[276]

Michel Verdon

Anthropologue, département d’anthropologie,
Université de Montréal

“Midwife or Toad ?
Philosophy and the Social Sciences.” *

In revue Philosophy of the Social Sciences, 15. 1985, pp. 53-63. A Sage Publication.

Introduction [53]

Defining a class of function statements [54]

A tautological and superfluous mechanism [57]

Explanatory or conceptual models ? [61]

References [63]

Introduction

The philosopher of social sciences is like a grammarian ; reading beyond individual performances, he will lay bare the implicit rules guiding the practice of social scientists and 'produce ... a reconstructed grammar' (p. x). This places him in a privileged position. Because social sciences are traumatized (i.e., regarded with low esteem) they are preparadigmatic and the philosopher-grammarian will be able to act as midwife in the creation of a consensual social science (i.e., a paradigmatic one). The book, needless to say, attempts to vindicate this presumption.

The demonstration opens up with the claim that a new (consensual) paradigm can only emerge from an understanding of the 'deep structure' of explanatory models. An explanation presupposes both causal linkages between two or more phenomena, and reference to a mechanism recognized as plausible by the scientific community to 'explain' the causal connection(s). By studying these various mechanisms—or, to use van Parijs's terminology, by making an inventory of the various 'patterns of intelligibility' accepted by a community of social scientists—the philosopher can generate a morphology of explanations which will guide him to the discovery of the 'deep structure of the social sciences' and enable him to act as a midwife in the delivery of a new paradigm.

Van Parijs then recognizes that 'there is one pattern which is of paramount importance' in the social sciences. He calls it 'the action pattern' and defines it 'by the class of mechanisms which generate an action on the basis of the subjective situation of an agent' (p. 19). Explanations of this kind are actional explanations and the key question is : are actional explanations the only legitimate type of explanation in the social sciences, or are there other 'patterns of intelligibility' equally plausible to the community of social scientists ? To answer this question van Parijs turns his attention to functional explanations. If it can be proved that they also constitute legitimate patterns of intelligibility, their understanding will further the philosopher's search for the 'deep structure'.

How are we then to define function statements, and what mechanism (or pattern of intelligibility) transforms function statements into functional explanations') Van Parijs regards function explanations as consequence explanations which 'causally ascribe the presence of an item, not to its consequences, but to a dispositional property of the situation in which it appears' (p. 26). This definition, however, does not eliminate some consequence statements which cannot be accepted as functional explanations, so that we must further distingish 'hard core' function statements from other consequence statements.

This, van Parijs achieves by noting that 'hard core' function statements relate to persistent features, the consequences of which are good (and the feature [54] persisting because the consequences are good) (p. 45). 'Hard core' function statements thus presuppose a mechanism of optimization, typically found in evolutionary theories. Looking first at evolutionary mechanisms based on natural selection the author concludes that they, like cybernetic mechanisms, can only account for a tiny fraction of functional explanations. Another evolutionary mechanism, that of reinforcement (shortened to 'R-evolutionary mechanism') seems more promising. In the language of behaviourist psychology from which it is borrowed, reinforcement corresponds, 'in its most elementary form,... to operant conditioning' (p. 96), a process by which new habits are formed because of enhanced satisfaction. Optimality in the R-evolutionary theory is therefore to be assessed in terms of chances of satisfaction, and not chances of reproduction.

Using three-dimensional graphics and catastrophe theory van Parijs then reviews some theories in linguistics, social anthropology and marxist sociology to demonstrate that they find their ultimate pattern of intelligibility in terms of R-evolutionary mechanisms. The R-evolutionary mechanism thus legitimates functional explanations. Functional explanations thus belong to the 'deep structure' of social sciences and, by understanding the mechanism upon which they are founded, the philosopher of social sciences can now act as midwife in the emergence of a paradigm.

Van Parijs has written a coherent book, confident that he is addressing the right problems in the right manner. To the non-initiated, he also seems to master all the intricacies of evolutionary theory, in its most abstract and formalized expressions. To the philosopher and student of social sciences the book may thus appear as extremely convincing and, for this reason, it deserves fuller comments than is normally allocated in a book review. When scrutinized through its internal logic and through the social phenomena it seeks to explain, the book may lose much of its magic. Let us see what remains then of van Parijs's endeavour.

