Dan Sperber (1997) Methodological individualism and cognitivism in the social sciences. Unpublished English version of “Individualisme méthodologique et cognitivisme.” In R. Boudon, F. Chazel & A. Bouvier (eds.), Cognition et sciences sociales (Presse Universitaires de France), 123-136.

“I would like to contrast two interpretations, a weak one and a strong one, of the notion of methodological individualism, and two interpretations, a weak one and a strong one, of the notion of cognitivism. This double contrast determines four ways in which one might choose to be at the same time a methodological individualist and a cognitivist in the social sciences. One way, where both positions are adopted with a weak interpretation, is of little interest. I will argue that another way, where both positions are adopted with a strong interpretation, is incoherent. I will compare the two other possibilities…”

Dan Sperber

I would like to contrast two interpretations, a weak one and a strong one, of the notion of methodological individualism, and two interpretations, a weak one and a strong one, of the notion of cognitivism. This double contrast determines four ways in which one might choose to be at the same time a methodological individualist and a cognitivist in the social sciences. One way, where both positions are adopted with a weak interpretation, is of little interest. I will argue that another way, where both positions are adopted with a strong interpretation, is incoherent. I will compare the two other possibilities.

According to methodological individualism (whichever way you interpret the notion), social phenomena can be adequately explained by showing that they are the outcome of individual behaviors. Methodological individualism is standardly contrasted with holism, according to which social phenomena can be explained only by invoking the behavior or the properties of entities which are irreducibly supra-individual, such as culture or institutions.

It has less been noticed that methodological individualism contrasts also, at least in principle, with another conceivable view that could be called “infra-individualistic.” According to infra-individualism, an adequate explanation of social phenomena should invoke the behavior or properties of infra-individual entities. Infra-individualistic explanations are hardly ever used in the social sciences (Gabriel Tarde being an interesting historical exception – see Karsenti 1993). Such explanations are standard, on the other hand, in epidemiology, this other science of collective phenomena. A good explanation of an infectious epidemic focuses on the behavior of pathogenic agents, bacteria or viruses, and on that of the antibodies produced by the immune system. Individuals (whose behavior is also, of course, relevant to explanation) are the locus at which many of these infra-individual processes take place, but they are not the agents of these processes. Another example of an infra-individualist approach is provided by the neo-Darwinian view of culture put forward by Richard Dawkins (1976, 1982; for a discussion, see Sperber 1996, chapter 5). According to Dawkins, culture is made up of units of information he calls “memes”, which replicate inside and across individuals in the manner of viruses.

It is generally taken for granted in current cognitive psychology that most mental processes are unconscious. Should unconscious mental processes be considered individual or infra-individual in the sense of having the individual as locus, but not as agent? It depends on how you specify the notion of an individual.

What are the individuals of methodological individualism? Are they human organisms as studied by biology and naturalistic psychology? Are they persons, conscious subjects, the characters common sense psychology is about? From the point of view of a naturalistic psychologist or philosopher (e.g. Dennett 1991), personhood is not a given; it is better seen as the product of a process of socialization (as indeed many anti-individualistic, anti-naturalistic anthropologists and sociologists have insisted).

Methodological individualists give an important explanatory role to involuntary and unforeseen consequences of voluntary behaviors. But what about involuntary behaviors? And what about involuntary aspects of voluntary behaviors? In other terms, should one pay attention to all aspects of all human behaviors, or should one grant a uniquely important explanatory role to human actions? And those individual behaviors that end up being invoked in the explanation of social phenomena, how should they themselves be explained? By drawing on scientific psychology with all its resources, or, essentially, in terms of the reasons of the agents?

In its strong interpretation, which is also the most fruitful and the most widespread (from Weber and Hayek to Boudon and Elster), methodological individualism enjoins that social phenomena be explained by showing how they result from individual actions, which themselves are to be explained by the reasons of the actors. It is methodological individualism so understood that I want to discuss. When I use the term “individualism” without further precision, it must be understood in this strong sense. On the other hand, the naturalistic view I am arguing for is individualistic in the weak sense. Naturalism grants a causal role only to entities and properties the natural character of which is manifest. Naturalism cannot but conflict with standard holism, which grants a causal role to supra-individual entities the place of which in nature is wholly mysterious. Naturalism should also be at odds with individualism in the strong sense, for the only organisms, and not persons, or actors, are manifestly natural entities.

