[Cohen, G. A. (1982), Functional explanation: Reply to Elster, Political Studies 28 (1):129-135]

[ Books | Articles | Reviews | Index | The main menu ]


[start of page 129]


University College, London

I THANK Jon Elster for his generous review, and for his criticisms, not all of which I accept. I shall not respond here to every criticism with which I disagree, but I do want to comment on what I think are misconceived objections to my chapters on functional explanation. Having done so, I shall offer reservations on the extent to which my 'sometimes uncertain grasp of economic theory' (p. 122E 1 ) led me into error.

I grant that my defence of a functionally construed historical materialism is only partly successful, but I reject the methodological criticisms Elster directs against it. I believe, moreover, that there is no viable alternative construal of the central claims of historical materialism, so that if my defence fails, historical materialism fails. Hence the cost incurred by Marxism, if I am wrong, is considerable. That is no reason for thinking that historical materialism, in the version I favour, is true, but I should like the cost of its falsehood - if it is false - to be acknowledged, something which, as I shall explain, Elster is reluctant to do.

1. In Marx's theory, as I present it, history is the growth of human productive power, and economic structures (sets of production relations) rise and fall according as they enable or impede that growth. Alongside a society's economic structure there exists a superstructure, of non-production relations, notably legal and political ones. The superstructure typically consolidates and maintains the existing economic structure, and has the character it does because of the functions it fulfils.

Historical materialism's central claims are that

(1) the level of development of the productive forces in a society explains the nature of its economic structure, and
(2) its economic structure explains the nature of its superstructure.

I take (1) and (2) to be functional explanations, because I cannot otherwise reconcile them with two further Marxian theses, namely that

(3) the economic structure of a society is responsible for the development of its productive forces, and
(4) the superstructure of a society is responsible for the stability of its economic structure.

(3) and (4) entail that the economic structure has the, function of developing the productive forces, and the superstructure the function of stabilizing the economic structure. These claims do not (as Elster rightly insists, and as he sometimes recognizes I realize) by themselves entail that economic structures and superstructures are explained by the stated functions: x may be functional for y even though it is false that x exists because it is functional for y. But (3) and (4), in conjunction with (1) and (2), do force us to treat historical materialist explanation as functional. No other treatment preserves

* I am indebted to Annette Barnes, Grahame Lock and Arnold Zuboff, for pointing out infelicities in an earlier version of this paper.

1 'E' indicates a reference to Elster's review, and 'C' indicates a reference to my book.
[end of page 129, start of page 130]

consistency between the explanatory primacy of the productive forces over the economic structure and the massive control of the latter over the former, or between the explanatory primacy of the economic structure over the superstructure and the latter's regulation of the former.

2. In a number of works2 Elster has criticized undisciplined uses of functional explanation by Marxist and other social theorists, on grounds almost all of which are also (independently) developed in my book. He is impressed by how thoughtlessly the device of functional explanation is invoked, without adequate evidence, in the apparent belief that if an item has a function, then, ipso facto, it also has a functional explanation. I am impressed by how essential it is to vindicate the functional- explanatory device, as opposed to its misuses, if historical materialism is to be defended. These contrasting motivations should not generate a confrontation, and I complain that Elster is wrong to mount one. I accept that I have failed to substantiate the truth of some of my functional-explanatory claims, but since, as I shall show, I am not guilty of methodological or conceptual error, the principal part of Elster's critique is a failure.

3. Sometimes Elster more or less correctly reports my intentions, as in the 'more charitable' interpretation of them on pp. 127-8E. But his reports - not, as he claims, my intentions - are inconsistent, and some of them are quite unfounded.

He begins by assigning to me a 'general methodological position, an attempt to vindicate functional explanation in the social sciences as sui generis, i.e. reducible neither to causal nor to intentional explanation'(p. 121 E). This description is misleading in one respect and false in another.

The description is misleading because one would never gather from it that I hold, as does Elster, that there is successful functional explanation in biological science: much of our dispute is over the extent to which social science can emulate biological science in this regard. My strategy is to vindicate functional explanation as a device by arguing, controversially, that universally accepted biological explanations are functional explanations. I then argue that explanations of similar pattern apply in social and historical studies. So it is an important constituent of my position that functional explanation is not peculiar to social science.