DEFINING A CLASS OF FUNCTION STATEMENTS

In van Parijs's own words any explanation presupposes at least two classes, namely the class of phenomena causally linked in the theory, and a contrast class. In his book, the class of function statements is contrasted to that of actional statements. [1] To define his class of function statements van Parijs adduces a number of examples, among which the following two figure prominently : the physiological statement that 'the function of the heart... is to circulate blood' (p. 34) and the anthropological one that 'the ceremonial customs of the Andaman Islanders [are explained] by their function of maintaining feelings of dependence on the social group and of attachment to it' (p. 35). Both statements, according to van Parijs, imply that the consequence (of the heartbeat, or ritual) is the welfare of the whole (organism or society).

Upon closer examination, however, we may be forced to disagree. What do physiologists mean by saying that 'the function of the heart is to circulate blood' ? In my opinion, this is but an euphemistic way of speaking of the heart's work, of the activity it performs. Anatomical research revealed that the heart is a muscle which, by expanding and contracting, pumps the blood that passes through its [55] ventricles. Before concluding that the blood thus pumped was circulated, Harvey had to demonstrate anatomically and experimentally that the blood could not seep from one ventricle to the other through the septum, that valves in the veins and arteries constrained the blood from flowing in any direction, and so on. This physiological statement thus presupposes (a) the anatomical identification of a given structure, the heart, and (b) the discovery of what the heart does. To constitute an explanation this causal connection between the heart and the circulation of blood required a 'pattern of intelligibility' which Harvey did provide and which is purely mechanical (based on simple principles of physics). The statement is therefore an explanation because of previous experimental work which disclosed a physical mechanism, and not because it implies a 'welfare view', let alone a 'cybernetic explanation'. No one would deny that, should the heart stop circulating blood (stop 'functioning', i.e., stop doing its work) the organism would die. But to use this consequence to assert that any physiological function statement about the heart implies that the heart works to keep the organism alive and well is logically unwarranted. In fact, a healthy heart may go on circulating blood in a sick body.

We can thus safely conclude that the pattern of intelligibility behind the physiological statement is neither a 'welfare view' nor a cybernetic mechanism, but a purely mechanical one arrived at through experimentation. Further along, when van Parijs needs to disprove the idea that cybernetic mechanisms could legitimate functional explanations, he no longer uses the example of the heart but rather invokes sweating and shivering, two homeostatic processes to bring back body temperature to its normal state. [2]

Now, let us examine Radcliffe-Brown's theory of ritual, classified together with the physiological example as a typical function statement. If the physiological statement amounts to saying that 'the work of the heart is...' we must then ask, 'what work does a ritual do ?' A ritual may certainly have many social or psychological implications but, as such, it does not do anything since, by its very definition, it is an activity, namely the ritualized worshipping of supernatural entities. In the performance of this activity individuals get together and one may well assume that this interaction produces a psychological state of 'togetherness' which enhances social solidarity and cohesion. There is, however, a major difference.

In the physiological example we assumed that anatomical research had already identified an organ, the heart ; in the anthropological example what would then be its equivalent ? Individuals ? Groups ? Let us leave the question open, and state simply that individuals get together to perform ritualized worshipping. If we claim that the function of the heart is to circulate blood, we must also assume that this activity (circulation of blood) is not performed indiscriminately by any other organ and that it comes to form, together with considerations Delating to its [56] anatomical structure, part of the very physiological definition of the heart. To complete the equivalence, can we say that 'ritualized worshipping' is part of the very definition of individuals or groups ? Obviously not. But individuals or groups may be complex organs, like the liver, and perform many different activities. We could then say : ritualized worshipping is only one function of individuals, or groups ; their other functions include adjudication, production, and so on. This, unfortunately, does not make sense either. The heart performs activities which the liver cannot perform, and vice versa, but any group, or adult individual, can perform all the activities mentioned. Insofar as there is a division of labour in society (as in physiological functions) we should rather say : adjudication, or ritualized worshipping, is the function of some individuals, or groups. These individuals or groups thus distinguish themselves as separate 'organizational' entities (or 'organs') through the activity they perform. What, then, does Radcliffe-Brown's statement amount to ?

It implies, first, that in the performance of a given activity (ritual) some individuals distinguish themselves as a 'ritual group'. The function (or work) of this ritual group is therefore ritual. The statement in this form is absurdly tautological but we can only expect as much if the work defines the organ. To by-pass this tautology we would have to introduce other criteria in the definition of group. We might then find that groups defined through filiation (lineages, in the conventional literature) perform these rituals. We could then state : the function of lineages in this or that society is to perform rituals. Already, we would have a statement more closely equivalent to the physiological one, but this is not the kind of statement that van Parijs has selected, because it would not meet his definition of a 'hard-core' function statement.