The individualistic approach focuses on the reasons and the rationality of social actors. To this extent, methodological individualism is concerned with mental representations, their grounds, and their role in behavior, hence with cognition. Quite generally, the social sciences, which study, among other things, beliefs, ideologies, technical knowledge, etc. have, trivially, a cognitive dimension. Research programmes focused on this aspect of the social and cultural domain can be called “cognitivist” in a weak sense of the term. Here however, I will use “cognitive” and ‘cognitivist” in the stronger sense that these terms have in the cognitive sciences strictly understood.

The hard core of the cognitive sciences is the ambition to provide a naturalistic and mechanistic explanation of mental processes. An explanation is mechanistic when it analyses a complex process as an articulation of more elementary processes, and it is naturalistic to the extent that there are good reasons to think that these more elementary processes could themselves be analyzed in a mechanistic fashion down to the level where their natural character would be wholly evident.

Almost all branches of psychology today call themselves “cognitive”. When the term is taken in its weak sense, the only objection is that it is used redundantly. After all, every branch of psychology is, at least in part, about mental representations. If, on the other hand, “cognitive” is understood in the strong sense, then the fact is that many research programmes in current psychology do not aim at providing mechanistic and naturalistic explanations, and are not, therefore strictly cognitive, whatever their other merits.

In the social sciences, mechanistic explanations are not infrequent – many individualistic explanations, in fact, can be seen as mechanistic -, but they hardly ever are naturalistic. To develop a truly naturalistic approach, bridges should be built between the social sciences and neighboring disciplines: biology, ecology, and, particularly, psychology in its naturalistic aspects. There are such links between demography and biology, or between human geography and ecology. Links to cognitive psychology, on the other hand, are rare, and either remain programmatic, or are limited to the study of very specialized topics, such as the evolution of languages or the classification of colors.

No one (or so I hope) is thinking of integrating the social sciences within the cognitive sciences. One might, much more modestly, wish to add to the list of current research programmes in the social sciences a programme partly grounded in the cognitive sciences and aimed at providing truly naturalistic explanations of social facts. Such a programme has rarely been articulated, let alone put into practice (but see Atran, 1990; Boyer, 1993; Hirschfeld, 1996; Hirschfeld & Gelman, 1994; Hutchins, 1995; Sperber, 1985b, 1987, 1996; Tooby & Cosmides, 1992). I have argued that such a programme should take the form of an epidemiology of representations.

Let us go back to epidemiology. Is an epidemic of influenza a social phenomenon? At least marginally, yes. Such an epidemic, not only has social effects, which is true also of a purely natural calamity such as an earthquake, it also has social causes; enabling causes such as the concentration and the daily movements of the population, and inhibiting causes such as health policy. Individualists could reasonably maintain that the social aspects of an epidemic of influenza can be dealt with without difficulty from an idividualistic standpoint. The risks incurred or avoided by individuals are an outcome of their actions, which themselves are to be explained by their reasons.

In the case of influenza, the division of labor between the infra-individualistic epidemiologist and the individualistic sociologist is easy enough. The division of labor might seem even easier in a case such as that of tobacco addiction. Tobacco addiction is an epidemiological phenomena, the effects of which are physiological and the causes of which are behavioral. Tobacco addiction is caused by human action, in particular reiterated acts of smoking. The sociologist might then study the behavioral causes of tobacco addiction, and the epidemiologist its effects on health, and the cause/effect relationship.

The practical syllogism that leads smokers to light a cigarette is sad and boring: I desire the kind of pleasure provided by smoking; smoking will satisfy this desire; hence I smoke. Sometimes, smoking has additional, or alternative, more “sociological” reasons such as the desire to imitate others, or the desire to give oneself a composure. Still, in general, the causal factor that best predict the moment at which a smoker lights a cigarette is the fall of the level of nicotine in his or her blood below a certain threshold. This cause, which has nothing to do with the calculus of reasons, is much more interesting, in explaining tobacco addiction, than the practical syllogism of the smoker, and much more general than occasional “sociological” reasons.

The structure of an interesting explanation of tobacco addiction is provided by a complex feedback loop which articulates physiological, psychological and economic factors. Smokers smoke because they feel the desire to smoke. They feel the desire to smoke because their level of nicotine falls. The fall in nicotine level causes smokers to smoke because, in smokers, this level is maintained at an average level high enough to cause physiological dependence. This level is maintained at a high enough level because smokers smoke. From the description of this psycho-physiological mechanism of self-perpetuation of a need, one moves without difficulty to the social economy of tobacco addiction: demand bolsters supply, supply bolsters demand, and so on.

