Contra Penner, What’s Really Wrong with Functional Explanations

Contra Penner, What’s Really Wrong with Functional Explanations Abstract Hans Penner offered a prominent line of criticism of social functionalism in the study of religion. Penner took exception to social-functional claims when these are presented as explanations. Following Carl Hempel, he argued that such explanations very often commit the logical fallacy of affirming the consequent. I argue that Penner was wrong about social functionalism. I draw on the work of some contemporaries of Penner who rejected Hempel’s model of scientific explanation and defended the propriety of social-functional explanation, particularly in connection with Marxian theory, and argue that rather than being functionalism’s Achilles’ heel, explanation is its foundation. I then apply my position to a recent book by Craig Martin that advocates for social functionalism in the study of religion. I conclude that, because it marginalizes explanation, the book offers little in the way of reasons to think that religious phenomena serve social functions. No one has any quarrel with a functional explanation where the mechanism is actually shown to be at work. The hard question is whether one can ever be justified in setting forward a functional explanation even in the absence of a specific mechanism. (Elster 1980, 126) INTRODUCTION I NEVER MET Hans Penner, although our academic careers overlapped for a few short years.1 Had we chanced to meet I might have thanked him for introducing me to the murky but fascinating topic of social functionalism in religious studies. That introduction came by way of his essay “What’s Wrong with Functional Explanations?” which I encountered at about the midpoint of my graduate training (Penner 1999). It was some years after Penner passed away that I arrived at the conviction that he was wrong about social functionalism. There are possible worlds in which he lives on long enough that I have a chance to change his mind about the subject. For reasons I will mention later, I estimate that I am successful in few, if any, of those worlds. In this article, I want to lay out the case for thinking that Penner’s formulation of the issues around social functionalism is deficient and offer what I think is a superior formulation. The position I will offer as a replacement for his is not my invention: to begin staking out the territory, I will basically be positioning the analytical Marxists of the late twentieth century as a corrective to Penner’s source, Carl Hempel. What results will support the idea that there is something wrong with functional explanations only in a modest way. I do not think that there is something so wrong with functional explanations that it is always a mistake to advance them or take them to be true. But neither do I think that social functionalism is clearly a promising research methodology for religious studies. I will begin by briefly rehearsing what social functionalism is in general terms, with reference to the role this doctrine has played in the history of religious studies. I will then exegete Hempel and Penner on the topic. I will then turn to the analytical Marxists, who were contemporaries of Penner’s but whose work I think he simply failed to appreciate. Discussions among three of these figures—G. A. Cohen, John Elster, and Richard Miller—expose the problem on which I eventually want to focus. I will conclude by describing that problem in my own terms, with reference to recent literature in which social functionalism appears with some prominence—literature that suggests to me that we scholars of religion have yet to make our peace with this part of our theoretical history. CLASSICAL SOCIAL FUNCTIONALISM In its classical form, social functionalism about religion is the position that the question of why religious phenomena are present within particular social groups can be answered by way of a description of their social effects. As a formally described “live option” in the academic study of religion, it originated around the turn of the twentieth century, flourished in the mid-century literature, was subject to close examination and strenuous criticism during these same decades, and declined notably after the 1960s. Social functionalism was given canonical expression in two texts from Emile Durkheim, Rules of Sociological Method (1895) and The Elementary Forms of Religious Life (1912). Here is a key passage from the latter text: There can be no society that does not experience the need at regular intervals to maintain and strengthen the collective feelings and ideas that provide its coherence and its distinct individuality. This moral remaking can be achieved only through meetings, assemblies, and congregations in which the individuals… reaffirm in common their common sentiments. Such is the origin of ceremonies that, by their object, by their results, and by the techniques used, are not different in kind from ceremonies that are specifically religious. (Durkheim [1912] 1995, 429; quoted in Godlove 2005, 111ff.) This kind of thinking is grounded on an analogy between social systems and biological organisms, an analogy freely admitted by such early defenders of functionalism as Alfred Radcliffe-Brown (Radcliffe-Brown 1935, 178; see also Merton 1957, 21).2 Just as the continued existence of an organism requires that certain effects be accomplished—that oxygen, say, be delivered to the organism’s cells—so too, the continued existence of a social system requires that certain effects be accomplished. And just as the body’s need for oxygen delivery is satisfied by the proper operation of certain organs within the body, the needs of a social entity are satisfied through the performance of certain kinds of regularized, coordinated activity on the part of its members. And thus, we can speak of the function of a religious ritual, say, in much the same way as we can speak of the function of the heart. So, for example, Radcliffe-Brown: Rites can be seen to be the regulated symbolic expression of certain sentiments. Rites can therefore be shown to have a specific social function when, and to the extent that, they have for their effect to regulate, maintain and transmit from one generation to another sentiments on which the constitution of the society depends. (Radcliffe-Brown 1945, 35) Classical social functionalism was committed to the idea that social functions have explanatory value, that certain social phenomena, in other words, can be explained by their functions. This, I think, is how to parse Durkheim’s reference to the origin of ceremonies above; and elsewhere in The Elementary Forms he states, after arguing that the very existence of a clan requires some way for its members to think of themselves as a unity and that clan emblems make such thinking possible, that “both the institution of the emblem and its place in the group’s life are thus explained” (Durkheim 1995, 235). From the beginning of its rise to prominence in the study of religion, social functionalism was accompanied by a body of critical reflection. Even in the classical texts of the early twentieth century, advocates of social functionalism acknowledged that the responsible deployment of the paradigm required attention to a number of distinct issues. I want to mention three of these. First, almost from the very beginning, social functionalism was premised on the possibility of avoiding appeals to teleology. Classical functionalists were interested primarily in those social effects of, for example, religious rituals that were not the effects that participants intended their actions to have.3 And early defenders of social functionalism recognized that teleological language could not be used in the description of purported social-functional arrangements without risk. The risk at issue was that of tacitly supposing that social groups are something like collective agents, capable of organizing their internal workings for preconceived purposes (much later, Jonathan Friedman would waggishly characterize Roy Rappaport’s functionalism as “Hegelian ecology” (Friedman 1979). Durkheim’s language in the first quote above simply transposes folk psychology to the level of the social group by suggesting that rituals take place because societies are aware of their own needs, aware that the regular occurrence of rituals will satisfy those needs, and act to institute the rituals so that their needs will be satisfied. And Durkheim did not provide a translation of this teleological language into terms innocent of collective agency. But as social functionalism developed as a method of analysis, the need for such translation became evident, so that by 1935 Radcliffe-Brown recommended the excision of teleology from social-functional explanations in part through the substitution of “necessary conditions of existence” for Durkheim’s “needs” (Radliffe-Brown 1935, 394).4 The project of social functionalism even in its classical form, in other words, was to make it possible for social phenomena to be explained by their consequences without appeal to teleological reasoning.5 Second, classical social functionalism was not uniformly committed to the position that the function of a social phenomenon explains that phenomenon’s origin. Although Durkheim made precisely such a claim in The Elementary Forms, this was simple carelessness on his part. Earlier, in The Rules of Sociological Method, he had distinguished claims about the causes of a social phenomenon from claims about its function, arguing that the “complete explanation” of social phenomena required both; but in The Elementary Forms, as Terry Godlove has noted, Durkheim simply interwove causal and functional strands of explanation (Godlove 2005, 166 n33; see also Giddens 1971, 90ff.). If not the origin of religious phenomena, then, what do functional claims explain? The question is not always easily answered, but one fairly accessible response is that functional claims explain the continued, regularized presence of religious phenomena within a particular social group; in other words, whatever the origin of these phenomena, they persist because of the function they serve.6 This sort of thinking is also behind the distinction, expressed by Radcliffe-Brown and later explored by Carl Hempel, between “historic-genetic” and social-functional explanations; it is also the grounds for the common accusation that social-functionalist treatments of cultural phenomena give short shrift to historical considerations (see Hempel 1965, 309; Turner and Maryanski 1979, 109–13). Third, at least some classical functionalists acknowledged the contestable nature of the basic premise of social-functional analysis. I think Radcliffe-Brown was pointing to the tip of a theoretically important iceberg when he acknowledged that the claim that social groups are organized into functionally unified systems is not an observation but a hypothesis.7 He acknowledged further that this hypothesis was controversial and that some of his contemporaries—he cited Robert Lowie in particular—rejected it, on the grounds that it imputed too high a degree of order to the products of collective human activity (Radcliffe-Brown 1935, 401). And in his lengthy midcentury analysis of social functionalism, Robert Merton offered the following remark concerning this “postulate of the functional unity of society”: This conception, in its recent formulations, was developed by social anthropologists, that is, by men primarily concerned with the study of non-literate societies. In view of what [Paul] Radin has described as “the highly integrated nature of the majority of aboriginal civilizations”, this assumption may be tolerably suitable for some, if not all, non-literate societies. But one pays an excessive intellectual penalty for moving this possibly useful assumption from the realm of small non-literate societies to the realm of large, complex and highly differentiated societies. (Merton 1957, 28, italics original) Merton went on to call particular attention to the pernicious effects of the assumption of functional unity for the analysis of religion: specifically, that it directs attention away from nonfunctional, socially disruptive effects of religion and from “internecine conflicts among religious groups,” for “all this abundantly known material is ignored in favor of illustrations drawn from the study of religion in non-literate society” (Merton 1957, 29). But for my purposes, a more general claim by Merton is more important: that however useful the idea of social-functional integration may be as a hypothesis guiding the ethnographic study of particular social arrangements, the idea that contemporary societies form functionally integrated wholes is a hypothesis to be confirmed through examination of specific cases rather than a secure deliverance of anthropological research.8 I will return to this point below; for the moment, I want to note that within classical functionalism itself is to be found awareness of the dependence of functionalism on the supposition that there are such things as social systems, and awareness that this supposition is not obviously true (on this point, see also McCauley and Lawson 1984, 378–81). To my mind, these three issues add up to something. It seems to me that classical social functionalism was always vulnerable to the worry that to think that large-scale social phenomena can be explained by their consequences is simply to extend teleological thinking, and perhaps agential thinking, beyond its proper domain, or to use the words of Kinglsey Davis, that functionalism is a “crank method” of social analysis.9 PENNER’S CRITICISM Critical discussions of functionalism in the middle of the twentieth century by Richard Braithwaite, Robert Merton, and others formed the background of Carl Hempel’s 1959 essay, “The Logic of Functional Analysis” (Hempel 1965). In this essay, Hempel charged functional explanations with the fallacy of affirming the consequent, or arguing If A, then B; B, therefore A. Hempel famously advocated the deductive-nomological or “covering law” model of (nonstatistical) scientific explanation, and his charge against functionalism is a product of his attempt to make sense of functional explanations on this model. On the covering-law model, an explanation consists of two components: first, a description of the conditions under which the phenomenon to be explained occurs (“whenever A is the case, B occurs”), and second, a statement that entails that these conditions are satisfied (“A is the case”). The existence of the phenomenon follows deductively from the conjunction of these statements. In a case of successful explanation, “the outcome described in the explanandum was to be expected in view of the antecedent circumstances and the general laws listed in the explanans. More precisely, the explanation may be construed as an argument in which the explanandum is deduced from the explanans” (Hempel 1965, 299). Hempel’s observation was that functional explanations almost never work on the deductive-nomological model. Functional explanations involve two empirical observations: that a specific set of rituals, say, are performed within a society, and that an effect of those rituals—social stability, say—obtains. Reconstructed on the “D-N” model, a functional explanation would be a set of claims from the conjunction of which the occurrence of the ritual could be deductively inferred. The crucial question thus is: what would need to be conjoined to the statement that social stability has been achieved in order for the occurrence of the ritual to follow deductively? Trivially, only statements that entail that the only way social stability could be achieved is through the performance of the ritual, so that whenever social stability obtains, the ritual has been performed. In formal terms, the argument if A then B; B; therefore A is invalid; this argument commits the fallacy of affirming the consequent. But the argument B if and only if A; B; therefore A is valid, its validity purchased by way of the specification that A must be the case in order for B to be the case. Hempel noted that functional explanations do not typically claim that the phenomena they are concerned to explain are the only possible mechanism for the production of the effects they produce; and yet, as he understood them, functional explanations are valid only when this is the case (Hempel 1965, 313).10 Functional explanations thus have a logical problem. If you are familiar with this charge, it is likely due to Penner’s influence. Penner repeated Hempel’s claim that the problem of functional explanations is a logical problem in publications that spanned his career, beginning with the 1971 essay “The Poverty of Functionalism” in History of Religions.11 The latest mention that I have been able to identify is in Penner’s last book, Rediscovering the Buddha (2009), where he briefly described functionalism as a theory that had long ago been “thoroughly dismantled as logically invalid” (Penner 2009, 207). It seems that Penner carried the torch of Hempel’s criticism to the end of his days, in a determined effort to bring illumination to us all. I take Penner’s discussion of 1989, excerpted and republished a decade later, as my source for his mature position on functional explanation. Penner’s main project in the essay was to apply Hempel’s critique of functionalism to the literature of religious studies. But in historical retrospect, the essay is also noteworthy for Penner’s treatment of the analytical Marxists. In the main, these contemporaries of Penner regarded the D-N model of scientific explanation as obsolete and did not make use of it in their work.12 But Penner himself remained committed to Hempel, and effectively retrojected elements from their discussions of functionalism back into a Hempelian framework. Penner’s application of Hempel is clear enough: an explanation of ritual that appeals to its consequences—social “maintenance”—“is invalid because it commits the fallacy of affirming the consequent. It asserts that ‘if y then z; z therefore y’” (Penner 1999, 256). What it would take to avoid this fallacy is the assertion that social maintenance requires the performance of the ritual—that the performance of the ritual is not just a sufficient, but also a necessary, condition for social maintenance. And Penner noted, correctly, that serious problems beset the project of arguing that religion, or some specific components of religion, are necessary conditions for satisfying some requirement of society. “No one has succeeded in showing that this is the case,” he remarked; “I do not know how it