Having avoided a first tautology, we can then ask : what are the implications of the performance of this activity (ritual) ? By bringing together individuals in ritualized worship, claims Radcliffe-Brown, collective rituals heighten their sense of 'togetherness' which, in turn, serves to maintain social cohesion. This statement does imply a 'welfare view' but no cybernetic mechanism whatsoever. There is no indication of any kind to the effect that rituals are precipitated by a social crisis (although some rituals are). [3]

In this example, a causal connection is made between a given activity (ritual) and a collective feeling (sense of togetherness) working as a source of social solidarity. The mechanism is therefore psychological. It could be a plausible explanation if it excluded other equally good consequences, or if a test existed to enable us to decide among competing explanations. But none of these conditions are met since we can hold a radically different view with equal plausibility. By bringing together many participants, collective rituals do increase interaction and, with it, the possibility (and actuality) of friction and conflict. One could thus surmise that people hold collective rituals to convince themselves periodically that, after all, they are much better off living isolated lives, dispersed in their small communities. One often hears that holidays make you appreciate your home routine ; why not assume that the disruption caused by collective rituals makes those who attend them appreciate their everyday routine better ? Both propositions have good consequences, are equally plausible but equally untestable. One of them, however (Radcliffe-Brown's), is questionable at a much deeper level.

Indeed, the individuals who left their dispersed communities to congregate in collective rituals already possessed enough sense of togetherness and social [57] solidarity to attend the ritual. Without a sense of solidarity, they would not have come in the first place. What do we then gain by claiming that the ritual 'heightens their sense of solidarity', since the very definition of ritual presupposes this solidarity ? The so-called 'function' of the ritual is therefore precontained in its definition ; in this respect, functional explanations in terms of social cohesion are tautological.

The physiological and the anthropological function statements thus have little in common. The latter is tautological, whereas the first one is not at all. In fact, the physiological equivalent of the anthropological statement would probably read like this : 'the function of the heart is first to pump blood so that, by the momentum imparted to the blood, the heart receives further impulses to pump the blood and therefore keeps it circulating, which circulation keeps the heart pumping...' ! Moreover, the physiological statement refers to a mechanical explanation which has been demonstrated experimentally whereas the anthropological statement supposes untestable psychological mechanisms. Finally, the anthropological statement is premissed on a 'welfare view' whereas the physiological statement is not, but neither of the two statements presupposes reference to cybernetic mechanisms.

We can thus safely conclude that van Parijs grounds his ambitious attempt to legitimate functional explanations on a class of function statements which have nothing in common. This, however, is only the least of his problems !

A TAUTOLOGICAL AND SUPERFLUOUS MECHANISM

To discard consequence statements which are not function statements, van Parijs introduces more strictures in the definition of his class of hard core statements : '... what is decisive in making the explanation of the presence of a persistent feature by its consequence a function explanation, is the fact that the consequences are good ... according to some criterion' (p. 46). This leads him to conclude :

If what makes a function statement explanatory has something to do with what makes it a function statement, these two suggestions may provide important elements of the underlying mechanism we want to characterize. [p. 44.]

In Radcliffe-Brown's theory of rituals, the mechanism explaining the causal connection between ritual and social cohesion was a psychological mechanism already precontained in the very definition of the phenomenon to be explained. This prompted us to label this type of explanation 'tautological'. Van Parijs' general thesis, in fact, is 'morphologically' similar. In his search for the mechanism which makes function statements explanatory van Parijs discovers that the mechanism which achieves this is one whereby chances of satisfaction are optimized, where the social actors derive a greater satisfaction (and hence a greater subjective good) from a given practice. This explanatory mechanism, on van Parijs' own admission, was precontained in the very definition of hard core function statements. To define his class of hard core function statements, moreover, van Parijs eliminated all consequence statements in which a feature could not be said to exist because its consequences were good. To the non-philosopher, van Parijs' argument amounts to a 'tautological-type functional explanation of functional explanations'. If to the 'sophisticated' philosopher it is otherwise, it needs to be demonstrated more clearly.

One could perhaps argue that tautological explanations are not legitimate patterns of intelligibility to connect causally two or more phenomena but that van Parijs' thesis is not concerned with the causal linkage between two or more phenomena. The thesis is absurd, but let us accept it for the sake of convenience. [58] We will then ask : what mechanism does van Parijs discover, which renders function statements explanatory ? The answer is : the 'R-evolutionary mechanism' or, the behaviourist notion that social practices emerge because they optimize a population's chances of satisfaction at equilibrium point. Assuming that one has clearly and unambiguously defined what 'a population at equilibrium point' means and that the definition is accepted by all social scientists, let us focus on the first part of the statement. The R-evolutionary mechanism implies that the greater the chances of satisfaction produced by a new practice, the greater the reinforcement, and the greater the ultimate likelihood that the practice will be adopted and persist. At equilibrium point the new practice maximizes satisfaction and thus appears to the population concerned, whether consciously or subconsciously, as a greater good.

