense of the causal connection leading from supernatural cause to natural effect (for a philosophical account, see Fales 2010)? We develop an epidemiology of religious representations concerning petitionary prayer and supernatural causation, on the basis of the concept of theological incorrectness and cognitive constraints on the formation and dissemination of religious beliefs. As a point of departure, we discuss a series of original experiments on the psychology of prayer conducted by Justin L. Barrett, a cognitive psychologist and leading figure in the cognitive science of religion (Barrett 2001).

Theological correctness and incorrectness

In the newly emerged field of Cognitive Science of Religion (CSR), researchers have brought findings from cognitive and developmental psychology and cognitive anthropology to bear on the explanation of religious ideas and practices (for a recent overview, see Barrett 2007; Barrett and Lanman 2008). CSR researchers argue that the formation and dissemination of religious representations is channelled by a number of domain-specific cognitive systems that are stable across cultures and that emerge in early childhood (e.g. Spelke and Kinzler 2007). In particular, Barrett has argued that a strong inclination to detect agency in the natural environment is a normal part of human psychology. The evolved mental module responsible for this inclination, which he termed the Hyperactive Agency Detection Device, makes us predisposed towards believing in supernatural entities such as invisible ancestors, immaterial spirits, animals that can change shape, ghosts, holy mountains, etc. In a review article of the field, Barrett takes as the main tenet of CSR the thesis that much of what is typically called religion may be understood as the natural product of aggregated ordinary cognitive processes’ (Barrett 2000: 29).[3]

One interesting finding that has emerged from this cognitive research is that people use different versions of the same religious concept under different cognitive load demands (Barrett and Keil 1996; Barrett 1999). Barrett observed that orthodox theology typically dictates properties of supernatural beings that are highly counterintuitive and that strain our cognitive resources, e.g., omnipotence, omniscience, eternal existence. When questioned about their opinions and given some time to reflect, people profess to accept official theology, but when they are engaged in ‘online’ tasks, applying religious concepts in practice, they make tacit assumptions that violate official theology. Instead, believers tend to fall back on more intuitive and anthropomorphic versions of supernatural beings. Barrett (1999) has coined the term ‘theological correctness’ to describe this phenomenon. The related concept of theological incorrectness describes the tendency of believers to stray from official theology if the latter is too cognitively burdensome (Slone 2004). Boyer (2001: 285) has called this the ‘tragedy of the theologian’. For example, while Christian doctrine stipulates that God can attend any number of events at the same time, people are caught reasoning as if God answers one prayer and then shifts his attention to the next. To take another example, D. Jason Slone (2004) has argued that the Calvinist doctrine of predetermination is ‘maximally counterintuitive’, because it leaves no room at all for human free will; hence, over time, it yielded to a conception of God that is more consonant with our intuitions.

The practice of prayer faces a number of other theological paradoxes that are shared by all religious traditions entertaining the notion of supernatural omniscience (Johnson 2005; Bering and Johnson 2005), in particular the three great monotheistic religions. First, praying to an all-knowing being such as God appears rather pointless, because God is supposed to be aware of my problems in any case. Second, if we make the additional assumption that God is morally perfect, and assuming that I request something morally good, God would have already done what I demand from Him anyway. But then in what sense can praying have any effect (Stump 1979)? If people consistently paid heed to these theological doctrines, they would not be engaged in petitionary prayer. Nevertheless, religious believers who assent to this official theology cannot help but inform God about their problems, just as they would do with ordinary social actors who have no full epistemic access to their inner mental life. According to Barrett, ‘[t]he simplification of concepts from the theological to the religious level appears to consist of a systematic distortion of features such that they more closely resemble intuitive ontological assumptions’ (Barrett 2000: 30).

In this paper, we focus on another form of cognitive tension that arises from traditional theology with respect to prayer. If God is an all-powerful Being, prima facie He may perform actions in almost any way He chooses.[4] Normal human agents are known to have limited causal powers: they have a spatially located body they can control to effect physical causation (but with limited strength and dexterity), they can communicate through speech within a certain range, as well as through indirect means (like cell phones or signalling), etc. With a supernatural agent like God, things are different. Lay religious believers generally do not receive a lot of instruction regarding the mode of operation of the divine being to which they direct their requests (Barrett 2001: 268). A similar problem applies to any supernatural being who is conceived as very powerful and capable of action in the natural world. As the theory of theological incorrectness suggests, however, the praying habits of ordinary believers need not reflect the theological doctrine of omnipotence.

Modes of supernatural causation

In an interesting paper on petitionary prayer among North American Protestants, Barrett (2001) applies the theory of theological correctness to divine omnipotence in Christianity. Christian theology does not specify how God precisely affects changes in the world. Barrett predicts that the silence of orthodox theology with respect to God’s preferred modus operandi leaves a ‘theological vacuum’ (Barrett 2001: 268). To fill this theological vacuum, present-day philosophers of religion have come up with sophisticated accounts of divine causation, such as, e.g., God acting on the level of quantum indeterminacies (Murphy 1995). However, folk conceptions of (divine) causation exhibit less explanatory depth, but rather rely on a vague notion of an underspecified generative link between cause and effect (De Smedt and De Cruz in press). Relying on developmental and cognitive psychological theories, Barrett (2001) argues that humans intuitively make a distinction between three modes of causation: mechanical, biological and psychosocial[5]. Accordingly, God may act either mechanically (materializing or removing physical objects, influencing physical processes), biologically (affecting the health of living beings, e.g., healing a person or striking him with disease), or psychosocially (influencing psychological states, e.g., relieving my pain, giving me the strength to face an ordeal, making someone fall in love with me). Past research has established that believers intuitively conceive of God as being located in some distant place (viz. Heaven), even if their official theology dictates otherwise (Barrett and Keil 1996). Furthermore, believers know from experience that normal social actors are bad at mechanical ‘action at a distance’, whereas they are good at ‘affecting psychological states [...] at a distance’ (Barrett 2001: 260). Because religious believers imagine God as a ‘human-like agent far away’ (Ibid.), they think He can be expected to act psychosocially rather than mechanically, thus ignoring the doctrine of divine omnipotence in practice. Indeed, in a series of experiments with Protestant subjects, Barrett found a praying preference among his subjects for psychosocial instead of mechanical or biological acts. Subjects were presented with a number of fictitious scenarios describing a troublesome predicament in which divine help would be welcome. For each scenario, they were presented with a mechanistic, a psychological, and a biological solution, and asked to rate how likely they would pray for that particular solution. Barrett found that subjects preferred the psychological (M = 5.77) solution over either the biological (M = 4.93) or mechanistic one (M = 4.23), with a significance level of 0.001.

Barrett correctly points out that psychosocial action is less constrained in the sense that it does not require physical contact, and that the intuitions of even young infants are sensitive to this fact (Woodward, Phillips, and Spelke 1993; Spelke, Phillips, and Woodward 1995). However, psychosocial action reaches its own limits at distances not far beyond those of mechanical action. Leaving aside modern telecommunication technology, psychosocial action by way of speech and body language is limited to a few (tens of) meters. Admittedly, some methods exist for extending the radius of psychosocial action – waving from a distance, sending a courier, emitting smoke signals – but these are limited in speed and efficiency. Similar extensions are available for physical and biological action as well, e.g., throwing a missile, sending a drug or poison. At small and medium distances, psychosocial action is very efficient indeed, but beyond that point it soon reaches limits comparable to those of mechanical action (which, of course, is related to the fact that psychosocial action is really a special case of mechanical action). In other words, at great distance, where God presumably resides, a normal social actor would not be any more efficient in effecting psychosocial action than mechanical action. On the other hand, we already possess indirect evidence that people can conceive of psychosocial action at a large distance. After all, prayer typically consists of a silent mental act directed at God. If people believe that God listens to their prayers, then they believe that God can be reached by purely psychosocial means. Maybe people just expect that God will return psychosocial requests with psychosocial actions?