could be shown to be the case without circular argument” (Penner 1999, 262). A good bit of his exposition of the historical literature revolves around what he regarded as failed attempts to solve this logical problem by various means, including invoking “functional equivalents or a set of social units taken jointly as a class of sufficient conditions” (Penner 1999, 262ff.). Penner’s brief engagement with the analytical Marxists takes place in the last few pages of his essay, embedded within his treatment of the idea that societies are “self-regulating systems.” (Penner 1999, 266). That idea, which Penner discusses as it appears in work by Ernest Nagel and Robert Brown, I have already mentioned: societies are “self-regulating systems” just in case they are organized in ways that make possible causal connections between the actions of persons and the conditions of human actions, such that the consequences of actions themselves shape future actions of the same type in ways that fall into persistent patterns. Penner appropriates from Brown the criticism that this sort of approach needs to specify “what property is being maintained in a steady state,” “what are the internal variables and can they or their effects be measured,” and “what are the external conditions which are assumed to be constant” (Penner 1999, 266ff.). What matters to Penner about this subject is that failure to answer these questions amounts to failure to answer the Hempelian criticism: In a self-regulating model, if the unit is a necessary condition for maintaining the system, we will then have to specify the exact relations which it maintains; since it would be most difficult to uphold the notion that a religious belief system, for example, functions to maintain all the relations in a social system. On the other hand, if we state that the unit is a sufficient condition, we then fail to explain exactly how the religious unit functions to satisfy the requirement. (Penner 1999, 267) It was at this point that Penner turned to Jon Elster, positioning one specific component of their discussions as confirmation of his conclusion regarding self-regulating systems. Elster had argued, in Ulysses and the Sirens, that functional explanations assume a “causal feedback loop” such as figures in the operation of self-regulating systems and that some species of functional explanations “fallaciously infer” the existence of such loops from the available evidence. The following year, Elster wrote a review of the book that inaugurated the analytical Marxist movement, G. A. Cohen’s Karl Marx’s Theory of History: A Defence. Penner cited the following sentences from this review: “I believe that I have seen no other mechanism that comes close to being for sociology what natural selection is for biology, even if this is not, to repeat, to say that it comes very close. Cohen, however, does not even attempt to provide such a mechanism, which is why I believe that his enterprise must be judged a failure” (Elster 1980, 127; cited in Penner 1999, 268ff.). Without offering his own gloss on Elster’s words, Penner noted his agreement, bringing the essay to a close shortly thereafter. I have chosen a different passage from Elster’s review as the epigraph for this article, because that passage is key to my argument that Penner’s problem with functional explanations was different from the problem with which Elster was concerned. But before I offer an account of the context for Elster’s remarks, I want to say how I think Penner understood the passage he cited. Recall that on the deductive-nomological model, explanations are supposed to make it possible for the occurrence of the explanandum to be deductively inferred. It is this requirement that generates Penner’s problem: only the claim that a ritual is a necessary condition for social maintenance would allow the occurrence of the ritual to be deductively inferred from the fact that social maintenance exists. I surmise from his discussion of Elster that Penner understood the role of “causal feedback loops” and “mechanisms” in a specific way: he regarded these as elements of any respectable case for the necessity of religious phenomena for social effects. Thus, the problem with functional explanations from which descriptions of such “mechanisms” are lacking is that they are formally invalid. As we have seen, absent an argument for necessity, it simply does not follow deductively from the existence of the social effect that the religious phenomenon has occurred. But such are not the terms in which Elster understood the problem. ANALYTICAL MARXISTS ON SOCIAL FUNCTIONALISM In this section I want to lay out what I think Penner missed in the discussions of his contemporaries. This will require retrieving some crucial terminology from Cohen—terminology that never, to my knowledge, appears in Penner’s work. A clearer sense of Cohen’s position will make it possible for me to say what Elster’s criticism of Marxian social functionalism amounted to. But after this, I will consider another figure from this period, Richard Miller, who offered what I regard as an effective rejoinder to Elster. It is Miller’s understanding of the real issue with functional explanation that I will take into the final section of this article. The project of Karl Marx’s Theory of History was to construct an intellectually respectable form of Marxist social theory using the tools of analytical philosophy (Cohen originally termed what he was doing “non-bullshit Marxism,” but cooler heads prevailed).13 On Cohen’s reconstruction, functional explanation is central to Marx’s theory of history, such that defense of that theory required a defense of the propriety of functional explanation. Cohen devoted two chapters to this project, one on functional explanation in general and one on functional explanation in Marx’s texts. I want to attend to Cohen’s treatment of the D-N model with some care. On the concluding pages of his chapter on functional explanation in general, Cohen attempted a defense of functional explanation against Hempel’s criticisms. But it is rather odd, and potentially misleading, that he chose to do so. For this engagement with Hempel was an addendum to his own account of functional explanation, which took place on entirely different terms—Cohen’s primary interlocutor in the chapter was not Hempel but Larry Wright (Wright 1973). His remarks in broaching the subject of the D-N model indicate fairly clearly that Cohen was assuming the viability of that model just for the sake of argument.14 In fact, before setting out his defense of functional explanation, he signaled his rejection of Hempel’s model, appealing to one of what was by then a standard set of counterexamples in which claims that satisfy the model’s requirements simply do not explain: “If Hempel were right, the law that mammals exist only if there is oxygen in the atmosphere, in conjunction with the fact that mammals exist, would explain the fact that there is oxygen in the atmosphere; and that consequence of the theory is quite unacceptable” (Cohen 2000, 273).15 It seems clear that Penner considered Cohen’s attempt to rehabilitate functional explanations in Hempel’s terms inadequate; in the same footnote in which he praised Cohen’s book as “brilliant,” he also wrote, “I do not believe that he succeeds in overcoming Hempel’s basic arguments against the logic of [functional] explanations” (Penner 1999, 255 n12). But it is not clear that it mattered to him that Cohen himself rejected the D-N model. To recall my opening remarks, what makes me despair regarding the (counterfactual) possibility of changing Penner’s mind about functional explanation is that his commitment to Hempel seems to have been unshakeable. By the mid-1970s, the shortcomings of the D-N model as an account of scientific explanation in general had been documented by a prodigious literature (see Salmon 1989), which I think explains why Cohen relegated his engagement with Hempel to the equivalent of a footnote. But even in 2003 Penner remarked, “I have yet to find any demonstration that Hempel was wrong in his interpretation of functional explanations… no one has come up with a schema that allows the argument to go through validly” (Penner 2003, 396). I do not know what, beyond awareness of the relevant critical literature, might have prompted him to consider alternatives to Hempel’s model for understanding how functional explanations are supposed to work. Perhaps familiarity with the approach to explanation inaugurated in the late 1960s by Gil Harman, which I will describe shortly, might have done the trick; perhaps not. I want to focus on the part of Cohen’s account that generated Elster’s reservations. Cohen drew attention to a feature of functional explanations that had troubled scholars from the earliest stages of the enterprise and was prominently represented in Merton’s essay on the subject. Functional explanations suppose that there is some “causal feedback mechanism” that closes the circle, as it were, between a phenomenon’s causes and its effects, so that the phenomenon’s disposition to cause the effects in question can itself become part of the explanation of why that phenomenon is actual. And a longstanding complaint against social-functional explanations—one that we have encountered—has been that while they must suppose that there is some such mechanism, they typically have no detailed account of its constituent parts or its operation. Cohen’s contribution to this discussion centers on his notion of an elaboration. Cohen understood functional explanations to imply the existence of some straightforwardly causal explanation of whatever the functional explanation explains. He cited some familiar examples of elaborations from both everyday discourse and the history of science. One way to elaborate a functional claim is to invoke agency, to say that some agent desired that the effect in question be regularly brought about, knew that the phenomenon under consideration has a disposition to bring about that effect, and therefore intentionally caused the phenomenon to occur regularly. Another type of elaboration is appeal to the dynamics of natural selection—to say, for example, that a functional explanation of the existence of the human heart holds in virtue of the straightforwardly causal dynamics that figure in Darwin’s theory. Cohen (somewhat controversially) regarded elaborations of functional explanations as more complete versions rather than replacements, versions that make clear why it is legitimate for these to invoke the notion of function in the first place. Thus, in his view, “Darwin’s theory is not a rival of functional explanation, but, among other things, a compelling account of why functional explanations apply in the biosphere” (Cohen 2000, 271).16 Now it was part of Cohen’s project to develop plausible elaborations for Marxian functional explanations, which he thought Marxists had generally failed to provide. But more importantly for present purposes, he also claimed that social-functional explanations might be respectable without any elaboration. “We can rationally hypothesize functional explanations even when we lack an account which, like Darwin’s, shows how the explanations work,” he argued. And this is particularly important in social science and history, “For functional explanations in those spheres often carry conviction in the absence of elaborative content” (Cohen 2000, 286). To actually have an elaboration of a functional explanation is to “know how it works,” and so is certainly to be desired; but Cohen claimed that actually having an elaboration is not necessary for either advancing or accepting a functional explanation. Cohen’s remarks on the subject are underdeveloped; I think it cannot be denied that he left the discussion in an unsatisfactory state. Elster’s main criticism of Cohen is aimed directly at the issue of the absence of elaborations, or in his terminology descriptions of “mechanisms,” in social-functional explanations. But what I want to focus on is just what problem Elster saw with this absence. Elster accepted the viability of functional explanations when the underlying “general mechanisms” could be specified. “I do not, of course, quarrel with the use of functional explanation in biology,” he wrote in his review. “Here natural selection provides a general mechanism that creates a presumption that beneficial consequences of structure or behaviour explain their own causes. Cohen does not, however, provide any similar mechanism for functional explanation in the social sciences, and therefore his argument cannot succeed” (Elster 1980, 125ff.). What Elster does not say is that what makes functional explanations in biology acceptable is the fact that the theory of natural selection delivers deductive validity. And it should, I think, have caught Penner’s attention that Elster followed Cohen in thinking that agential elaborations also make for viable functional explanations. For on the D-N model, explanations invoking desires, feelings, or decisions are only valid if these bring about the actions they purportedly explain of necessity, and Elster did not commit himself to this controversial position. In fact, in relation to none of the viable general mechanisms he could imagine—“conscious choice, artificial selection, natural selection, reinforcement and absorbing Markov chains”—did Elster raise the issue of deductive validity (Elster 1980, 127). Rather, the issue he saw is indicated above. The availability of a general mechanism creates a presumption that a functional explanation might apply in the domain in question; and some such presumption is needed for a functional explanation to be advanced with justification. Elster offered an analogy to round off his criticism of Cohen: “In the pseudo-sociology of the 19th century there was general agreement that there must be in societies some analogy to the cell, the hard question being whether this should be taken to be the individual, the family, the firm and so on. Today, of course, we simply reject the presumption of there being a social cell; and in the same way we should simply reject the presumption that beneficial consequences explain their causes” (Elster 1980, 127). Elster’s view, then, was not that functional explanations are fallacious unless they involve valid deductive inferences. Rather, his claim was that there is not good reason to think that, in the social realm, there is a set of general mechanisms from which elaborations could be constructed for the sorts of functional explanations championed by Cohen. And I am inclined to think that what Elster regarded as evidence for this claim was just that a long history of attempts to describe such mechanisms had not delivered anything resembling success. I am also inclined to think that what stands behind this thinking on Elster’s part is the longstanding worry I mentioned above, that functional explanation in the social realm involves the fundamental mistake of applying teleological or agential thinking to a domain where it does not belong. Although I agree with Elster that (to use Cohen’s terminology) elaborated social-functional explanations are more complete and therefore preferable to unelaborated ones, I will not give him the last word on this issue. The philosopher who most effectively defended functional explanation on this point is Richard Miller, who followed his Analyzing Marx (1984) with a major treatise on the nature of explanation in the natural and social science, Fact and Method (1987). At greater length than Cohen, Miller developed an alternative to Hempel’s understanding of scientific explanation, which he called the “causal theory of explanation” (Miller 1987, 120).17 And he used this alternative to address the issue of the legitimacy of functional explanations in the social sciences. Miller’s position was that at least some functional explanations are really “shorthand” for longer explanations that make no appeal to functions—functional explanation in biology, for example, he understood in this way. But the functional explanations he considered interesting were social-functional explanations. Like Cohen, he understood that such explanations suppose that a phenomenon’s disposition to cause a particular effect is somehow linked to the proximate causes of that phenomenon, and, like Cohen, he considered this the central challenge to the legitimacy of social-functional explanation. “It might seem,” he remarked, “that explanations appealing to social functions are as mysterious as appeals to inherent goals in organic evolution, unless the appeal to the effects of the functions is, similarly, derivative. And yet, an investigator appealing to an objective social function often seems not to have even a sketch of an underlying non-functional explanation” (Miller 1987, 119). Miller argued, as had Cohen, that “the functional explainer need not have a sketch of the causal links connecting function and phenomenon explained” to advance a respectable functional explanation (Miller 1987, 122). He offered a lengthier defense of this claim than Cohen had. Miller based his defense on an observation about the working methods of natural scientists: these, he claimed, regularly postulate the existence of unknown mechanisms as components of their hypotheses, and the fact that such a postulation may say almost nothing about the details of the mechanism in question does not doom the explanation in which it figures to failure. Such postulations occur with particular frequency when scientific hypotheses cross divisions between levels of description. So, for example, “a metallurgist explaining a fracture as due to crystallization under repeated stress is committed to the existence of molecular interactions, governed by intermolecular forces, connecting stress with crystallization and crystallization with fracture. Still, the metallurgist need have no idea what this molecular story is, and, indeed, until recently, metallurgists had none” (Miller 1987, 122).18 The principle Miller extracts from these considerations is this. Our ordinary practices of assessing scientific explanations permit these explanations to contain “gaps,” areas where a hypothesis claims that some process operates without being able to say very much about how it operates. So, we should be willing to permit functional explanations to remain silent regarding the details of the mechanisms that link dispositions to proximate causes, simply because “such permissiveness is typical of principles linking different levels of causation, throughout the sciences” (Miller 1987, 122). At this point Miller states with clarity and force a