This paper outlines an anthropological approach to religious representations that is grounded in recent findings and hypotheses in cognitive psychology. The argument proceeds in four points. First, the main goal of this framework is to account for the recurrence of certain types of mental representations in religious systems. Recurrent features are not necessarily universal. They are the outcome of cognitive systems that make certain representations easier to acquire than others. Second, a cognitive approach must take into account the diversity of religious representations. It is argued here that religious systems bring together ontological assumptions, causal claims, episode types and social categories. These four “repertoires” may have different functional properties, and may therefore be acquired and represented in different ways. Third, universal features of tacit, intuitive systems may impose strong constraints on the variability of religious ideas. This is illustrated on the basis of ethnographic data. Finally, the type of representations one finds in religious belief-systems consists in conjectures, the cognitive salience of which is variable and should be evaluated in precise terms.
Contrary to other domains of anthropological study, theories of religious belief and action have not been much influenced so far by the remarkable development of cognitive science. True, there is in anthropology a subdiscipline known as “cognitive anthropology” (or “ethnosemantics”) which focuses on the cognitive aspects of cultural representations. This approach, however, has been so far limited to representations of the everyday world: biological taxonomies, classification of daily activities, kinship terminologies, etc.1 Religion, on the other hand, is relatively neglected in cognitive approaches. This is paradoxical, in view of the importance of religious belief and action in anthropological theory and practice. There are, obviously, some notable exceptions to this generalisation (see e.g. Dougherty 1985, Dougherty & Fernandez 1980, 1982, Lawson & Mc Cauley 1990). By and large, however, the study of religious belief and action is still conducted in the framework of anthropological theories which pay little if any attention to the findings and hypotheses of cognitive science.2
In this paper, I will present the elements of a possible cognitive approach to religious representations.3 This paper is therefore largely programmatic and partly speculative. The research programme, however, is not very far removed from actual cognitive research, and the speculation is mostly consistent with both anthropological and psychological research. My aim here is to show that a cognitive approach is possible, and to examine how it can account for the representation and transmission of religious representations. The theory is about religious ideas rather than “religions” in the broad sense. The aim is to describe the processes whereby subjects acquire, represent and transmit certain ideas and practices. The theory may not be sufficient to account for the social dynamics of religious movements or the historical development of religious doctrines. Such “macro-phenomena” of religious transmission are not directly within the scope of a cognitive theory.
In the following pages I will put forward four main hypotheses. The first one is that the goal of a cognitive approach is to account for the recurrence of certain features of religious representations in many different cultures. This general objective is often misconstrued, and I will try and give a formulation which avoids certain common misunderstandings. The second point is the cognitive diversity of religious ideas. In a given culture, the set of representations that constitute each individual’s religious ideas is distributed in several “repertoires”, which have different functional properties. Here I will identify four such cognitive repertoires, which are particularly important in the description of religious representations. The third hypothesis is about the cognitive constraints on the content and organisation of religious ideas. Far from being pure cultural constructions, religious systems appear to be strongly constrained by universal, probably innate properties of cognitive systems, especially those properties which govern people’s intuitive understanding of their everyday world. A fourth point is that we need a precise notion of cognitive salience in order to describe and explain the processes of acquisition and belief-fixation in the domain of cultural knowledge. For each of these four points, I will try to show that a cognitive approach provides a plausible alternative to classical anthropological theories of religion.
It may seem the obvious first step, in the construction of a general theory of religion, to enumerate the universal features which the theory will set out to explain. Indeed, this is very much what can be found in most anthropological attempts so far.4 Here, however, I will argue that this seemingly obvious way of proceeding is in fact mistaken and misleading, and that the search for universal is the main reason why anthropology has not produced a theoretically plausible account of religious ideas. In what follows, I will try to show how the problem of universals conceals other, important aspects of the problem.
The main starting point of the theory outlined here is that, in the variety of cultural systems of religious belief, there is a notable recurrence of certain precise themes or ideas, and that this recurrence ought to be explained. These themes or ideas are not universal, but they constitute a repertoire most elements of which can be found in most cultures, in one form or another. To take but a few examples, it is assumed in many (but not all) cultures that a non-physical component of persons survives after death, to become an invisible intentional being, endowed with perceptions, beliefs and intentions. In the same way, it is assumed in many (but not all) cultures that certain people are especially likely to receive direct inspiration or messages from extra-natural agencies, like gods and spirits. In many (but not all) cultures it is admitted that performing certain ritual recipes in the exact way and order prescribed can bring about changes in physical states of affairs, through causal mechanisms which are presumed but not observable. Such features are widespread in many cultures, yet they are not necessarily present in all of them. Each feature is present in most cultures, each culture has many of those features in its set of religious ideas, yet none of them should be taken a priori as universal.
In cultural anthropology, the recurrence of certain religious ideas is not explained in a satisfactory way, for the simple reason that is not explained at all. The few features I mentioned above are well-known to most students of religion. Their recurrence in different cultures, however, is not considered an object of scientific inquiry, for two symmetrical reasons.
A first obstacle is the pre-theoretical, instinctive form of relativism that is somehow intrinsic to anthropological investigation. Because anthropologists are professionally trained to detect and emphasize cultural differences, they naturally under-estimate the recurrence of similar ideas in different cultures. Moreover, when this recurrence is noticed, it is often treated as a deceptive appearance, which conceals underlying differences. It is widely assumed that apparently similar beliefs cannot really be similar, because they occur in different “cultural contexts”. This idea, however, is vague enough to contain both a trivial truth and a profound fallacy. Take for example the widespread idea that the gods are so remote that one cannot communicate with them except through the channel of inspired mediums. Obviously, this idea can take on very different “meanings” in different cultures; more precisely, it carries rather different implications, for those who think that the gods have a direct influence on the living’s well-being, and those who think that they do not. The idea that certain people are privileged “channels”, however, is the same idea in both contexts. We need a theory that could account for the fact that this idea is so widespread, while others are not.
Conversely, the general study of religious ideas is often hampered by the widespread idea, that “human nature”, the proper subject-matter of cultural anthropology, is manifest only in the universal features of the species. What is not universal in human cultures has therefore nothing to do with human nature. In other words, one assumes that there is a division, between cultural invariants on the one hand, which are explainable by various non-cultural factors (ecological, biological, psychological, etc.), and the rest. Cultural features which are not universal are ipso facto outside the influence of those various ecological, biological, psychological etc., factors. If certain traits of religious ideas are widespread but not universal, they are therefore considered outside the scope of a general theory of religion. Obviously, this argument rests on a confusion of levels, between processes and their outcome. That a universal process exists does not imply that its outcome will be the same in all possible circumstances. It is precisely the point of an explanatory theory to reduce diversity, and show in what manner diverse phenomena result from the encounter of general mechanisms, on the one hand, and a manifold of contingently diverse circumstances on the other.