The question remains whether, as Barrett supposes, mechanical action at a large distance strains our causal intuitions more than psychosocial action does, and is therefore intuitively less preferable. Although theoretically plausible, we think this hypothesis sits uncomfortably with the available empirical data on religious prayer practice and belief in supernatural causation. As for Christianity, one of the most common reasons for prayer are health issues (Schmied 2007; ap Siôn 2009), and Christian believers all over the world are firmly convinced that God can answer prayers by effecting a miraculous healing – a belief that has sometimes met with dire consequences (Peters 2008).[6] According to a 2004 survey by the National Center for Health Statistics (Barnes et al. 2004), 43 percent of the American adult population had prayed for their own health in the previous year. As is well known, the official procedure for beatification in the Roman Catholic Church requires a miraculous healing ‘from beyond the grave’.[7] The reliability of such miracle reports has stirred controversy for centuries (Hume 2000 [1748]; Peters 2008; Dawkins 2006), and ever since Francis Galton sparked the great prayer-gauge debate at the end of the 19th century (Galton 1872; Zaleski and Zaleski 2005; Mullin 2008; Park 2008), the alleged therapeutic effects of petitionary prayer have received attention from medical researchers (for a recent example, see Benson et al. 2006), and even the effect of prayer on plant growth has been investigated (for an overview, see Francis and Evans 1995).

Belief in supernatural causation of a biological and mechanical sort is a recurrent feature of religions across the world, as is the practice of appeasing, petitioning and asking favours from supernatural beings (Zaleski and Zaleski 2005). In his worldwide survey of 186 different cultures, Murdock (1980) reports that, in every single one of them, illness and misfortune are attributed to the actions of supernatural beings. Typical requests for non-psychosocial action include praying – or performing rituals – for rainfall or good harvest, for averting natural disasters, for protection on the battlefield, for pregnancy, etc. This solid tradition of mechanical and biological interventions is also reflected in virtually all religious scriptures, where supernatural beings are portrayed as capable of performing all kinds of non-psychological feats (next to psychosocial ones). Many examples from the Old and New Testaments are well-known: God parting the Red Sea, tearing down the walls of Jericho, appearing in a burning bush, healing king Hezekiah and the prophetess Miriam, and of course, impregnating a virgin and physically coming down to earth as Jesus.

In sum, this pervasiveness of non-psychosocial forms of divine intervention is problematic for Barrett’s (2001) claim that people find such actions too counterintuitive and that therefore they intuitively  expect forms of psychosocial action from God.

The plausibility of supernatural causation

Is it plausible that people prefer one type of supernatural causation over another, as Barrett suggests? Research has shown that believers do not show any particular interest in the causal mechanisms by which supernatural beings effect changes in our world, and instead like to focus on the motivations and intentions of these entities (Boyer and Bergstrom 2008: 121). But there may be other ways in which believers develop tacit preferences for certain modes of supernatural causation. In particular, both cognitive and external constraints may influence the formation and dissemination of religious representations. Researchers in the theological incorrectness framework have been mainly occupied with the ways in which ordinary cognition drives people away from theologically correct doctrines, but they have paid relatively little attention to the way in which external factors, relating to the regularities and vicissitudes of the natural world, can effect similar constraints. In his book on theological incorrectness, D. Jason Slone even writes that CSR researchers should not be much concerned with whether religious representations ‘refer to external realities’:

The content-claims of religions are peripheral to the actual object of study in the cognitive science of religion. [...] Whether or not gods exist makes little or no difference at all to the study of brain mechanisms that are involved in the production of religious thought and the performance of religious actions (Slone 2004: 47).

Slone is right that external reality does not affect the innate brain mechanisms responsible for religious beliefs (although of course it does so indirectly via the mechanism of natural selection), but we think it is self-evident that external reality must in some way influence the specific representations that these cognitive mechanisms give rise to. If we want to develop a realistic epidemiological model of (religious) representations (Sperber 1996), we have to take into account both our cognitive architecture and the external stimuli on the basis of which our cognitive apparatus is operating.

In the case of petitionary prayer, our epidemiology of representations must be informed by the relevant scientific knowledge on supernatural causation, in particular that there is no convincing evidence for the efficacy of petitionary prayer or other forms of supernatural intervention in the natural world (mechanistic, psychological or otherwise). The most extensive and careful studies of petitionary prayer have not shown statistically significant benefits of petitionary prayer (Benson et al. 2006: 378-381; Park 2008; Hines 2003; Matthews, Conti, and Christ 2000).[8] If we lived in a world where prayer were regularly answered (or some prayers, or maybe only those of some religious creeds), this would clearly make a difference on the formation and dissemination of prayer practices and beliefs. For example, if the prayers of a certain religious creed were to work reliably for curing disease, we could imagine the news to spread like wildfire, and surely that religion would rapidly win new converts. There is no historical evidence for the existence of such a religious tradition. In light of the lack of scientific and historical evidence for the efficacy of prayer, let us therefore assume the null hypothesis, viz. soliciting the help of supernatural beings for bringing about natural effects is not efficacious. Even if one wants to leave the door open for supernatural causation, one should at least acknowledge that, for most religious believers, the events following prayer are not a simple function of the desires expressed through it.

To see how this is bound to affect the pattern of prayer beliefs and practices, consider the following scenario.[9] Suppose I am cast up on a desert island with no food, no drinking water and no prospect for help. I know that ships pass by occasionally, but I have no idea when the next ship is due. If I direct a prayer to God and ask Him to materialize food and water before my eyes, I will wind up being disappointed. By contrast, if I pray to God to grant me the strength to endure my ordeal until the next ship passes by, my predicament is a little different. If indeed I manage to survive until the next ship comes by, I might be tempted to attribute my rescue to God. We propose that people prefer the latter form of prayer not because of the psychosocial nature of the request, but rather because the effects asked for are all but indistinguishable from a natural course of events (i.e. from the range of possible outcomes that are likely to occur on a purely natural account). Note that the two characterizations need not concur. If I pray for rainfall (for drinking water) or for a ship to come by, I am clearly requesting physical acts from God, but this time of a far more subtle sort than in the case of food materialization.[10] Maybe clouds were already packing together, and a ship might already have been on its course to my island. In these cases, no less than in the psychosocial example, it is impossible to distinguish divine action from the contingent and natural unfolding of events. Has God really intervened on my behalf, for example by subtly steering the ship’s course, or influencing ocean currents, or manipulating the captain into taking a different route? There is no way to ascertain this, as there is no way to know for sure whether I would in any case have survived a few days without food.