crucial point that is absent from Cohen’s discussion, a point that will be a centerpiece of my remarks in the last part of this article. That point is that what legitimates the postulation of an unknown and undescribed mechanism in the natural sciences is that postulation’s explanatory value. Space permits only the most cursory discussion of the apparatus on which this claim depends. But some treatment of the topic is essential for my purposes. In 1965 Gil Harman published an essay titled “The Inference to the Best Explanation” in The Philosophical Review. Harman’s purpose was to argue that explanations involve a form of inference that is distinct from both the inductive and deductive forms that had dominated discussions to that point. The core of his argument was that it is part of our ordinary cognitive practices to take certain claims to be true just because we take them to explain things that are the case: “In making this inference one infers, from the fact that a certain hypothesis would explain the evidence, to the truth of that hypothesis.” But inference to the best explanation involves assessing the value of explanations, preferring as candidates for truth that explanation that we rate most highly: “One infers, from the premise that a given hypothesis would provide a ‘better’ explanation for the evidence than would any other hypothesis, to the conclusion that the given hypothesis is true” (Harman 1965, 89). Harman inaugurated a series of discussions aimed at identifying just what makes some explanations better than others—discussions that have hardly concluded. Miller’s book is a participant in these discussions. Space does not permit an exegesis of Miller on the subject, and neither do my purposes require it. What is important for me to note is that Miller held that what makes for a successful explanation—an explanation that is deserving of assent—is not that it culminates in a valid deductive inference, but that it explains better than its rivals. Here is Miller applying this point to social-functional explanations and thus making the point that was absent in Cohen’s treatment: Virtually every important functional explanation in social anthropology is supported by a case with these outlines: “Here is an enduring pattern of behavior, on the part of intelligent, appropriately informed people. Its existence regularly violates these people’s individual, immediate interests. No strong state apparatus is mobilized to maintain it. Yet it endures. The best explanation of why it endures is that it serves a long-term, large-scale function of preserving certain features of the society.” …The functional explainer is relying on the principle that persistent, complex behavior violating individuals’ immediate interests requires an explanation. And he or she is arguing that this explanatory need cannot be satisfied without appealing to objective social functions. (Miller 1987, 125) In other words, there is good reason to think that the sort of “general mechanisms” to which Elster referred exist, even if little or nothing can be said about them in detail, when their postulation makes for a better explanation than others on offer. I note that Miller’s position is something like the reverse of Elster’s: where Elster thought that descriptions of such mechanisms are required to create the presumption that a social-functional explanation might apply, Miller claimed that the high value of a social-functional explanation will create a presumption that some such mechanisms exists.19 The lesson I draw from the analytical Marxists here is that functionalism’s problem is not one of logical validity: it is a mistake to reject a social-functional explanation just because it cannot deliver a valid deductive inference. Rather, the problem is that the notion of social function presupposes a rather exotic arrangement of causes and effects within a society, and thus a burden of proof attaches to claims to the effect that a phenomenon serves a social function. The one who sees no reason to think that any such arrangement exists (this is how I understand Elster’s position) will not accept social-functional claims as candidates for truth. But there is a source of reasons to think that there are such arrangements, and that is the value of the explanations in which they figure. The viability of the notion of social function itself, then, is inextricably tied to the utility of that notion for the project of explanation in the social sphere. Explanation is not the Achilles’ heel of social functionalism but its foundation. MARTIN’S CRITICAL INTRODUCTION I want to apply this conclusion to a recent entry in the functionalist tradition within religious studies. Craig Martin’s A Critical Introduction to the Study of Religion originated, according to Martin’s description, as an attempt to rewrite parts of Peter Berger’s functionalist classic of 1967, The Sacred Canopy, for a contemporary student audience; and Martin states clearly in his first chapter that functionalism, conjoined to the “hermeneutics of suspicion,” defines his approach to the study of religion (Martin 2014, 8). Martin fuses the older functionalist tradition with more recent work by such authors as Bruce Lincoln and Pierre Bordieu. The premise of the book, then, is that one way the “critical” study of religion can proceed at present is by way of social functionalism. At least on the surface, there is no mystery about what Penner would have thought of A Critical Introduction. I think it likely that he would have extended to Martin the verdict he issued (in Rediscovering the Buddha) on the idea that Mauss’s analysis of gift-giving should be replaced by an account of “how gifts function as a variable for the satisfaction of needs in a particular political/economic system”: “we know that this theory, known as functionalism, has been thoroughly dismantled as logically invalid” (Penner 2009, 256ff.). It should be apparent that such will not be my own judgment. But what is particularly interesting about the contrast between the Hempelian and post-Hempelian reconstructions of the notion of social function is that the latter does more than simply clear functionalism of an accusation levied by the former. On the post-Hempelian reconstruction, functionalism has a problem that Hempel’s reconstruction does not capture. I have hinted at this problem above; to bring it fully into view, I want to sketch a different, and less dismissive, attitude that Penner might have taken to Martin’s book. One noteworthy feature of A Critical Introduction is that the idea that claims about social functions are explanations plays only a marginal role in the book. That idea emerges into the clearest focus in Martin’s first chapter. But that focus is not a sharp one (I will provide some support for this claim in a moment), and after this first chapter the topic of explanation basically disappears. The theoretically important material in the remainder of the book is presented as claims about “how society works,” “how religion works,” and how religious and social phenomena can be used by persons seeking to promote their own interests or the interests of their class. Arguably, claims of these sorts are connected to the project of explaining religion. But as there are no places in those chapters where Martin returns to the theme of explanation, the connections are not described. I do want to document the commitment to social-functional explanation that appears in Martin’s first chapter. He introduces the notion by way of references to Freud: “Why do religions exist? For Freud, the answer is simple: because something is necessary to alleviate the friction created by social repression. Religion in society is like the oil in an engine that keeps it from overheating”; “Freud would say that all of those things are illusions people believe in simply because they are necessary to relieve psychological pressure and thereby keep society running” (Martin 2014, 9). And shortly afterward Martin commits himself to functionalism as a species of explanation: alongside other projects, “we will look for a functionalist explanation for the data under consideration—if this myth is not true, for instance, does it continue to be told because it serves some sort of social function for the community that tells it?” (Martin 2014, 11). Since Martin does not discuss functional explanation in the abstract, the examples he provides are the book’s only source of information regarding how he understands the method. Take, for example, his treatment of an observation by the anthropologist Max Gluckman: According to the Barotse, if an individual does not pay the doctor then the medicine given for the illness will magically make the illness return. However, from an outsider’s perspective we can utilize a hermeneutic of suspicion and look for a functionalist explanation: if we assume that the claim is, in fact, not true, why might people in this community make such a claim? This one is easy to answer: if people believe this they will be more likely to pay their doctor’s bills for fear of getting sicker. This is a myth that serves the function of ensuring that doctors will be paid. (Martin 2014, 12, italics original) Here Martin positions a claim about the consequences of a myth’s acceptance as an answer to the question of why people continue to assert its truth. The absence of any “causal feedback mechanism” that would turn those consequences into causes is particularly evident in this example; Martin offers neither any account of such a mechanism nor any indication that his functional explanation presupposes one. A similar softness of focus is on display in his subsequent rejection of the “strong functionalist” position that “society works like a machine, and that religion is an essential part that has one and only one function in the machine,” in preference for the “weak functionalist” position that holds that “religious traditions can serve various social functions but without completely reducing religion to those social functions.” Martin positions “weak functionalism” as a position that makes it possible to think that “if it turns out that the supernatural claims under consideration are true . . . our functional explanation can still be true”—so, for example, that “Even if it is true that the Barotse doctor’s medicine will make the patient sick if she does not pay, it is still true that this belief functions to ensure that the doctor will be paid” (Martin 2014, 14, italics original). The thought here seems to be that the social consequences of a true claim explain why people assert it. It is not clear whether Martin means to suggest that in this case the functional explanation is the only correct one, such that the truth of the myth has no explanatory value in relation to its assertion, or (which is a more interesting position) that multiple explanations of the same social practice can be true—that the Barotse assert the myth because of its truth and they assert it because of its social consequences. But a choice between these positions does not arise, because Martin does not say how the myth’s social effects explain why certain of the Barotse assert its truth. The conclusion I draw from these observations, together with the disappearance of the theme of explanation following the first chapter, is that explanation is not central to social functionalism as it is presented in A Critical Introduction, its stated commitment to functional explanation notwithstanding. The book is clearly committed to the project of identifying the social functions that religious phenomena can serve. But it is not consistently concerned with using claims about social functions to explain religion.20 I think this aspect of the book’s character might provide grounds for something other than outright dismissal by the devotee of Hempel’s D-N model, for Hempel’s criticism was centered on the dynamics of functional explanation. So far as I can see, the model provides no grounds for dismissing simple claims to the effect that a phenomenon has a social function: when such claims are not presented as explanations, it seems to me that the devotee of Hempel should have no objection to the possibility of their being true. In fact, Hempel himself left the door open to the truth of social-functional claims at the conclusion of his canonical essay, writing that when the problems of social-functional explanation are fully in view, “functional analysis” might perfectly well persist as “a program of inquiry aimed at determining the respects and the degrees in which various systems are self-regulating” (Hempel 1965, 330).21 Imagining Penner at his most charitable, then, I can envision him arguing that a fundamental theoretical confusion detracts from only the first part of A Critical Introduction, and that the bulk of the book, where the notion of social function is not clearly tied to explanation, is thereby clear of this defect. But here my position motivates a different reservation. I have argued that the issue with the notion of social functions for the post-Hempelians has to do with reasons for thinking that there exist mechanisms embedded within the social order that make consequences into causes. And I agree with Miller that it is reasonable to accept that some such mechanism exists when positing its existence makes for the best explanation of the phenomena in question. On this view, the absence of tidy accounts of causal feedback mechanisms in those sections of Martin’s book where explanation is in view is not in itself a flaw: it is to be expected that such mechanisms will be, as it were, invisible to the naked eye, and so knowable only through postulation. But what is grounds for criticism is the soft focus on the theme of explanation and absence of any engagement with the question of whether social-functional explanations are better explanations than the alternatives. For if claims to the effect that a religious phenomenon serves a social function is not part of a “best explanation,” it is not clear what reason there is to think that the causal feedback mechanisms that convert effects into causes exist, and thus it is not clear what reason there is to think that there is any functioning going on in the first place. Let me apply these reflections to the example of the Barotse myth. Consider the question of whether one should accept Martin’s functional explanation, that certain of the Barotse repeat the myth because a person who believes it to be true will be more likely to pay their doctors. On my position, one would have reason to believe that the explanation is true if it is better than the available alternatives. I see nothing standing in the way of such a claim, but Martin does not make it—he does not claim that the social-functional explanation is better than the available alternatives, only that it is a possibly true explanation. The result is that this case offers an example of social-functional explanation, but not an example of a reason to think that such an explanation is true. So far as I can discern, this state of affairs holds throughout the book. The disappearance of the theme of explanation after the first chapter means the absence in the remainder of the book of engagement with the question of whether there is any reason to think that religion serves social functions. I do not doubt that it would be possible for the theoretical devices that Martin appropriates from Berger, Lincoln, Bordieu, and others—legitimation, authority, habitus, and so on—to be presented as resources for the explanation of religion; in fact, it makes considerable sense to me to think of these devices as possible components of the “feedback mechanisms” that social-functional explanations entail. And one possible application of such a presentation might be to a case for the viability of social functionalism in the study of religion, on the grounds that it can generate explanations that are at least as good as the alternatives. But as it stands, the book contains no such presentation and makes no such case; and, while it makes clear that a variety of functional claims can be and have been made regarding religious phenomena, it never really says why anyone should think any of these are true.22 So where Penner’s understanding of functional explanation would likely motivate the outright dismissal of Martin’s approach but might be compatible with a grudging acceptance, my own position brings into view a different problem, which is an absence in the book of arguments for the viability of social functionalism as a methodology for the study of religion. CONCLUSION I think it undeniable that social functionalism is now but a shadow of what it once was in our field; and my impression is that for the most part, where the language of social functionalism hangs on in our literature, its connection to the project of explaining religion is tenuous at best. It is possible that Hempel’s criticism of functional explanation, and Penner’s work in disseminating this criticism, has had something to do with this, because so far as I can see Penner’s position more or less entailed that scholars of religion should avoid functional explanations altogether unless they were in a position to argue that religious phenomena are necessary for the social effects they produce. But my disagreement with Penner motivates a different set of recommendations for those drawn to social-functionalist thinking. My first recommendation is that scholars who are inclined to argue that religious phenomena serve social functions should cast their claims precisely as explanations, whether of the religious phenomena themselves or of other actual phenomena,23 and should not feel an obligation to satisfy the requirements of the covering-law model. To be respectable, social-functionalist explanations do not need to be deductively valid; they just need to be, plausibly, at least as good as the available alternatives. And my second recommendation follows closely behind the first: scholars who, for whatever reason, are unwilling to put their social-functionalist claims forward as not just explanations but very good explanations should either resign themselves to making unsupported assertions or avoid functionalist language altogether. These recommendations are motivated by a rather cranky (but hardly original) observation on my part. It seems to me that in the historical literature, the link between social functionalism and explanation was almost never as firm as it should have been, and that where social functionalism persists, it often does so in the form of unsupported assertions and half-formed insinuations (had I chanced to meet Penner, a good opening gambit before trying to change his mind might have been to commiserate with him about this). I think it would be an improvement to the literature of our field if these were either converted into genuine attempts at explanation or eliminated. Frankly, it is not clear to me either that the waning of social functionalism has been a great loss for the study of religion or that its complete disappearance would be. If at one time prominent scholars hoped that social functionalism might be the key to explaining religion, it seems to me the foundations of this hope were always somewhat mysterious, and the results of a century’s work suggest they were never very firm.24 And (although this is a topic for another day) if social functionalism has been useful for a different project, that of criticizing religion, its utility can be overstated. Where the question is what is to be made of religion in relation to ethical or political values—knowing, for example, whether religion abets or obstructs various kinds of oppression—it can be helpful to know about the effects that religious phenomena have, the interests they serve, and the uses that are made of them. But claims about such matters are conceptually independent of claims about social function, and one can have perfectly good reason to think that such claims are true without thinking that they figure in the operation of any self-regulating systems.25 Whether there are any such systems is still, I take it, an open question and an interesting one. But it is a question that matters for fewer purposes than some have thought. Radcliffe-Brown’s attention was captured by Robert Lowie’s suspicion that social-functionalist thinking amounted to an unwarranted imposition of order on the chaos of collective life. I will not take the time to translate Lowie’s words into my own idiom, but I affirm the spirit of the remark with which he brought his Primitive Society to a close. To that planless hodge-podge, that thing of shreds and patches called civilization, its historian can no longer yield superstitious reverence. He will realize better than others the obstacles to infusing design into the amorphous product; but in thought at least he will not grovel before it in fatalistic acquiescence but dream of a rational scheme to supplant the chaotic jumble. (Lowie 1920, 441) Footnotes 1 An early version of the arguments of this article was presented in November 2014 at a session of the American Academy of Religion Philosophy of Religion unit. I am grateful to session participants for useful feedback and also to Nancy Frankenberry, Craig Martin, Jeffrey Stout, and two anonymous reviewers for the JAAR for comments on various drafts. 