Do people ever engage in “magical thinking” ?

Would you enjoy your cocktail less, if it came in a glass labelled “vomit”?

One solid result of cognitive psychology, or so it would seem, is that most people, regardless of education, opinion or personality, can be induced to think in magical terms given the appropriate stimuli and conditions. People will be reluctant to don a sweater if told that it used to belong to Adolf Hitler. They resist drinking from a glass of water in which an experimenter has briefly dunked a plastic cockroach. There is a great variety of such effects, initially demonstrated by Paul Rozin and Carol Nemeroff and replicated by many others, including Paul Harris in developmental studies.

This was salutary news for cultural anthropologists, who suspected that there was something deeply wrong with the notion that magical thinking was a prerogative of the Other, either quasi-naked people with bones through their noses, or less exotic peasants and barbarians with “pre-logical” mentality. So – we now know that we all are that Other, so to speak.

But does magical thinking actually exist? Do the experiments actually show it in action?



All a matter of interpretation



The empirical literature is solid. The results are not at issue here. People can be induced to make “magical” choices, the phenomenon is stable, resists rationalization, is found in most human cultures and is not confined to early development. The notion of an exotic primitive mentality is dead and gone. No problem there.

The problem lies in the cognitive computational description of the processes engaged. Consider a paradigmatic case, that of people who prefer to drink from a glass A labeled ‘H2O’ than a glass B labelled ‘vomit’. They have seen that water from the same pitcher was poured into both glasses, in some versions of the protocol they even wrote the labels themselves and stuck them on the glasses… yet they feel more comfortable drinking from one than the other. In such studies, the participants readily accept that there is no real difference between the two glasses, and that the notion of a magical connection is indeed absurd. Yet their choices are predictably swayed towards A.

Hence the apparently uncontroversial conclusion that somehow they do hold the magical belief suggested by the experimental protocol, albeit in a tenuous way.

But who is doing the believing?

It seems to me that this last inference is based on a rather vague and probably misleading view of the cognitive processes engaged, notably that:

[1] the mind includes a central belief-box where the organism’s current beliefs are stored and combined to produce new inferences;

[2] decision-making is the outcome of stored representations in that belief-box combined with a hierarchy of the organism’s goals, presumably stored in some “current preference box” buffer.

If these are valid assumptions, then the participant’s behavior (choosing glass A) requires a decision based on a preference (glass A is better than glass B) which itself entails a belief like “there is something bad about the contents of glass B” — so we can interpret the behavior as evidence for experimentally-induced magical belief.

But that is a rather odd description of belief and decision-making. In an alternative description of cognitive architecture – one so obviously true that I need not defend it here – the premises are rather different:

[1] The mind does not include a central belief-box but a (probably rather large) number of belief-adjudicating modules, automatically activated by the similarities in contents and scope among any n-tuplet of beliefs produced by different domain-specific modules;

[2] Decision-making is the outcome of current goal competitions, informed by competition between those belief-adjudicating systems.

Under these really straightforward assumptions, the participant’s behaviour suggests an interpretation that differs from the standard story.

First, what happens when people see a glass labelled “vomit” is that some threat-detection modules are automatically activated, as the label matches one of their input conditions – a cue indicating a substance dangerous to ingest. Other pieces of conceptual information, e.g. “The label truly represents the contents”, or on the contrary “I wrote and stuck this label on the glass myself”, “This is all a game suggested by the experimenter”, etc., do not enter in the threat-detection module’s processing because they simply do not match any of its input formats.

Second, the current state of the subject’s decision-making process can be described as follows: Most mental systems and modules have no input in the particular choice of glasses, because they are designed to focus on other matters. [ii] Some higher belief-adjudicating modules, which do process information like “I wrote and stuck this label on the glass myself”, etc., yield no preference for A or B, as they entail that both glasses are the same; [iii] one small set of modules (the threat-detection one and its daughters) is signaling a preference for avoiding glass A.

Since decisions are swayed by whatever competitive edge (however small) one plan has over alternatives, the participant proceeds to choose glass B.

Under this interpretation, neither the organism nor indeed any part of the organism can be described as holding the belief “there is poison in glass A”. This is true even of the threat-detection module, which need only convey information like “the word ‘poison’ denotes potential threats”, without any inference about the contents of the glass.