Subtle divine action

While the examples offered above (rainfall, the passing of a ship, perseverance) involve events that are indistinguishable from the natural course of events, it is important to note that such events are not guaranteed to take place in nature’s normal course. It would be strange to pray for an outcome that can be reliably expected to occur in any case, even if that would involve supernatural intervention of a sort that is perfectly indistinguishable from the course of nature. For example, I will not pray to God to make objects fall to the floor when I drop them, because I can reliably expect gravity to do the job. In other words, if the event prayed for is certain to take place regardless, why bother praying? In the general discussion of his prayer study, Barrett (2001: 268) suggests that, in addition to psychological causation, people may also prefer modes of divine action that are ‘ambiguous’ (see also Barrett and Lanman 2008: 116). The point is echoed by Hood, Hill, & Spilka (2009: 47) in their discussion of religious attribution: ‘situations involving high ambiguity and high threat may have the greatest likelihood of calling forth religious explanations’. This in turn is reminiscent of Bronislaw Malinowski’s famous observation that the Trobriand islanders make more extensive use of magical ritual in situations involving higher uncertainty and unpredictability:

While in the villages on the inner lagoon fishing is done in an easy and absolutely reliable manner by the method of poisoning, yielding abundant results without danger and uncertainty, there are on the shores of the open sea dangerous modes of fishing and also certain types in which the yield greatly varies according to whether shoals of fish appear beforehand or not. It is most significant that in the lagoon fishing, where man can rely completely on his knowledge and skill, magic does not exist, while in the open-sea fishing, full of danger and uncertainty, there is extensive magical ritual to secure safety and good results (Malinowski 1925/1992: 30-31).

The notion of causal ambiguity is an important one, but it needs to be explicated more clearly. For example, while Barrett contrasts ambiguous causation with forms of ‘mechanistic causation’, we think that mechanistic causation can be effected in ambiguous ways as well. We can distinguish at least three kinds of settings in which supernatural causality is rendered subtle and unascertainable.

(i)                supernatural agents may interfere with or manipulate complex, stochastic processes in which causal relations are difficult to assess, e.g., weather phenomena, natural disasters, luck on a battlefield, success in chance and sport games.

(ii)              they may influence natural processes that are either invisible or difficult to observe directly, and whose causal determinants are poorly understood, e.g., being cured from or stricken with disease, conceiving a child.[11]

(iii)            they may act as partial causes in conjunction with natural causes (Lupfer, Tolliver, and Jackson 1996: 388-389), in a way that makes it difficult to disentangle the respective contributions, e.g., giving me the strength to win a duel, helping me finish an exam, supporting a bridge that is on the brink of collapse.

In all these cases, we have no full epistemic access to the causal relations and causal antecedents responsible for the effect, which allows our minds to (partly) attribute the events to supernatural agency. To put it in another way, explanations in terms of supernatural causation are parasitic upon types of events whose natural occurrence is uncertain and whose nexus of causal antecedents is not fully transparent , i.e. not open to epistemic access. Note that, although the supernatural mode of action may be subtle and unascertainable, this does not mean that the alleged effect attributed to supernatural intervention is unimpressive. For example, recovery from a lethal disease or rainfall for the crops are quite tangible and spectacular results for the people who benefit from them. In general, it is well-known that many religious believers show no hesitation in attributing spectacular natural disasters (or the averting thereof) to supernatural intervention.

When pope John Paul II was shot and severely wounded in an assassination attempt in 1981, he believed that the ‘motherly hand’ of Our Lady of Fátima ‘guided the bullet’s path’, enabling the pope to stop ‘at the threshold of death’ (Stanley 2000). Any form of robust physical or biological intervention would have sufficed to protect the pope, but he believed that the Virgin Mary interfered in this particular, subtle way: not by preventing the gunman to shoot in the first place, or by directing the bullet away from his body altogether, but by steering its course ever so slightly so that he, though severely hurt, just managed to survive the assassination attempt.

Similarly, people have no difficulties with praying for biological interventions per se, such as curing from arthritis or cancer, but they are unlikely to pray for an amputated limb to grow back. Likewise, praying for rainfall to ensure good harvest seems an acceptable praying habit, but asking God to materialize full-grown crops before your very eyes seems a little unusual, even if God is omnipotent. The point is summoned well in Le Jardin d'Épicure by Anatole France (France 1900: 204), in which the French writer visits Lourdes with a companion who, upon seeing all the braces and crutches hanging there as evidence of healing, remarks: ‘A single wooden leg would have been quite more convincing.’[12]

By its very nature, psychosocial causation (e.g., possessing courage or mental strength, being relieved from anger or depression, etc.) is typically more complex and less observable compared with many forms of biological and physical causation. For that reason, it is plausible that the ‘preference’ for subtle modes of causation often translates into requests for psychosocial causation, as Barrett’s research suggests. However, whereas Barrett’s (2001) account entails that forms of biological and physical causation are counterintuitive in general and hence less preferred, our hypothesis predicts that people have no cognitive difficulty in conceiving such acts and hence are not hesitant to request them, provided that their mode of causality takes on subtle and unascertainable forms. Note that, while we argue that people regularly pray for supernatural interventions that de facto involve physical and biological causation, we do not claim that people bother to cognitively represent the precise modus operandi, either when making their requests (through prayer or ritual) or when attributing some subsequent event to a supernatural actor answering their prayer. As Boyer and Bergstrom wrote, people are less interested in the precise causal mechanisms of supernatural intervention than they are in its effects and in the agents responsible for them: 

People assume that the ancestors or gods are involved in various occurrences (bad crops, illness, death, etc.) but generally do not bother to represent in what way they bring about those states of affairs (Boyer and Bergstrom 2008: 121).

For example, Roman Catholics have elaborate beliefs about the intercessory capacities and responsibilities of different patron saints (i.e., the specializations of the saints concerning occupational hazards, particular nationalities, ethnic or social groups, and particular illnesses and afflictions), but they do not bother to represent the specific causal mechanisms employed by their protectors.[13]  Often, a superficial similarity to something in the saint’s life seems to suffice (e.g., Veronica who allegedly wiped Jesus’ face as he bore the cross to Golgotha became patron saint of photographers). This lack of interest in causal understanding aligns well with the idea developed by Dan Sperber and Pascal Boyer that religious explanations are ‘relevant mysteries’ (Sperber 1996: 73; Boyer 2001: 14): they do not so much explain events in terms of more simple and familiar processes, but instead they make use of salient and evocative mysteries. In particular, religious beliefs are psychologically satisfying because they allow people to attribute natural occurrences to agents, and to explain them in terms of their folk psychology (motivations and intentions)—such explanations give a feeling of epistemic satisfaction, even if they are superficial (Wegner 2003)

The psychology of supernatural causation

How does this preference for subtle divine action develop and play out psychologically? Janssen, De Hart & Den Draak (1990) and Barrett (2004) have drawn attention to the rather indeterminate and abstract way in which religious people often describe the effects of their prayers (support, blessing, trust, etc.), appearing in stark contrast to the concrete needs which typically occasion their prayers (e.g., serious illness). An interesting suggestion for this discrepancy is hinted at by Janssen et al., but not further developed in their paper: ‘It could be argued that people adapt the intended effects to the experienced effects, accepting a principal discrepancy between needs and effects’ (Janssen, De Hart, and Den Draak 1990: 105). Indeed, even if the intended effect of my prayer is something like sudden and full recovery from illness, it is plausible that people will eventually lower their expectations. For example, people could still pray for eventual recovery while accepting a longer period of illness, or they could resign to their predicament and ask God to support them psychologically. Consider the famous Serenity prayer written by the Protestant theologian Reinhold Niebuhr:

God grant me the serenity / To accept the things I cannot change / Courage to change the things I can / And wisdom to know the difference.

