[Elster, Jon (1990), Merton's Functionalism and the Unintended Consequences of Action, in Jon Clark, Celia Modgil and Sohan Modgil, eds., Robert Merton: Consensus and Controversy, London New York: Falmer Press, pp.129-35]

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

Merton's Functionalism and the Unintended Consequences of Action

Jon Elster

[start of page 129]

Robert Merton's article on 'Manifest and Latent Functions' (1957a:19-84) is an acknowledged classic of modern sociology. I do not think it merits that status. It created as much confusion as it dispelled, or more. It does not seem to have generated much good empirical research, by Merton or by anyone else. In this article I do not try to document the latter claim in anything like a satisfactory fashion. For the 'anyone else' part of that claim, the reader is referred to Campbell (1982), who asserts that the concepts of manifest and latent functions, while usually referred to in sociological textbooks, are little used in actual research. For the 'Merton' part of the claim, the reader is referred to the brief discussion below of Merton's work in the sociology of science.

Merton's early article on 'The Unanticipated Consequences of Purposive Social Action' (1936b) is probably less famous than the study of manifest and latent functions, but has better claims to the status of a classic. Ilere Merton uses what is in my opinion the appropriate language for social scientific analysis-the causal-cum intentional terminology of intended and unintended consequences. It is a little gem of an article, identifying several mechanisms through which the actual effects of behaviour may deviate from the intended ones. Although he does not discuss what I believe to be the most important mechanism, viz. the tendency for people to act on wrong assumptions about what other people will do (Elster, 1978: Ch. 5), the article is perfectly lucid and free of ambiguity. In this respect it contrasts favourably with the later, more famous article, which goes beyond intentions and causes to introduce functions as a separate category.

'Manifest and Latent Functions' does make some important contributions. It contains an impressive demolition of earlier forms of functionalism, based on the postulate of the functional unity of society, the postulate of universal functionalism and the postulate of indispensability. Also he makes it very clear that he does not rest his case for functionalism in sociology on dubious analogies from biology. Finally,

[end of page 129, start of page 130]

'Manifest and Latent Functions' remains worth reading for its extensive discussion of American machine politics, which almost amounts to an article within the article. Although I don't think Merton succeeds in imposing a consistent and plausible structure on this problem, the sheer richness and suggestiveness of his analysis invite further reflection.

On the critical side, the main weakness of the article is that the task of functional analysis is never made fully clear. In particular, Merton never slates in so many words whether the task of functional analysis is to explain social phenomena or, more modestly, to identify and describe phenomena that might otherwise be overlooked. Clearly, manifest functions-intended and recognized consequences of action-have explanatory power. The question on which most of the following discussion is focused is whether latent functions-the unintended and unrecognized effects of action-can explain the actions of which they are the consequences.

Commenting on Merton's discussion of the Hawthorne experiment, G. A. Cohen writes: 'He identifies a function in the Hawthorne experiment, but fails to note it is not a function which explains why the experiment took place' (1978: 283). Cohen then adds in a footnote: 'It might be claimed in Merton's defence that he was concerned only with identifying functions of social patterns and institutions, not with functionally explaining them. This is a highly implausible reading, and if it is correct, then we may object that in an article recommending the study of functions Merton neglected their explanatory significance'.

I shall argue that the non-explanatory reading, while not 'highly' implausible, is probably less natural than the explanatory one. We may observe, nevertheless, that in a recent article devoted to Merton's distinction between manifest and latent functions, Cohn Campbell (1982: 41-2) does not cite explanation among Merton's justifications for the paradigm. Instead he cites:

1 the 'desire to eliminate what [Merton] saw as a widespread confusion between the subjective categories of disposition and the objective ones of consequence';

2 the claim 'that the distinction would aid both "systematic observation and later analysis" by directing observation towards "salient elements of a situation" and preventing the "inadvertent oversight of these elements"'; and

3 the idea 'that the distinction "directs attention to theoretically fruitful fields of inquiry", by which he means that it is not enough for spciologists to "confine themselves to the study of manifest functions"'.

Referring to an obscure footnote in Merton (1957a: 71n90), Crothers (1987a: 71) similarly argues that functional analysis must not be confused with explanation, the latter being causal rather than functional.