2 The idea that social functionalism hangs on this sort of analysis was not universally accepted (see, for example, Davis 1959, 768). 3 “[T]he point of functional explanation has normally been to show that there are ‘reasons’ for the existence and continuance of social institutions that are quite distinct from the reasons actors might have for whatever they do” (Giddens 1987, 68). 4 Nicos Mouzelis, at least, recalls this feature of functionalism in its classical form: “When theoretically sophisticated functionalists talk about social needs or functional requirements, they simply mean conditions of existence” (Mouzelis 1995, 132). 5 In Rethinking Social Theory, Roger Sibeon claims that one of the cardinal sins of sociological thinking is “functional teleology,” which he describes as “an invalid form of analysis involving attempts to explain the causes of social phenomena in terms of their effects, where ‘effects’ refers to outcomes or consequences viewed as performances of ‘functions.’” The problem with this mode of analysis is that “in the absence of a demonstration of intentional and successful planning by actors somewhere, sometimes, it is a teleological fallacy to attempt to explain the causes of phenomena in terms of their effects.” It is not clear to me whether Sibeon is unaware that some functionalists understood functional explanations to involve no invocations of teleology or whether he thinks that the two cannot be disentangled. He is on firm ground, however, in claiming that many invocations of functionalism in the literature seem not to avoid teleological thinking (Sibeon 2004, 6). 6 This supposition explains the following passage by Radcliffe-Brown: “The scanty information I have been able to gather suggests that the lineage and joint-family organization of some parts of India is losing something of its former strength and solidarity and that what we should expect as the inevitable accompaniment of this, a weakening of the cult of ancestors, is also taking place” (Radcliffe-Brown 1945, 38). In other words, the functionalist position predicts that religious phenomena will disappear once the effect that it is their function to produce is no longer socially necessary. 7 “The function of a particular social usage is the contribution it makes to the total social life as the functioning of the total social system. Such a view implies that a social system…has a certain kind of unity, which we may speak of as a functional unity. We may define it as a condition in which all parts of the social system work together with a sufficient degree of harmony or internal consistency, i.e. without producing persistent conflicts which can neither be resolved nor regulated. This idea of the functional unity of a social system is, of course, a hypothesis. But it is one which, to the functionalist, it seems worthwhile to test by systematic examination of the facts” (Radcliffe-Brown 1935, 181). 8 Hempel noted his agreement with this result of Merton’s analysis. He described as “the general hypothesis of self-regulation” the idea that “within certain limits of tolerance or adaptability, a system of the kind under analysis will—either invariably or with high probability—satisfy, by developing appropriate traits, the various functional requirements (necessary conditions for its continued adequate operation) that may arise from changes in its internal state or in its environment” (Hempel 1965, 317). Hempel had reservations concerning the extent to which claims as to the self-regulating character of particular social systems could be empirically tested; but more importantly, he regarded it as crucially important that it not be forgotten by social thinkers that the hypothesis of self-regulation is a hypothesis. A central task to which social-functional analysis could be applied, for him, was to determine whether there are any self-regulating social systems—“to ascertain to what extent such phenomena can be found, and can be represented by corresponding laws” (Hempel 1965, 325). 9 “Some functionalists have seemed unwilling to discipline their language or test their propositions, taking instead the easy path of verbal tapestry. As a result, sociologists who have learned the techniques of empirical research come to feel that functionalism is a crank method, and they are encouraged in this by functionalists themselves who say that they are engaged in non-causal analysis, whatever that is” (Davis 1959, 768). 10 Note that there have been such claims in the literature. For example, Bronisław Malinowski suggests: “Thus magic fulfills an indispensable function within culture. It satisfies a definite need which cannot be satisfied by any other factors of primitive civilization” (cited in Merton 1957, 34). 11 Other publications of Penner’s that display his commitment to Hempel’s model are Penner 1989 (anthologized as Penner 1999) and Penner 2003. 12 See Salmon 1989 for an overview of the fortunes of Hempel’s model over the course of the period in question. 13 Cohen recounted the history of analytical Marxism in his Introduction to the 2000 edition of Karl Marx’s Theory of History. While eschewing the “more aggressive” designation of “non-bullshit Marxism,” he did not entirely repent of its spirit: “In fact, there exists Marxism which is neither analytical nor bullshit, but, once such (as we may designate it) pre-analytical Marxism encounters analytical Marxism, then it must either become analytical or become bullshit” (Cohen 2000, xxvi). 14 “A consequence explanation relates to a consequence law in whatever way explanations relate to law. It is not our task to say what that relation is. But it may be clarifying to exhibit how consequence laws figure in explanations on the supposition that the correct answer to the question here suspended is Hempel’s.” Cohen’s mammal example is drawn from a paragraph on the same page in which his remarks “trespass into assessment of the D-N model itself” (Cohen 2000, 272ff.). 15 On “famous counterexamples” to Hempel, including Bromberger’s flagpole example, see Salmon 1989, 47–50. 16 The general principle operative here is that “the sceptic misconstrues as various alternatives to functional explanation what are in fact various more complete forms functional explanations may take” (Cohen 2000, 271). 17 As Miller describes it, in such a theory “an unanalyzed notion of causation is used to describe how a successful explanation works.” The novelty of this approach, in other words, is that it begins with a broadly inductive understanding of how we assess the goodness of explanations, and then allows anything to be termed a “cause” that could figure in successful explanations (Miller 1987, 120). 18 Miller credits Garfinkel 1981 for this example. 19 To be sure, there are mysteries here, such as how an explanation containing the sort of gaps that Miller describes could possibly be better than an explanation containing no gaps at all. Here I have space only for some trivial examples that indicate, I hope, that there is no simple answer to this question. In some cases, we are perfectly happy to accept “gappy” explanations, as for example when I accept that you hit me because you were angry with me without having any idea how exactly mind-body causality works. And in other cases, a gappy explanation might be better than either no explanation at all or a highly implausible explanation; so, for example, an unelaborated social-functional explanation might be better than “gapless” explanations involving phlogiston, fairies, or Geist. 20 Places where Martin clearly positions the results of his analyses as explanations are not entirely absent in the book, but they are quite rare and often not clearly functionalist in character. Here is one example, drawn from Martin’s discussion of the origins of “Hinduism”: “Why posit a unified identity where none exists? . . . In the end, the fabrication of a timeless Hindu identity appears to have been a power play born out of racial, ethnic, and cultural hatred of some Indians towards Muslims in their midst” (Martin 2014, 156f). 21 See also Mouzelis: “As Merton pointed out long ago, once it is made clear that the functional or dysfunctional consequences exerted by a social item on the social whole must not be used as a cause explaining how this item came about, then it is perfectly legitimate to ask questions about whether or not this item strengthens or weakens the social whole” (Mouzelis 1995, 132). 22 Having advanced this criticism of A Critical Introduction, it is only fair to note that as it is a text written for undergraduates, there are limits to the extent to which such a book can reasonably be expected to engage some of the more elusive methodological issues its proposals raise. That being said, I do think that the book represents something of a missed opportunity to make undergraduates aware of some of the complexities of functionalism as a method of investigation. 23 In a 1975 essay Roger Cummins attempted a defense of functionalism in part by arguing that Hempel and others had misunderstood what functional explanations explain. What they explain, according to Cummins, is not the existence of the phenomena that function, but rather the existence of whatever those phenomena function to make actual. So, for example, “appeal to the function of the contractile vacuole in certain protozoans explains how these organisms are able to keep from exploding in fresh water,” rather than explaining why those organisms have contractile vacuoles in the first place (Cummins 1975, 751; I am grateful to Nancy Frankenberry for this reference). I do not follow Cummins in thinking that functional explanations must be construed in this way. But certainly, one can explain a phenomenon—a social condition, say—by postulating something that functions to make it actual. When that something is a religious phenomenon, such an explanation will generate a reason to think that phenomenon serves a social function if that explanation is better than the available alternatives. Thus, it is possible to argue that a religious phenomenon serves a social function without at the same time attempting to explain that phenomenon. 24 In making this remark, I have particularly in mind the exchange between Roy Rappaport and Jonathan Friedman regarding Rappaport’s Pigs for the Ancestors (Rappaport 1984). The massive size and careful argumentation of the work notwithstanding, Friedman was simply unconvinced that Rappaport had shown that pig-slaughtering rituals among the Maring-speaking Tsembaga functioned to regulate the Tsembaga relation to their environment. “The cybernetic analysis of Maring society is unnecessary to account for phenomena that are due to more mundane mechanisms,” Friedman remarked. “The fact that the slaughter of pigs is ritualised has, I think, to do with the nature of Maring as well as other tribal social relations and not with a regulatory function” (Friedman 1979, 258). 25 In commenting on a draft of this article, Martin has defined weak functionalism as “a scholarly approach [that] demonstrates how elements of culture are utilized by subjects—either intentionally or unintentionally—in ways that produce social effects (through legitimation, mystification, naturalization, and so on) that serve the interests of some subject positions in a society over others” (personal communication). I think that my concluding remarks are in accord with this definition, as it does not seem to me that it makes use of the notion of social function. REFERENCES Cohen G. A . 2000 . Karl Marx’s Theory of History: A Defence . Expanded ed. Princeton, NJ : Princeton University Press . Cummins Robert . 1975 . “ Functional Analysis .” The Journal of Philosophy 20 : 741 – 65 . Google Scholar Crossref Search ADS Davis Kingsley . 1959 . “ The Myth of Functional Analysis as a Special Method in Sociology and Anthropology .” American Sociological Review 24 : 757 – 72 . Google Scholar Crossref Search ADS Durkheim Emile . [1912] 1995 . The Elementary Forms of Religious Life . Translated by Karen E. Fields . New York : The Free Press . ——. [1895] 2013 . Rules of Sociological Method and Selected Texts on Sociology and its Method . Basingstoke : Palgrave MacMillam . Elster Jon . 1980 . “ Cohen on Marx’s Theory of History .” Political Studies 28 : 121 – 28 . Google Scholar Crossref Search ADS Friedman Jonathan . 1979 . “Hegelian Ecology: Between Rousseau and the World Spirit .” In Social and Ecological Systems , edited by P. C. Burnham and R. F. Ellen , 253 – 70 . New York : Academic Press . Garfinkel Alan . 1981 . Forms of Explanation: Rethinking the Questions in Social Theory . New Haven, CT : Yale University Press . Giddens Anthony . 1971 . Capitalism and Modern Social Theory . Cambridge : Cambridge University Press . ——. 1987 . Social Theory and Modern Sociology . Stanford, CA : Stanford University Press . Godlove Terry . 2005 . Teaching Durkheim . New York : Oxford University Press . Harman Gilbert . 1965 . “ The Inference to the Best Explanation .” The Philosophical Review 74 : 88 – 95 . Google Scholar Crossref Search ADS Hempel Carl . 1965 . “ The Logic of Functional Analysis .” In Aspects of Scientific Explanation and Other Essays in the Philosophy of Science . New York : The Free Press . Lowie Robert H . 1920 . Primitive Society . New York : Boni and Liveright . Martin Craig . 2012 . A Critical Introduction to the Study of Religion . Bristol, CT : Equinox . McCauley Robert N. , and E. Thomas Lawson . 1984 . “ Functionalism Reconsidered .” History of Religions 23 : 372 – 81 . Google Scholar Crossref Search ADS Merton Robert . 1957 . “ Manifest and Latent Functions .” In Social Theory and Social Structure , edited by Robert Merton , 19 – 84 . New York : Free Press . Miller Richard . 1984 . Analyzing Marx: Morality, Power and History . Princeton, NJ : Princeton University Press . ——. 1987 . Fact and Method: Explanation, Confirmation and Reality in the Natural and the Social Sciences . Princeton, NJ : Princeton University Press . Mouzelis Nicos . 1995 . Sociological Theory: What Went Wrong? Diagnosis and Remedies . London : Routledge . Penner Hans . 1971 . “ The Poverty of Functionalism .” History of Religions 11 : 91 – 97 . Google Scholar Crossref Search ADS ——. 1989 . Impasse and Resolution: A Critique of the Study of Religion . New York : Peter Lang . ——. 1999 . “ What’s Wrong with Functional Explanations ?” In Language, Truth, and Religious Belief , edited by Nancy Frankenberry and Hans Penner , 246 – 70 . Atlanta : Scholars Press . ——. 2003 . “ On Self-Regulating Systems, Cannibals, and Cogs that Turn No Wheels: A Response to Joel Sweek .” Method & Theory in the Study of Religion 15 : 390 – 408 . Crossref Search ADS ——. 2009 . Rediscovering the Buddha: The Legends and Their Interpretations . Oxford : Oxford University Press . Radcliffe-Brown A. R . 1935 . “ On the Concept of Function in Social Science .” American Anthropologist , New Series 37 : 394 – 402 . Google Scholar Crossref Search ADS ——. 1945 . “ Religion and Society .” Journal of The Royal Anthropological Institute of Great Britain and Ireland 75 : 33 – 43 . Crossref Search ADS Rappaport Roy . 1984 . Pigs for the Ancestors: Ritual in the Ecology of a New Guinea People . 2nd ed . New Haven, CT : Yale University Press . Salmon Wesley . 1989 . Four Decades of Scientific Explanation . Minneapolis : University of Minnesota Press . Sibeon Roger . 2004 . Rethinking Social Theory . London : Sage Publications . Turner Jonathan H. , and Alexandra Maryanski . 1979 . Functionalism . Melo Park, CA : The Benjamin/Cummings Publishing Co . Wright Larry . 1973 . “ Functions .” The Philosophical Review 82 : 139 – 68 . Google Scholar Crossref Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of the American Academy of Religion. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of the American Academy of Religion Oxford University Press

Contra Penner, What’s Really Wrong with Functional Explanations

Loading next page...