Instead of asking God to bring about a certain state of affairs directly, I ask him to influence my psychological attitude regarding my predicament. As we argued above, this mode of action (condition (iii), as a partial cause cooperating to a natural effect) is more difficult to ascertain than straightforward intervention on God’s part. Moreover, note that the very structure of the Serenity prayer can accommodate a certain amount of failure. If I succeed in changing whatever it is that I wanted to change, this is because God granted me the courage to do so. On the other hand, if I fail to change whatever it is that I tried to change and eventually I have to give up, this is because God granted me the serenity to realize that I could not change it anyway (and, of course, to know the difference).

In general, religious believers who request supernatural interventions that are subtle and indistinguishable from the natural course of events, will have a better chance of finding themselves in a situation in which they can attribute the events in question to God answering their prayers. A better chance, that is, compared with those who expect divine firework displays, in the sense of manifest violations of the natural order. As Nicholas Humphrey wrote, people must have known all along that full-fledged and palpable miracles just don’t happen: ‘They must have known these sobering truths because time and again they and their fellow human beings must have come slap up against the evidence for them’ (Humphrey 1995: 54). It seems reasonable to expect that people adapt their wishes and expectations to such sobering experiences, and that they won’t keep praying for unambiguous and robust supernatural acts if the results are invariably disappointing. In other words, such a process of psychological self-correction will steer believers away from demanding divine acts of a robust kind, instead fostering a psychological preference for modes of action that are ambiguous and less vulnerable to clear refutation (Boudry and Braeckman 2011).

This is not to say that religious believers who have ‘learnt’ not to pray for robust divine action will never find themselves in a situation in which their prayers remain apparently ‘unanswered’. If we maintain the null hypothesis on the efficacy of prayer, even prayers for subtle modes of divine action will appear to be answered at most occasionally.[14] Indeed, the history of prayer itself bears witness to the simmering doubts about its efficacy, as Zaleski & Zaleski noted:

the sheer abundance of devout tracts exhorting the faithful to pray often, pray fervently, and pray with confidence in achieving desired results suggests that belief in the efficacy of prayer has always needed some degree of shoring up (Zaleski and Zaleski 2005: 333).

The frequency of apparent confirmation will depend on the probability function of the event type in question (e.g., spontaneous remission of a disease, rainfall after a drought), but in any case, psychological research has established that beliefs may become entrenched even on the basis of a small number of apparent confirmations. People are prone to confirmation bias (Nickerson 1998), which means that they pay attention to and remember confirmations of a favoured hypothesis, while they quickly discard or explain away adverse evidence. More specifically, the psychological literature on cognitive dissonance and motivated reasoning (Kunda 1990; Aronson 1992; Tavris and Aronson 2008) suggests that, when firmly held beliefs are confronted with apparent failure, people rely on a repertoire of rationalizations. As Barrett noted in a later discussion of his prayer study (Barrett 2004), the whole idea of addressing a request to a whimsical or possibly reluctant supernatural agent suggests some ways of accommodating failure: ‘prayer commonly assumes the possibility that a request could be approved, denied, or put off until a later date’ (Barrett 2004: 71). In such a setting, a prayer that appears to have failed may in fact be granted ‘on a different timetable’ or just be denied for some good reason that we mortals cannot fathom. In this way, as Barrett (2001: 74) notes, ‘negative evidence rarely threatens belief in God’. Other rationales are possible for occasional failure: ‘My prayer did not fit into God’s plan’ – ‘Too much answered prayers would spoil us’ – ‘My praying ritual was not performed correctly’ – ‘The purpose of prayer is ‘to construct the soul, not to instruct God’ (Augustine)[15] – ‘God cannot be coerced through prayer’ (for a related discussion, see Martin 2004: 118-121). If believers are highly committed to their faith, we can expect them to be motivated to maintain belief in the goodness and omnipotence of God (e.g. Kushner 1981), and to explain failure by what Evans-Pritchard has termed ‘secondary elaborations’ (see below).[16]

The persistence of the confirmation bias ensures that even a small number of apparent successes may succeed in outweighing the instances of apparent failure (Barrett 2004: 74). In the case of prayers for robust divine action, however, there is little or no room for apparent confirmations. Thus, if we confine ourselves to personal experiences regarding the efficacy of prayer, the point is not that disappointment and cognitive dissonance will never arise for those who prefer subtle modes of action on God’s part, but that those who demand robust actions will always encounter failures that occasion the need for some rationalization, and will never encounter occasional personal successes to compensate for the failures. The confirmation bias needs some events to be biased towards.

Causal intuitions and folk physics

Besides this form of psychological adjustment, one may argue that there are purely cognitive factors at play in the preference for subtle modes of causation. Research in developmental psychology suggests that, from early infancy, humans possess an intuitive core knowledge about spatio-temporal objects, which is sometimes designated as folk physics. Looking time experiments with children reveal a number of such implicit assumptions: 1) objects move as bounded and discrete wholes (cohesion principle ), 2) objects move along continuous and connected paths (continuity principle), 3) objects do not interact at a distance (contact principle) (Spelke 1994; Spelke and Kinzler 2007). It is not surprising that natural selection has endowed us with an intuitive grasp of these basic spatio-temporal principles, because they are apparent in the physical environment on which our ancestors depended for survival, at least at the scale of medium-sized objects.

Is it plausible that people dislike robust forms of supernatural causation simply because these violate deeply-engrained causal intuitions and folk physics? For example, the instant materialization or displacement of desired objects before my very eyes would violate the principle of continuity. Or, to give another example, if a physical obstacle were to suddenly disintegrate as a result of my prayer to remove it, this would surely violate the principle of cohesion. The problem with this argument is that precisely such breaches in the fabric of our natural world by supernatural agents often form the subject of religious narratives, legends and holy writ. The Biblical stories about a burning bush not being consumed by the flames, or about Jesus walking on water are clearly violations of the basic principles of folk physics. Furthermore, many religious traditions contain stories that seem to exploit violations of precisely the intuitive principles mentioned above. For example, Christian monks and saints are often believed to possess the ability of bilocation, i.e. being physically present at two different places at the same time, thus violating the principles of cohesion and continuity (Nickell 1993: 216-219). Similar stories are to be found in Buddhism, Hinduism, Neo-paganism, shamanism and many other religious traditions. Likewise, popular stories about psychokinesis and action at a distance in religion and magic clearly violate the contact principle. Examples from the Christian tradition include Jesus transforming water into wine, calming a storm on a boat with his frightened disciples, or cursing a fig tree and causing it to wither away.[17]

If the preference for subtle supernatural causation stemmed from the fact that robust causation violates causal intuitions and folk physics, we should expect to find the same cognitive bias in religious mythology and legends. What we find, however, is that firework displays by God are pervasive in distant hearsay and religious mythology, whereas they are hardly ever the subject of personal experience or first-person eyewitness testimony (see also for example Evans-Pritchard 1965 [1937]: 195-201). How religious believers make sense of this discrepancy is itself an interesting empirical question: maybe they believe that robust supernatural events befall only very exceptional people, or are confined to holy history, or that they no longer occur because of the lack of faith in modern times. In any case, it seems that the preference for subtle and ambiguous divine action in petitionary prayer cannot simply result from the constraints of causal intuitions.