In a nutshell, an experiment like this is indeed revealing – but not of magical thinking. It shows that any change of preferences induced by some modular processes, somewhere in the mind, is sufficient to sway decision-making when all else (i.e. the preferences induced by other modules) is neutral with regard to that decision. Which is interesting, and should be studied, but does not require magical beliefs.

Now the same applies, I would suggest, to other, non-experimental cases of apparent magical beliefs. People tell you that if they paint a spider on a photograph of their neighbour, said neighbour will certainly get sick and thereby pay for his misdemeanours. It seems to me that all this can occur without any part of the mind being committed to a magical belief.

The inner voice of anthropological conscience

Now I hear the voice of classical anthropological wisdom shouting me down, with cries of “What on earth are you talking about? How can you deny the existence of magical beliefs when people (in some cases) actually tell you ‘Yes, I do believe in magic, I believe that the spider picture will make my neighbour sick’”?

The voice is loud but the argument is not compelling.

There is certainly a difference between people who choose glass A and say “but I know it really is silly” and those who choose glass A and say “I think there is actually some bad magic in glass B”. There is a difference in explicit discourse, but note that there is no difference in either their intuitions or in their behavior.

This would suggest that the explicit discourse is not what triggers the intuitions or behaviors, but is an interpretation of one’s own behaviors. That is, once we make choice that seem “magical” (as a result of the processes described above), we may have to justify them to others and ourselves. In some cultural contexts, like Rozin’s subjects, we can say “it’s silly but I can’t help it”. In other contexts, we can draw on a culturally salient model of “magic” to say “it does make sense and I like others know that it happens”. Given that there must be many occasions when our modular systems yield choices we cannot really justify, there is ample ecological space for the latter kind of discourse to become culturally stable.

The models developed by cognitive scientists to explain “magical connections” (see in particular Jesper Sørensen’s detailed analysis) are just as valid in the present scenario as before. But magical belief is not really necessary or sufficient for these models.

So the primitive Other was really like us all along. It is not that we have weird ideas too, but jut that they have well-designed adapted minds too.


  • Warren Winter 19 May 2011 (22:37)

    Would your modular interpretation not yield the prediction that explicit discourse containing magical explanations could moderate these intuitions, however they are generated? Given how robust nocebo effects are, I would be surprised if the expectations induced by such discourse wouldn’t amplify these visceral reactions.

  • Nick Connolly 20 May 2011 (00:59)

    Is the social aspect of these example being ignored? It is not magical thinking, for example, to be reluctant to be seen drinking from a glass marked ‘vomit’. Consider also the case of the jumper previously owned by Hitler – or a killer etc. Imagine a person who was eager to wear such a jumper, not because of magical thinking but because they identified with the beliefs of the owner and owning memrobilia was a way of indulging their interests. We would find a person who intentionally wore such a jumper repugnant because we assume that they are wearing it as symbolic act – an endorsement of original owner’s views. Naturally most people would therefore not wish to unintentionally engage in a similar act. This wouldn’t be magical thinking but rather some sort of social semiotic thinking.

  • Paulo Sousa 20 May 2011 (03:57)

    Salut, Pascal. Quite interesting piece, but let raise some doubt here. You said: Now the same applies, I would suggest, to other, non-experimental cases of apparent magical beliefs. People tell you that if they paint a spider on a photograph of their neighbour, said neighbour will certainly get sick and thereby pay for his misdemeanours. It seems to me that all this can occur without any part of the mind being committed to a magical belief. Now I hear the voice of classical anthropological wisdom shouting me down, with cries of “What on earth are you talking about? How can you deny the existence of magical beliefs when people (in some cases) actually tell you ‘Yes, I do believe in magic, I believe that the spider picture will make my neighbour sick’”? However, in these anthropological cases, I think it is much more interesting to characterize the representations involved as reflective beliefs a la Sperber than to avoid the label ‘belief’. There are two levels of representations involved here (the second one being optional, and, if present, subject to differences in terms of cultural elaboration): a) spider painted in neighbor’s picture causes neighbor sickness b) a magical mechanism of causation (e.g., something a la Frazer’s homeopathic principle) For the first level, it is clear that it is much better to use a notion of reflective belief — how would one explain the linguistic and non-linguistic behavior motivated by this type of representation? For the second level, when it exists, then theoretical propriety of the usage of the notion of reflective belief will depend on the level of cultural elaboration. If you just want to claim that in most cases there is no representation like “b” or that in most cases these b-representations are not elaborated enough to be called belief in any reasonable sense, then I would probably agree. Paulo

