Magic and inference

I must confess a predilection for the anthropological and psychological writing of the mid-twentieth century, when anthropologists were still trying to explain culture and the principles of the cognitive revolution in psychology were first being worked out.  It was in the course of my most recent historical dalliance that I came across E. E. Evans-Pritchard’s 1933 article “The intellectualist (English) interpretation of magic”, which occasions the present musing:

Two models of magical associations have been proposed.  Frazer proposed that magical associations are the result of the application of a cognitive rule.  Evans-Pritchard countered that magical associations are too selective to be the result of such a rule.  In computational terms, Evans-Pritchard's proposal is that magical associations are represented by a look-up table.  Yet the occasional generalization and extension of magical associations suggests that at least in some cases inheritors of such traditions are seeking to capture the look-up table with a rule.

Edward Evan Evans-Pritchard (1902–1973) was an anthropologist most famous for his ethnography, Witchcraft, Oracles, and Magic among the Azande (1937) in which he argued that primitive peoples are no less rational than anyone else, …

…, that even apparently bizarre ideas about witchcraft are, if not correct, at least intellectually respectable, when understood in context.  Evans-Pritchard’s article, written after he had completed his fieldwork among the Azande but before he wrote Witchcraft, offers a window into the intellectual background of his famous book.  In the article, he reviews the theories of magic proposed by Edward Tylor and James Frazer, the latter of whom argued that magical practices were based on an intellectual mistake, the fallacious conclusions of which were subsequently protected from disconfirmation.  Frazer argued that primitives made accurate observations about the world, but erred in reasoning about them using the principle of similarity (that similar objects can causally influence each other) and the principle of contiguity (that two objects that have once been in contact ever after causally influence each other).  This faulty reasoning led them to carry out magical practices.

Evans-Pritchard pointed out that Frazer’s psychological explanation was too general, that it could not account for the selectivity with which magical associations were made: "One is not surprised that a Greek peasant can see a resemblance between the colour of gold and the colour of jaundice but the problem is why he should associate these two things together in magical performances when he does not associate them together in other situations and why he associates these two particular things and not other things which have the same qualities of colour.  (p. 308)"

Some other account must be given of the selectivity of magical associations.  Evans-Pritchard argued (a la Durkheim) that the association was a social fact, existing prior to the individual who learned it and providing the conditions for the individual’s mental representation:
"We must not say that a Greek peasant sees that gold and jaundice have the same colour and that therefore he can use the one to cure the other,  Rather we must say that because gold is used to cure jaundice colour associations between them becomes established in the mind of a Greek peasant. (p. 308)"
Evans-Pritchard’s observation is undeniably correct, and this way of construing the relation between cultural traditions and individuals’ mental representations is fundamental to the epidemiology of representations outlined by Dan Sperber and extended, in the field of religion, especially by Pascal Boyer.  I believe the Boyeran account of the magical manipulation of gold to cure jaundice would, very roughly, go as follows: the association between gold and jaundice is highly communicable because it is conceptually simple (largely intuitive), slightly puzzling (the procedure attributes to gold a causal property that gold does not intuitively have), and relevant in environments where gold is available and jaundice is endemic.  As it should, this account explains how, given that a particular gold–jaundice association is socially available, that association is apt to be transmitted widely enough to become cultural.

Yet in the course of my own research on magical and mantic uses of Bibles, I found a significant number of cases where people took existing associations and extended or generalized them.  I believe other historians and anthropologists studying magical practices have found the same phenomenon.

There may be something to be gained by looking at this problem computationally.  Frazer’s claim was that magical associations were the result of the application of general rules of thought; Evans-Pritchard’s counter was that magical associations were far too selective to be the result of a general rule of thought.  In computational terms, Frazer’s claim was that magical associations could be mentally represented by a rule; Evans-Pritchard’s claim was that they were so selective that they could be mentally represented only in the form of a look-up table.  On the whole, Evans-Pritchard’s account must carry the day, and I think Boyer and others have been correct to build on this foundation.  But would we not be on solid cognitive ground to surmise that those who encounter such arbitrary associations tacitly—and sometimes explicitly—seek to find a rule to describe the look-up table?  It seems to me that one of our primary cognitive operations must be to seek greater compression—a more parsimonious representation—of our own mental representations.  The reuse of symbolic associations, especially their generalization, may well reflect this basic cognitive operation at work.


  • comment-avatar
    Olivier Morin 3 November 2008 (11:57)

    Very interesting point. I agree on the whole, and I’d like to suggest (although I cannot speak on his behalf) that you may find support in the theories of Boyer himself. In the Boyerian account of supernatural concepts, minimal counter-intuitiveness is a key feature. By definition, a minimally counter-intuitive concept is still largely an intuitive concept, which means that it is compatible with an intuitive theory that we have about the world. These theories, in the Boyerian account, are highly structured, rule-based, and above all, inferentially rich. This inferential richness is part of the reason why (minimally counter-) intuitive concepts are attractive in the first place. Another source of support for a slightly more Frazerian view of magic is the study of disgust mechanisms by Paul Rozin and Jon Haidt. Psychological mechanisms of disgust appear to exhibit some of the properties of magical thinking that Frazer speculated about: the two laws of Sympathy, i.e. similarity and contiguity. People refuse to drink from a glass of water in which the extremity of three yards of rope have been dipped, the extremity of which is bathed in a bucket of urine. This looks like a striking illustration of the principle of contiguity. Likewise, the fact that people refuse to wear a sweater that may have once been worn by Mussolini. This of course does not invalidate the Durkheimian (Maussian) objection: the mere application of these cognitive principle is not sufficient (may not even be necessary) to define magic. Still, they seem to play an interesting part there and elsewhere.