The cultural transmission of prayer practices

Up to this point, we have described the formation of prayer beliefs and practices on the level of individual psychological mechanisms. We can now take this approach one step further and outline the ways in which larger cultural trends may emerge from these psychological processes and intuitions. In Dan Sperber’s epidemiological model of culture (Sperber 1990; 1996), which has had a formative influence on CSR, our shared cognitive make-up provides constraints through which the dissemination of representations is channelled. Small selection pressures in the transmission of beliefs, aggregated over many transmissions, will give rise to larger cultural trends. Liénard and Boyer (2006) and McCauley and Lawson (2002) have applied this approach to religious rituals, and we can now apply it to belief patterns about supernatural causation. Assuming that the facts of nature are the same for all, it is reasonable to expect that the disillusion with robust modes of divine causation will become part of the collective experience of religious communities (see also Barrett 2004: 70-74). Belief in the present-day feasibility and reliability of robust supernatural deeds, as opposed to distant hearsay and historical narratives, are unlikely to thrive if they are to compete with beliefs in subtle divine causation.

There are several modes of cultural transmission in which representations about supernatural causation may be passed on and disseminated in a religious community. Cultural transmission can be conceptualized as a flow of information between or within generations: horizontal, oblique or vertical cultural transmission.

(i)                vertical transmission (Cavalli-Sforza and Feldman 1981): children are initiated in prayer and other religious practices by their parents;

(ii)              oblique transmission (Cavalli-Sforza and Feldman 1981): this is similar to vertical transmission, though in this case learners do not acquire information about supernatural beings from their parents but from any member of the older generation (older family members, elders, priests);

(iii)            horizontal transmission (Cavalli-Sforza and Feldman 1981): people of the same generation acquire prayer practices from each other; 

On another level, cultural selection can be modelled in terms of biases that underlie the transmission between people:

(iv)             conformist bias (Henrich and Boyd 1998): members of a religious community tend to adopt the notions on supernatural causation that are most commonly held within that community;

(v)               prestige bias (Henrich and Gil-White 2001): (candidate) members of a religious community adopt particular ideas on supernatural agency because of the (perceived) prestige of some of that community’s affiliates holding such ideas.

Religious representations thus transmitted may include (a) assumptions about the things believers may reasonably expect from supernatural beings, (b) implicit conceptions regarding the modus operandi of supernatural beings, (c) historical narratives and received accounts of answered prayers, constituting templates for future interaction with supernatural beings, (d) explicit theological solutions for why supernatural beings act in certain ways but not in others. Obviously, as Sperber and Claidière (2008) observe, all learning is biased by content. In this case, it is not unlikely that the cultural transmission of prayer practices and individual experiences with prayer interact to favour beliefs in subtle and unascertainable modes of supernatural causation, i.e., to ask for divine actions that are (almost) indistinguishable from the natural course of events.

This general outline of pathways of cultural assimilation, which is based on theoretical models of cultural transmission, leaves many questions unanswered. In the context of this paper, however, we merely want to draw attention to the point that not every member of a religious community need go through the same stages of personal disappointment of the kind described by Humphrey above. By means of explicit or implicit religious instruction, cultural selection forces may supplement and reinforce intrapersonal selection to the same effect. To put it bluntly, I may refrain from asking spectacular displays by God either because I have myself experienced the disappointment following such requests, or because others have instructed me not to make them. For instance, theoretical modelling (Henrich and Boyd 1998) indicates that conformism is an adaptive strategy in a broad range of ecological circumstances: it is often advantageous to adopt a culturally transmitted trait (such as refraining from eating certain mushrooms) without attempting to verify the quality of the acquired cultural trait (e.g., trying out the mushroom anyway and becoming violently ill). A similar mechanism may be at work in the social transmission of petitionary prayer: one learns from one’s parents, elders or peers that it is no use asking for the regeneration of lost limbs or the appearance of food out of nowhere.

To illustrate this effect of accumulated experience on religious beliefs, consider a second study by Barrett, in which he uses the same scenarios but substituted God with a ‘comparably endowed super-agent’ (Barrett 2001: 264). Subjects were asked to consider a futuristic supercomputer (Uncomp) with comparable God-like powers, but ‘components physically located all over the earth’. The term ‘praying’ was replaced by ‘asking for’. Barrett found that subjects in the Uncomp group did not have any preference for either mode of action and explains this in terms of the non-locality of Uncomp. Our account suggests a different explanation: subjects had no experience with the fictitious Uncomp, whereas of course practising believers had plenty of experience with and culturally transmitted ideas about directing prayers to God. Whereas religious believers have had ample opportunity to adjust their conception of God’s modus operandi on the basis of experience, and have also been exposed to a long tradition of religious believers with similar experiences, none of this holds for Uncomp. A futuristic computer with God-like powers is simply a fictitious character invented by the experimenter, which, from the perspective of Protestant believers, is quite different from God. Thus, subjects have no reasons not to accept the stipulated omnipotence of Uncomp at face value, which explains their relative lack of preference for either mode of action.[18]

Note that there may be still other ways in which certain praying practices (or patterns of belief) may be conducive to a process of self-validation, and thus possess an ‘advantage’ in terms of cultural dissemination. First, belief in the biological healing powers of supernatural beings may achieve cultural success in virtue of the fact that genuine faith on the part of the person afflicted may engender a placebo effect, the result of which may afterwards be attributed to God’s help. Second, if I pray to God to give me the strength to face a difficult ordeal, my act of praying and my faith in God may increase my self-confidence and reduce stress levels, resulting in a form of self-fulfilling prophecy. In addition, we have to take into account a self-validating selection effect regarding the people who, other things being equal, are still around to recount their miraculous healing or rescue, as Nicholas Humphrey’s Law of the Efficacy of Prayer makes clear: ‘In a dangerous world there will always be more people around whose prayers for their own safety have been answered than those whose prayers have not.’[19]

Different religious traditions

Thus far, we have mainly focused on the Christian tradition, describing how the psychology of supernatural causation plays out in the practice of petitionary prayer directed to a single divine being. In cultures with very different supernatural beliefs and religious practices, however, we might expect to find similar preferences for subtle modes of action. An interesting case study is provided by Evans-Pritchard’s (1965 [1937]) landmark anthropological investigation of magic and witchcraft among the Azande in Sudan. The Azande believe that some members of their community are witches who possess the mystical power to injure and kill other individuals, to harm their crops, to make houses collapse, etc. Indeed, the Azande invariably attribute death, disease and other forms of misfortune to the malignant actions of witches, thus making no distinction between different modes of causation (psychosocial vs. biological or mechanical). As Evans-Pritchard noted, however, the Zande people are certainly not unaware of the natural causes leading to such events. Interestingly, they believe that the mystical cause of witchcraft acts through a chain of natural causes, making a subtle contribution as a ‘co-operating cause’ (Evans-Pritchard 1965 [1937]: 72). For instance, when the Azande attribute the collapse of a mud house to witchcraft, they know very well that, as it happens, a colony of termites had been gnawing through the pillars of the house and undermining its foundations. Although they accept this natural cause of the event, they insist that only witchcraft explains why this particular house collapsed at that particular moment. Again, we see that the actions of the supernatural agent – in this case, a human being endowed with supernatural powers – are believed to contribute as a partial cause to some event, which renders the supernatural modus operandi subtle and virtually imperceptible. As Evans-Pritchard notes, ‘[t]he attribution of misfortune to witchcraft does not exclude what we call its real causes but is superimposed on them and gives to social events their moral value’ (Evans-Pritchard 1965 [1937]: 73). Why don’t witches just make the house collapse at one stroke, instead of acting indirectly and through the efforts of termites? On the other hand, how would the belief that witches may destroy houses at a single stroke fare in the Zande community, compared with the belief that they consistently employ the services of ants or other seemingly natural causes?