In such an explanation one has a complex causal chain linking a variety of causal processes, some internal to individual organisms, others taking place in the environment. Among internal causes, one has mental causes. Among mental causes, one has beliefs, desires, and practical syllogisms leading to actions. Beliefs, desires, and practical inferences may be described as “reasons”. To do so is to draw attention to the fact that these mental states and processes can be evaluated from a normative point of view: they are open to a judgment of rationality.

The individualistic approach in sociology typically explains action in terms of reasons. In this, it may be seen as responding to a legitimate interest. An explanation in terms of reasons appeals to a form of intelligibility which is at the basis of our mutual interactions. Reasons can be argued about, they help determine responsibilities. From a naturalistic point of view however, reasons should be of interest to us not qua reasons, but qua causes among other causes. They are not a priori more interesting than other causes.

Epidemiology in the widest sense is the study of the way in which various conditions become distributed in population. The conditions considered may be physiological and pathological as in classical epidemiology; they could also be mental and “normal”.

A human population is inhabited by a much wider population of mental representations of all kinds: beliefs, values, techniques, projects, intentions and so on. These mental representations are distributed in the brains of individuals. Behaviors are caused by mental representations. The behavior of an individual, for instance walking or speaking, may be perceptible to other individuals, or it may leave perceptible traces, for instance footsteps or writing. I will call such perceptible behaviors and traces “public productions”. The public productions of an individual may provide an input to the mental processes of other individuals, causing them to construct their own mental representations. These representations can in turn result in public productions, which can trigger the construction of yet other mental representations in other individuals, and so on. A human group is thus crisscrossed by a mesh of causal chains where mental and environmental links alternate. Everything social, I would argue, is caught in that mesh.

Some of these causal chains have a stabilizing affect (see also Millikan 1984). They bring about the distribution, in the group members’ minds, of mental representations which are, if not identical, at least similar to one another, and the distribution, in the environment of the group, of familiar and reliable artifacts. Thus compelling arguments induce their hearers, once convinced, to convince others. Walkers open a trace, which other walkers follow, and others again, progressively inscribing in the landscape a path. Every stabilized social phenomenon, be it described as a social practice, or as a cultural representation, or as an institution, is the outcome of such processes of distribution.

A description of social facts from this epidemiological point of view is both mechanistic and naturalistic. Complex processes are decomposed into chainings of elementary processes. Some of these elementary processes are to be studied by ecology, other elementary processes are to be studied by cognitive psychology. In other terms the epidemiology of representations acknowledges causal powers only in material phenomena: environmental and mind/brain processes.

A methodological individualist should, it seems, grant that a causal explanation of the distribution of representations in human groups, if it were possible, would exhaust the causal explanation of the subject matter of a social sciences. To the extent that such an explanation would treat the reasons of social actors as mental causes of their actions, it would encompass an explanation in terms of reasons. Nevertheless, the individualist might maintain that, should such a project of causal explanation ever be effectively developed, it would make manifest the primordial character of reasons. At best, therefore, an epidemiology of representations would converge, after a painstaking detour through the narrow path of naturalism, with methodological individualism. If so, why should sociologists pay much attention to the cognitive sciences?

There are many examples, however, where psychological factors other than reasons play an essential role in the stabilization of social-cultural phenomena. I have just considered the case of tobacco addiction. Elsewhere, I have discussed the case of the folk classifications, of beliefs, and of oral traditions (see Sperber 1985b, 1990, 1994, 1996) and I will evoke them briefly before concluding. But first, I will first discuss the case of historical linguistics (see also Salmon 1995). It should provides a test case since it meets the conditions in principle most favorable for an individualistic explanation.

Historical linguistics is one of the most accomplished social sciences. It has been, for a long time, individualistic in the weak sense, that is, it tries to explain those collective phenomenon called languages and their evolution as the combined effect of a vast number of individual speech behaviors. There is no human activity where will manifests itself in a richer and subtler fashion than in speech. Through their choice of words, of grammatical forms, of language registers, and so on, speakers achieve a great variety of goals. They also modify in a locally imperceptible way the distribution of the linguistic data which will determine the language acquired by the next generation. The evolution of language is thus the involuntary cumulative effect of countless voluntary individual actions. It should be a perfect case where methodological individualism in the strong sense could demonstrate its fecundity. Conversely, if the strong individualistic approach contributes only marginally to the explanation of the evolution of language, then its explanatory powers are not as good as advertised.

There is no doubt, I repeat, that the micro-events which together bring about the macro-evolution of language are voluntary actions. There is no doubt that this evolution is an involuntary effect of these actions. How might the strong individualistic approach possibly fail explain such an ideally tailored case? It could fail because, voluntary actions always have involuntary aspects the cumulative causal effects of which may be greater than those of the voluntary aspects.