  • Gordon Ingram 20 May 2011 (09:35)

    I am with you on the need to recognise the Other in our own community. In my experience, evolutionary/cognitive anthropologists actually tend to be much better at this than traditional social/cultural anthropologists, due to the latter’s commitment to the idea that every culture is unique. I am more sceptical on the value of invoking “beliefs” (in any normal sense of the word) to explain the Rozin/Nemeroff type of result. I would not like to drink from a cup labelled VOMIT, not because I believe that anything bad would happen to me, but because I believe it would make me feel bad (a pretty true belief, I should imagine). This is classical, Pavlovian conditioning – negative response aversion – and I don’t see the need to move beyond this level of explanation here. Or can you point to a study along these lines that contradicts a Pavlovian prediction? Where things get much more interesting is in the potential for social validation to override this kind of response aversion. E.g., many Catholics might not like to drink red liquid from a glass labelled BLOOD, but they have no problem with taking the Eucharist. Is this because they have a genuine reflective belief that this is a good thing to do, or simply because it is a socially validated practice? An interesting research question, surely. It’s also interesting that the negative affect generated by the idea of drinking blood might originally have served as a commitment device during the period of early Christian persecution, before familiarity drained it of its power. And as Nick Connolly points out above, in cases where the behaviour is not socially validated, then adopting it can look like making a statement of some kind, which could add to the motivation not to do it. Has observability ever been manipulated in these experiments? It seems to me that there is a [b]huge[/b] difference between the essentially reactive (aversive) behaviour of not drinking from a glass labelled VOMIT, and Paulo’s much more proactive, goal-driven example of using a spider charm to effect neighbour sickness (which is much closer I think to Frazer’s conception of magical thinking). In the latter case a belief (=belief in the efficacy of what is doing) may be doing more of the work, but then again it may be that this kind of repressed aggression just makes people feel better! I’m not sure if there’s anything really [b]wrong[/b] with your “selecting between multiple beliefs” approach, but rather than making this idea of “belief” do all the cognitive work, I would contend that we will get a much fuller explanation of real-world phenomena like the Eucharist and spider curses if we can somehow integrate affective, reflective and social-cognitive levels of explanation. And I believe that “belief” belongs firmly on the reflective level!

  • Don Gardner 20 May 2011 (09:53)

    Pascal’s post–especially the last line–reminded me of Mounce’s ingenious thought experiment (‘Understanding a primitive society’ [[i]Philosophy[/i] (1973) 48:347-62]; roughly, a Wittgensteinian critique of anthropological treatments of the magico-religious). Mounce invites us to imagine being asked to push a pin into the eye of a good–let’s say, photographic–portrait of one’s mother. Nobody, he guesses, could do this without some misgivings. Yet, it is obvious that many would overcome their qualms (especially if the inducements were appropriate) in the light of what they (reflectively) believe about causal relations among physical objects. But now, Mounce asks, continuing with his grizzly example, if one had complied with the request and then learned that one’s mother had lost her sight in that very eye, who could escape the thought that pushing in the pin had — somehow — caused her to do so? (The abductive suggestion, it seems, simply could not be kept at bay.) Finally, Mounce asks, who amongst us could then bring him- or herself to push the pin into the other eye of the mother’s portrait? This and connected examples (such as swallowing a day’s worth of one’s saliva ordinarily versus storing it–no matter how hygienically–and downing it a few days later) show that sometimes the psychology underlying particular actions and reactions (including refusals to act) can be very complicated, but do they speak one way or the other about the sub-personal mechanisms involved? Likewise, traditional anthropological approaches to magic (mostly symbolist/functionalist, even when they are dressed up in more a la mode outfits) and their shortcomings seem to me to have little implication about sub-personal psychological processes. (This, even if the use of the label “magical” in experimental settings would be likely to figure in such processes). I was also a little mystified by Pascal’s statement that the difference between people who choose glass A and say “but I know it really is silly” and those who choose glass A and say “I think there is actually some bad magic in glass B” is a matter of what he characterises as “explicit discourse, rather than their intuitions or behavior.” The speech acts here are very different indeed, and the psychology underpinning them would likewise be expected to be quite different, I’d have thought. Of course, there are probably similarities too, but that just indicates–once again–that differences don’t have to be large to be decisive (psychologically, interpersonally, socio-historically).