There is also, more seriously, an outright falsehood in Elster's description. For I do not hold, to put it as he does later on (p. 125E), that 'functional explanation [is] a separate explanatory category on a par with causal explanation'. This attribution flatly ignores, and contradicts, my explicit statement (p. 250C, etc.) that functional explanation is a variety of causal explanation. It is, in my view, causal explanation of a special type, which is why it deserves a special name, but it is not explanation of some non-causal type. Elster should at least acknowledge that this is my declared position, even if he wrongly thinks that I am untrue to my own declaration.

Still, it might seem that there is an irreducibility claim, needing more careful statement than Elster himself has provided, to which I am indeed committed. The reader can decide that for himself when, after I have given a brief statement of the sort of causal explanation I think functional explanation is, the character of the irreducibility claim will be more clear.

Every causal explanation mentions one or more causally relevant features, which contribute to the explanation of what is explained. In my account a functional explanation is a causal explanation in which a certain sort of dispositional fact is a causally relevant feature. Where what is to be explained is an event of type E, the dispositional fact is that if E occurs, it brings about some consequence F. When E occurs (partly) because if E occurs it brings about F, we have a functional explanation

2 For example, J. Elster, Logic and Society, (New York, Wiley, 1978); Ulysses and the Sirens (Cambridge, Cambridge University Press, 1979) and 'Marxism, Functionalism and Game Theory', in Marxist Perspectives, forthcoming.
[end of page 130, start of page 131]

of E 3. 'Birds developed hollow bones because hollow bones facilitate flight' is an explanation of the functional type, with E being possession of hollow bones by birds and F being the facilitation of flight, which is a consequence of E.

An example of functional explanation in social science is that scale of production is large in a certain industry because large scale reduces costs in that kind of industry. If one now asks, how does the putatively explanatory functional fact, that large scale reduces costs, explain large scale, then the answer will be different in different cases. Sometimes it will be that the functional fact selects in favour of firms which for accidental reasons expand their scale, sometimes that wise planners recognize the functional fact and act accordingly, and sometimes both elements will figure in what I call the elaboration. But the differently elaborated explanations all depart from the same functional fact, and that is why I consider them all functional explanations.

Whether one should now say that functional explanation is in one case reducible to something like natural selection and in another to a story about human intentions is of no official interest to me, and depends entirely on what is intended by the highly ambiguous term 'reducible', which should not be used without explication. If Fido's being a dog is reducible to his being a collie (or whatever kind of dog he is), then functional explanation is reducible in the stated fashion; and if not, then it is not. Whether something's being a member of a genus reduces to its being whatever species of that genus it is is an issue on which I have no obligation or desire to pronounce.4

4. I say, then, that a functional explanation is one in which a functional fact figures. It may be offered with or without an elaboration, that is, an account of how the functional fact contributes to explaining what it does.

I also say that whenever a functional explanation is true, there is some true elaboration of it (p. 271C), or, in the language preferred by Elster and others, 5 there is some relevant mechanism at work. I then claim (p. 272C) that it is sometimes rational to have confidence in a functional explanation in advance of having a good idea of what the mechanism may be. Elster rejects this epistemic claim (p. 127E), and he charges that it contradicts my insistence elsewhere (pp. 255 ff., 283C) that one may not infer from the fact that x is functional that the existence or nature of x is explained by its function(s). He thinks that in developing this point I 'make out a good case against functional explanation', whereas I was simply indicating widely ignored constraints which functional explanations must meet. With analogous logic one could say of a critic of the fallacy 'post hoc ergo propter hoc' that he had 'made out a good case against' causal explanation.

The asserted inconsistency is between my critique of slipshod functional-explanatory practice and my willingness to hypothesize functional explanations in the absence of knowledge of mechanisms. To show why the inconsistency charge fails, and to display the oversight which led Elster to make it, let us take the social mobility example he introduces on p. 126E. Four claims need to be distinguished here:

(5) The mobility occurring in society s at time t has favourable consequences for class domination.

3 Or, more strictly, a consequence explanation of E, but the distinction between functional explanation and consequence explanation, on which see pp. 263-4C, is of no polemical relevance, and I shall speak of functional explanation throughout this paper.