This “probabilistic” approach to recurrent features in religious systems has important consequences for the type of model that is supposed to account for religious ideas. Here I must draw a broad distinction between two types of accounts that can be put forward in the explanation of the emergence of recurrent features in a population of organisms. I will label these accounts generative and selective. Given a series of recurrent features, a generative model posits an underlying mechanism such that, if it is present, it will provide a sufficient explanation for the occurrence of these features. Take for example the fact that, in all tigers, the anatomical structure of the retractile claws is exactly similar, A sufficient explanation of the recurrence is provided by models of genotype inheritance combined with models of embryological development. Selective models, on the other hand, account for recurrent features by positing (i) a set of underlying mechanisms which are necessary yet insufficient to produce the recurrence, and (ii) a set of inputs such that, given the underlying mechanisms, they will produce the recurrence. This type of explanation would be necessary if we want to explain, for example, why tigers have retractile claws rather than non-retractile ones. In such a case, an evolutionary explanation will typically invoke (i) a series of random mutations, providing the input to (ii) a fitness-maximizing mechanism (natural selection), the combination of which provides a sufficient explanation. The fitness-maximizing mechanism, on its own, provides necessary, yet insufficient conditions for the recurrence.
It should be obvious, from this example, that generative and selective models can be invoked in order to explain the same recurrence, seen from a different point of view. This does not mean, however, that choosing between these two explanatory schemes is a matter of convenience or subjective preference. The notion of a “point of view” is to be understood here in a precise sense, as implying a set of clearly defined goals for the explanation. Given such a precisely defined objective, choosing a generative account, where a selective one would have been pertinent, may generate considerable problems. Biological theory for instance would be in serious trouble if it tried to explain evolutionary trends by generative models, e.g. explaining the evolution of vertebrates by some underlying mechanisms that “pushed” fish out of the water and made them become reptiles. The evolutionary scenarios developed in biology do not require any generative models of evolutionary trends, and in fact exclude them explicitly.
Applied to cultural ideas, the notion of a selective model means that, given certian circumstances and a variety of mental representations entertained by a population of subjects, some of those representations are more likely than others to be stored in the subjects’ memories and communicated to other subjects. Many anthropological models are, implicitly or explicitly, based on this evolutionary metaphor,5 and treat the transmission of ideas as a function of their “survival” value. A set of constraints is posited, such that it will make it more likely for certain representations to “survive” (i.e. be memorized and transmitted) than for others. This is the starting point of various models of cultural evolution (Cavalli-Sforza & Feldman 1981, Cavalli-Sforza 1986), “gene-culture co-evolution” (Lumsden & Wilson 1981) or “dual inheritance” (Boyd & Richerson 1985). These various theories share the premise that the recurrence of cultural traits must be explained by selection processes rather than by generative mechanisms. They include precise hypotheses about the transmission or diffusion of cultural material. The point of departure is that, given a random input of cultural traits at generation Gi, a process can be described, such that its operation on the input will increase the relative probability of certain traits appearing at generation Gi+1 (and of course decrease the probability of other traits). Selective models generally focus on transmission processes as the main cause of recurrence.
Another, qualitative approach that is based on a selective stance is Sperber’s notion of an “epidemiology of ideas” (1985, 1991). The argument is based on an analogy, to the effect that, by and large, the relation of anthropology to psychology can be construed as similar to that of epidemiology to physiology. While physiology puts forward hypotheses about, e.g. the way various viruses or germs may affect the body’s functioning, epidemiology on the other hand is concerned with the ways in which diseases spread. In much the same way, psychology is concerned, among other things, with the acquisition or representation of certain ideas or beliefs. Anthropology, on the other hand, specializes in observing the spread of ideas or beliefs. It may focus on either short-lived epidemics, like fashions, or more stable endemic infections like traditions. An important consequence of this notion of “epidemiology” is that psychological processes are directly pertinent to anthropological theory. In the same way as such physiological aspects have direct consequences for the spread of a given disease in given circumstances, the psychological processes of representations and transmission are bound to affect the patterns of cognitive “epidemics”.
The ideas presented in this paper are, by and large, consistent with the assumptions of these selective models. I will claim that micro-processes of cognition and interaction impose strong constraints on the diffusion and transmission of religious assumptions, thereby leading to the recurrence of ideas observed in the religious domain. Before proceeding to this point, however, we must examine the types of mental representations which are the object of such selective processes of cultural transmission.
In anthropological descriptions, religious representations are generally presented as constituting shared, integrated, explicit, context-free general propositions, e.g. “the spirits dwell in the rivers”, “the ancestors are invisible”, “only shamans can negotiate with the spirits”, etc. As many authors have observed, such descriptions in terms of “collective representations” are extremely ambiguous, as far as actual cognitive processes are concerned.6 Obviously, a cognitive approach to religious ideas should provide precise answer to the questions concerning the actual representations involved, rather than work on the basis of ad hoc descriptions. In the following pages, I will try to put forward the elements of such an approach. My first contention is that a cognitively realistic description of religious ideas will have to take into account the cognitive diversity of the religious domain. That is to say, the representations we call “religious” belong to different domains and may have different cognitive properties. I will then examine the possible contributions of cognitive science to the study of the different types of representations in question.
Religion should not be construed as an isolated domain. This point is generally accepted by most anthropologists, who consider it a mistake to study the religious life of a given culture as though it was isolated from other domains of social life. More recently, most anthropologists have come to agree that such a division would be illegitimate for a deeper reason, namely, because the very category “religion” may well be a Western construction, of limited validity in the study of different cultures. Here I will not enter such debates; I will accept as a tentative characterisation of the domain, the fact that the cultural representations we are focusing on concern unobservable extra-human entities and processes. Such a practical definition is more or less what anthropologists have in mind when they talk about “religion”, and will be largely sufficient for the purposes of the present argument.
In the general discussions concerning the coherence (or lack of coherence) of the notion of “religion”, one aspect is generally left aside, that of the cognitive unity of the religious domain. Theories of religion, in either anthropology or sociology, are generally based on the implicit premise, that all religious ideas are acquired and represented in the same way. This assumption is itself based on a more general principle, following which most mental representations are acquired and represented in the same way. Whether based on an empiricist or rationalist epistemology, anthropological models treat all cultural representations as functionally similar. Against this view, however, it must be noted that most recent advances in experimental psychology, especially in the domain of cognitive development, insist on the functional specificity of certain types of representations.
In a parallel fashion, I will argue that the representations concerning extrahuman entities and processes belong to different repertoires, which have different functional properties. Here I will consider four such repertoires, in which the most important elements of religious belief-systems can be found. They are the ontological, the causal, the episode and the social roles repertoires, respectively, I will try to show that describing and explaining people’s religious ideas consists in describing what is included, in a given culture, in each of these repertoires, and explaining how their contents are gradually made plausible. Before turning to this point, however, let me give a succint description of the typical contents of each repertoire.