  • Don Gardner 20 May 2011 (10:34)

    Gordon’s comment was not up when I posted mine. While he makes many very good points, I’m not sure about why we should see a “a [b]huge[/b] difference between the essentially reactive (aversive) behaviour of not drinking from a glass labelled VOMIT, and … using a spider charm” to make a neighbour sick, unless his point is about their respective cultural currency. Also, his mention of Frazer’s views reminded me that I’d intended to give a plug for the careful treatment of the heterogeneity of the thoughts and practices we anthropologists tend to bring under the rubric “magic” to be found in chapter 9 of Skorupski’s [i]Symbol and theory[/i] (a 1976 work that, sad to say, has been much misrepresented in the anthropological literature).

  • Gordon Ingram 20 May 2011 (11:57)

    Don, I guess the main difference between the vomit example and the spider example is that you can’t really explain the latter in terms of simple conditioning (unless the agent has directly witnessed enough examples of the spider curse “working”, which I doubt). So it requires a more complex explanation in terms of prospective planning and imagination. Some aspects of human behaviour are pretty complicated, of course, but other aspects (when Pascal’s “higher belief-adjudicating modules” don’t get involved) are just slightly more elaborate versions of what dogs etc do, and that is the distinction I was trying to draw. I totally agree with your first comment in terms of the importance of not neglecting the distinction between people who don’t drink from the glass but know that it’s “silly”, and those who don’t drink from it but think it’s “bad magic”. There may be no difference in behaviour in this particular experimental paradigm, but translated to other contexts, the difference in cognition may well lead to real differences in behaviour, due to the potential for blocking other reasons for action. If I am sick and I believe there is medicine in the glass, I can easily inhibit the aversive response to the label (or indeed the bad taste); but someone who sincerely believes that the label indicates bad magic may not inhibit this response so easily. This clearly can have tragic real-world consequences, as in cases where religious people refuse to allow medical treatment for their sick children. Finally I must apologise to Pascal as on re-reading your original post, I realise that I mistook the entire thrust of it, which was exactly to argue (as I did) that implicit beliefs [b]don’t[/b] do all the work! I think that what threw me off was your reference to “belief-adjudicating systems”, which made me think that you were modelling the lower-level modules in terms of implicit beliefs or similar. But there is still the problem that the case of the spider curse may not be assimilable to the case of the vomit drinking: it does seem to be a more genuine case of “magical thinking” in terms of a plan that is motivated by a belief in a particular action’s efficacy.

  • Benson Saler
    Benson Saler 21 May 2011 (17:48)

    In my understanding of Pascal’s blog, he distinguishes between what we can read off and infer of a person’s behaviors and intuitions on the one hand and the cognitive processes that underlie and account for those behaviors and intuitions on the other hand. While some of the former may strike us as instantiations of “magical thinking,” Pascal argues that we need not invoke such thinking in order to account for the latter. Indeed, where there are stable cultural traditions that include a theory of magic, such theories are more likely to constitute ways of explaining or justifying our intuitions and behaviors to ourselves and to others than of causing them. That position strikes me as plausible, especially in light of Pascal’s explication of it. As such, it makes an interesting contrast to the position espoused by the anthropologist Richard Shweder in his paper, “Likeness and Likelihood in Everyday Thought: Magical Thinking in Judgments about Personality” (Current Anthropology 18:637-58, 1977). Shweder argues that what is basic to magical thinking is the apperception of likeness and the giving of greater weight to apperceptions of similarity than to apperceptions of differences. Magical thinking, so conceived, Shweder holds, is different than probabilistic thinking (as understood, say, with reference to the four-celled contingency table). Magical thinking, Shweder maintains, is how most of us think much of the time: it is universal among humankind. Probabilistic thinking, in contrast, is limited to small numbers of people who have undergone the rigorous tutoring necessary to acquire it and who employ it in limited contexts. In short, Shweder argues, magical thinking is a common form of thought in our species regardless of socio-cultural differences whereas probabilistic thought is evinced mainly by scientists qua scientists in contemporary Western societies. Shweder attempts to support his thesis by drawing on a number of experiments undertaken by psychologists — a rather rare thing for a cultural anthropologist to do in 1977 (and, alas, a still rather rare thing for most cultural anthropologists to do today, although I think that things are improving, if only slowly so far). In any case, Shweder’s argument is an example — perhaps the best at the time of his essay — of an effort to posit and support universals of thought, particularly with respect to so-called magical thinking. His approach, however, lacks certain important elements in our present-day understandings of cognitive architecture.