Interestingly, a similar conception of indirect supernatural causation through natural causes also appears in Christian theology, such as in the work of Thomas Aquinas on the distinction between primary and secondary causes. According to Aquinas, causation by God (primary causation) and natural causes (secondary causation) operate on entirely different levels: God is the cause of all causes, the cause of the world's existence; He does not work apart from secondary causes, or in addition to them (for a discussion, see Johnson 1996). Rather, God works through the acts of finite agents: ‘the act of being is what secondary agents [creatures] produce through the power of the primary agent [God]’ (Aquinas 1258-1264: book 3, chapter 66, par 7).

The Azande also regularly consult oracles about the various threats of witchcraft and about the courses of action to take in their life. In the benge, one of the most respected Zande oracles, a poison is administered to a fowl, following a number of elaborate preparations, and a question is put to it. The oracle is believed to provide a yes/no answer depending on whether the animal survives or dies. After observing a series of such divinations, Evans-Pritchard notes that there is no objective way to predict whether or not the bird will die, given the amount of poison or the size of the fowl. To all intents and purposes, the fate of the fowl is a matter of chance.[20] As we have argued before, precisely such stochastic processes, to which human beings have no epistemic access, are psychologically optimal for belief in supernatural causation. Because they allow for regular confirmations, they are conducive to self-validation, a process that is augmented by the kinds of questions and the way these are typically phrased. Among the Azande, apparent oracular failures are explained away or reinterpreted by ‘evasive secondary elaborations’ (Evans-Pritchard 1965 [1937]: 319) that are provided for by the belief system itself: improper preparation of the poison, the violations of taboos, interference of witchcraft or evil magic, refusal of the oracle to give the right answer. Elsewhere we have argued that when the causal relations in a belief system are underspecified and the effects are ambiguous, this engenders subtle forms of inferential circularity, rendering the belief system impervious to adverse evidence (Boudry and Braeckman 2011).


In many religious traditions, in particular the three monotheistic faiths, supernatural beings are conceived of as very powerful or even omnipotent agents, who are able to act in almost any way they like. Experimental findings suggest, however, that religious believers typically do not respect such counterintuitive theological doctrines when they are engaged in everyday religious practice, even though they may endorse them when explicitly questioned and when given some time to reflect. The related concepts of theological correctness and theological incorrectness have sparked a renewed interest in the psychology of petitionary prayer, which is still a relatively unexplored domain. Barrett’s (2001) study on prayer is a very welcome exception, but his account suffers from a conceptual problem: the characteristic distance range of psychosocial as opposed to mechanistic action is largely similar, which complicates his argument about an intuitive preference for supernatural causation of the former kind.

On an empirical level, the hypothesis defended by Barrett sits uncomfortably with the fact that beliefs in supernatural beings acting mechanically and biologically are pervasive in the Protestant tradition he investigated, and indeed across cultures (Johnson 2005; Murdock 1980). The account of theological incorrectness we have put forward is informed by our best knowledge about the efficacy of prayer (leading us to adopt the null hypothesis). Viewed in that light, we saw that people will develop preferences for supernatural interventions that are subtle and indistinguishable from the natural course of events. These different modes of action need not be cognitively represented as such by believers – as indeed believers do not care much about modes of supernatural action – but they may emerge from the psychological mechanisms of self-correction and the proposed epidemiology of prayer practices.

Our account suggests that people stray from orthodox theology not only because of the way their minds work, but also in virtue of what the world looks like (and the interaction between both). In an epidemiology of religious representations, both our innate cognitive make-up and the structure of external reality impose selective pressure on representations. In particular, we argued that, given that the causal structure of our world is partly inscrutable, beliefs in subtle and unascertainable modes of supernatural causation will be compelling and cognitively appealing because they are more susceptible to occasional confirmation and less vulnerable to repeated disconfirmation. Psychological mechanisms of self-correction and basic principles of folk physics will steer believers away from beliefs in robust and palpable forms of divine action (Boudry and Braeckman 2011), an effect that is further enforced by cultural transmission. Future research may extend this approach to the attribution of supernatural agency in general (i.e., not related to prayer), for example in terms of divine punishment and retribution for moral transgression. It will be interesting to know whether, in such cases as well, believers are attracted to gods that move in mysterious ways.



We would like to thank Johan Braeckman, Stefaan Blancke, Helen De Cruz, John Teehan, the editors of Religion and several anonymous reviewers for their helpful suggestions and comments, which have substantially improved the paper. This work was supported by the Research Foundation Flanders (FWO) and by Ghent University [grant number COM07/PWM/001].



ap Siôn, Tania. 2009. Ordinary Prayer and the Empirical Study of Prayer Cards. Rural Theology 7/1: 17-31.

Aquinas, Thomas. 1258-1264. Summa Contra Gentiles. Retrieved from: http://dhspriory.org/thomas/ContraGentiles3a.htm.

Aronson, Elliot. 1992. The Return of the Repressed: Dissonance Theory Makes a Comeback. Psychological Inquiry 3/4: 303-311.

Atran, Scott. 2002. In Gods We Trust : The Evolutionary Landscape of Religion. Oxford: Oxford University Press.

Augustine, A. 1872/73. Oeuvres Complètes. Paris: Librairie de Louis Vives.

Barnes, Patricia M., Eve Powell-Griner, Kim McFann, Richard L. Nahin. 2004. Complementary and Alternative Medicine Use among Adults: United States, 2002. Seminars in Integrative Medicine 2/2: 54-71. DOI 10.1016/j.sigm.2004.07.003

Barrett, Justin L. 1999. Theological Correctness: Cognitive Constraint and the Study of Religion. Method & Theory in the Study of Religion 11: 325-339.

________. 2000. Exploring the Natural Foundations of Religion. Trends in Cognitive Sciences 4/1: 29-34.

________. 2001. How Ordinary Cognition Informs Petitionary Prayer. Journal of Cognition and Culture 1/3: 259-269.

________. 2004. Why Would Anyone Believe in God? Walnut Creek (Calif.): AltaMira Press.

________. 2007. Cognitive Science of Religion: What Is It and Why Is It? Religious Compass 1/6: 768-786.

Barrett, Justin L., F. C. Keil. 1996. Conceptualizing a Nonnatural Entity: Anthropomorphism in God Concepts. Cognitive Psychology 31/3: 219-247.