 
/lp/ou_press/contra-penner-what-s-really-wrong-with-functional-explanations-qpVXC30h9o
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the American Academy of Religion. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
0002-7189
eISSN
1477-4585
D.O.I.
10.1093/jaarel/lfy002
Publisher site
See Article on Publisher Site

Abstract

Abstract Hans Penner offered a prominent line of criticism of social functionalism in the study of religion. Penner took exception to social-functional claims when these are presented as explanations. Following Carl Hempel, he argued that such explanations very often commit the logical fallacy of affirming the consequent. I argue that Penner was wrong about social functionalism. I draw on the work of some contemporaries of Penner who rejected Hempel’s model of scientific explanation and defended the propriety of social-functional explanation, particularly in connection with Marxian theory, and argue that rather than being functionalism’s Achilles’ heel, explanation is its foundation. I then apply my position to a recent book by Craig Martin that advocates for social functionalism in the study of religion. I conclude that, because it marginalizes explanation, the book offers little in the way of reasons to think that religious phenomena serve social functions. No one has any quarrel with a functional explanation where the mechanism is actually shown to be at work. The hard question is whether one can ever be justified in setting forward a functional explanation even in the absence of a specific mechanism. (Elster 1980, 126) INTRODUCTION I NEVER MET Hans Penner, although our academic careers overlapped for a few short years.1 Had we chanced to meet I might have thanked him for introducing me to the murky but fascinating topic of social functionalism in religious studies. That introduction came by way of his essay “What’s Wrong with Functional Explanations?” which I encountered at about the midpoint of my graduate training (Penner 1999). It was some years after Penner passed away that I arrived at the conviction that he was wrong about social functionalism. There are possible worlds in which he lives on long enough that I have a chance to change his mind about the subject. For reasons I will mention later, I estimate that I am successful in few, if any, of those worlds. In this article, I want to lay out the case for thinking that Penner’s formulation of the issues around social functionalism is deficient and offer what I think is a superior formulation. The position I will offer as a replacement for his is not my invention: to begin staking out the territory, I will basically be positioning the analytical Marxists of the late twentieth century as a corrective to Penner’s source, Carl Hempel. What results will support the idea that there is something wrong with functional explanations only in a modest way. I do not think that there is something so wrong with functional explanations that it is always a mistake to advance them or take them to be true. But neither do I think that social functionalism is clearly a promising research methodology for religious studies. I will begin by briefly rehearsing what social functionalism is in general terms, with reference to the role this doctrine has played in the history of religious studies. I will then exegete Hempel and Penner on the topic. I will then turn to the analytical Marxists, who were contemporaries of Penner’s but whose work I think he simply failed to appreciate. Discussions among three of these figures—G. A. Cohen, John Elster, and Richard Miller—expose the problem on which I eventually want to focus. I will conclude by describing that problem in my own terms, with reference to recent literature in which social functionalism appears with some prominence—literature that suggests to me that we scholars of religion have yet to make our peace with this part of our theoretical history. CLASSICAL SOCIAL FUNCTIONALISM In its classical form, social functionalism about religion is the position that the question of why religious phenomena are present within particular social groups can be answered by way of a description of their social effects. As a formally described “live option” in the academic study of religion, it originated around the turn of the twentieth century, flourished in the mid-century literature, was subject to close examination and strenuous criticism during these same decades, and declined notably after the 1960s. Social functionalism was given canonical expression in two texts from Emile Durkheim, Rules of Sociological Method (1895) and The Elementary Forms of Religious Life (1912). Here is a key passage from the latter text: There can be no society that does not experience the need at regular intervals to maintain and strengthen the collective feelings and ideas that provide its coherence and its distinct individuality. This moral remaking can be achieved only through meetings, assemblies, and congregations in which the individuals… reaffirm in common their common sentiments. Such is the origin of ceremonies that, by their object, by their results, and by the techniques used, are not different in kind from ceremonies that are specifically religious. (Durkheim [1912] 1995, 429; quoted in Godlove 2005, 111ff.) This kind of thinking is grounded on an analogy between social systems and biological organisms, an analogy freely admitted by such early defenders of functionalism as Alfred Radcliffe-Brown (Radcliffe-Brown 1935, 178; see also Merton 1957, 21).2 Just as the continued existence of an organism requires that certain effects be accomplished—that oxygen, say, be delivered to the organism’s cells—so too, the continued existence of a social system requires that certain effects be accomplished. And just as the body’s need for oxygen delivery is satisfied by the proper operation of certain organs within the body, the needs of a social entity are satisfied through the performance of certain kinds of regularized, coordinated activity on the part of its members. And thus, we can speak of the function of a religious ritual, say, in much the same way as we can speak of the function of the heart. So, for example, Radcliffe-Brown: Rites can be seen to be the regulated symbolic expression of certain sentiments. Rites can therefore be shown to have a specific social function when, and to the extent that, they have for their effect to regulate, maintain and transmit from one generation to another sentiments on which the constitution of the society depends. (Radcliffe-Brown 1945, 35) Classical social functionalism was committed to the idea that social functions have explanatory value, that certain social phenomena, in other words, can be explained by their functions. This, I think, is how to parse Durkheim’s reference to the origin of ceremonies above; and elsewhere in The Elementary Forms he states, after arguing that the very existence of a clan requires some way for its members to think of themselves as a unity and that clan emblems make such thinking possible, that “both the institution of the emblem and its place in the group’s life are thus explained” (Durkheim 1995, 235). From the beginning of its rise to prominence in the study of religion, social functionalism was accompanied by a body of critical reflection. Even in the classical texts of the early twentieth century, advocates of social functionalism acknowledged that the responsible deployment of the paradigm required attention to a number of distinct issues. I want to mention three of these. First, almost from the very beginning, social functionalism was premised on the possibility of avoiding appeals to teleology. Classical functionalists were interested primarily in those social effects of, for example, religious rituals that were not the effects that participants intended their actions to have.3 And early defenders of social functionalism recognized that teleological language could not be used in the description of purported social-functional arrangements without risk. The risk at issue was that of tacitly supposing that social groups are something like collective agents, capable of organizing their internal workings for preconceived purposes (much later, Jonathan Friedman would waggishly characterize Roy Rappaport’s functionalism as “Hegelian ecology” (Friedman 1979). Durkheim’s language in the first quote above simply transposes folk psychology to the level of the social group by suggesting that rituals take place because societies are aware of their own needs, aware that the regular occurrence of rituals will satisfy those needs, and act to institute the rituals so that their needs will be satisfied. And Durkheim did not provide a translation of this teleological language into terms innocent of collective agency. But as social functionalism developed as a method of analysis, the need for such translation became evident, so that by 1935 Radcliffe-Brown recommended the excision of teleology from social-functional explanations in part through the substitution of “necessary conditions of existence” for Durkheim’s “needs” (Radliffe-Brown 1935, 394).4 The project of social functionalism even in its classical form, in other words, was to make it possible for social phenomena to be explained by their consequences without appeal to teleological reasoning.5 Second, classical social functionalism was not uniformly committed to the position that the function of a social phenomenon explains that phenomenon’s origin. Although Durkheim made precisely such a claim in The Elementary Forms, this was simple carelessness on his part. Earlier, in The Rules of Sociological Method, he had distinguished claims about the causes of a social phenomenon from claims about its function, arguing that the “complete explanation” of social phenomena required both; but in The Elementary Forms, as Terry Godlove has noted, Durkheim simply interwove causal and functional strands of explanation (Godlove 2005, 166 n33; see also Giddens 1971, 90ff.). If not the origin of religious phenomena, then, what do functional claims explain? The question is not always easily answered, but one fairly accessible response is that functional claims explain the continued, regularized presence of religious phenomena within a particular social group; in other words, whatever the origin of these phenomena, they persist because of the function they serve.6 This sort of thinking is also behind the distinction, expressed by Radcliffe-Brown and later explored by Carl Hempel, between “historic-genetic” and social-functional explanations; it is also the grounds for the common accusation that social-functionalist treatments of cultural phenomena give short shrift to historical considerations (see Hempel 1965, 309; Turner and Maryanski 1979, 109–13). Third, at least some classical functionalists acknowledged the contestable nature of the basic premise of social-functional analysis. I think Radcliffe-Brown was pointing to the tip of a theoretically important iceberg when he acknowledged that the claim that social groups are organized into functionally unified systems is not an observation but a hypothesis.7 He acknowledged further that this hypothesis was controversial and that some of his contemporaries—he cited Robert Lowie in particular—rejected it, on the grounds that it imputed too high a degree of order to the products of collective human activity (Radcliffe-Brown 1935, 401). And in his lengthy midcentury analysis of social functionalism, Robert Merton offered the following remark concerning this “postulate of the functional unity of society”: This conception, in its recent formulations, was developed by social anthropologists, that is, by men primarily concerned with the study of non-literate societies. In view of what [Paul] Radin has described as “the highly integrated nature of the majority of aboriginal civilizations”, this assumption may be tolerably suitable for some, if not all, non-literate societies. But one pays an excessive intellectual penalty for moving this possibly useful assumption from the realm of small non-literate societies to the realm of large, complex and highly differentiated societies. (Merton 1957, 28, italics original) Merton went on to call particular attention to the pernicious effects of the assumption of functional unity for the analysis of religion: specifically, that it directs attention away from nonfunctional, socially disruptive effects of religion and from “internecine conflicts among religious groups,” for “all this abundantly known material is ignored in favor of illustrations drawn from the study of religion in non-literate society” (Merton 1957, 29). But for my purposes, a more general claim by Merton is more important: that however useful the idea of social-functional integration may be as a hypothesis guiding the ethnographic study of particular social arrangements, the idea that contemporary societies form functionally integrated wholes is a hypothesis to be confirmed through examination of specific cases rather than a secure deliverance of anthropological research.8 I will return to this point below; for the moment, I want to note that within classical functionalism itself is to be found awareness of the dependence of functionalism on the supposition that there are such things as social systems, and awareness that this supposition is not obviously true (on this point, see also McCauley and Lawson 1984, 378–81). To my mind, these three issues add up to something. It seems to me that classical social functionalism was always vulnerable to the worry that to think that large-scale social phenomena can be explained by their consequences is simply to extend teleological thinking, and perhaps agential thinking, beyond its proper domain, or to use the words of Kinglsey Davis, that functionalism is a “crank method” of social analysis.9 PENNER’S CRITICISM Critical discussions of functionalism in the middle of the twentieth century by Richard Braithwaite, Robert Merton, and others formed the background of Carl Hempel’s 1959 essay, “The Logic of Functional Analysis” (Hempel 1965). In this essay, Hempel charged functional explanations with the fallacy of affirming the consequent, or arguing If A, then B; B, therefore A. Hempel famously advocated the deductive-nomological or “covering law” model of (nonstatistical) scientific explanation, and his charge against functionalism is a product of his attempt to make sense of functional explanations on this model. On the covering-law model, an explanation consists of two components: first, a description of the conditions under which the phenomenon to be explained occurs (“whenever A is the case, B occurs”), and second, a statement that entails that these conditions are satisfied (“A is the case”). The existence of the phenomenon follows deductively from the conjunction of these statements. In a case of successful explanation, “the outcome described in the explanandum was to be expected in view of the antecedent circumstances and the general laws listed in the explanans. More precisely, the explanation may be construed as an argument in which the explanandum is deduced from the explanans” (Hempel 1965, 299). Hempel’s observation was that functional explanations almost never work on the deductive-nomological model. Functional explanations involve two empirical observations: that a specific set of rituals, say, are performed within a society, and that an effect of those rituals—social stability, say—obtains. Reconstructed on the “D-N” model, a functional explanation would be a set of claims from the conjunction of which the occurrence of the ritual could be deductively inferred. The crucial question thus is: what would need to be conjoined to the statement that social stability has been achieved in order for the occurrence of the ritual to follow deductively? Trivially, only statements that entail that the only way social stability could be achieved is through the performance of the ritual, so that whenever social stability obtains, the ritual has been performed. In formal terms, the argument if A then B; B; therefore A is invalid; this argument commits the fallacy of affirming the consequent. But the argument B if and only if A; B; therefore A is valid, its validity purchased by way of the specification that A must be the case in order for B to be the case. Hempel noted that functional explanations do not typically claim that the phenomena they are concerned to explain are the only possible mechanism for the production of the effects they produce; and yet, as he understood them, functional explanations are valid only when this is the case (Hempel 1965, 313).10 Functional explanations thus have a logical problem. If you are familiar with this charge, it is likely due to Penner’s influence. Penner repeated Hempel’s claim that the problem of functional explanations is a logical problem in publications that spanned his career, beginning with the 1971 essay “The Poverty of Functionalism” in History of Religions.11 The latest mention that I have been able to identify is in Penner’s last book, Rediscovering the Buddha (2009), where he briefly described functionalism as a theory that had long ago been “thoroughly dismantled as logically invalid” (Penner 2009, 207). It seems that Penner carried the torch of Hempel’s criticism to the end of his days, in a determined effort to bring illumination to us all. I take Penner’s discussion of 1989, excerpted and republished a decade later, as my source for his mature position on functional explanation. Penner’s main project in the essay was to apply Hempel’s critique of functionalism to the literature of religious studies. But in historical retrospect, the essay is also noteworthy for Penner’s treatment of the analytical Marxists. In the main, these contemporaries of Penner regarded the D-N model of scientific explanation as obsolete and did not make use of it in their work.12 But Penner himself remained committed to Hempel, and effectively retrojected elements from their discussions of functionalism back into a Hempelian framework. Penner’s application of Hempel is clear enough: an explanation of ritual that appeals to its consequences—social “maintenance”—“is invalid because it commits the fallacy of affirming the consequent. It asserts that ‘if y then z; z therefore y’” (Penner 1999, 256). What it would take to avoid this fallacy is the assertion that social maintenance requires the performance of the ritual—that the performance of the ritual is not just a sufficient, but also a necessary, condition for social maintenance. And Penner noted, correctly, that serious problems beset the project of arguing that religion, or some specific components of religion, are