4 I mischievously report that in Ulysses and the Sirens Elster says that 'there are basically three modes of explanation in science: the causal, the functional and the intentional' (p. viii). This suggests that he holds an irreducibility thesis much stronger than any which attracts me. But this footnote is mischievous, since the quoted statement, though prominently placed, misdescribes Elster's considered position, which emerges on p. 128E, and on which I shall comment in section 4 below.

5 For reasons too complex to display here, I think the familiar 'mechanism' terminology is unfortunate, but it is well entrenched, and I shall sometimes myself use it here.
[end of page 131, start of page 132]

(6) Mobility occurs in society s at time t because it has favourable consequences for class domination. (7) Whenever mobility would have favourable consequences for class domination, mobility occurs. (8) It is in virtue of mechanism m that the favourable consequences of mobility in society s at time t explain that case of mobility.

Elster and I agree that (5) is an insufficient reason for asserting (6), and that a large amount of social theory is therefore founded on a fallacy. But he thinks I fall into the fallacy myself when I license assertion of (6) when nothing like (8) can (yet) be defended. He thinks that one may move from (5) to (6) through (8) alone. But he has ignored an alternative route from (5) to (6), expounded in the section on 'Confirmation' in my Chapter IX, on which he does not comment. The alternative route is via (7). For if it were shown that, quite generally, the 'best brains'(p. 126E) are sucked into the upper class just when that would have a stabilizing effect, and not otherwise, then, I claim, we should be justified in asserting (6) even if we could not (yet) specify a mechanism. We defend a functional explanation of a particular functional item by showing that whenever an item would have that function it appears.6 Such supporting evidence may be hard to find, and it may therefore be hard to confirm the functional explanation. But that does not touch the point of principle, which is that a confirming general pattern might be discernible, while we remain in the dark regarding mechanisms. The general principle, here applied to functional explanation, is that we may be confident that a caused b in a given context because of appropriately parallel cases in other contexts, even if we do not know how a caused b.7 We do not have to know the mechanism to be confident of a functional explanation because we do not have to know the mechanism to be confident of a causal explanation.

What Elster says I 'presumably' mean (p. 126E) by the passage he quotes is therefore a misdescription of my position. I am true to our shared view that merely to identify a function is not to provide a functional explanation. I confess to no methodological error in my account of functional explanation in general, nor, consequently, in my presentation of a functionally construed historical materialism. I regret that I did not motivate the thesis that economic structures are functionally explained as well as I should have liked to,8 but my affirmation of it was not a product of fallacious thinking.

5. I say that functional explanations are sometimes elaborated with reference to human purposes, and sometimes with reference to processes of selection which lack intentionality. Elster complains that representing functional explanation as a genus of which intentional and other elaborations are species 'obscures the vital distinction between short-term and long-term [functional] explanations'(p. 128E). But why should I reject that important distinction, of which Elster rightly makes so much in Ulysses and the Sirens? There is no reason for me not to welcome a treatment of elaboration improved by Elster's work, which clarifies those differences between kinds of mechanism which make them more or less probable candidates in elaboration of functional explanations with different temporal references.

But Elster would say that, in offering this response, I persist in classifying explanations in a way which is not optimal. My contrast was between functional and non-functional causal explanations,9 whereas Elster thinks 'causal vs. intentional explanations is a more fruitful dichotomy'(p. 128E).

6 For the sake of clarity I oversimplify here. For the complexities see pp. 265C, 272-7C.
7 For a non-functional-explanatory example, see p. 286C.
8 Elster is rightly more generous to the other central claim, which explains the superstructure in terms of the base (p. 125E), though I think he exaggerates the extent to which, in fact and in my presentation, the fit of superstructure to base is intentionally wrought.
9 Not, as Elster says (p. 128E), between functional and causal explanations, but I am here ignoring that misrepresentation, which was dealt with in section 2 above.
[end of page 132, start of page 133]

Since causal explanations are explanations by reference to causes, one might reasonably suppose that intentional explanations are explanations by reference to intentions. If so, Elster's classification is inherently questionable, for explanations of actions and other phenomena by reference to intentions arguably are, as Elster might agree, themselves causal, with intentions in the role of causes. I think, however, that Elster means by 'intentional explanations' explanations of the formation of intentions, or explanations in which the formation of intention is a component, and these conceivably are not causal.