  • Don Gardner 22 May 2011 (04:58)

    Isn’t one of the crucial differences here (inadequately signalled by the talk of “sub-personal” versus “acts” in my original response) that between cognitive architectural states and processes and those we advert to or presuppose within the “intentional stance”? Pascal’s original post tended to sideline this distinction (see his use of “holding…a belief”, for example, which is sometimes applied to a module and at others to a person). My point might be clearer, perhaps, if we invoke Strawson’s reactive attitudes (e.g. when we respond, immediately and automatically, with anger or disgust to something); these mostly engage phylogenetically ancient affect programs directly (let us say), but normally (and necessarily) remain attitudes of the person whose acts evince them. In this context, it might be right to say that it seems that the distinction between the cognitive and ecological features of representations often needs to be applied “inside the skin.” Fred’s hunger and the belief that there is good food in the larder won’t lead to the usual outputs/behaviours/acts if, for example, they subsist along with a belief that his mortal enemy has hidden a taipan in the larder. There seems little reason to expect “magical” to be relevant to any of the belief boxes the cognitive psychologist posits, for it is mostly a term of evaluative art of comparatively recent invention. Interestingly (and this is a point Dan Sperber’s [i]Rethinking symbolism[/i]) made a long time ago), most of the orthodox cultural anthropology of religion and magic depends precisely upon taking for granted special classes of (“symbolic”) belief, posited of agents without cognitive compunction.

  • Hillary Lenfesty 22 May 2011 (22:14)

    Thanks, Pascal- interesting post. The two examples you gave of magical thinking, the non-vomit filled glass labeled “vomit” and putting the spider on the photo of a neighbor, both fall under a general category of ‘misbeliefs’, but seem to originate from different cognitive processes. In the case of the former, it is indeed not silly to make these kind of “Type II” errors when the threatening stimulus is at hand: better to avoid the threat than approach it, even if it’s not really going to harm you. But this kind of behaviour does not really require any kind of (reflective) reasoning about causes and effects. Yes, it is true that drinking the “vomit” (or “cyanide” in the case of the original study) could CAUSE you to become ill. But- and perhaps this is what Paulo is getting at- this kind of avoidant behaviour to threats does not require any kind of reflection on the causal process. All that the mechanism needs to function properly is a certain kind of input to elicit a certain kind of behaviour. In the case of ‘homeopathic’ magic, e.g., believing that a spider will harm a representation of the person, “magical thinking” is here is perhaps of a qualitatively different kind- i.e., more reflective. Some people may actually believe that the representation (e.g. photograph, or fingernail paring, or lock of hair) actually is or at least embodies the person who is the target of the intended harmful action. If they believe that the representation IS the person then it can simply be called action. But if it is the case that the representation is not believed to be the actual person, but simply a representation of them (i.e., the person putting the spider on the photo admits it is JUST a photo), there is a sort of reflective belief process that allows Person A (with the spider) to believe that dislocated action upon the representation of Person B will cause an effect on that person “who is located at a distance” (in Frazer’s own words). This is in fact impossible, but it’s not impossible that we frequently initiate causes that give results which we cannot see. E.g., I win the lottery and make some phone calls to some banks who will deposit part of the winnings into my mother’s bank account- I thus believe that the money will materialize to her without me even seeing it happen. So, this kind of reasoning dealing with unseen causes or effects is something we deal with every day, and when it occurs in ‘magical’ rituals it may very well simply be an extension of this reasoning process. Having a false ‘belief’ about the presence of poison or vomit in bottle a harmless liquid is a ‘misbelief’ that manifests itself from an entirely different cognitive process (the threat-detection system).

  • Hillary Lenfesty 22 May 2011 (22:29)

    Gordon- I hadn’t read your posts before I posted mine, but now I see you also made the point about aversive (avoidant) behavior v. pro-active behaviour. In my thesis I am currently writing an section on the difference between passive vs. active encounters with ‘essences’. I’m glad you picked up on this point, too!