necessary conditions for satisfying some requirement of society. “No one has succeeded in showing that this is the case,” he remarked; “I do not know how it could be shown to be the case without circular argument” (Penner 1999, 262). A good bit of his exposition of the historical literature revolves around what he regarded as failed attempts to solve this logical problem by various means, including invoking “functional equivalents or a set of social units taken jointly as a class of sufficient conditions” (Penner 1999, 262ff.). Penner’s brief engagement with the analytical Marxists takes place in the last few pages of his essay, embedded within his treatment of the idea that societies are “self-regulating systems.” (Penner 1999, 266). That idea, which Penner discusses as it appears in work by Ernest Nagel and Robert Brown, I have already mentioned: societies are “self-regulating systems” just in case they are organized in ways that make possible causal connections between the actions of persons and the conditions of human actions, such that the consequences of actions themselves shape future actions of the same type in ways that fall into persistent patterns. Penner appropriates from Brown the criticism that this sort of approach needs to specify “what property is being maintained in a steady state,” “what are the internal variables and can they or their effects be measured,” and “what are the external conditions which are assumed to be constant” (Penner 1999, 266ff.). What matters to Penner about this subject is that failure to answer these questions amounts to failure to answer the Hempelian criticism: In a self-regulating model, if the unit is a necessary condition for maintaining the system, we will then have to specify the exact relations which it maintains; since it would be most difficult to uphold the notion that a religious belief system, for example, functions to maintain all the relations in a social system. On the other hand, if we state that the unit is a sufficient condition, we then fail to explain exactly how the religious unit functions to satisfy the requirement. (Penner 1999, 267) It was at this point that Penner turned to Jon Elster, positioning one specific component of their discussions as confirmation of his conclusion regarding self-regulating systems. Elster had argued, in Ulysses and the Sirens, that functional explanations assume a “causal feedback loop” such as figures in the operation of self-regulating systems and that some species of functional explanations “fallaciously infer” the existence of such loops from the available evidence. The following year, Elster wrote a review of the book that inaugurated the analytical Marxist movement, G. A. Cohen’s Karl Marx’s Theory of History: A Defence. Penner cited the following sentences from this review: “I believe that I have seen no other mechanism that comes close to being for sociology what natural selection is for biology, even if this is not, to repeat, to say that it comes very close. Cohen, however, does not even attempt to provide such a mechanism, which is why I believe that his enterprise must be judged a failure” (Elster 1980, 127; cited in Penner 1999, 268ff.). Without offering his own gloss on Elster’s words, Penner noted his agreement, bringing the essay to a close shortly thereafter. I have chosen a different passage from Elster’s review as the epigraph for this article, because that passage is key to my argument that Penner’s problem with functional explanations was different from the problem with which Elster was concerned. But before I offer an account of the context for Elster’s remarks, I want to say how I think Penner understood the passage he cited. Recall that on the deductive-nomological model, explanations are supposed to make it possible for the occurrence of the explanandum to be deductively inferred. It is this requirement that generates Penner’s problem: only the claim that a ritual is a necessary condition for social maintenance would allow the occurrence of the ritual to be deductively inferred from the fact that social maintenance exists. I surmise from his discussion of Elster that Penner understood the role of “causal feedback loops” and “mechanisms” in a specific way: he regarded these as elements of any respectable case for the necessity of religious phenomena for social effects. Thus, the problem with functional explanations from which descriptions of such “mechanisms” are lacking is that they are formally invalid. As we have seen, absent an argument for necessity, it simply does not follow deductively from the existence of the social effect that the religious phenomenon has occurred. But such are not the terms in which Elster understood the problem. ANALYTICAL MARXISTS ON SOCIAL FUNCTIONALISM In this section I want to lay out what I think Penner missed in the discussions of his contemporaries. This will require retrieving some crucial terminology from Cohen—terminology that never, to my knowledge, appears in Penner’s work. A clearer sense of Cohen’s position will make it possible for me to say what Elster’s criticism of Marxian social functionalism amounted to. But after this, I will consider another figure from this period, Richard Miller, who offered what I regard as an effective rejoinder to Elster. It is Miller’s understanding of the real issue with functional explanation that I will take into the final section of this article. The project of Karl Marx’s Theory of History was to construct an intellectually respectable form of Marxist social theory using the tools of analytical philosophy (Cohen originally termed what he was doing “non-bullshit Marxism,” but cooler heads prevailed).13 On Cohen’s reconstruction, functional explanation is central to Marx’s theory of history, such that defense of that theory required a defense of the propriety of functional explanation. Cohen devoted two chapters to this project, one on functional explanation in general and one on functional explanation in Marx’s texts. I want to attend to Cohen’s treatment of the D-N model with some care. On the concluding pages of his chapter on functional explanation in general, Cohen attempted a defense of functional explanation against Hempel’s criticisms. But it is rather odd, and potentially misleading, that he chose to do so. For this engagement with Hempel was an addendum to his own account of functional explanation, which took place on entirely different terms—Cohen’s primary interlocutor in the chapter was not Hempel but Larry Wright (Wright 1973). His remarks in broaching the subject of the D-N model indicate fairly clearly that Cohen was assuming the viability of that model just for the sake of argument.14 In fact, before setting out his defense of functional explanation, he signaled his rejection of Hempel’s model, appealing to one of what was by then a standard set of counterexamples in which claims that satisfy the model’s requirements simply do not explain: “If Hempel were right, the law that mammals exist only if there is oxygen in the atmosphere, in conjunction with the fact that mammals exist, would explain the fact that there is oxygen in the atmosphere; and that consequence of the theory is quite unacceptable” (Cohen 2000, 273).15 It seems clear that Penner considered Cohen’s attempt to rehabilitate functional explanations in Hempel’s terms inadequate; in the same footnote in which he praised Cohen’s book as “brilliant,” he also wrote, “I do not believe that he succeeds in overcoming Hempel’s basic arguments against the logic of [functional] explanations” (Penner 1999, 255 n12). But it is not clear that it mattered to him that Cohen himself rejected the D-N model. To recall my opening remarks, what makes me despair regarding the (counterfactual) possibility of changing Penner’s mind about functional explanation is that his commitment to Hempel seems to have been unshakeable. By the mid-1970s, the shortcomings of the D-N model as an account of scientific explanation in general had been documented by a prodigious literature (see Salmon 1989), which I think explains why Cohen relegated his engagement with Hempel to the equivalent of a footnote. But even in 2003 Penner remarked, “I have yet to find any demonstration that Hempel was wrong in his interpretation of functional explanations… no one has come up with a schema that allows the argument to go through validly” (Penner 2003, 396). I do not know what, beyond awareness of the relevant critical literature, might have prompted him to consider alternatives to Hempel’s model for understanding how functional explanations are supposed to work. Perhaps familiarity with the approach to explanation inaugurated in the late 1960s by Gil Harman, which I will describe shortly, might have done the trick; perhaps not. I want to focus on the part of Cohen’s account that generated Elster’s reservations. Cohen drew attention to a feature of functional explanations that had troubled scholars from the earliest stages of the enterprise and was prominently represented in Merton’s essay on the subject. Functional explanations suppose that there is some “causal feedback mechanism” that closes the circle, as it were, between a phenomenon’s causes and its effects, so that the phenomenon’s disposition to cause the effects in question can itself become part of the explanation of why that phenomenon is actual. And a longstanding complaint against social-functional explanations—one that we have encountered—has been that while they must suppose that there is some such mechanism, they typically have no detailed account of its constituent parts or its operation. Cohen’s contribution to this discussion centers on his notion of an elaboration. Cohen understood functional explanations to imply the existence of some straightforwardly causal explanation of whatever the functional explanation explains. He cited some familiar examples of elaborations from both everyday discourse and the history of science. One way to elaborate a functional claim is to invoke agency, to say that some agent desired that the effect in question be regularly brought about, knew that the phenomenon under consideration has a disposition to bring about that effect, and therefore intentionally caused the phenomenon to occur regularly. Another type of elaboration is appeal to the dynamics of natural selection—to say, for example, that a functional explanation of the existence of the human heart holds in virtue of the straightforwardly causal dynamics that figure in Darwin’s theory. Cohen (somewhat controversially) regarded elaborations of functional explanations as more complete versions rather than replacements, versions that make clear why it is legitimate for these to invoke the notion of function in the first place. Thus, in his view, “Darwin’s theory is not a rival of functional explanation, but, among other things, a compelling account of why functional explanations apply in the biosphere” (Cohen 2000, 271).16 Now it was part of Cohen’s project to develop plausible elaborations for Marxian functional explanations, which he thought Marxists had generally failed to provide. But more importantly for present purposes, he also claimed that social-functional explanations might be respectable without any elaboration. “We can rationally hypothesize functional explanations even when we lack an account which, like Darwin’s, shows how the explanations work,” he argued. And this is particularly important in social science and history, “For functional explanations in those spheres often carry conviction in the absence of elaborative content” (Cohen 2000, 286). To actually have an elaboration of a functional explanation is to “know how it works,” and so is certainly to be desired; but Cohen claimed that actually having an elaboration is not necessary for either advancing or accepting a functional explanation. Cohen’s remarks on the subject are underdeveloped; I think it cannot be denied that he left the discussion in an unsatisfactory state. Elster’s main criticism of Cohen is aimed directly at the issue of the absence of elaborations, or in his terminology descriptions of “mechanisms,” in social-functional explanations. But what I want to focus on is just what problem Elster saw with this absence. Elster accepted the viability of functional explanations when the underlying “general mechanisms” could be specified. “I do not, of course, quarrel with the use of functional explanation in biology,” he wrote in his review. “Here natural selection provides a general mechanism that creates a presumption that beneficial consequences of structure or behaviour explain their own causes. Cohen does not, however, provide any similar mechanism for functional explanation in the social sciences, and therefore his argument cannot succeed” (Elster 1980, 125ff.). What Elster does not say is that what makes functional explanations in biology acceptable is the fact that the theory of natural selection delivers deductive validity. And it should, I think, have caught Penner’s attention that Elster followed Cohen in thinking that agential elaborations also make for viable functional explanations. For on the D-N model, explanations invoking desires, feelings, or decisions are only valid if these bring about the actions they purportedly explain of necessity, and Elster did not commit himself to this controversial position. In fact, in relation to none of the viable general mechanisms he could imagine—“conscious choice, artificial selection, natural selection, reinforcement and absorbing Markov chains”—did Elster raise the issue of deductive validity (Elster 1980, 127). Rather, the issue he saw is indicated above. The availability of a general mechanism creates a presumption that a functional explanation might apply in the domain in question; and some such presumption is needed for a functional explanation to be advanced with justification. Elster offered an analogy to round off his criticism of Cohen: “In the pseudo-sociology of the 19th century there was general agreement that there must be in societies some analogy to the cell, the hard question being whether this should be taken to be the individual, the family, the firm and so on. Today, of course, we simply reject the presumption of there being a social cell; and in the same way we should simply reject the presumption that beneficial consequences explain their causes” (Elster 1980, 127). Elster’s view, then, was not that functional explanations are fallacious unless they involve valid deductive inferences. Rather, his claim was that there is not good reason to think that, in the social realm, there is a set of general mechanisms from which elaborations could be constructed for the sorts of functional explanations championed by Cohen. And I am inclined to think that what Elster regarded as evidence for this claim was just that a long history of attempts to describe such mechanisms had not delivered anything resembling success. I am also inclined to think that what stands behind this thinking on Elster’s part is the longstanding worry I mentioned above, that functional explanation in the social realm involves the fundamental mistake of applying teleological or agential thinking to a domain where it does not belong. Although I agree with Elster that (to use Cohen’s terminology) elaborated social-functional explanations are more complete and therefore preferable to unelaborated ones, I will not give him the last word on this issue. The philosopher who most effectively defended functional explanation on this point is Richard Miller, who followed his Analyzing Marx (1984) with a major treatise on the nature of explanation in the natural and social science, Fact and Method (1987). At greater length than Cohen, Miller developed an alternative to Hempel’s understanding of scientific explanation, which he called the “causal theory of explanation” (Miller 1987, 120).17 And he used this alternative to address the issue of the legitimacy of functional explanations in the social sciences. Miller’s position was that at least some functional explanations are really “shorthand” for longer explanations that make no appeal to functions—functional explanation in biology, for example, he understood in this way. But the functional explanations he considered interesting were social-functional explanations. Like Cohen, he understood that such explanations suppose that a phenomenon’s disposition to cause a particular effect is somehow linked to the proximate causes of that phenomenon, and, like Cohen, he considered this the central challenge to the legitimacy of social-functional explanation. “It might seem,” he remarked, “that explanations appealing to social functions are as mysterious as appeals to inherent goals in organic evolution, unless the appeal to the effects of the functions is, similarly, derivative. And yet, an investigator appealing to an objective social function often seems not to have even a sketch of an underlying non-functional explanation” (Miller 1987, 119). Miller argued, as had Cohen, that “the functional explainer need not have a sketch of the causal links connecting function and phenomenon explained” to advance a respectable functional explanation (Miller 1987, 122). He offered a lengthier defense of this claim than Cohen had. Miller based his defense on an observation about the working methods of natural scientists: these, he claimed, regularly postulate the existence of unknown mechanisms as components of their hypotheses, and the fact that such a postulation may say almost nothing about the details of the mechanism in question does not doom the explanation in which it figures to failure. Such postulations occur with particular frequency when scientific hypotheses cross divisions between levels of description. So, for example, “a metallurgist explaining a fracture as due to crystallization under repeated stress is committed to the existence of molecular interactions, governed by intermolecular forces, connecting stress with crystallization and crystallization with fracture. Still, the metallurgist need have no idea what this molecular story is, and, indeed, until recently, metallurgists had none” (Miller 1987, 122).18 The principle Miller extracts from these considerations is this. Our ordinary practices of assessing scientific explanations permit these explanations to contain “gaps,” areas where a hypothesis claims that some process operates without being able to say very much about how it operates. So, we should be willing to permit functional explanations to remain silent regarding the details of the mechanisms that link dispositions to proximate causes, simply because “such permissiveness is typical of principles linking different levels of