In his most general statement of the 'central orientation of functionalism' Merton refers to 'the practice of interpreting data by establishing their consequences for larger structures in which they are implicated' (1957a: 46-7). It is hard to see what interpreting could mean other than explaining. Further on he argues that the latent functions of conspicuous consumption are 'gratifying and go far toward explaining the continuance of the pattern' (ibid.: 58, emphasis supplied); or, still further on, that they 'help explain the persistence and the social location of the pattern of conspicuous consumption' (ibid.: 69; emphasis supplied). He observes that labelling the Hopi rain dance a superstitious practice of primitive folk 'in no

[end of page 130, start of page 131]

sense accounts for the group behaviour' (ibid.:64; emphasis supplied), and adds that 'given the concept of latent function, however, we are reminded that this behavior may perform a function for the group'. By implication this latent function does 'account for' the practice in question.

Other statements are less explicitly explanatory, but point in the same direction. Thus he refers to 'consequences [of a structure] which may provide basic social support for the structure' (ibid.: 72n91). With respect to the political machine he claims that 'the functional deficiencies of the official structure generate an alternative (unofficial) structure to fulfill existing needs somewhat more effectively' (ibid.: 73). The first statement suggests that consequences are important in maintaining (and thus explaining) institutions, the second that they are instrumental in creating (and thus explaining) them. At a more general level he makes two provisional assumptions: that 'persisting cultural forms have a net balance of functional consequences either for the society considered as a unit or for subgroups sufficiently powerful to retain these forms intact' (ibid.: 32), and that 'when the net balance of the aggregate of consequences of an existing social structure is clearly dysfunctional, there develops a strong and insistent pressure for change' (ibid.: 40). These statements amount to saying that because dysfunctional elements tend to disappear, persisting forms can be assumed to be functional and to persist because they are functional. I return to this argument. Here it is cited only as evidence for Merton's explanatory intentions.

We cannot conclude, however, that Merton intended all his analyses to have explanatory import. His comments on the Hawthorne experiment, for example, do not lend themselves to this reading; nor would it be a substantively plausible account (Cohen, 1978: 257-8). Here functional analysis - like much of what goes for 'dialectics'-seems to be little more than the discovery of surprising, paradoxical and counterintuitive consequences or, perhaps, the identification of valuable consequences (see below). His general statements about 'drawing attention to' the unintended and unrecognized effects of social phenomena also point in this direction. It is inherently satisfactory for the mind to discover that good sometimes leads to evil and evil to good; that less can be more and more can be less; or that individual rationality can bring about collective irrationality and vice versa. And it is easy to mistake the thrill of discovery for the thrill of explanation, which is what seems to have happened to Merton between the 1936 essay on unanticipated consequences and the 1949 article on manifest and latent functions. This comment, of course, is speculative and ad hominem. No part of my case rests on it.

Before I say more about the issue of explanation, I want to discuss some other ambiguities in Merton's paradigm of functional analysis. The classification of consequences in intended versus unintended and recognized versus unrecognized suggests four categories rather than merely two. Also, we have to ask, as Merton rarely does: intended by whom? recognized by whom? These questions, in turn, are related to another: functional for whom?

In any given item of functional analysis, there are two groups of individuals involved: those who engage in practices we want to explain and those who benefit from these practices, i.e. those for whom they are in some sense functional.1 The question of intention arises only for the first group, the question of recognition may arise for both. The two groups may coincide, overlap or be totally disjoint. In the last case, the possibility arises that the effects are unintended by those who produce them but recognized by those who benefit from them. If, moreover, the latter have

[end of page 131, start of page 132]

the power to sustain the activities from which they benefit, we have the ingredients of a filter explanation (Nozick, 1974: 21-2; Elster, 1984: 30). Although he has no place for this kind of explanation in his general dichotomous scheme, Merton tacitly admits it when he refers to activities that have a net balance of functional consequences for subgroups sufficiently powerful to retain these forms intact (ibid.: 32).2 It is hard to see how power to maintain the beneficial activities would be relevant unless the benefits were also recognized by the beneficiaries.