The ontological repertoire is the set of assumptions people entertain about the existence of non-observable entities. This catalogue will include ideas about there being, e.g., a distant impersonal Creator somewhere in the skies, water-spirits near ponds and rivers, invisible ancestors lurking in the darkness of the forest, etc. This catalogue of ideas is called ontological because it consists of elementary assumptions about what sorts of things there are in the world.
The causal repertoire is a catalogue of ideas and assumptions about causal connections between the entities described in the ontological repertoire, on the one hand, and observable events and states of affairs on the other. Thus, a causal repertoire may include assumptions like “gods get angry if no sacrifice is performed” or “reciting this formula will guarantee good crops”. It is important to note that such a repertoire may include generalisations of this kind, as well as propositions concerning singular events or states of affairs, like “So and so got disease X because he did not observe prohibition Y”, or “we had no crops last year because of such and such’s witchcraft”.
The episode repertoire consists of descriptions of a certain range of event-types, which arc connected to the ideas contained in the ontological and causal repertoire. In order to describe a religion, one must identify a certain set of actions and interactions which are deemed to be of particular types. Ritual performance, obviously, is the most important type of religious episodes. Again, it is important to note that the representations involved may be about singular objects, e.g. memories of ritual X being performed last week, or about generalised types, like the list of things to do in order to perform a certain rite in a proper way.
The social roles repertoire is a catalogue of representations concerning differences between people. In this we will include people’s ideas concerning their priests, shamans, or religious specialists, but also ideas concerning other differences which are relevant in religious action. For instance it will include people’s ideas about gender where relevant, about growth and maturation where an adult-child opposition is concerned, people’s ideas about the effects of initiation, etc. All these ideas are used to characterize, sometimes categorize social actors, either in abstracto or as particular persons.
In order to make these explanations more intuitively clear, let me take an example of their concrete application, concerning the religious ideas of the Fang of Cameroon (see Boyer 1986, 1990 for more detail). In the following sections I will use this example to illustrate some of the general claims I want to make about religious ideas. Although the specific contents of the “repertoires” are, obviously, particular to Fang culture, it must be stressed that the general conclusions I will put forward are in fact based on much broader ethnographic comparisons, which cannot be included in this paper. For the sake of simplicity, I will introduce each theoretical statement as though it had been developed in order to explain the particular Fang data. It must be kept in mind that it is in fact of general relevance for the acquisition and representation of religious ideas.
Ontological ideas. Let me first take the repertoire of ontological ideas, which seems to be organized around three main points: a pair of distant personified gods, the ghosts (bekong) and the spirits (minkugu). In Fang mythology, there are two creator-gods. Mebeghe is the name of a god understood to be the creator of all natural things while Nzame is supposed to be at the origin of most cultural techniques and social institutions. It must be stressed that the narratives concerning the origins are not the object of much attention or speculation. Nzame and Mebeghe are remote gods. Their powers are not really invoked or used in the explanation of natural or social occurrences, although there are individual variations in this domain. The role of Nzame is sometimes conceived as that of an impersonal, purposeless fate. It must be pointed out, however, that such a notion of contingency is alien to the Fang intellectual climate, as to that of most African societies. If salient events are to be explained at all, they have to be explained in terms of goals and intentions; this is precisely the type of explanations provided by the concepts of ghosts and spirits.
The term bekong could best be translated as “ghosts”. After death a person’s “shadow” is supposed to become a wandering spirit, generally malevolent until it is given appropriate funeral rites. It is then said to dwell in ghost-villages, and is supposed to protect the villagers. Correct performance of traditional rites is indispensable, lest ghosts may “throw” illnesses of various kinds at the living. Ghosts have beliefs, desires, feelings, emotions, and generally all the non-physical characteristics of humans. In Fang discourse, however, they are generally treated as a kind (“the ghosts do this”, “the ghosts want that”, etc.) rather than as individuals.
The mystical personnel also includes rather mysterious creatures called minkugu, generally glossed as “spirits”. These are not clearly identified as ghosts. They are described as smaller, not related to village clans and lineages in a defined way, and rather uncanny, They too can “throw” illnesses, and some specialists say their remedies are given by the minkugu rather than the ghosts. Although most people insist that there is a difference in kind between bekong and minkugu, their ideas concerning the exact differences in powers, appearance, etc., are extremely vague.
Causal connections. The repertoire of causal connections includes many general connections between the entities described above and various classes of events, which however are generally less salient then representations of singular episodes. It is for instance admitted that the ghosts can trigger various kinds of misfortune if the living do not perform appropriate traditional rites. People who hold such ideas, however, rather focus on memories of singular occurences, in which a certain disease was diagnosed as “thrown” by the ghosts, and then successfully cured. In the same way, the fact that illness is commonly caused by witchcraft is stated in very vague and general terms. People, however, have precise memories of many singular cases, with the problem, the diagnosis and the outcome. The same remark applies to magical charms and recipes. While their general efficacy is stated only in the vaguest terms, people have numerous accounts of the successes or failures of precisely identified recipes.
Episodes. These representations concern categories of actions, notably ritual actions, related to the entities of the ontological repertoire. The repertoire includes names for complex scripted actions. To take but one example, nku melan is an initiation ritual during which neophytes are shown the skulls of their ancestors, usually concealed in special shrines. The main characteristic of such actions is that they are represented as rigidly scripted. They consist in a list of sub-actions which, from the actors’ viewpoint, must be performed in the appropriate way and in the appropriate order, by the appropriate specialist. Although actual performance may display great variations, the participants are generally unaware of those changes or consider them insignificant.
Social roles. There are of course many categories designating types of people. Here I will only mention the subset that is directly pertinent to religious ideas. Some of these categories are supposedly descriptive, identifying some persons by virtue of the rituals they have learned to perform. Thus, a diviner is called mod ngam, “divination-man” or a story-teller specialized in myths is a mbomm-vet, “harp player”. A crucial category is that of ngengang (“healer”, in fact many types of ritual specialists). Most rituals connected with questions of magical connections, and relationships with the ghosts, require the intervention of a ngengang. Such activity-based categories, however, presuppose another classification, this one in terms of unobservable qualities. It is impossible in Fang society to be ascribed any religious role without being considered a beyem, that is, a person who carries an invisible organ called evur. Every living person either is or else is not a beyem, there is no intermediate point. There is, however, no way of telling for sure whether any given person is or is not one. This is a matter of conjectures, based on the person’s behaviour and ritual successes. The category mimmie (“simple folk”) designates people who have no evur. Another important category is that of ntuban nlot (“pierced head”) people, that is, people who have undergone specific initiation rites. Again, since these initiations are shrouded in secrecy, it is rather difficult to tell whether any given individual is or is not a member of that category.