  • comment-avatar
    Kate Devitt 3 November 2008 (14:53)

    ”It seems to me that one of our primary cognitive operations must be to seek greater compression—a more parsimonious representation—of our own mental representations. The reuse of symbolic associations, especially their generalization, may well reflect this basic cognitive operation at work.”
    I’m sure you’re right that generalization is a basic cognitive operation. Humans can’t resist turning associations into patterns and attaching causal explanations to them regardless of justification. Perhaps magical thinking and non-magical thinking share the same functional role?-And only via hindsight do we judge the former illogical and the latter rational?
    One way to consider the question of functional role is by examining the implications of magical thinking on belief. That is, if a practitioner held that their magical beliefs were ontologically grounded, then we might suppose that they function just as belief, thus are no different to ordinary thought. However, what if magical thinking was more like pretence and less like normal belief? Suppose magical beliefs went into the possible world box (PWB) along with imaginings, counterfactual reasoning and pretence (see Nichols and Stich, 2003)? We know that fiction can draw real emotions, so magical thinking could still be rational, depending on its purposes. A defensible use for magical thinking might be for psychological state management, e.g. western occultism, feng shui or talismans. The use of fictional or arbitrarily meaningful symbols doesn’t stop them having an effect on one’s mood if that’s part of the game. Cognitive behavioral therapy uses counterfactual reasoning for precisely the same purposes. Recent work in sport psychology reveals the competitive benefits of imaginative ritual before performance.
    In summary, there’s a good chance that magical thinking is no different to normal thinking, including generalizations and the attempt to create more parsimonious mental representations. There’s also the possibility that some (if not all) magical thinking is different to normal thinking. One way magical thinking might be functionally distinct is if the representations are treated more like imaginings and less like belief.
    Kate Devitt
    Reference: Nichols, S. & Stich, S. (2003) Mindreading: An Integrated Account of Pretence, Self-Awareness and Understanding Other Minds. OUP.

  • comment-avatar
    Paulo Sousa 3 November 2008 (15:31)

    Hey Olivier and Brian (nice to hear from you, my old Michigan friend)!
    Perhaps it is important to explicate more distinctions in this discussion. First, there is more to magic than Frazer’s principles. According to Frazer’s two types of magic, a change C produced in an object O1 may produce a similar change C’ in another object O2 if :
    A) In Homeophatic magic, if O1 and O2 are similar in some respects. Example: A man injures an image of his enemy (O1) in order to injure his enemy (O2);
    B) In contagious magic, if O1 and O2 have been in previous contact. Example: A man injures something that was in previous contact with his enemy like the clothes of the enemy (O1) in order to injure his enemy (O2). Now, for intellectualists, these principles suppose some action at a distance between the two objects (“a material medium of some sort which, like the ether of modern physics, is assumed to unite distant events or objects and to convey impressions from one to the other: in one case a similarity between the two events or objects; in another a previous contact.” Frazer). There is at least another type of reasoning involved in magic:
    (C) In contagious transfer of propeties, by putting two objects in contact, a property P of an object O1 is transferred to another object O2. Ex: A man chews the seeds of a very fertile plant and spits it on his son so that his son grows as fast as the plant. The belief here is that by putting the two objects in contact, a property P (grow fast) of an object O1 (the seeds) is transferred to another object O2 ( the child). Objects O1 and O2 need not to be similar or to have been in previous contact (the seeds need not be similar to the child or have been in previous contact with the child).
    The examples given by Olivier are related to this last principle.

  • comment-avatar
    Paulo Sousa 3 November 2008 (15:55)

    My second point. I think there are three different levels in what Brian is saying (?). First, a criticism to Frazer is that, if distant things are supposed to be influencing each other simply by the fact that they share some similarity or previous contact, then we would have a completely chaotic understanding of the world. Let’s call this position the completely unconstrained magic. Second, supposing that there are some additional cognitive constraints establishing magical associations, then there may different levels of constraints. I think that Brian is claiming that these constraints are more general than what would be implied simply by a look-up table.

  • comment-avatar
    Olivier Morin 4 November 2008 (11:42)

    Here is a prediction that might be derived from your hypothesis.
    If, as you suggest, computational constraint is what brings people to try and compress their magical beliefs into a more compact format – more rule-based, less like a look-up table – you would expect magical belief systems to become more coherent, that is, easier to deduce on the basis of rules and principles (however loose and vulnerable to exception these rules and principles might be), when they are found in oral cultures. In cultures where magical beliefs are supported by written sources, the computational constraint is much less stringent, so that it would do no harm to entertain magical beliefs that are mere lists of (more or less pointless) associations.
    One example, unfortunately not a magical one, might be taboo (I have to thank Nicolas Baumard for this last observation). As Mary Douglas observed, some taboo system seem to have a coherence of some kind. Others are much more difficult to explain, and a few, like one category of Jewish Mitzvot, are deliberately pointless. As Rabbis say, they escape reason. Is it a coincidence that these particular taboos happen to be found in a highly litterate tradition?

  • comment-avatar
    guest guest 9 November 2008 (22:33)

    Thanks for the interesting article. Dan Howitt