The other category missing in Merton's paradigm is that of benefits which are intended but unrecognized. This case arises mainly when the benefits are unrecognized by the beneficiaries, as in paternalism. It is clearly of less theoretical interest than those which involve behaviour maintained by their unintended consequences. Merton would perhaps count it as a manifest function, since the effects will usually be recognized by those who intentionally produce them. This does not affect my main point, however, which is that for some explanatory purposes we need to distinguish between actors and beneficiaries. Filter explanations, in particular, rest on this distinction.

When Merton says that dysfunctions create a 'strong and insistent pressure for change', he must imply that these are somehow recognized. On the other hand, he does assert the existence of 'latent dysfunctions' (ibid.: 51n) which, presumably, are not recognized. I do not see how these statements can be reconciled. To bring out the point, consider two definitions of the Marxian concept of alienation (Elster, 1985: 74 ff.), as either sense of a lack of meaning or lack of a sense of meaning. One can see how the former 'dysfunction' could set up a pressure for change, provided that the victims not only perceive the problem but correctly identify its cause. The latter, however, cannot act as a causal agent in social change. For another example, consider the belief in witches. Although these tend to be dysfunctional by inducing conformity and stifling innovation (Thomas', 1973: 643-4), they provide an element of stability rather than of change. And if there can be stable bad situations, there can also be unstable good ones, like cooperation in a Prisoner's Dilemma.

Moreover, when the outcome but not the cause of a dysfunction is perceived, as is often the case if the causal chains are protracted or complex, we should not expect any pressure for change; or, if there is pressure for change, we should not in general expect it to be successful in doing away with the problem. Environmental and economic crises often defy understanding: they generate change, but no solutions. Hence we see that in addition to the distinction between recognition by actors and recognition by beneficiaries, we need to distinguish between recognition of the consequences themselves and recognition of the causal responsibility for these consequences.

Merton rightly emphasizes the need to identify the alternatives which are excluded by the activity which is subject to functional analysis or explanation (1957a: 57). But he does not draw the full implications from this insight. For one thing, he does not make it clear whether an activity, to be functional, must also be optimal within the set of alternative arrangements. For another, he does not explain what determines the realization of one among several equi-functional (or perhaps equi-optimal) arrangements. Merton suggests, for instance, that the rain ceremonial of the Hopi can be 'accounted for' in terms of its contribution to group unity. Now one might think, first, that an optimal common activity would be one that both contributed to group unity and achieved its manifest purpose, thus avoiding waste and disappointment. If he believes that the lack of success of the

[end of page 132, start of page 133]

manifest purpose was essential to fulfilment of the latent function, he should say so and explain why; if not, he should explain the absence of another arrangement with a larger 'net balance' of benefits. Second, one might ask why the rain ceremonial, rather than any of the innumerable other joint activities one could imagine, was realized as a 'means' (ibid.) to group unity.

A final set of ambiguities has been so well charted by Cohn Campbell that I need only cite from his discussion:

Altogether there are at least four different meanings which Merton gives to the manifest-latent distinction. There is, first of all, that presented in the explicit formulation, i.e. the contrast between conscious intention and actual consequence. Secondly... Merton himself comes to use the dichotomy to refer to the commonsense knowledge (or sometimes the perspective of another discipline) and sociological understanding. Thirdly, there is the usage which equates manifest with the formal and official aims of organizations and latent with the purposes fulfilled by unofficial or illegal ones. Finally, there is the suggestion that manifest and latent relate to different levels of bnderstanding, with the former equal to apparent or surface meaning while the latter concerns the deeper or underlying reality of the phenomenon in question. (1982: 33)
The first distinction is used in Merton's discussion of the Hopi; the second in his discussion of conspicuous consumption; the third in the discussion of the political machine; while the last is present throughout. Combined with the other ambiguities and inconsistencies that ~ have been discussing, this confusion creates something close to conceptual chaos.

Let me now confront the problem of explanation more directly. The central task for any explanation-by-consequences is to provide a mechanism by which the consequences uphold or maintain the behaviour one wants to explain. In the absence of some kind of feedback from effect to cause, explanation by consequence remains a totally mysterious notion.3 It is metaphysically impossible for an event to be explained by another event that occurs at a later time. Merton's only reference to mechanisms is his suggestion that dysfunctional entities tend to disappear. Else-where (Elster, 1983: 61ff.) I have discussed a similar argument (proposed by Stinchcombe, 1975) at some length, concluding that it does not work. It amounts essentially to asserting that non-dysfunctional institutions, including as a special case the functional ones, maintain themselves over time, because they do not set up any pressure for change. The argument is vulnerable in at least three respects.