Now since I never discuss explanations of the formation of intentions, it is hard to see how our classifications can compete. And Elster's claim that his is superior is further vitiated by his failure to specify the interest relative to which the superiority is asserted. Internally coherent classifications are assessable only relative to the interests they are meant to serve. If the interest is the general metaphysical one in what distinguishes human behaviour, then my classification, which was not directed at that interest, is inapt. My narrower aim was to reconstruct historical materialism. In that exercise I found it fruitful to emphasize the shared functional character of explanations which in other respects may be contrasted with one another. I would not deny that the contrasts are of paramount importance within the more general perspective from which Elster inappropriately criticizes me.

6. My contention that the condition of Marxism may be such that its adherents are justified in putting forward large unelaborated functional-explanatory claims will be more credible if scientists in other domains have, at a certain stage, been in a similar intellectual position. And I think natural historians were in that position before Darwin advanced the subject. He showed how functional facts about the equipment of species contribute to explaining why they have that equipment. Now I claim, and Elster must and does deny, that, before Darwin discovered the chance variation/natural selection mechanism, the belief that species had the useful characters they did because they were useful was already justified. The belief was certainly widely held, 10 by men who had no idea how to elaborate it, and by others, like Lamarck, who had what proved to be an unworkable idea of how to elaborate it. But was pre-Darwinian functional-explanatory belief justified?

Elster's answer to this question will be found on p. 126E. I reply:

(1) There are two pre-Darwinian functional explanatory beliefs to be considered, first, that species have the features they do because they are ecologically adaptive, and second, an entailment of the first, that species have the features they do because the y are in some way adaptive or life-enhancing.

(2) The first, more specific belief, was, as Elster says, refuted by Darwin. He also says it was shown to be unjustified: 'the leap from the analysis of consequences to a consequence explanation was quite arbitrary'. But that is an unpersuasive diagnosis of where the pre-Darwinian error lay. It surely lay not in the inference Elster deplores but in an insufficiently discriminating analysis of consequences to begin with, and so in the premise of that inference. Had perceptive pre-Darwinians distinguished between the ecological and reproductive maximands they might have found evidence, while still in the dark about mechanisms, that the reproductive maximand was the controlling factor. To be sure, it might be unreasonable to expect them to have made the needed distinction ahead of discovery of the mechanism conferring relevance on it. But if it is unreasonable, then the specific false belief was justified: they had all the evidence they

10 I am here going beyond the claim made by Frankfurt and Poole, and by Boorse, and rightly emphasised by them against analyses of the concept of function which draw upon Darwin, that before Darwin it was already thought that species' equipment had functions. I maintain that it was already thought that the equipment was there because of its functions. See H. Frankfurt and B. Poole, 'Functional Explanation in Biology', British Journal for the Philosophy of Science (1966); and C. Boorse, 'Wright on Functions', Philosophical Review (1976).
[end of page 133, start of page 134]

could and can reasonably be expected to have gathered. But I need not rest my case there.

(3) For I would add that the more general pre-Darwinian belief, that the features of species are explained as in some way life-enhancing, was vindicated by Darwin, and I do not see how one could deny that it was justified.

It is a truism that scientific progress often enjoins revisions in hypotheses which make them true under more or less modest restatement, and the advances have no tendency to impugn, retrospectively, the justification with which the original beliefs were held. Hence Elster overplays the distinction between ecological and reproductive adaptation.

I need claim only that Lamarck and others were justified in believing the less specific functional explanations they formulated. But I would in fact go beyond that. I do not think an innocent reader of Lamarck can avoid the conclusion that he knew that the utility of features explained their presence. It is implausible to deny that he knew that, and not implausible to suggest that historical materialism may be in its Lamarckian state.

For centuries men knew the functional facts of natural history, and they rejoiced or despaired at the apparent impossibility of accounting for them in a fully scientific way: hence the prestige of the argument from design for the existence of God. It took time and genius to solve the riddle of natural history. Something similar might be true of society. It is a fallacy to think that because natural history found its deliverance, Marxism will too, and I am amazed at the more than 'somewhat unfair' suggestion (p. 127E) that I am guilty of such an argument by analogy. But the biological case, in historical perspective, counsels greater caution than Elster is disposed to exercise.