In fact, the cognitive analysis of speech shows what mix of voluntary and involuntary processes produce any utterance (see Levelt 1989). The choice of words for instance hangs in part on the intentions of the speaker, in part on syntactical constraints, and in part on the degree of relative accessibility of lexical items, a degree which changes as speech proceeds. Complex considerations of economy – economy of articulation for the speaker, and anticipated economy of processing for the hearer – affect in an involuntary manner the form of any utterance. The evolution of language turns out to be the involuntary effect of both intentional and unintentional aspects of speech, and, actually, more of the latter than of the former. Historical linguistics is individualistic, but not in the strong sense of the term.

Historical linguistics is the very example of an epidemiological discipline that neither ignores nor privileges actors’ reasons, and that takes into account all aspects, physiological and psychological, conscious and unconscious, of individual behaviors in order to explain their cumulative social effects.

One might object: historical linguistics is a peculiar discipline, with its own specialists and its own methods. Its explanations draw on factors and causal mechanisms which, for the most part, are specific to language. Historical linguistics does not, therefore, provide a good model for the social sciences in general. All this is true, but precisely because of its very specificity, the case of historical linguistics provides a quite general lesson. Imagine a sociological methodology which would sort a priori, among cognitive factors, those that may be invoked by sociology – say, reasons – and those that are to be studied by psychology alone. Such a methodology would commit to special disciplines, e.g. historical linguistics, whole dimensions of social life. It would end up defining the proper domain of sociology as the very domain in which the favored methodology happens to be fruitful. Surely, no mature empirical science should be defined by its instrument.

More important still, the case of historical linguistics tends to cast doubt on the very possibility of a homogeneous explanation of social facts. In an epidemiological perspective, one should not expect that the model which best explains a specific social phenomena should automatically apply to other social phenomena. Epidemiology is a discipline rich in conceptual tools but without a general theory. This is so because epidemiology studies the combined effects of quite diverse micro-mechanisms, different combinations of micro-mechanisms being at work in different cases. Compare, for instance, the case of tobacco addiction and that of malaria. In an epidemiology of representations, similarly, one should expect the causal explanation of different social phenomena to invoke each time different articulations of different environmental and psychological micro-mechanisms. There is no a priori reason, in particular, to either discount or put all one’s bet on the explanatory relevance of rational actions.

What is true of historical linguistics is true of the social domain in general. I’ll give but one more quick example. Consider the myths of an oral culture. Two types of reasons have been invoked to explain that the members of such a culture believe in their myths. In the intellectualists tradition, from James Frazer to Robin Horton, it is argued that myths provide such a culture with the best available answers to questions about life, death, group identity, power and so on. I have stressed another kind of reason to believe in myths (see Sperber 1985a, chapter 2). Individuals are born and grow in a society where these myths are already there. They hear them from the mouth of elders whom they have good reason to trust. Myths are accepted, therefore, in virtue of a sensible “argument from authority.” Should some individuals actually find a myth grossly unbelievable, they might also find it quite unbelievable that the elders and almost everybody else around them are all grossly mistaken in their belief. A sensible degree of intellectual modesty might suggest that they follow common opinion rather than their own ratiocination.

Whether one explains belief in myths by invoking one or the other type of reasons – the value of myths as answers to the great questions, or the authority attributed to those from whom one hears the myths – or even if one were to invoke simultaneously both kinds of reasons, one would be far from having explained why myths have the striking kind of contents that they have. If individuals are motivated in their beliefs by the search for answers to great questions, and if they can be satisfied with the kind of answer myths provide, then why, in the wide open space of possible answers, only such a very narrow range (well explored in comparative mythology) is being exploited across cultures? The argument from authority is also compatible, of course, with just any kind of content. To explain the characteristic content of myths, one must therefore turn to psychology and in particular to two of its branches, the psychology of memory and that of “naive theories”.

In an oral culture, human memory filters the contents that are likely to stabilize. This filtering is not just quantitative, it is also qualitative. Thus stories are better remembered than descriptions. Amongst stories, those with a “good form” are better remembered. Among stories with a good form, those involving certain types of characters and of objects are better remembered, and so on. Themes optimal for human memory are rooted in naive theories which themselves are rooted in genetically determined cognitive dispositions (see Boyer 1994, Sperber 1985b, 1990). The contents of myths are only weakly selected by the mechanism of rational evaluation. They are strongly selected by other cognitive mechanisms which operate inside individuals, but unbeknownst to them.