1 The basic premise, that non-dysfunctional institutions remain stable over time, is false. Adaptively neutral phenomena are often subject to drift.

2 Even if dysfunctional institutions tend to disappear, there is no presumption that they will be replaced by functional ones. Perhaps disequilibrium is a more fundamental feature of societies than equilibrium (Nelson and Winter, 1982).

3 There may be many non-dysfunctional alternatives to a given dysfunctional institution. We would like to know which of them will emerge and at what time, not simply that one of them ultimately will. Are the functional ones more likely to emerge than those which are merely neutral from the adaptive point of view? And if so, why?

Merton has nothing to say about the two most frequently cited mechanisms by which behaviours or institutions could be upheld by their consequences: natural

[end of page 133, start of page 134]

selection and reinforcement (see van Parijs, 1981, for a thorough discussion). Evolutionary biology and animal psychology have charted these mechanisms in the utmost detail, but it remains to be shown that either can be used to support functional explanation in the social sciences. In any case, for the present exegetical purposes there is no need to discuss them at further length, since they seem very far in spirit from Merton's paradigm. In particular, both mechanisms rest on the notion that alternatives to an unsatisfactory state are generated randomly, whereas Merton argues more obscurely that they are generated by a 'pressure for change'.

I conclude by a brief comment on the language of functions in Merton's sociology of science. I shall first quote two passages at some length, and then suggest that they may give us a key to Merton's main interest in functions:

Multiple discoveries can thus be seen to have several and varied functions for the system of science. They heighten the likelihood that the discovery will be promptly incorporated in current scientific knowledge and will so facilitate the further advancement of knowledge. They confirm the truth of the discovery (although on occasion errors have been independently arrived at). They help us detect a problem which I have barely and far from rigorously formulated, to say nothing of having solved it: How to calculate the functionally optimum amount of redundancy in independent efforts to solve scientific problems of designated kinds, such that the probability of the solution is approximately maximized without entailing so much replication of effort that the last increment will not appreciably increase that probability. (Merton, 1973: 80)

[C]onsidered in its implications for the reward system, the Matthew effect [the award of disproportionate credit to senior authors for joint work or multiple discoveries] is dysfunctional for the careers of individual scientists who are penalized in the early stages of their development, but considered in its implications for the communication system, the Matthew effect, in cases of collaboration or multiple discoveries, may operate to heighten the visibility of new scientific communications. This is not the first instance of a social pattern's being functional for certain aspects of a social system and dysfunctional for certain individuals within that system. (ibid.: 447-8)

It is hard to believe that Merton here is using 'function' in an explanatory sense. The idea that redundancy or the disproportionate attention to seniority are explained by the benefits these phenomena confer on the growth of knowledge is too implausible to entertain. Merton is simply calling attention to certain consequences of behavioural patterns in science. The question is, why would he want to do this? One answer, mentioned earlier, could be that the consequences are surprising, counterintuitive or paradoxical. In the present case this does not seem to be the case. It is pretty obvious that the chances of a problem being solved increase with the number of scholars working on it, but that each new scholar creates a smaller increment of success (at least beyond a certain number). The Matthew effect owes its fame, I believe, more to the lucky choice of a phrase than to any surprising insights it has yielded.

Another, more plausible answer is that the identification of consequences can be important for policy reasons. Research councils need to know about the optimal degree of redundancy of funding. Journal editors might want to resist claims for more democracy in science, e.g. for the alphabetical listing of authors. In Merton's discussion of the political machine the policy aspects also form an important theme, and serve as basis for a warning to politicians against replacing the machine with another system that does not deliver the same goods.

What, then, are we to conclude? Does Merton simply restate Mandeville's insight, that 'those who can enlarge their view, and will give themselves the leisure of

[end of page 134, start of page 135]

gazing at the prospect of concatenated events, may, in a hundred places, see Good spring up and pullulate from Evil, as naturally as chickens do from eggs'? Or does he claim, more ambitiously, that the evil we observe in the cause can be explained by the good we see in the effect? I have to report that after many readings and re-readings I still do not know. Perhaps Merton has been playing the oldest game of all, that of suggesting the strong claim without actually sticking his neck out and embracing it explicitly? More charitably and more plausibly, does his laudable love for rich empirical detail and for fine conceptual distinctions make it difficult for him to stand back from his material and perceive the larger structure of the argument? The force of the 1936 essay on unanticipated consequences, written before his mind was uncluttered with the formidable array of facts that he has since mastered and assimilated, might support this answer.