causation, throughout the sciences” (Miller 1987, 122). At this point Miller states with clarity and force a crucial point that is absent from Cohen’s discussion, a point that will be a centerpiece of my remarks in the last part of this article. That point is that what legitimates the postulation of an unknown and undescribed mechanism in the natural sciences is that postulation’s explanatory value. Space permits only the most cursory discussion of the apparatus on which this claim depends. But some treatment of the topic is essential for my purposes. In 1965 Gil Harman published an essay titled “The Inference to the Best Explanation” in The Philosophical Review. Harman’s purpose was to argue that explanations involve a form of inference that is distinct from both the inductive and deductive forms that had dominated discussions to that point. The core of his argument was that it is part of our ordinary cognitive practices to take certain claims to be true just because we take them to explain things that are the case: “In making this inference one infers, from the fact that a certain hypothesis would explain the evidence, to the truth of that hypothesis.” But inference to the best explanation involves assessing the value of explanations, preferring as candidates for truth that explanation that we rate most highly: “One infers, from the premise that a given hypothesis would provide a ‘better’ explanation for the evidence than would any other hypothesis, to the conclusion that the given hypothesis is true” (Harman 1965, 89). Harman inaugurated a series of discussions aimed at identifying just what makes some explanations better than others—discussions that have hardly concluded. Miller’s book is a participant in these discussions. Space does not permit an exegesis of Miller on the subject, and neither do my purposes require it. What is important for me to note is that Miller held that what makes for a successful explanation—an explanation that is deserving of assent—is not that it culminates in a valid deductive inference, but that it explains better than its rivals. Here is Miller applying this point to social-functional explanations and thus making the point that was absent in Cohen’s treatment: Virtually every important functional explanation in social anthropology is supported by a case with these outlines: “Here is an enduring pattern of behavior, on the part of intelligent, appropriately informed people. Its existence regularly violates these people’s individual, immediate interests. No strong state apparatus is mobilized to maintain it. Yet it endures. The best explanation of why it endures is that it serves a long-term, large-scale function of preserving certain features of the society.” …The functional explainer is relying on the principle that persistent, complex behavior violating individuals’ immediate interests requires an explanation. And he or she is arguing that this explanatory need cannot be satisfied without appealing to objective social functions. (Miller 1987, 125) In other words, there is good reason to think that the sort of “general mechanisms” to which Elster referred exist, even if little or nothing can be said about them in detail, when their postulation makes for a better explanation than others on offer. I note that Miller’s position is something like the reverse of Elster’s: where Elster thought that descriptions of such mechanisms are required to create the presumption that a social-functional explanation might apply, Miller claimed that the high value of a social-functional explanation will create a presumption that some such mechanisms exists.19 The lesson I draw from the analytical Marxists here is that functionalism’s problem is not one of logical validity: it is a mistake to reject a social-functional explanation just because it cannot deliver a valid deductive inference. Rather, the problem is that the notion of social function presupposes a rather exotic arrangement of causes and effects within a society, and thus a burden of proof attaches to claims to the effect that a phenomenon serves a social function. The one who sees no reason to think that any such arrangement exists (this is how I understand Elster’s position) will not accept social-functional claims as candidates for truth. But there is a source of reasons to think that there are such arrangements, and that is the value of the explanations in which they figure. The viability of the notion of social function itself, then, is inextricably tied to the utility of that notion for the project of explanation in the social sphere. Explanation is not the Achilles’ heel of social functionalism but its foundation. MARTIN’S CRITICAL INTRODUCTION I want to apply this conclusion to a recent entry in the functionalist tradition within religious studies. Craig Martin’s A Critical Introduction to the Study of Religion originated, according to Martin’s description, as an attempt to rewrite parts of Peter Berger’s functionalist classic of 1967, The Sacred Canopy, for a contemporary student audience; and Martin states clearly in his first chapter that functionalism, conjoined to the “hermeneutics of suspicion,” defines his approach to the study of religion (Martin 2014, 8). Martin fuses the older functionalist tradition with more recent work by such authors as Bruce Lincoln and Pierre Bordieu. The premise of the book, then, is that one way the “critical” study of religion can proceed at present is by way of social functionalism. At least on the surface, there is no mystery about what Penner would have thought of A Critical Introduction. I think it likely that he would have extended to Martin the verdict he issued (in Rediscovering the Buddha) on the idea that Mauss’s analysis of gift-giving should be replaced by an account of “how gifts function as a variable for the satisfaction of needs in a particular political/economic system”: “we know that this theory, known as functionalism, has been thoroughly dismantled as logically invalid” (Penner 2009, 256ff.). It should be apparent that such will not be my own judgment. But what is particularly interesting about the contrast between the Hempelian and post-Hempelian reconstructions of the notion of social function is that the latter does more than simply clear functionalism of an accusation levied by the former. On the post-Hempelian reconstruction, functionalism has a problem that Hempel’s reconstruction does not capture. I have hinted at this problem above; to bring it fully into view, I want to sketch a different, and less dismissive, attitude that Penner might have taken to Martin’s book. One noteworthy feature of A Critical Introduction is that the idea that claims about social functions are explanations plays only a marginal role in the book. That idea emerges into the clearest focus in Martin’s first chapter. But that focus is not a sharp one (I will provide some support for this claim in a moment), and after this first chapter the topic of explanation basically disappears. The theoretically important material in the remainder of the book is presented as claims about “how society works,” “how religion works,” and how religious and social phenomena can be used by persons seeking to promote their own interests or the interests of their class. Arguably, claims of these sorts are connected to the project of explaining religion. But as there are no places in those chapters where Martin returns to the theme of explanation, the connections are not described. I do want to document the commitment to social-functional explanation that appears in Martin’s first chapter. He introduces the notion by way of references to Freud: “Why do religions exist? For Freud, the answer is simple: because something is necessary to alleviate the friction created by social repression. Religion in society is like the oil in an engine that keeps it from overheating”; “Freud would say that all of those things are illusions people believe in simply because they are necessary to relieve psychological pressure and thereby keep society running” (Martin 2014, 9). And shortly afterward Martin commits himself to functionalism as a species of explanation: alongside other projects, “we will look for a functionalist explanation for the data under consideration—if this myth is not true, for instance, does it continue to be told because it serves some sort of social function for the community that tells it?” (Martin 2014, 11). Since Martin does not discuss functional explanation in the abstract, the examples he provides are the book’s only source of information regarding how he understands the method. Take, for example, his treatment of an observation by the anthropologist Max Gluckman: According to the Barotse, if an individual does not pay the doctor then the medicine given for the illness will magically make the illness return. However, from an outsider’s perspective we can utilize a hermeneutic of suspicion and look for a functionalist explanation: if we assume that the claim is, in fact, not true, why might people in this community make such a claim? This one is easy to answer: if people believe this they will be more likely to pay their doctor’s bills for fear of getting sicker. This is a myth that serves the function of ensuring that doctors will be paid. (Martin 2014, 12, italics original) Here Martin positions a claim about the consequences of a myth’s acceptance as an answer to the question of why people continue to assert its truth. The absence of any “causal feedback mechanism” that would turn those consequences into causes is particularly evident in this example; Martin offers neither any account of such a mechanism nor any indication that his functional explanation presupposes one. A similar softness of focus is on display in his subsequent rejection of the “strong functionalist” position that “society works like a machine, and that religion is an essential part that has one and only one function in the machine,” in preference for the “weak functionalist” position that holds that “religious traditions can serve various social functions but without completely reducing religion to those social functions.” Martin positions “weak functionalism” as a position that makes it possible to think that “if it turns out that the supernatural claims under consideration are true . . . our functional explanation can still be true”—so, for example, that “Even if it is true that the Barotse doctor’s medicine will make the patient sick if she does not pay, it is still true that this belief functions to ensure that the doctor will be paid” (Martin 2014, 14, italics original). The thought here seems to be that the social consequences of a true claim explain why people assert it. It is not clear whether Martin means to suggest that in this case the functional explanation is the only correct one, such that the truth of the myth has no explanatory value in relation to its assertion, or (which is a more interesting position) that multiple explanations of the same social practice can be true—that the Barotse assert the myth because of its truth and they assert it because of its social consequences. But a choice between these positions does not arise, because Martin does not say how the myth’s social effects explain why certain of the Barotse assert its truth. The conclusion I draw from these observations, together with the disappearance of the theme of explanation following the first chapter, is that explanation is not central to social functionalism as it is presented in A Critical Introduction, its stated commitment to functional explanation notwithstanding. The book is clearly committed to the project of identifying the social functions that religious phenomena can serve. But it is not consistently concerned with using claims about social functions to explain religion.20 I think this aspect of the book’s character might provide grounds for something other than outright dismissal by the devotee of Hempel’s D-N model, for Hempel’s criticism was centered on the dynamics of functional explanation. So far as I can see, the model provides no grounds for dismissing simple claims to the effect that a phenomenon has a social function: when such claims are not presented as explanations, it seems to me that the devotee of Hempel should have no objection to the possibility of their being true. In fact, Hempel himself left the door open to the truth of social-functional claims at the conclusion of his canonical essay, writing that when the problems of social-functional explanation are fully in view, “functional analysis” might perfectly well persist as “a program of inquiry aimed at determining the respects and the degrees in which various systems are self-regulating” (Hempel 1965, 330).21 Imagining Penner at his most charitable, then, I can envision him arguing that a fundamental theoretical confusion detracts from only the first part of A Critical Introduction, and that the bulk of the book, where the notion of social function is not clearly tied to explanation, is thereby clear of this defect. But here my position motivates a different reservation. I have argued that the issue with the notion of social functions for the post-Hempelians has to do with reasons for thinking that there exist mechanisms embedded within the social order that make consequences into causes. And I agree with Miller that it is reasonable to accept that some such mechanism exists when positing its existence makes for the best explanation of the phenomena in question. On this view, the absence of tidy accounts of causal feedback mechanisms in those sections of Martin’s book where explanation is in view is not in itself a flaw: it is to be expected that such mechanisms will be, as it were, invisible to the naked eye, and so knowable only through postulation. But what is grounds for criticism is the soft focus on the theme of explanation and absence of any engagement with the question of whether social-functional explanations are better explanations than the alternatives. For if claims to the effect that a religious phenomenon serves a social function is not part of a “best explanation,” it is not clear what reason there is to think that the causal feedback mechanisms that convert effects into causes exist, and thus it is not clear what reason there is to think that there is any functioning going on in the first place. Let me apply these reflections to the example of the Barotse myth. Consider the question of whether one should accept Martin’s functional explanation, that certain of the Barotse repeat the myth because a person who believes it to be true will be more likely to pay their doctors. On my position, one would have reason to believe that the explanation is true if it is better than the available alternatives. I see nothing standing in the way of such a claim, but Martin does not make it—he does not claim that the social-functional explanation is better than the available alternatives, only that it is a possibly true explanation. The result is that this case offers an example of social-functional explanation, but not an example of a reason to think that such an explanation is true. So far as I can discern, this state of affairs holds throughout the book. The disappearance of the theme of explanation after the first chapter means the absence in the remainder of the book of engagement with the question of whether there is any reason to think that religion serves social functions. I do not doubt that it would be possible for the theoretical devices that Martin appropriates from Berger, Lincoln, Bordieu, and others—legitimation, authority, habitus, and so on—to be presented as resources for the explanation of religion; in fact, it makes considerable sense to me to think of these devices as possible components of the “feedback mechanisms” that social-functional explanations entail. And one possible application of such a presentation might be to a case for the viability of social functionalism in the study of religion, on the grounds that it can generate explanations that are at least as good as the alternatives. But as it stands, the book contains no such presentation and makes no such case; and, while it makes clear that a variety of functional claims can be and have been made regarding religious phenomena, it never really says why anyone should think any of these are true.22 So where Penner’s understanding of functional explanation would likely motivate the outright dismissal of Martin’s approach but might be compatible with a grudging acceptance, my own position brings into view a different problem, which is an absence in the book of arguments for the viability of social functionalism as a methodology for the study of religion. CONCLUSION I think it undeniable that social functionalism is now but a shadow of what it once was in our field; and my impression is that for the most part, where the language of social functionalism hangs on in our literature, its connection to the project of explaining religion is tenuous at best. It is possible that Hempel’s criticism of functional explanation, and Penner’s work in disseminating this criticism, has had something to do with this, because so far as I can see Penner’s position more or less entailed that scholars of religion should avoid functional explanations altogether unless they were in a position to argue that religious phenomena are necessary for the social effects they produce. But my disagreement with Penner motivates a different set of recommendations for those drawn to social-functionalist thinking. My first recommendation is that scholars who are inclined to argue that religious phenomena serve social functions should cast their claims precisely as explanations, whether of the religious phenomena themselves or of other actual phenomena,23 and should not feel an obligation to satisfy the requirements of the covering-law model. To be respectable, social-functionalist explanations do not need to be deductively valid; they just need to be, plausibly, at least as good as the available alternatives. And my second recommendation follows closely behind the first: scholars who, for whatever reason, are unwilling to put their social-functionalist claims forward as not just explanations but very good explanations should either resign themselves to making unsupported assertions or avoid functionalist language altogether. These recommendations are motivated by a rather cranky (but hardly original) observation on my part. It seems to me that in the historical literature, the link between social functionalism and explanation was almost never as firm as it should have been, and that where social functionalism persists, it often does so in the form of unsupported assertions and half-formed insinuations (had I chanced to meet Penner, a good opening gambit before trying to change his mind might have been to commiserate with him about this). I think it would be an improvement to the literature of our field if these were either converted into genuine attempts at explanation or eliminated. Frankly, it is not clear to me either that the waning of social functionalism has been a great loss for the study of religion or that its complete disappearance would be. If at one time prominent scholars hoped that social functionalism might be the key to explaining religion, it seems to me the foundations of this hope were always somewhat mysterious, and the results of a century’s work suggest they were never very firm.24 And (although this is a topic for another day) if social functionalism has been useful for a different project, that of criticizing religion, its utility can be overstated. Where the question is what is to be made of religion in relation to ethical or political values—knowing, for example, whether religion abets or obstructs various kinds of oppression—it can be helpful to know about the effects that religious phenomena have, the interests they serve, and the uses that are made of them. But claims about such matters are conceptually independent of claims about social function, and one can have perfectly good reason to think that such claims are true without thinking that they figure in the operation of any self-regulating systems.25 Whether there are any such systems is still, I take it, an open question and an interesting one. But it is a question that matters for fewer purposes than some have thought. Radcliffe-Brown’s attention was captured by Robert Lowie’s suspicion that social-functionalist thinking amounted to an unwarranted imposition of order on the chaos of collective life. I will not take the time to translate Lowie’s words into my own idiom, but I affirm the spirit of the remark with which he brought his Primitive Society to a close. To that planless hodge-podge, that thing of shreds and patches called civilization, its historian can no longer yield superstitious reverence. He will realize better than others the obstacles to infusing design into the amorphous product; but in thought at least he will not grovel before it in fatalistic acquiescence but dream of a rational scheme to supplant the chaotic jumble. (Lowie 1920, 441) Footnotes 1 An early version of the arguments of this article was presented in November 2014 at a session of the American Academy of Religion Philosophy of Religion unit. I am grateful to session participants for useful feedback and also to Nancy Frankenberry, Craig Martin, Jeffrey Stout, and two anonymous reviewers for the JAAR for comments on various drafts. 