  • Konrad Talmont-Kaminski 22 May 2011 (23:42)

    My problem with ‘magical thinking’ is rather different from the one that Pascal brings up. The phrase appears to suggest that there exists a special kind of thinking which is responsible for supernatural beliefs and which is in some way separate from normal, nonmagical thinking. The idea is, of course, nonsense – as I am sure Pascal would agree – given that supernatural beliefs are just byproducts of normal cognitive mechanisms. Yet, something like the line of thought seems to have quite a bit of popularity with ideas such as ‘the God spot’ raising their ugly head every now and again. So, that is why I’ve generally avoided the term like the plague.

  • Jesper Soerensen 27 May 2011 (11:05)

    I believe Pascal’s blog post is right on the spot. This week we had a small informal seminar on magical thinking at the RCC in Aarhus with developmental psychologist Cristine Legare (Austin, Texas) who is doing excellent experimental work on magic (amongst other things). One of the things that struck me is how the category of ’magical thinking’ or ’magical beliefs’ works as a category in experimental and developmental psychology. It seems to be a dustbin category containing almost everything not included in standard Judeo-Christian religion or deemed ’rational’. Thus not only the robust results of Rozin and colleagues, but beliefs in Santa Claus, the Tooth Fairy and Easter Bunny are described as ’magical’ even if it, from a cognitive and anthropological perspective is difficult to see why these prima facie should distinguished from beliefs in angels, a resurrected Christ or other ’religious’ beliefs. This underscores Pascal’s point that categorization of something as ’magical’ does not help explain the cognitive processes underlying observed behavior – in fact by precluding practices such as the Eucharist, healing by touch or even ingestion of the daily vitamin pill, we might actually miss how fundamental cognitive processes underlies a number of observed behavior. My suggestion in this regard is to focus in particular on how ritualized behavior elicits certain types of cognitive processing usually described as ’magic’ – in short the non-functionality of ritualized behavior redirects cognitive attention towards the perceptual details of the actions. This then enables the embedding of the ritual into a pragmatic action sequence by relating these features to a purported goal by means of relations of similarity, contiguity and contagion (for a more detailed analysis, see: Sørensen 2007a, 2007b, Nielbo & Sørensen 2011). That being said, the responses to Pascal’s blog raise an important issue. Several comments have raised the problem of beliefs. Pascal’s original argument that the behavioral responses found in the experiments do not require the suggestion of any belief-states, but merely activates a number of domain-specific inferences seems to me to be a parsimonious explanation. However, as several have pointed out, doing so will not explain why such stimuli are produced in the first place. Unearthing a doll full of pins under your door might elicit a intuitive response of fear and predation, but producing the doll, ritually sticking pins into some body parts and thereafter burying it under our neighbors doors-step surely requires some sort of stabilized belief about the likely, possible or merely claimed effect. This belief is likely to combine socially transmitted idea about particular actions and their effect with certain perceptual features of the actions that makes this effect more ’believable’ and emotionally satisfying (e.g. sticking pins into a doll and not a pillow). We should therefore distinguish between the cognitive processes underlying the production of ’magical stimuli’ and those activated when observing these. Of course these will be connected on a pragmatic level, as the cognitive effect of magical stimuli, i.e. the activation of particular types of inferences, all else being equal will enhance the relevance of beliefs motivating the construction of the stimuli thus stabilizing some and not other cultural practices. Further, as Don rightly points out, abductive reasoning establishing connections between seldom events (sticking a pin in a photo of your mother and her subsequent loss of eye-sight) and such well-established phenomena as confirmation bias and reactions to cognitive dissonance will also help stabilize culturally transmitted beliefs in the effect of certain ritual actions. If we can unearth how these processes interact we might be able to address the problem succinctly posed by Malinowski almost 100 years ago: why do people who in most endeavors act totally rational in their pragmatic pursuits in some situations combine these with magical and ritual behavior? A related question is how why rituals that on most accounts explicitly negate any direct causal intuitions as to their effect, are so regularly claimed and represented to be particular efficacious and powerful in bringing about certain real world effects? Nielbo, K. & Sørensen, J. (2011). Spontaneous Processing of Functional and Non-functional Action Sequences. Religion, Brain and Behavior 1(1), 18-30 Sørensen, J. (2007a). A Cognitive Theory of Magic. Lanham: AltaMira Press Sørensen, J. (2007b). Acts that Work: A Cognitive Approach to Ritual Agency. Method and Theory in the Study of Religion 19, 281-300