This presentation of Fang ideas differs from ordinary ethnographic presentations of religious ideas, in that the assumptions are presented as fragmented bits of information, with no indication of the “models” or “systems” they belong to.7 This is mainly in order to avoid the “intellectualist” or “theologistic” fallacy, which takes it for granted that all cultural representations are included in coherent, systematic models. The distinction made between those repertoires is of course on an analytical nature. The actual representations which are studied by anthropologists combine elements from several or all of these repertoires. For instance, people have ideas about ancestor-ghosts which combine ontological aspects (about the existence of such beings), causal ones (about their possible role in human affairs), some identification of episodes (e.g. the rituals that may be performed to placate them) and of social categories (concerning the type of specialists who perform those rituals).
The rationale for the division between those four categories is that the elements from different repertoires are likely to behave in functionally different ways, in the acquisition and fixation of belief, This is because each of these repertoires is in fact the extension to religious matters of conceptual structures and assumptions which can be found also in other, non-religious domains. Everyday knowledge comprises, among many other things, (i) a set of ontological assumptions, about what kinds of objects there are in the world and how different they may be; (ii) a set of principles on which causal connections, in various domains of experience, are evaluated; (iii) some ways of dividing the continuous flow of action into discrete episodes; (iv) some general representations about the possible differences between persons. My main contention here is twofold: (i) that some implicit assumptions and principles from non-religious knowledge are carried over in religious representations; (ii) that these principles and assumptions play a crucial role in the acquisition and transmission of religious representations.
Furthermore, this “fragmented” description is more realistic from the point of view of acquisition. In most societies people do acquire their religious ideas in a fragmented, disjointed and often inconsistent form. They are faced with a mass of utterances and actions, as well as comments about those singular occurences. As I said above, they are seldom if ever faced with systematic presentations of the local stock of religious presentations. Utterances are made because they have a precise point, actions are performed with a definite purpose. Now producing a consistent description of the religious world is very rarely the point of people’s utterances or the purpose of their actions. Only tangentially do they serve such a purpose. An obvious consequence is that we must have a precise description of the processes whereby people build more general models on the basis of such fragmented material.
All this leads us to the main question posed at the beginning of this paper. Some features of religious ideas appear to be particularly recurrent. A cognitive approach is based on the assumption that this recurrence may be at least partly explained by cognitive constraints on the acquisition and transmission of cultural representations. In other words, the theory would assume that certain representations, or combinations of representations, are optimally learnable or memorable. It is therefore no surprise that such features become recurrent properties of religious systems, This condition of “optimal learnability” is the fundamental point of a selective model of religious representations. In the next section I will try and present some possible directions for a cognitive approach to this problem.
Most anthropological theories are based on the assumption that cultural knowledge is only weakly constrained by cognitive processes, and that all its important aspects are culturally transmitted. Cultural anthropology, however, does not seem to have a very precise account of how this cultural transmission is taking place. The discipline generally adopts what Bloch (1985: passim) calls the “anthropological theory of cognition”, that is, the idea that people brought up in a culture are given a ready-made conceptual scheme, which is absorbed, as it were, in a mysterious way that is never described. As Bloch points out, this account of acquisition is obviously insufficient in the case of simple everyday concepts, like names for natural kinds and artefacts; one might think, a fortiori, that it is rather implausible as an account of the transmission of complex cultural ideas. This “theory of cognition” includes two particularly implausible assumptions. One is that cultural transmission is, by and large, a passive process. Minds are conceived as containers of ideas, which are more or less empty at the onset of cultural acquisition, and are gradually filled with whatever ready-made products are given by “the culture”. The other assumption is that this filling process is simple. Both assumptions, however, fly in the face of all the experimental evidence available in the domain of concept acquisition and belief-fixation (see e.g. Markman 1989 for a general survey).
Against this widespread notion of cultural transmission, I will contend that universal properties of human minds are likely to impose strong cognitive constraints on the range and organisation of cultural representations that can be transmitted from generation to generation. To be more precise, my hypothesis is that in each of the four repertoires described above, the religious assumptions are constrained, in non-trivial ways, by the assumptions that govern the non-religious or everyday representations of the repertoire. In other words, people’s representations about (religious) ontologies are constrained by their ordinary ontology, their ideas about (religious) causation are constrained by ordinary causal representations, and so on. In the following pages I will first survey the type of psychological data and theories on which this assumption is based. This will be done mainly by considering some aspects of children’s conceptual development; developmental studies provide particularly clear illustrations of the ways in which those assumptions are represented and constrain the range of beliefs subjects can entertain about a given object. I will then apply this kind of method to one of the four repertoires mentioned above, that of religious social roles.
A number of recent studies in conceptual development tend to shed light on processes which are directly relevant to the questions concerning religious ideas. In order to understand this, it may be of help to describe the theoretical background of those studies and hypotheses. In a “classical”, Piagetian understanding of conceptual development, the child is assumed to apply general learning heuristics to a variety of experienced phenomena. The developmental stages identified in Piagetian frameworks are mainly described in formal, non-domain specific terms. At each stage the child is described as applying certain structural procedures, which do not depend on the type of phenomena concerned. These formal procedures are gradually modified on the basis of experience, leading to the next stage, at which point other formal procedures can be applied, again across conceptual domains. For instance the child’s causal thinking develops from an “animistic” stage, in which all causal connections are attributed to intentions, to a later differentiation of intentional and mechanistic causation (see e.g. Piaget 1954, 1974, Laurendau & Pinard 1962, and a critique in Leslie 1979, Carey 1985, Leslie & Keeble 1987).
There are some reasons, however, to doubt the relevance of such structural cross-domain principles. First, cognitive studies tend to show that there are important functional differences between cognitive domains. To take but a few examples, the way subjects represent and remember faces is very different from the treatment of other types of visual stimuli; the way they make inferences about living kinds differs from the processes focused on artefacts or other non-living objects. Notions of causal connections may be very different, depending on the animate or inanimate nature of the objects concerned, from a very early age. Instead of describing human minds as “general processors” endowed with some all-purpose cognitive structures fed by perceptual inputs, psychological studies uncover domain-specific structures, which apply only to a limited domain of external phenomena. Furthermore, most recent research in developmental processes tends to show that such specialised structures appear very early in conceptual development. The way children reason, at any stage, depends on the conceptual domain to which the intellectual operations are applied. This type of research indicates that children’s conceptual development proceeds on the basis of strong theoretical presumptions, concerning the type of properties and generalities to be expected in different ontological domains (Keil 1979, 1987, Atran 1989). The presumptions concern large ontological domains such as physical objects, artefacts, living kinds, persons. They constitute “naïve” theories of the domains in question (R. German 1990).
To take a simple example, even 4 year olds seem to make instance-based inductive generalisations in a different way, whether the objects concerned are artefacts or exemplars of living kinds (Gelman & Markman 1986: 203–205, Gelman & Markman 1987, passim, Keil 1986, Gelman & Coley 1991). Even small children assume that members of a living kind share undefined properties, that make them similar in spite of superficial differences. This assumption has a variety of consequences. Children for instance seem to make spontaneous inductive generalisations over natural kinds on the basis of single exemplars, even when their biological knowledge is minimal. Moreover, they seem to “project” certain properties rather than others. While the property of having certain organs is spontaneously projected from an exemplar to the kind in general, certain surface properties, such as weight, are not projected (Gelman & Markman 1986, 1987, Gelman 1988). Also, children find unnatural the suggestion that members of a living kind might be “transformed” into members of another kind, whereas such transformations are considered possible in artefacts (Keil 1986passim, 1989: 183–215).