2 The idea that social functionalism hangs on this sort of analysis was not universally accepted (see, for example, Davis 1959, 768). 3 “[T]he point of functional explanation has normally been to show that there are ‘reasons’ for the existence and continuance of social institutions that are quite distinct from the reasons actors might have for whatever they do” (Giddens 1987, 68). 4 Nicos Mouzelis, at least, recalls this feature of functionalism in its classical form: “When theoretically sophisticated functionalists talk about social needs or functional requirements, they simply mean conditions of existence” (Mouzelis 1995, 132). 5 In Rethinking Social Theory, Roger Sibeon claims that one of the cardinal sins of sociological thinking is “functional teleology,” which he describes as “an invalid form of analysis involving attempts to explain the causes of social phenomena in terms of their effects, where ‘effects’ refers to outcomes or consequences viewed as performances of ‘functions.’” The problem with this mode of analysis is that “in the absence of a demonstration of intentional and successful planning by actors somewhere, sometimes, it is a teleological fallacy to attempt to explain the causes of phenomena in terms of their effects.” It is not clear to me whether Sibeon is unaware that some functionalists understood functional explanations to involve no invocations of teleology or whether he thinks that the two cannot be disentangled. He is on firm ground, however, in claiming that many invocations of functionalism in the literature seem not to avoid teleological thinking (Sibeon 2004, 6). 6 This supposition explains the following passage by Radcliffe-Brown: “The scanty information I have been able to gather suggests that the lineage and joint-family organization of some parts of India is losing something of its former strength and solidarity and that what we should expect as the inevitable accompaniment of this, a weakening of the cult of ancestors, is also taking place” (Radcliffe-Brown 1945, 38). In other words, the functionalist position predicts that religious phenomena will disappear once the effect that it is their function to produce is no longer socially necessary. 7 “The function of a particular social usage is the contribution it makes to the total social life as the functioning of the total social system. Such a view implies that a social system…has a certain kind of unity, which we may speak of as a functional unity. We may define it as a condition in which all parts of the social system work together with a sufficient degree of harmony or internal consistency, i.e. without producing persistent conflicts which can neither be resolved nor regulated. This idea of the functional unity of a social system is, of course, a hypothesis. But it is one which, to the functionalist, it seems worthwhile to test by systematic examination of the facts” (Radcliffe-Brown 1935, 181). 8 Hempel noted his agreement with this result of Merton’s analysis. He described as “the general hypothesis of self-regulation” the idea that “within certain limits of tolerance or adaptability, a system of the kind under analysis will—either invariably or with high probability—satisfy, by developing appropriate traits, the various functional requirements (necessary conditions for its continued adequate operation) that may arise from changes in its internal state or in its environment” (Hempel 1965, 317). Hempel had reservations concerning the extent to which claims as to the self-regulating character of particular social systems could be empirically tested; but more importantly, he regarded it as crucially important that it not be forgotten by social thinkers that the hypothesis of self-regulation is a hypothesis. A central task to which social-functional analysis could be applied, for him, was to determine whether there are any self-regulating social systems—“to ascertain to what extent such phenomena can be found, and can be represented by corresponding laws” (Hempel 1965, 325). 9 “Some functionalists have seemed unwilling to discipline their language or test their propositions, taking instead the easy path of verbal tapestry. As a result, sociologists who have learned the techniques of empirical research come to feel that functionalism is a crank method, and they are encouraged in this by functionalists themselves who say that they are engaged in non-causal analysis, whatever that is” (Davis 1959, 768). 10 Note that there have been such claims in the literature. For example, Bronisław Malinowski suggests: “Thus magic fulfills an indispensable function within culture. It satisfies a definite need which cannot be satisfied by any other factors of primitive civilization” (cited in Merton 1957, 34). 11 Other publications of Penner’s that display his commitment to Hempel’s model are Penner 1989 (anthologized as Penner 1999) and Penner 2003. 12 See Salmon 1989 for an overview of the fortunes of Hempel’s model over the course of the period in question. 13 Cohen recounted the history of analytical Marxism in his Introduction to the 2000 edition of Karl Marx’s Theory of History. While eschewing the “more aggressive” designation of “non-bullshit Marxism,” he did not entirely repent of its spirit: “In fact, there exists Marxism which is neither analytical nor bullshit, but, once such (as we may designate it) pre-analytical Marxism encounters analytical Marxism, then it must either become analytical or become bullshit” (Cohen 2000, xxvi). 14 “A consequence explanation relates to a consequence law in whatever way explanations relate to law. It is not our task to say what that relation is. But it may be clarifying to exhibit how consequence laws figure in explanations on the supposition that the correct answer to the question here suspended is Hempel’s.” Cohen’s mammal example is drawn from a paragraph on the same page in which his remarks “trespass into assessment of the D-N model itself” (Cohen 2000, 272ff.). 15 On “famous counterexamples” to Hempel, including Bromberger’s flagpole example, see Salmon 1989, 47–50. 16 The general principle operative here is that “the sceptic misconstrues as various alternatives to functional explanation what are in fact various more complete forms functional explanations may take” (Cohen 2000, 271). 17 As Miller describes it, in such a theory “an unanalyzed notion of causation is used to describe how a successful explanation works.” The novelty of this approach, in other words, is that it begins with a broadly inductive understanding of how we assess the goodness of explanations, and then allows anything to be termed a “cause” that could figure in successful explanations (Miller 1987, 120). 18 Miller credits Garfinkel 1981 for this example. 19 To be sure, there are mysteries here, such as how an explanation containing the sort of gaps that Miller describes could possibly be better than an explanation containing no gaps at all. Here I have space only for some trivial examples that indicate, I hope, that there is no simple answer to this question. In some cases, we are perfectly happy to accept “gappy” explanations, as for example when I accept that you hit me because you were angry with me without having any idea how exactly mind-body causality works. And in other cases, a gappy explanation might be better than either no explanation at all or a highly implausible explanation; so, for example, an unelaborated social-functional explanation might be better than “gapless” explanations involving phlogiston, fairies, or Geist. 20 Places where Martin clearly positions the results of his analyses as explanations are not entirely absent in the book, but they are quite rare and often not clearly functionalist in character. Here is one example, drawn from Martin’s discussion of the origins of “Hinduism”: “Why posit a unified identity where none exists? . . . In the end, the fabrication of a timeless Hindu identity appears to have been a power play born out of racial, ethnic, and cultural hatred of some Indians towards Muslims in their midst” (Martin 2014, 156f). 21 See also Mouzelis: “As Merton pointed out long ago, once it is made clear that the functional or dysfunctional consequences exerted by a social item on the social whole must not be used as a cause explaining how this item came about, then it is perfectly legitimate to ask questions about whether or not this item strengthens or weakens the social whole” (Mouzelis 1995, 132). 22 Having advanced this criticism of A Critical Introduction, it is only fair to note that as it is a text written for undergraduates, there are limits to the extent to which such a book can reasonably be expected to engage some of the more elusive methodological issues its proposals raise. That being said, I do think that the book represents something of a missed opportunity to make undergraduates aware of some of the complexities of functionalism as a method of investigation. 23 In a 1975 essay Roger Cummins attempted a defense of functionalism in part by arguing that Hempel and others had misunderstood what functional explanations explain. What they explain, according to Cummins, is not the existence of the phenomena that function, but rather the existence of whatever those phenomena function to make actual. So, for example, “appeal to the function of the contractile vacuole in certain protozoans explains how these organisms are able to keep from exploding in fresh water,” rather than explaining why those organisms have contractile vacuoles in the first place (Cummins 1975, 751; I am grateful to Nancy Frankenberry for this reference). I do not follow Cummins in thinking that functional explanations must be construed in this way. But certainly, one can explain a phenomenon—a social condition, say—by postulating something that functions to make it actual. When that something is a religious phenomenon, such an explanation will generate a reason to think that phenomenon serves a social function if that explanation is better than the available alternatives. Thus, it is possible to argue that a religious phenomenon serves a social function without at the same time attempting to explain that phenomenon. 24 In making this remark, I have particularly in mind the exchange between Roy Rappaport and Jonathan Friedman regarding Rappaport’s Pigs for the Ancestors (Rappaport 1984). The massive size and careful argumentation of the work notwithstanding, Friedman was simply unconvinced that Rappaport had shown that pig-slaughtering rituals among the Maring-speaking Tsembaga functioned to regulate the Tsembaga relation to their environment. “The cybernetic analysis of Maring society is unnecessary to account for phenomena that are due to more mundane mechanisms,” Friedman remarked. “The fact that the slaughter of pigs is ritualised has, I think, to do with the nature of Maring as well as other tribal social relations and not with a regulatory function” (Friedman 1979, 258). 25 In commenting on a draft of this article, Martin has defined weak functionalism as “a scholarly approach [that] demonstrates how elements of culture are utilized by subjects—either intentionally or unintentionally—in ways that produce social effects (through legitimation, mystification, naturalization, and so on) that serve the interests of some subject positions in a society over others” (personal communication). I think that my concluding remarks are in accord with this definition, as it does not seem to me that it makes use of the notion of social function. REFERENCES Cohen G. A . 2000 . Karl Marx’s Theory of History: A Defence . Expanded ed. Princeton, NJ : Princeton University Press . Cummins Robert . 1975 . “ Functional Analysis .” The Journal of Philosophy 20 : 741 – 65 . Google Scholar Crossref Search ADS Davis Kingsley . 1959 . “ The Myth of Functional Analysis as a Special Method in Sociology and Anthropology .” American Sociological Review 24 : 757 – 72 . Google Scholar Crossref Search ADS Durkheim Emile . [1912] 1995 . The Elementary Forms of Religious Life . Translated by Karen E. Fields . New York : The Free Press . ——. [1895] 2013 . Rules of Sociological Method and Selected Texts on Sociology and its Method . Basingstoke : Palgrave MacMillam . Elster Jon . 1980 . “ Cohen on Marx’s Theory of History .” Political Studies 28 : 121 – 28 . Google Scholar Crossref Search ADS Friedman Jonathan . 1979 . “Hegelian Ecology: Between Rousseau and the World Spirit .” In Social and Ecological Systems , edited by P. C. Burnham and R. F. Ellen , 253 – 70 . New York : Academic Press . Garfinkel Alan . 1981 . Forms of Explanation: Rethinking the Questions in Social Theory . New Haven, CT : Yale University Press . Giddens Anthony . 1971 . Capitalism and Modern Social Theory . Cambridge : Cambridge University Press . ——. 1987 . Social Theory and Modern Sociology . Stanford, CA : Stanford University Press . Godlove Terry . 2005 . Teaching Durkheim . New York : Oxford University Press . Harman Gilbert . 1965 . “ The Inference to the Best Explanation .” The Philosophical Review 74 : 88 – 95 . Google Scholar Crossref Search ADS Hempel Carl . 1965 . “ The Logic of Functional Analysis .” In Aspects of Scientific Explanation and Other Essays in the Philosophy of Science . New York : The Free Press . Lowie Robert H . 1920 . Primitive Society . New York : Boni and Liveright . Martin Craig . 2012 . A Critical Introduction to the Study of Religion . Bristol, CT : Equinox . McCauley Robert N. , and E. Thomas Lawson . 1984 . “ Functionalism Reconsidered .” History of Religions 23 : 372 – 81 . Google Scholar Crossref Search ADS Merton Robert . 1957 . “ Manifest and Latent Functions .” In Social Theory and Social Structure , edited by Robert Merton , 19 – 84 . New York : Free Press . Miller Richard . 1984 . Analyzing Marx: Morality, Power and History . Princeton, NJ : Princeton University Press . ——. 1987 . Fact and Method: Explanation, Confirmation and Reality in the Natural and the Social Sciences . Princeton, NJ : Princeton University Press . Mouzelis Nicos . 1995 . Sociological Theory: What Went Wrong? Diagnosis and Remedies . London : Routledge . Penner Hans . 1971 . “ The Poverty of Functionalism .” History of Religions 11 : 91 – 97 . Google Scholar Crossref Search ADS ——. 1989 . Impasse and Resolution: A Critique of the Study of Religion . New York : Peter Lang . ——. 1999 . “ What’s Wrong with Functional Explanations ?” In Language, Truth, and Religious Belief , edited by Nancy Frankenberry and Hans Penner , 246 – 70 . Atlanta : Scholars Press . ——. 2003 . “ On Self-Regulating Systems, Cannibals, and Cogs that Turn No Wheels: A Response to Joel Sweek .” Method & Theory in the Study of Religion 15 : 390 – 408 . Crossref Search ADS ——. 2009 . Rediscovering the Buddha: The Legends and Their Interpretations . Oxford : Oxford University Press . Radcliffe-Brown A. R . 1935 . “ On the Concept of Function in Social Science .” American Anthropologist , New Series 37 : 394 – 402 . Google Scholar Crossref Search ADS ——. 1945 . “ Religion and Society .” Journal of The Royal Anthropological Institute of Great Britain and Ireland 75 : 33 – 43 . Crossref Search ADS Rappaport Roy . 1984 . Pigs for the Ancestors: Ritual in the Ecology of a New Guinea People . 2nd ed . New Haven, CT : Yale University Press . Salmon Wesley . 1989 . Four Decades of Scientific Explanation . Minneapolis : University of Minnesota Press . Sibeon Roger . 2004 . Rethinking Social Theory . London : Sage Publications . Turner Jonathan H. , and Alexandra Maryanski . 1979 . Functionalism . Melo Park, CA : The Benjamin/Cummings Publishing Co . Wright Larry . 1973 . “ Functions .” The Philosophical Review 82 : 139 – 68 . Google Scholar Crossref Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of the American Academy of Religion. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Journal

Journal of the American Academy of ReligionOxford University Press

Published: Sep 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off