Barrett, Justin L., J. A. Lanman. 2008. The Science of Religious Beliefs. Religion 38/2: 109-124. DOI 10.1016/j.religion.2008.01.007

Benson, Herbert, J. A. Dusek, J. B. Sherwood, P. Lam, C. F. Bethea, W. Carpenter, S. Levitsky, P. C. Hill, D. W. Clem, M. K. Jain. 2006. Study of the Therapeutic Effects of Intercessory Prayer (Step) in Cardiac Bypass Patients: A Multicenter Randomized Trial of Uncertainty and Certainty of Receiving Intercessory Prayer. American Heart Journal 151/4: 934-942.

Bering, Jesse M., Dominic D.P. Johnson. 2005. "O Lord You Perceive My Thoughts from Afar": Recursiveness and the Evolution of Supernatural Agency. Journal of Cognition and Culture 5: 118-142.

Boudry, Maarten, Johan Braeckman. 2011. How Convenient! The Epistemic Rationale of Self-Validating Belief Systems. Philosophical Psychology  DOI 10.1080/09515089.2011.579420

Boyer, Pascal. 1994. The Naturalness of Religious Ideas : A Cognitive Theory of Religion. Berkeley (Calif.): University of California Press.

________. 2001. Religion Explained: The Evolutionary Origins of Religious Thought. New York (N.Y.): Basic Books.

Boyer, Pascal, Brian Bergstrom. 2008. Evolutionary Perspectives on Religion. Annual review of anthropology 37: 111-130.

Cavalli-Sforza, Luigi Luca, Marcus W. Feldman. 1981. Cultural Transmission and Evolution: A Quantitative Approach. Princeton, NJ: Princeton University Press.

Dawkins, Richard. 2006. The God Delusion. London: Bantam.

De Smedt, J., H. De Cruz. in press. The Cognitive Appeal of the Cosmological Argument. Method & Theory in the Study of Religion

Evans-Pritchard, Edward E. 1965 [1937]. Witchcraft, Oracles and Magic among the Azande. Oxford: Clarendon press.

Fales, Evan. 2010. Divine Intervention: Metaphysical and Epistemological Puzzles. London: Routledge.

France, A. 1900. Le Jardin D'épicure. Paris: Calmann Lévy.

Francis, Leslie J., Thomas E. Evans. 1995. The Psychology of Christian Prayer: A Review Ofempirical Research. Religion 25/4: 371-388.

Galton, Francis. 1872. Statistical Inquiries into the Efficacy of Prayer. Fortnightly Review 12: 125-135.

Henrich, Joseph, Robert Boyd. 1998. The Evolution of Conformist Transmission and the Emergence of between-Group Differences. Evolution and human behavior 19/4: 215-241.

Henrich, Joseph, Francisco J. Gil-White. 2001. The Evolution of Prestige: Freely Conferred Deference as a Mechanism for Enhancing the Benefits of Cultural Transmission. Evolution and human behavior 22/3: 165-196.

Hines, Terence. 2003. Pseudoscience and the Paranormal (2nd Ed.). Amherst, NY: Prometheus Books

Hood, Ralph W. , Peter C. Hill, Bernard Spilka. 2009. The Psychology of Religion: An Empirical Approach. New York: The Guilford Press.

Howard-Snyder, Daniel, Paul K. Moser. 2002. Divine Hiddenness: New Essays. Cambridge Cambridge University Press.

Hume, David. 2000 [1748]. An Enquiry Concerning Human Understanding. Oxford: Clarendon Press.

Humphrey, Nicholas. 1995. Soul Searching : Human Nature and Supernatural Belief. London: Chatto & Windus.

Janssen, Jacques, Joep De Hart, Christine Den Draak. 1990. A Content Analysis of the Praying Practices of Dutch Youth. Journal for the Scientific Study of Religion 29/1: 99-107.

Johnson, Dominic D.P. 2005. God’s Punishment and Public Goods. Human Nature 16/4: 410-446.

Johnson, Elizabeth A. 1996. Does God Play Dice? Divine Providence and Chance. Theological Studies 57: 3-18.

Kunda, Ziva. 1990. The Case for Motivated Reasoning. Psychological Bulletin 108/3: 480-480.

Kushner, Harold S. 1981. When Bad Things Happen to Good People. New York: Avon Books.

Lienard, Pierre, Pascal Boyer. 2006. Whence Collective Rituals? A Cultural Selection Model of Ritualized Behavior. American Anthropologist 108/4: 814.

Lupfer, Michael B., Donna Tolliver, Mark Jackson. 1996. Explaining Life-Altering Occurrences: A Test of the 'God-of-the-Gaps' Hypothesis. Journal for the Scientific Study of Religion 35/4: 379-391.

Malinowski, Bronislaw. 1925/1992. Magic, Science, and Religion and Other Essays. Prospect Heights, IL: Waveland Press.

Martin, Luther H. 2004. Petitionary Prayer. Cognitive Considerations. In Religion in Cultural Discourse. Essays in Honor of Hans G. Kippenberg on the Occasion of His 65th Birthday, ed. B. Luchesi and K von Stuckrad. Berlin: Walter de Gruyter, 115-126.

Matthews, William J. , Jim Conti, Theodore Christ. 2000. God’s Hmo: Prayer, Faith, Belief and Physical Well-Being’. Skeptic 8/2: 64-66.

McCauley, Robert N., E. Thomas Lawson. 2002. Bringing Ritual to Mind: Psychological Foundations of Cultural Forms. Cambridge Cambridge University Press.

Mullin, Robert Bruce. 2008. Science, Miracles and the Prayer-Gauge Debate. In When Science and Christianity Meet, ed. D. C. Lindberg and R. L. Numbers. Chicago and London: University of Chicago Press, 203-224.

Murdock, George Peter. 1980. Theories of Illness: A World Survey. Pittsburgh University of Pittsburgh Press.

Murphy, N. 1995. Divine Action in the Natural Order: Buridan’s Ass and Schrödinger’s Cat. In Chaos and Complexity: Scientific Perspectives on Divine Action, ed. Russell R.J., N. Murphy and A.R. Peacocke. Vatican City State: Vatican Observatory Publications 325-357.

Murray, Michael J. 1993. Coercion and the Hiddenness of God. American Philosophical Quarterly 30/1: 27-38.

Nickell, Joe. 1993. Looking for a Miracle: Weeping Icons, Relics, Stigmata, Visions & Healing Cures. New York: Prometheus Books.

Nickerson, Raymond S. 1998. Confirmation Bias: A Ubiquitous Phenomenon in Many Guises. Review of General Psychology 2: 175-220.

Park, Robert L. 2008. Superstition : Belief in the Age of Science. Princeton, N.J.: Princeton University Press.

Peters, Shawn Francis. 2008. When Prayer Fails: Faith Healing, Children, and the Law. Oxford: Oxford University Press.

Schellenberg, John L. 2006. Divine Hiddenness and Human Reason: Cornell University Press.

Schmied, Gerhard 2007. God Images in Prayer Intention Books. Implicit Religion 5/2: 121.

Slone, D. Jason. 2004. Theological Incorrectness: Why Religious People Believe What They Shouldn't. New York: Oxford University Press, USA.

Spelke, Elizabeth S. 1994. Initial Knowledge: Six Suggestions. Cognition 50: 431-445.

Spelke, Elizabeth S., K. D. Kinzler. 2007. Core Knowledge. Developmental Science 10/1: 89-96.

Spelke, Elizabeth S., Ann Phillips, Amanda L.  Woodward. 1995. Infants’ Knowledge of Object Motion and Human Action. In Causal Cognition: A Multidisciplinary Debate, ed. Dan Sperber, David Premack and Ann James Premack, 44–78.