As for 'intentional explanations' in physics (p. 126E), I do not accept that, on my account, they would have been justified. To say why not would be to tell a very long story, and here I must content myself with the dogmatic assertion that principles like that of least-time and least-effort are not, despite appearances, amenable to restatement in the canonically functional-explanatory form exhibited at pp. 260-1C.

7. Commenting on my construal of the 1859 Preface (p. 125E), Elster says that for me its crucial thesis is that production relations variously further and fetter productive development, and he agrees that the Preface says so. He also says that I 'add', 'what with some plausibility can also be imputed to Marx, that this furthering or fettering explains the emergence or disappearance of the production relations'.

I find this characterization doubly unsatisfactory. First, I do not say, or think, that the first, nearly truistic thesis, is crucial. Who would deny that some kinds of economic structure are good and some bad for productive growth, at a given stage of productive development? The crucial thesis is the one Elster says I 'add', and my second objection is that to describe it as having 'some [exegetical] plausibility' is a serious understatement. After describing an historical phase in which relations become fetters on forces, Marx says, 'Then begins an epoch of social revolution'. The only possible reading of this is that the revolution begins then because the relations have become fetters. And I supply ample further textual and other argument on pp. 136-50C which proves that Marx held what Elster officially allows may be attributed (only) 'with some plausibility'.

Elster is reluctant to concede that the Marxist theory of society is functional explanatory because he thinks that, if so, it is untenable, and he wants himself to sustain a kind of Marxism. In a forthcoming article on 'Marxism, Functionalism and Game Theory' he criticizes the functional tendency in the Marxism of Marx and others, and he urges that game theory replace functional explanation as Marxism's methodological helpmeet.

But Elster's recommendation is unacceptable. Game theory cannot pretend to do what functional explanation proposes to do for historical materialism, because it has no bearing on its central theses (claims (1) and (2) on p. 129 above).

Elster puts game theory to deft use in a discussion of the dialectics of class struggle

[end of page 134, start of page 135]

for which I have great admiration. But no one would say - and certainly I did not - that how individuals and classes conduct themselves in class struggle may be functionally explained. Functional explanation applies to the long-term outcomes of class struggle, which must conform to the central theses of historical materialism. Historical materialism must say, I argued on p. 149C, that 'the class which rules through a period, or emerges triumphant after epochal conflict, is the class best suited, most able and disposed, to preside over the development of the productive forces at the given time'. Game theory can illuminate the vicissitudes of conflict, and the strategies pursued in it, but it cannot give a Marxist answer to the question why epochal conflict is settled one way rather than another. Marxism's answer to that may be false, but I cannot envisage a non-functional answer to it which is Marxist.

8. Finally, some remarks on the extent to which my comparative ignorance of economics weakened my book (p. 122E).

Certainly I should have paid attention to the literature on economic analogies to natural selection, of which Elster provides a brilliant overview in Ulysses and the Sirens, Ch. III.5.

As for my treatment of the concept of increase in productivity, I am intrigued by Elster's assertion that it is either too formalized, or not formalized enough. Perhaps the assertion is correct, but it would be good to know what its grounds are.

Finally, Elster is certainly right that, contrary to what is said in the last sentence of the book, Chapter V, and also Appendix I, do presuppose the labour theory of value. That last sentence does not, however, reflect 'uncertain grasp of economic theory', but an absentmindedness which is corrected in a forthcoming second impression. Chapter V and Appendix I differ from the rest of the book in being largely just expository of Marx, rather than also defensive. I should have said so, and I should have qualified the last sentence accordingly. I have long thought the labour theory of value false,11 and I was not wishing to commit myself to it in expounding Marx's theory of fetishism, which does, indeed, presuppose it.

My own view is that the doctrine of commodity fetishism is largely false, but that there is deep truth in the idea of the fetishism of capital. Elements for a theory of capital fetishism free of commitment to the labour theory of value will be found on pp. 105 -7C and 122-3C, and in section VIII of the article mentioned in footnote 11 here.

11 Partly for conventional reasons, and partly for the more eccentric ones given in my 'Labour Theory of Value and the Concept of Exploitation', Philosophy and Public Affairs (1979).
[end of page 135]

[Cohen, G. A. (1982), Functional explanation: Reply to Elster, Political Studies 28 (1):129-135]

[ Books | Articles | Reviews | Index | The main menu ]