In such conditions, the explanation of myths must appeal to a variety of factors, some ecological, such as the volatility of verbal representations in a society without writing, and others psychological. Among psychological factors, some have to do with conscious processes or at least processes capable of consciousness, such as speculative reflection or decision making (about who to believe). Other psychological factors, the most important ones, have to do with essentially unconscious processes such as memorization, or the spontaneous formation of intuitive knowledge.

A plausible explanation of myth, just as a plausible explanation of the evolution of language, must be cognitive in the strong sense, and, therefore, it can be individualistic only in the weak sense.


To recapitulate. I have described two versions of individualism. Individualism in the weak sense contrasts only with holism. Individualism in the strong sense (where the “individual” means person, and behavior is explained in terms of reasons) contrasts not only with holism but also with a kind of infra-individualism for which processes which take place within an individual organism without being the action of a conscious agent have a major role to play in the explanation of social facts. I have also contrasted two versions of cognitivism. Cognitivism in the weak sense is the rather trivial acknowledgment that cognitive phenomena may play a role in the explanation of social facts. Cognitivism in the strong sense is the adoption of the mechanistic and naturalistic programme of cognitive sciences.

It goes without saying that, in both cases, the strong versions are more interesting: more open to challenge, of course, but also potentially much more fruitful. I have tried to show that it is impossible to opt simultaneously for both the strong versions of individualism and of cognitivism. Today’s individualistic sociology is individualistic in the strong sense and cognitivistic in the weak sense. This sociology has produced some splendid research, but, as any research programme, it also has limits, which might be narrower than is generally acknowledged. I have tried to show that a programme individualistic in the weak sense and cognitivist in the strong sense, in other terms mechanistic and naturalistic, might be of interest too.


Atran, Scott (1990). Cognitive Foundations of Natural History. Cambridge: Cambridge University Press.

Barkow, J., L. Cosmides & J. Tooby (Eds.) (1992) The Adapted Mind: Evolutionary Psychology and the Generation of Culture. New-York: Oxford University Press.

Boyer, Pascal (1994). The Naturalness of Religious Ideas. Berkeley: University of California Press.

Dawkins, Richard (1976). The Selfish Gene. Oxford: Oxford University Press.

Dawkins, Richard (1982). The Extended Phenotype, San Francisco, Freeman.

Dennett, Daniel (1991). Consciousness Explained. New York: Little, Brown, & Co.

Hirschfeld, L.A. (1996). Race in the making: Cognition, Culture and The Child’s Construction of Human Kinds. Cambridge, Mass.: MIT Press.

Hirschfeld, L. A. & S. A. Gelman (eds), (1994). Mapping the Mind: Domain pecificity in Cognition and Culture, New York: Cambridge University Press.

Hutchins, Edwin (1995). Cognition in the wild. Cambridge, Mass.: MIT Press.

Karsenty, Bruno (1993). Présentation. In Gabriel Tarde, Les Lois de l’Imitation, (first edition 1890). Paris: Kimé.

Levelt, Willem J. M. (1989). Speaking: From intention to articulation. Cambridge, Mass.: Bradford Books, M.I.T. Press.

Millikan, R.G. (1984). Language , thought, and other biological categories. Cambridge, Mass: MIT Press.

Salmon, Merrilee H. (1995). Causal explanations of linguistic behavior. Paper presented to the 10th International Congress of Logic, Methodology, and Philosophy of Science. Florence, August 1995.

Sperber, Dan (1985a). On Anthropological Knowledge. Cambridge: Cambridge University Press.

Sperber, Dan (1985b). Anthropology and psychology: towards an epidemiology of representations (The Malinowski Memorial Lecture 1984). Man (n.s.) 20, 73-89.

Sperber, Dan (1987). Les sciences cognitives, les sciences sociales et le matérialisme. Le Débat, 47, 105-115.

Sperber, Dan (1990). The epidemiology of beliefs. In Colin Fraser & George Gaskell (eds.) The social psychological study of widespread beliefs. Oxford: Clarendon Press, 25-44.

Sperber, Dan (1994). The modularity of thought and the epidemiology of representations. In L. A. Hirschfeld & S. A. Gelman (eds), Mapping the Mind: Domain specificity in cognition and culture, New York: Cambridge University Press, 39-67.

Sperber, Dan (1996). Explaining Culture: A Naturalistic Approach. Oxford Blackwell.

Tooby, J. & Cosmides L. (1992). The psychological foundations of culture. In J. Barkow, L. Cosmides & J. Tooby (Eds.) The Adapted Mind: Evolutionary Psychology And The Generation Of Culture. New-York: Oxford University Press.