The child also appears to develop, from the early stages of cognitive development, strong expectations concerning the behaviour of physical objects (Spelke 1988, 1990). For instance, the principles of continuity (objects move in continuous paths) and solidity (objects do not coincide in space) seem to be present in children as young as 4 months (Baillargeon 1987, Baillargeon & Hanko-Summers 1990).
Another domain which is the object of early theoretical development (roughly at the “pre-operational” stage in Piagetian terms) is that of mental processes. The child gradually develops a set of theoretical principles to do with the non-physical nature of mental entities such as thoughts, desires and dreams (Wellman & Estes 1986, Wellmann & Gelman 1988). Also, the child’s understanding of mental processes constitutes a naïve implicit theory, with causal assumptions concerning the relations between perceptions, thoughts and intentions (Astington, Harris & Orson 1988, Wellmann 1990, Perner 1991, Whiten 1991).
Such theories can be said to be “naïve” in that they are only partly congruent with adult notions, and constrain further conceptual development. The following are a series of tentative conclusions that can be drawn from this experimental work on such presumptions:
(i) they are domain-specific, and trigger functionally different cognitive processes, depending on the domain. In other words, acquiring knowledge does not imply applying an all-purpose, “theory-making” cognitive device to a variety of available stimuli. On the contrary, it implies applying significantly different cognitive heuristics to different domains;
(ii) they seem to develop spontaneously, independently of tuition or objective changes in the available information: for instance, children seem to shift from a “desire-based” naïve psychology to a “belief + desire” type of psychological explanation when they are about 4, without any changes in the kind of explanations or implicit explanations offered in their social environment. To put it very crudely, people learn much more than they have been taught;
(iii) such naïve conceptual presumptions are connected in a complex way to later theoretical development. In some cases, such as commonsense or “folk”-psychology, adult conceptions seem to flesh out the conceptual skeleton provided by naïve conceptions. They provide more material, more explanatory schemes, but never go against the spontaneous assumptions. In other cases, such as the acquisition of mathematics or physics, it is necessary for subjects to acquire counter-intuitive principles (such as e.g. the difference between force and motion, which even in schooled adults is often a domain of uncertainty).
(iv) The spontaneous assumptions seem to constitute cross-cultural universals; that is to say, as far as there is evidence, that evidence bears out the hypothesis that such naïve conceptual constructions are universal (see for instance Avis & Harris 1991 on Pygmy children’s theories of mind, Jeyifous 1985 on natural kinds and artefacts in Nigeria). This of course seems a direct consequence of point (iii). If such assumptions do not depend on cultural tuition, how could they differ from culture to culture?
To sum up, studies of concept acquisition seem to show that, from a very early age, a number of principles orient the subjects’ attention to certain aspects of experience, and constitute the basis of quasi-theoretical intuitive principles, on which empirical knowledge can be built. Most of these specific principles remain tacit, even in adults. Such data are extremely important for a study of cultural representations, and especially of religious ideas. They make it possible to put forward a more precise cognitive account of religious transmission.
My hypothesis is that the intuitive knowledge principles described here impose constraints on the content and organisation of religious ideas. Obviously, most religious assumptions focus on objects which seemingly violate commonsense assumptions. It can be shown, however, that in order to acquire these notions, subjects have to rely, implicitly, on the intuitive principles described above. In this framework, a religious idea would be described as cognitively optimal if (i) it contains an explicit violation of commonsense thinking and (ii) it makes implicit use of the intuitive principles of commonsense knowledge. The hypothesis is that religious representations which are cognitively optimal will be the most recurrent ones. Being easier to learn and memorise, they would have a greater “survival value”, in terms of cultural transmission, than other ideas. In the next section I will illustrate these hypotheses by describing the combination of intuitive and non-intuitive principles in the Fang case.
Let me now return to the Fang repertoire of social categories. Categories like ngengang (specialized healer) are connected to the umbrella concept beyem (persons having a magical capacity). Now the problem is that, as I said above, there is very little explicit discourse or implicit suggestions as to exactly what constitutes the difference between beyem and other people. The identification of a person as a beyem is based on two types of general features. There is, on the one hand, a series of observable traits, like the fact that a person is said to have undergone a certain initiation, that he or she performs certain rituals, etc. These elements, however, are neither necessary nor sufficient conditions for judging that someone is a beyem. People who do not have these traits are sometimes said to be beyem, and some persons who perform the rituals are not said to be beyem. The idea that a person is a beyem is based on the assumption that he or she has “something more” than the superficial features, something that all members of those categories have. No one can represent what it consists of, but it has to be there, otherwise, whatever one’s activities, one is not an exemplar of the category. External typical criteria are just indirect (and insufficient) evidence of the fact that people really belong to the category. This “essentialist” interpretation of the group of beyem makes it possible to understand both the vagueness of people’s statements about what makes beyem beyem, and the idea that any particular person either is or else is not one.
These features of the category “beyem” led me to put forward the hypothesis that such categories are represented in the same way as certain natural kind terms (Boyer 1990: 101–105, forthcoming (b)). The representation and use of a natural kind term always involves two types of general assumptions: (i) some assumptions about the typical features of the exemplars of the kind. Although most of these features are present in most exemplars, they are neither necessary nor sufficient conditions; (ii) the presumption of an underlying trait that is common to all exemplars of the kind (Schwarz 1979passim).
It is important to stress that the implicit assumption is about the existence of an underlying trait, not about what it consists of. Few people except biologists bother to represent what makes giraffes giraffes, although everyone does suppose that there is some such underlying trait. This implicit hypothesis is a necessary aspect of the everyday use of natural kind terms. The idea of an undefined common essence is a powerful cognitive mechanism, universally available to human minds; together with basic principles of tax-onomic ordering, it organizes biological knowledge in all cultures (see Atran 1987passim). There is good experimental evidence that the assumption of an underlying essence, together with the idea that typical traits are, precisely, only typical, is involved in the representation of living kinds at the earliest stages of cognitive development (Medin & Ortony 1989).
As I observed above, the features mentioned as typical of beyem do not constitute necessary and sufficient conditions for membership. At the same time, however, people suppose that there is some underlying feature that is common to all persons in the position. In other words, there is a presumption that persons occupying a certain position share an essence, although what the essence consists of is left undefined. People who are beyem are thus considered naturally different from others. In other words, the representation of the position seems to be based on the extension to social differences of spontaneous assumptions which prove extremely successful in dealing with the natural world.
To sum up, my contention here is that the way people acquire and represent the categories denoting certain types of position is very similar to the way they acquire and represent certain categories about the natural world. To be more specific, certain crucial assumptions spontaneously applied to natural discontinuities, concerning e.g. the presence of an unobservable essence or the plausibility of instance-based generalisations, are directly applied to those social domains.
