It is often held that Marxism embodies distinctive methodological doctrines which distinguish it from ‘bourgeois’ social science.footnote1 The difference has been characterized in various ways: Marxism is scientific and materialist, bourgeois theory ideological and idealist; Marxism is holistic, bourgeois theory is individualistic; Marxism is dialectical and historical, bourgeois theory is linear and static; Marxism is anti-empiricist and anti-positivist, bourgeois theory empiricist and positivist. These claims have differed considerably in substance, but the near consensus view has been that an irreconcilable methodological fissure divides Marxism from its rivals.footnote2 Recently this unanimity has been broken by a current of Marxist theory, sometimes labeled ‘analytical Marxism’, which categorically rejects claims for Marxism’s methodological distinctiveness.footnote3 In contrast to what has generally been maintained, authors such as Jon Elster, John Roemer, Adam Przeworski and G.A. Cohen have argued that what is distinctive in Marxism is its substantive claims about the world, not its methodology, and that the methodological principles widely held to distinguish Marxism from its rivals are indefensible, if not incoherent.

Perhaps the most striking example of the rejection of claims to Marxian methodological distinctiveness comes from those analytical Marxists who explicitly declare themselves proponents of ‘methodological individualism’, thereby endorsing a methodological position they attribute to sound social science, but one that virtually all Marxists have traditionally rejected.footnote4 As is well known, Marx inveighed against the ‘individualism’ of the classical economists and contractarian philosophers, heaping scorn on efforts to conceive individuals abstracted from social relations and on theories based upon the imputed choices of these ‘abstracted individuals’. And nearly all Marxists, whatever their differences, have accorded explanatory relevance to social ‘totalities’, in apparent opposition to the strictures of individualist forms of analysis. Furthermore, until quite recently, proponents of methodological individualism have been equally scornful of Marxism. Hayek and Popper, among others, have even promoted methodological individualism expressly as an alternative to Marxian explanatory practices. It is therefore ironic, to say the least, to maintain that what is worth taking seriously in Marx’s thought can be reconstructed in methodological individualist fashion; and that only by recasting Marxian explanations in this way can we save the ‘rational kernel’ (as Marx might have put it) of Marx’s thought from the indefensibility of so many of his own formulations and from the obscurantism that afflicts much of what has come to be identified as Marxism.

We are sympathetic to the idea that what is distinctive in Marxian theory is substantive, not methodological; and that as a science of society, the methodology adopted by Marxists ought to be just good scientific methodology. But methodological individualism is not good scientific methodology, even if, as we will show, some of the intuitions that motivate it are sound. The plausibility of Marxian methodological individualism depends, of course, on what methodological individualism is thought to be. Unfortunately, at the current stage of discussion, many of the obscurities that have always pervaded debates about methodological individualism are effectively reproduced in the Marxian context. One objective of this essay is to reduce this confusion by clarifying the stakes in claims for and against methodological individualism, both as these apply to the specific context of Marxian explanations and to social scientific explanations generally.

In the next section, we characterize methodological individualism by contrasting it with three other stances towards explanation in social science. This will be followed by a more intensive discussion of methodological individualism itself, suggesting that its reductionist ambitions cannot be fulfilled. Nevertheless we will argue, in the final section, that a practical implication of methodological individualism—that the microfoundations for macro-level theory should be elaborated—is timely and important, even if methodological individualism itself is not. Throughout this discussion, Jon Elster’s book, Making Sense of Marx, will be a central point of reference.footnote5 Elster is among the most insightful of Marxian methodological individualists, and this book represents the most sustained attempt within the Marxian tradition to defend methodological individualism. It is therefore a useful point of departure for an examination of the doctrine’s strengths, as well as its flaws.

Methodological individualism is a claim about explanation. It is the view that all social phenomena are best explained by the properties of the individuals who comprise the phenomena; or, equivalently, that any explanation involving macro-level, social concepts should in principle be reduced to micro-level explanations involving only individuals and their properties. In order to give methodological individualism a precise definition, it will be helpful to contrast it with three other possible views: atomism, radical holism and anti-reductionism. The first two of these positions, at least in their pure form, probably have no actual defenders, but they are implicit tendencies within social theory. Indeed, in debates over methodological individualism, disputants sometimes appear to confuse their opponents’ views with one or the other of these positions. Thus defenders of methodological individualism depict anti-reductionists as radical holists, and defenders of anti-reductionist positions sometimes regard methodological individualists as atomists. Therefore, in order to clarify the issues at stake, it will be useful to map out all four possibilities.

These methodological stances towards social scientific explanation differ in what they regard as explanatory. They can be distinguished on two dimensions: whether or not they regard the properties of and relations among aggregate social entities as irreducibly explanatory; and whether or not they regard relations among individuals as explanatory.footnote6 Aggregate social entities include such things as societies, groups, classes, organizations, nations, communities. Such entities have properties (e.g. inflation rates, institutional forms, distributions of income) and exist in a variety of relations to each other (e.g. relations between collectively organized classes). Individuals also have both properties (e.g. beliefs, abilities, resources) and exist in a variety of relations with other individuals (e.g. sibling relations, employer–employee relations, etc.). Taking these two dimensions together, we get the following typology of principles of explanation of social phenomena:

Atomism is a methodological stance which denies that relations—whether between individuals or between social entities—are ever genuinely explanatory. Consider any social phenomenon—for example, the transformation from feudalism to capitalism. An atomist would say that this transition can in principle be fully explained by causal processes strictly internal to individuals in the society in question. While interactions among these individuals matter for explaining the emergence of feudalism, the causal processes which govern the outcomes of such interactions are entirely intra-individual.footnote7 The atomist would insist, in other words, that only entities which are fully constituted non-relationally are explanatory. On the face of it, atomism seems plainly unsustainable. In our everyday lives we exist within a network of relations to other people—as parents, siblings, employers, customers, and so on. These relations appear to be explanatory, and also, it would seem, irreducible: being a parent, for instance, necessarily involves another individual, the child. But atomism is not quite so implausible as may at first appear. The atomist might argue that everything that seems explanatory about irreducible relations between individuals actually is explanatory only because of the corresponding (non-relational) psychological states of these individuals; that what matters explanatorily in, say, power relations between individuals is not an irreducible relation between these individuals, but their beliefs and desires, considered atomistically. If I believe you will punish me if I do x and you believe that I have these beliefs, then we will each act in particular ways. The apparent power ‘relation’ between individuals, the argument would go, is really no more than a set of reciprocal beliefs, and it is these beliefs, rather than any ‘objective relation’, which explains actions.