1 I am assuming, for simplicity, that all benefits are ultimately benefits to people, not to some abstract entity like 'society' or 'science' This may be a dubious assumption in some cases. In a society divided on ethnic, religious or similar grounds, an institution may benefit 'society' by ensuring a minimal level of integration even if all individuals would be better off were it to break up into two or more autonomous units. Also 'science' as a system of knowledge may well benefit from practices that provide few tangible benefits to identifiable individuals.

2 Elsewhere, however, Merton (1957a: 65) opposes the analysis of latent functions to analysis in terms of 'manipulation by powerful subgroups'.

3 G. A. Cohen (1978, 1982) disagrees with both statements. First, he argues, an explanation by consequences can be backed by a consequence law rather than by a mechanism. Second, the underlying mechanism (which must be assumed to exist but need not be exhibited for the explanation to be valid) can take the form of causally efficacious dispositions rather than of feedback loops. For the present purposes. these disagreements are of secondary importance.

[end of page 135, start of page 136]



I am always surprised by the ease with which the title 'classic' is granted - or refused - to social scientists by social scientists. While it may take centuries for a musical work to become a classic, as in the case of Charpentier's Medée, for a sociological paper it often takes only one or two years. The two texts of Robert Merton discussed by Professor Elster are somewhat older, but I have no opinion as to whether they have 'the status of a classic'. 1 know only that they are important, and I have the feeling that Elster is unfair to the text on 'Manifest and Latent Functions'.

Although I cannot go into detail here, it seems to me that Merton has shown in this text - and this is not a negligible achievement - that the concept of function can have a positive and fruitful use, but also a negative and dangerous one. While he does not use these exact words, he has made clear in the paper that the concept of function is acceptable when it is used as a kind of shorthand notation summarizing social processes which can also, though more awkwardly, be described without using the concept of function. To put this idea into the form of a methodological theorem: the notion of function is useful only if it is not indispensable. Thus, when Merton showed how, at the time he conducted his analysis, the Democratic Party machine met needs among the lower classes which were not satisfied by a public social security system of the type developed, say, in Germany or France, and how this also served the interests of the Party, then he gave an illuminating example in which the word 'function' is used in a purely nominal way. This example and many others suggest that function is an acceptable and useful concept when it is entirely dissociated from any functionalist view of society. This implies a sharp distinction-again using non-Mertonian vocabulary-between what might be called methodological functionalism and substantive functionalism. As I have tried to show elsewhere (Boudon, 1980: 195-202), Merton was probably the first sociologist to suggest clearly that the notion of function was useful and perfectly acceptable in the first context, and dangerous and unacceptable in the second.

As to Merton's notion of latent function, one may not like the concept, but it points, if not to the major function, at least to one of the major duties of the social sciences. Manifest functions are visible and do not need the social sciences to be detected. Latent functions are not only invisible but sometimes half-consciously hidden. Social actors have good reasons not to recognize their existence. Thus, because producers of collective goods, such as political parties or trade unions, are exposed to the free-riding of their potential members, they have to create or take

[end of page 136, start of page 137]

advantage of a variety of mechanisms in order to fulfil their manifest function more effectively. For example, to increase their organizational strength, trade unions may apply a range of indirect and direct pressures on non-members in order to force them to join. But these mechanisms will have to remain latent for obvious social reasons. No trade union wishes to admit or let it be widely known that it is forcing people to join its organization. A study such as Olson's Logic of Collective Action (1965), among others, confirms the usefulness of the notion of latent function.


[Ed. note: Elster refrained from writing a rejoinder as he felt there was nothing in Boudon's contribution with which he disagreed.]

[end of page 137]

[Elster, Jon (1990), Merton's Functionalism and the Unintended Consequences of Action, in Jon Clark, Celia Modgil and Sohan Modgil, eds., Robert Merton: Consensus and Controversy, London New York: Falmer Press, pp.129-35]

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