In the domain of religious positions, we find a set of ideas which have no natural referent; nor are they the object of any explicit tuition. My first claim is that, in such cases, the transmission process is in part assured by the fact that subjects spontaneously apply to the material at hand some assumptions which are highly salient in natural domains. My second claim is that the representation of certain religious positions as the consequence of inherent, unobservable natural properties has a very high “survival potential”, as it were, precisely because it makes use of assumptions carried over from core knowledge.
Many social categories related to religious performance seem to be construed in a way that is similar to the Fang beyem, that is, with an “essentialist” hypothesis. People know that the persons included in the category (shamans, priests, diviners, etc.) are actually different from the others, although they tend to take external criteria, like the performance of certain rituals, as only symptomatic of the underlying difference. This does not mean, obviously, that all social categories to do with religious performance are implicitly conceived as quasi-biological. The biological analogy is only a particularly salient way of expressing a deeper, “essentialist” principle, following which the difference between members of the category and other people must be postulated, cannot be observed, and usually is not reducible to external criteria. Also, this essentialist hypothesis makes it possible to understand the fact that in many groups, the idea of a transformation is absurd, as far as such religious social categories are concerned. For instance, take the case of a Fang diviner, who for some reason does not manage to convince people of his genuineness, and is subsequently declared to be a non-beyem. No-one in such a case would ever think that the person in question first was a beyem and then lost that quality. They would assume that the person had been a non-beyem all along, and wrongly identified as a beyem.
Surprisingly, there is very little reliable data on the ways in which social categories are actually represented, for instance on the way people interpret those revisions and re-identifications which inevitably happen in any social group. Because of the scarcity of data, it is difficult to evaluate the general relevance of the hypothesis of “social essentialism” put forward here. It is particularly difficult to understand in what way essential hypotheses can be combined with formal criteria for membership, for instance in churches, wher there exist formal criteria for priesthood. The question, whether such institutional criteria as ordination actually represented by members of a congregation as sufficient conditions for membership of the social category priest, is a moot one. Most studies of doctrinal religions take explicit theologies as their main object, and neglect to examine to what extent those theologies are congruent with people’s actual representations. In the absence of such data, however, the hypothesis of social essentialism provides at least an example of the type of result a cognitive study of religious roles may lead to, even if in this case the relevance of the hypothesis may be limited to traditional, non-scriptural religious systems.
Let me now turn to the organizing principles of the conceptual structures developed. As I mentioned above, anthropological descriptions often take it for granted that religious ideas invariably come in a theoretical format. It appears, however, that the sets of representations concerned cannot really be called “theoretical”, unless one has a rather metaphorical understanding of that term. To return to our example, the Fang ideas about ritual specialists (beyem for instance) are not really integrated in a theoretical format, for all the reasons listed above. For one thing, they are not consistent and they come in different formats (some ideas are about singular occurences and others about general principles, etc.). Differences between “theories” and the type of conceptual structures we are focusing on, can be apprehended in terms either epistemic or cognitive. From an epistemic viewpoint, some anthropologists have pointed out that the “theoretical” format is singularly inappropriate to a description of religious beliefs. Sperber for instance (1982 passim) contends that describing such ideas in terms of pro-positional attitudes is misleading, in that the objects of belief are not amenable to a propositional description. They can be called “semi-propositional” representations, in that they provide only fragmentary elements of propositional identification. Here, however, I will leave aside these epistemic questions and focus on the cognitive aspects of religious conceptual structures. In many respects, their most salient characteristics make the “theoretical” description rather inadequate. Here I will insist on an aspect that is generally ignored in anthropological models, namely the conjectural nature of these assumptions.
The conjectural nature of religious ideas is often a direct consequence of the way they are produced. The identification of social roles, for instance, inasmuch as it is based on assumptions of shared essence, naturally leads to the fact that every singular identification is a conjecture. Categories are thought to denote intrinsic differences between kinds or sorts of people. That such underlying differences exist is taken as true. That they apply to particular people hic et nunc is necessarily a matter of conjectures and non-demonstrative inferences. That the person really occupies the position in question (i.e., really has the natural properties posited) is only the most plausible conclusion, given the observable properties at hand, but corrections and re-identifications are always possible, and do in fact happen. The Fang who re-identify someone as not being a ngengang would say that although the person did have the relevant surface features, he/she was not really a ngengang from the beginning. The main point here is that there is no possibility to tell the difference on the basis of external, observable differences. And since there are no such differences, deciding one way or another is a matter of corrigible guesses and inferences.
That religious ideas command variable commitment is a commonplace of religious anthropology. Every anthropologist knows from experience that even the most basic tenets of religious systems may be represented as extremely convincing conjectures rather than intuitively obvious facts. Another aspect of this question, however, is less often commented on, namely the fact that, in any cultural environment, there is a constant competition between alternative religious interpretations of any event or state of affairs. Even in simple cases where only one religious framework is available, that framework inevitably leaves considerable latitude in the interpretation of any single event. A consequence is that, in most circumstances, any subject is provided with several interpretative schemes, the difference between which is a matter of plausibility or “naturalness”. For instance, the societies where illness is interpreted as the result of malevolent human witchcraft invariably have alternative etiological schemes, linking illness e.g. to purely contingent somatic disorders, to some powerful god’s intervention, etc. The only way of explaining why a certain causal scheme is (conjecturally) taken as relevant in a particular case is to explain what makes it intuitively more natural. A description of the acquisition and fixation of religious assumptions must take into account the fact that there are intuitively obvious differences in certainty between thoughts; moreover, those differences are not immutable, they may and do evolve. Thoughts may become more (or less) intuitively plausible after certain processing episodes. These aspects are crucial for the acquisition and representation of religious ideas.
It seems difficult to account for the transmission of the latter type of conceptual structures without some precise notion of cognitive salience. The point of such a notion would be (i) to give a more precise formulation of the intuitions concerning partial credal states, i.e. states in which a belief is held only partially, (ii) to provide an account of the strengthening (and weakening) of commitment, and (iii) to describe the type of conceptual structures that are based on assumptions of variable salience. Here, obviously, I cannot give a full-blown formulation of such a framework. Rather, I will try to give a more detailed description of the “brief which a theory of cognitive salience would have to fulfill.
Some notion of degree of commitment has always been a necessary ingredient in theories of subjective probability (cf. for instance Ramsey 1931, Carnap 1950). In such theories, having a partial commitment to a proposition “p” means believing that the probability of “p” being true is 1/x, where x<1. Such models, however, do not necessarily provide a good model for degrees of belief, or even for actual inductive reasoning.8 The salience of an assumption should be approached in terms that link subjective intuitions of credal states to functional properties of the assumption. Such models are available, for instance in the “framework for induction” put forward by Holland et al. (1986). The main consequence of such models is to define the cognitive salience of an assumption as the objective probability of its activation, given a certain set of conditions, notably given that a certain set of other representations are already activated.