Sperber, D., N. Claidière. 2008. Defining and Explaining Culture (Comments on Richerson and Boyd, Not by Genes Alone). Biology and Philosophy 23/2: 283-292.

Sperber, Dan. 1990. The Epidemiology of Beliefs. In The Social Psychological Study of Widespread Beliefs, ed. George Gaskell and Colin Fraser. Oxford: Clarendon Press, 25-44.

________. 1996. Explaining Culture: A Naturalistic Approach. Cambridge, Mass.: Blackwell.

Stanley, Alessandra. 2000. Vatican Discloses the 'Third Secret' of Fatima. The New York Times, May 14.

Stump, Eleonore. 1979. Petitionary Prayer. American Philosophical Quarterly 16/2: 81-91.

Swinburne, Richard. 2004. The Existence of God. Oxford: Clarendon Press.

Tavris, Carol, Elliot Aronson. 2008. Mistakes Were Made (but Not by Me) : Why We Justify Foolish Beliefs, Bad Decisions, and Hurtful Acts. London: Pinter and Martin.

Wegner, Daniel M. 2003. The Mind’s Self-Portrait. Annals of the New York Academy of Sciences 1001: 212-225.

Woodward, Amanda L., Ann Phillips, Elizabeth Spelke. 1993. Infants’ Expectations About the Motion of Animate Versus Inanimate Objects. In Proceedings of the Fifteenth Annual Conference of the Cognitive Science Society. Boulder, CO, 1087-1091.

Zaleski, Philip, Carol Zaleski. 2005. Prayer : A History. Boston, New York: Houghton Mifflin Co.



[1] Naturally, praying practices are complex and can take a diversity of forms (Zaleski and Zaleski 2005). In this paper, however, we confine ourselves to petitionary prayer and similar practices in religious traditions.

[2] The notion of modus operandi will be used to denote the purported causal mechanisms by which God effects changes in the world. As we shall see further on, according to Barrett (2001), humans intuitively categorize causes into three modes of causation: mechanical, biological and psychosocial.

[3] Although we have framed our discussion in terms of religion, it need not be confined to traditional religious categories. In some cases, the practice of directing requests to supernatural entities may be better classified as instances of ‘magic’ or ‘superstition’ rather than (institutionalized) religion (see below). This need not be problematic, as one of the tenets of CSR’s naturalistic approach is precisely a dissolution of sharp boundaries between such categories as religion, superstition, magic and pseudoscience. By explaining religious beliefs and practices as predictable by-products of our basic cognitive architecture, the CSR approach automatically brings religion closer to other expressions of human nature.

[4] Of course, God cannot perform any actions that either involve logical contradictions (creating a round triangle), violate his other attributes (committing a sin, since He is morally perfect), or that are self-defeating (creating a stone that He cannot lift). Further limitations may apply, but they are not relevant to our purpose, as people rarely pray for divine actions that involve such outright contradictions. In this context, we merely want to argue that there is no a priori reason to expect an omnipotent (or very powerful) being to perform only certain types of action as opposed to others.

[5] One may object against Barrett’s threefold typology that it does not provide an adequate typology of causation. What is important in this regard, however, is how people intuitively think about different modes of causation. There does seem to be a rough-and-ready intuitive difference between the three kinds of causation, and this is all Barrett needs (note that our own argument will not depend on this distinction).

[6] It might be argued that people think of disease as a psychosocial affliction, e.g., possession by an evil spirit. Although this may apply in particular cases, it certainly does not work as a general solution. Nowadays, in contrast to pre-scientific times, religious believers are often well aware of the biological nature of their illness (e.g., tumor, infection), but that does not stop some of them from praying to God and attributing their healing to supernatural intervention. In any case, the psychologization of supernatural causation does not work at all in the examples of straightforward physical causation that we will discuss next.

[7] The alleged miraculous healing from beyond the grave (by the candidate saint) has to be investigated by the Consulta Medica, a board of physicians appointed by the Vatican, to determine whether the recovery was sudden and permanent, and to rule out scientific explanations for the healing.

[8] In addition, supernatural causation engenders a number of philosophical problems. For example, Fales (2010) has recently argued that ‘theo-mundane’ causation can only be accepted at the cost of revising foundational scientific principles such as the law of energy conservation.

[9] This scenario is similar to one that has been used in Barrett’s experiment (Barrett 2001: 261).

[10] Note that a ship is normally navigated by a captain, so that its course may be construed as a function of the captain’s psychological states, and God as psychologically influencing the captain. But we can as well imagine God to cause a deserted raft to be washed ashore.

[11] Note that there is partial overlap between (i) and (ii). Although the causal determinants of pregnancy and illness are partly stochastic, there are also tractable causal factors that may simply be poorly understood (in the case of pregnancy: infertile periods, means and frequency of intercourse, natural fertility; in the case of illness: malnutrition, contact with infected persons, genetic disposition). The difference between both categories is one of degree: weather phenomena are more stochastic than impregnation, and chance games are (ideally) purely stochastic.

[12] In French, the quote reads ‘Une seule jambe de bois en dirait bien davantage’. The remark is often wrongly attributed to Anatole France himself, who in fact disagreed with his companion and insisted that a wooden leg would not have impressed him any more than a crutch.

[13] An agent need not even be involved in the causal chain of supernatural beliefs, as is witnessed by the various superstitious and magical beliefs in various cultures (see for example Evans-Pritchard 1965 [1937]: 82-83). For  example, it is completely unclear how placing a horse-shoe above the doorway could bring about luck or misfortune, but this does not seem to stop people from finding such causal beliefs perfectly credible (Slone 2004: 103-120).

[14] Of course, religious believers themselves need not entertain something like the null hypothesis. All we mean to say is that a certain state of affairs (the inefficacy of prayer) acts as a constraint on belief formation. The null hypothesis is just a way to describe that state of affairs.

[15] 'ut ipsa [man] construatur, non ut Deus instruatur' (Augustine 1872/73: Epistola CXL, caput XXIX, 69).

[16] In the wake of scientific evidence on the inefficacy of prayer, many liberal theologians have relinquished the idea of divine interventions in the natural world altogether. Notably, since the prayer-gauge debate in the 19th century, theologians have begun reinterpreting the value of prayer on a purely moral and spiritual level, purging it from any form of miraculous interventionism. (Mullin 2008).

[17] Researchers in CSR have argued that religious representations are successful precisely because they violate intuitive expectations about the ontological categories to which they belong. To be more precise, religious representations have to be ‘minimally counterintuitive’ (Boyer 1994; Atran 2002) to be more memorable, which means that they display a small number of violations against a background of intuitively expected properties.

[18] Another and even more simple explanation is that Uncomp is explicitly described as having components ‘physically located all over the earth’ (Barrett 2001: 264, our italics), whereas God presumably is not to be thought of as a physical entity. This difference by itself may account for the greater preference for physical causation in the Uncomp condition.

[19] The ‘law’ was posted on the EDGE Question Center, 2004, retrieved from http://www.edge.org/q2004/page2.html#humphrey

[20] The same point holds for the other oracles in the Zande belief system, the termite oracle (dakpa) and the friction oracle (iwa) (Evans-Pritchard 1965 [1937]: 352-386).