We cannot explain the actual transmission and fixation of cultural representations without a framework for cognitive salience. This should be so designed as to provide a precise account of the interaction of two types of mechanisms in the construction of conceptual structures. First, some assumptions are directly imported from intuitive knowledge, and are often domain-specific assumptions. Second, the material provided by any cultural environment provides the input for mechanisms of strengthening-weakening which are not domain-specific. They concern the general effects of confirmation, refutation and revision on the salience of assumptions.
To sum up, I have put forward in this paper a series of hypotheses that constitute only tentative elements of a cognitive framework for the explanation of religious ideas. The first one is that the main point of a cognitive approach is to account for the recurrence of non-trivial properties of religious representations. Although this goal has often been misconstrued in anthropological research, it is now becoming increasingly clear that we must take into account important findings and hypotheses of psychology and cognitive science, in order to go beyond the present state of anthropological descriptions. Another assumption is that we cannot understand the transmission of religious ideas if we do not take into account the functional specificities of different types of .mental representations. A third hypothesis is that intuitive knowledge structures impose constraints on religious representations. More precisely, if religious representations combine explicit violations of some intuitive principles, and implicit confirmation of other intuitive principles, then they are cognitively optimal. This means that they are more likely to be acquired, stored and transmitted than representations which do not include this particular combination of violated and confirmed assumptions. A final hypothesis is that the processes whereby religious ideas are made intuitively plausible to human minds are inductive processes of belief-fixation, which cannot be studied unless we have precise models of the strengthening of non-demonstrative inferences.
The main point of this paper was to show that recent advances in experimental psychology, notably in the field of conceptual development, can throw light on the processes whereby subjects develop intuitive understandings of religious notions, even in contexts where cultural transmission is fragmentary. Obviously, it is not possible to decide on the validity of such hypotheses, on the sole basis of the type of data produced in ordinary anthropological fieldwork. The various forms of “participant observation”, intuitive hypothesis-testing and informal interview techniques favoured by anthropologists constitute an indispensable grounding for the study of religious representations. However, they are not designed to provide the kind of fine grain description of mental representation that is the necessary prerequisite of a cognitive study. This will probably require a different mode of data-gathering, in which traditional fieldwork methods are completed with more constrained experimental studies. Such experimental studies may not solve all the traditional problems of the anthropology of religion. They will, however, provide a precise answer to a long-neglected question, that of the processes whereby certain types of ideas are made “natural” and intuitively obvious to human subjects in different cultural settings.
King’s College PASCAL BOYER
University of Cambridge
GB - Cambridge CB2 1ST
1 As Keesing points out, the cultural models described in modem cognitive anthropology “comprise the domain of (culturally constructed) common sense. They serve pragmatic purposes. They explain the tangible, the experiential […] the probable” (1987: 374).
2 See Boyer 1990: ch. 1, Lawson & McCauley 1990: 32–44, Boyer forthcoming (a) for detailed treatments of these questions.
3 The hypotheses presented in this paper constitute a summary of the main points of Boyer forthcoming (c).
4 See Morris 1987 for a detailed survey of anthropological theories of religion. Lawson & McCauley (1990: 122–123) give a succint description of the three types of universals (substantive, formal and functional) that can be posited in a cognitive theory.
5 This approach, of course, is not really new in anthropological theory, and the intellectual prestige of Darwinian theory led to many models of cultural evolution based on some notion of selection, before and after Tylor’s famous statement that “to the ethnographer, the bow and arrow is a species” (Tylor 1871 [I]: 7). See Ghiselin (1973 passim) and Ingold (198: 33–47) for an analysis of these theories and the multiple misunderstandings they often produced.
6 See for instance Harris & Heelas 1979 for a general survey of the ambiguities in the psychological implications of descriptions couched in terms of “collective representations”, and Boyer 1987 for a discussion of the problems generated by such descriptions in the study of cultural transmission.
7 The anthropological literature on religious ideas is of course more diverse, although the point generally holds (see Boyer 1987 passim for a detailed argument). for examples of monographs that take into account the “fragmentedness” or incompleteness of the input, see e,g, Keesing 1982, Toren 1987, 1988.
8 There is no space here to review the vast literature on inductive belief-fixation and the various problems generated by the application of “inductive logics” to actual reasoning. See Nisbett & Ross (1980) for a description of the “vividness” phenomena which make it necessary to have some notion of cognitive salience. Kahnemann, Slovic and Tversky eds. (1982) is a classical source on non-logical heuristics in the evaluation of subjective probabilities, and Osherson et al. (1986) on the relevance of such problems for argumentation theory. Holland et al. (1986) give a survey of the literature and put forward a detailed framework, from which most of my remarks are inspired.
Astington, J.W., Harris, P., & Olson, D.R. (eds.)
1988 Developing theories of mind, Cambridge: Cambridge University Press.
Atran, S.
1987 Ordinary constraints on the semantics of living kinds. A commonsense alternative to recent treatments of natural-object terms. Mind and Language 2: 27–63.
1989 Basic conceptual domains, Mind and Language 4: 5–16.
Avis, J. & Harris, P.L.
1991 Belief-desire reasoning among Baka children: evidence for a universal conception of mind, Child Development 62: 460–467.
Baillargeon, R.
1987 Young infants’ reasoning about the physical and spatial characteristics of a hidden object, Cognitive Development 2: 179–200.
Baillargeon, R. & Hanko-Summers, S.
1990 Is the object adequately supported by the bottom object? Young infants’ understanding of support relations, Cognitive Development 5: 29–54.
Bloch, M.
1985 From Cognition to Ideology, in R. Fardon (ed.), Power and Knowledge. Anthropological and Sociological Approaches, Edinburgh: Scottish Academic Press.
Boyd, R. & Richerson, P.J.
1985 Culture and the evolutionary process, Chicago: University of Chicago Press.
Boyer, P.
1986 The ‘Empty’ Concepts of Traditional Thinking. A Semantic and Pragmatic Description, Man (n.s.) 21: 50–64.
1987 The Stuff ‘Traditions’ Are Made Of. On the Implicit Ontology of an Ethnographic Category, Philosophy of the Social Sciences 17: 49–65.
1990 Tradition as Truth and Communication. A Cognitive description of traditional discourse, Cambridge: Cambridge University Press.
[forthcoming (a)]. Introduction: Cognitive Aspects of Religious Symbolism, in Boyer, P. (ed.), Cognitive Aspects of Religious Symbolism, Cambridge: Cambridge University Press.
[forthcoming (b)]. Pseudo-natural kinds, in Boyer, P. (ed.), Cognitive Aspects of Religious Symbolism, Cambridge: Cambridge University Press.
[forthcoming (c)]. The Naturalness of Religious Ideas: Outline of a Cognitive Theory of Religion, Berkeley/Los Angeles: University of California Press.
Carey, S.
1985 Conceptual Change in Childhood, Cambridge, Mass.: The M.I.T. Press.
Carnap, R.
1950 Logical foundations of probability, London: Routledgc and Kegan Paul.
Cavalli-Sforza, L.L.
1986 Cultural Evolution, American Zoologist 26: 845–855.
Cavalli-Sforza, L.L. & Feldman, M.W.
1981 Cultural transmission and evolution: a quantitative approach, Princeton, N.J.: Princeton University Press.
Dougherty, J.D.W. (ed.)
1985 Directions in Cognitive Anthropology, Urbana/Chicago: University of Illinois Press.
Dougherty, J.D.W. & J.W. Fernandez (eds.)
1980 Cognition and Symbolism, [special issue of] American Ethnologist, 7.
Dougherty, J.D.W. & J.W. Fernandez (eds.)
1982 Cognition and Symbolism II, [special issue of] American Ethnologist, 9.
Gelman, R.
1990 First principles organizing attention to and learning about relevant data: number and the animate-inanimate distinction as examples, Cognitive Science 14: 79–106.
Gelman, S.
1988 The development of induction within natural kind and artefact categories, Cognitive Psychology 20: 65–95.
Gelman, S. & J. Coley
1991 Language and categorization: the acquisition of natural kind terms, in Gelman, S. & Byrnes, J.P. (eds.) Perspectives on Language and Thought: Interrelations in Development, Cambridge: Cambridge University Press.
Gelman, S. & E. Markman
1986 Categories and Induction in Young Children, Cognition 23: 183–209.
1987 Young children’s inductions from natural kinds: the role of categories and appearances, Child Development 58: 32–41.
Ghiselin, M. T.
1973 Darwin and evolutionary psychology, Science 179: 964–968.
Harris, P. & Heelas, P.
1979 Cognitive processes and collective representations, European Journal of Sociology 20: 211–241.
Holland, J.H., K.J. Holyoak, R.E. Nisbett & P.R. Thagard
1986 Induction. Processes of Inference, Learning and Discovery, Cambridge, Mass.: The M.I.T. Press.
Ingold, T.
1986 Evolution and Social Life, Cambridge: Cambridge University Press.
Jeyifous, S.
1985 Atimodemo: Semantic Conceptual Development among the Yoruba, PhD Thesis, Cornell University.
Kahnemann, D., Slovic, P. & Tversky, A., (eds.)
1982 Judgment under uncertainty: heuristics and biases, Cambridge: Cambridge University Press.
Keesing, R.
1982 Kwaio Religion. The Living and the Dead in a Solomon Island Society, New York: Columbia University Press.
Keesing, R.
1987 Models, “folk” and “cultural”, in D. Holland & N. Quinn (eds.), Cultural Models in Language and Thought, Cambridge: Cambridge University Press.
Keil, F.C.
1979 Semantic and Conceptual Development, Cambridge, Mass.: Harvard University Press.
1986 The acquisition of natural kind and artefact terms, in A. Marrar & W. Demopoulos (eds.), Conceptual Change, Norwood, N.J.: Ablex.
1987 Conceptual development and category structure, in U. Neisser (ed.)., Concepts and Conceptual Development: Ecological and Intellectual Factors in Categorization, Cambridge: Cambridge University Press.
1989 Concepts, Kinds and Conceptual Development, Cambridge Mass.: The M.L.T., Press.
Laurendau, M. & A. Pinard
1962 Causal Thinking in the Child, New York: International Universities Press. Lawson, E.T. & R. McCauley
Laurendau, M. & A. Pinard
1990 Rethinking Religion. Connecting Culture and Cognition, Cambridge: Cambridge University Press.
Leslie, A.
1979 The Representation of Perceived Causal Connection, D. Phil, thesis, University of Oxford.
Leslie, A. & Keeble, S.
1987 Do six-months old infants perceive causality?, Cognition 25: 265–288.
Lumsden, C.J. & Wilson, E.O.
1981 Genes, Minds and Culture, Cambridge Mass.: Harvard University Press.
Markman, E.M.
1989 Categorization and Naming in Children: Problems of Induction, Cambridge, Mass.: The M.I.T. Press.
Medin, D.L., & Ortony, A.
1989 Psychological essentialism, in Vosniadou, S. & Ortony, A. (eds.), Similarity and Analogical Reasoning, Cambridge: Cambridge University Press
Morris, B.
1987 Anthropological theories of religion, Cambridge: Cambridge University Press.
Nisbett, R. & Ross, L.
1980 Human Inference. Strategies and shortcomings of social judgement, Englewood Cliffs, N.J.: Prentice-Hall.
Osherson, D.N., Smith, E.E. & Shafir, E.B.
1986 Some origins of belief, Cognition 24: 197–224.
Perner, J.
1991 Understanding the Representational Mind, Cambridge, Mass.: The M.I.T. Press.
1954 The child’s construction of reality, New York: Basic Books.
Piaget, J.
1974 Understanding causality. New York: Norton.
Ramsey, F.P.
1931 The foundations of mathematics, London: Kegan Paul.
Schwarz, S.P. (ed.)
1979 Natural Kind Terms, Cognition 7: 301–315.
Spelke, E.S.
1988 The origins of physical knowledge, in L. Weizkrantz (ed.), Thought without Language, Oxford: Oxford University Press.
1990 Principles of object perception, Cognitive Science 14: 29–56.
[forthcoming] Physical Knowledge in Infancy: Reflections on Piaget’s theory, in Carey, S., & Gelman, S. (eds.), Biology and Cognition.
Sperber, D.
1982 Apparently Irrational Beliefs, in M. Hollis & S. Lukes (eds.), Rationality and Relativism, Oxford: Basil Blackwell.
1985 Anthropology and Psychology: towards an epidemiology of representations, Man 20: 73–89.
1991 The epidemiology of beliefs, in C. Fraser (ed.), Psychological Studies of Widespread Beliefs, Oxford: Oxford University Press.
Toren, C.
1987 Children’s perceptions of gender and hierarchy, in G. Jahoda & I.M. Lewis (eds.), Acquiring Culture, London: Routledge.
1988 The continuity and mutability of tradition as process, Man n.s. 23: 696–717.
Tylor, E.B.
1871 Primitive Culture, London: John Murray.
Wellmann, H.M.
1990 The Child’s Theory of Mind, Cambridge, Mass.: The M.I.T. Press.
Wellmann, H., & Estes, D.
1986 Early understanding of mental entities: a re-examination of childhood realism, Child Development 57: 910–923.
Wellmann, H.M. & Gelman, S.
1988 Children’s understanding of the non-obvious, in Sternberg, R.J. (ed.), Advances in the Psychology of Human Intelligence (vol. 4), Hillsdale N.J.: Lawrence Erlbaum Associates.
Whiten, A., (ed.)
1991 Natural Theories of Mind: the Evolution, Development and Simulation of Everyday Mindreading, Oxford: Blackwell.