I have argued earlier that a 'welfare view' does not imply a cybernetic explanation. While acknowledging that an 'R-evolutionary mechanism' is not a cybernetic one, I cannot see how it differs significantly from a 'welfare view' since, after all, the R-evolutionary mechanism has been introduced to explain how

not only is the feature present because its presence has certain consequences, not only are these consequences good, but the feature is present because these consequences are good according to some criterion. [p. 46.]

The 'welfare view' does not say otherwise ; in fact, this very quotation could serve to define the 'welfare view' if we added (as is implicit in the R-evolutionary theory) 'because these consequences are good for the global population according to some criterion'. [4] Having discarded the 'welfare view' as an inadequate pattern of intelligibility to legitimate function statements van Parijs reintroduces a variant of the welfare view to legitimate function statements ! But even if we were to be shown convincingly that the R-evolutionary mechanism differed significantly (from the point of view of a pattern of intelligibility) from a welfare view, we would still wonder what it adds to our understanding of social problems ? Since my own competence lies within the field of social anthropology, I will examine some of the anthropological examples used in the elaboration of the proof.

Collective rituals then come back on stage, particularly Durkheim's theory that' "the real function" ... of religious ritual is to tighten the bonds between the individuals and the social group and so to invigorate the natives' minds' (p. 138). The latter part of the statement does not appear to me particularly Durk-heimian, but this is no place to review Durkheim. In essence, despite van Parijs' efforts to introduce an important distinction, Durkheim's and Radcliffe-Brown's theories are more or less synonymous. This theory, we are told, is in need of an [59] R-evolutionary explanation : 'religious ritual is present in the given context because, compared to its absence, its presence maximizes the natives' chances of satisfaction...' (p. 138). But, surely, this so-called mechanism does not add anything to Durkheim's theory ! It does not lead to important reinterpretations, to new questions. It does not solve old problems or enable us to derive new, unsuspected, corollaries. Nor does it transform Durkheim's theory into a testable and more 'scientific' one. From the perspective of a social anthropologist (not endowed with the X-ray vision of a philosopher of social sciences), the R-evolutionary explanation of collective rituals is utterly superfluous. It is even pernicious, in that it attempts to revive a theory which younger generations of anthropologists have long considered dead, and long buried.

Elsewhere, van Parijs turns his new explanatory tool to a different question, the alleged correlation 'between a society's sex ratio and the practice of polygamy' (p. 133), explaining that

In a society in which the sex ratio is about one to one, the chances of satisfaction structurally associated with the various possible values of the behavior variable may be given by the continuous line in the following diagram.... Monogamy, in such a context, because it minimizes the likelihood of frustration, corresponds to the peak of the functional landscape. [p. 134.]

Concretely, one draws a 'curve of chances of satisfaction' corresponding to a 1:1 sex ratio, and the peak of the curve will represent the practice providing optimality, or maximizing the chances of satisfaction. If the practice observed does correspond to the peak of the curve, one will then observe special cultural rationalizations to reinforce and perpetuate the practice.

There is no need to ask what kind of matrimonial practice would correspond to a 1 :1.7438 sex ratio, and how this is to tally with the particular age composition of the population ! What I want to dwell upon is the crude simplicity of the instance selected. The causal connection suggested is completely erroneous and not recognized by any serious writer on the topic of polygyny. If van Parijs had consulted the most recent and most important statement on the question, namely Goody's Production and Reproduction, he would have found a complex model linking matrimonial practices to modes of production, political control and the devolution of property. The mechanisms involved in the explanation would then be of two kinds : (a) those explaining the causal connection between the various phenomena which, in van Parijs' classification, probably correspond to actional explanations of some kind ; and, (b) the demographic mechanisms used to generate polygyny, namely differential age at marriage for men and women, the compulsory remarriage of widows, and so on. Such a complex model would no doubt have left no room for van Parijs' R-evolutionary explanation. Instead, van Parijs chose a false and ultra-simple theory to fit his crudely simplistic behaviourist reductionism, explaining polygyny in terms of satisfaction/frustration.

But let us give him the benefit of the doubt and assume that his mechanism could handle more complex models, and that this is just what he does in the case of matrilateral cross-cousin marriage. Levi-Strauss suggested long ago (1949) that the greater frequency of matrilateral cross-cousin marriages could be explained by the fact that they required a greater number of exchanging units and would thus serve to integrate larger populations. Translated into van Parijs' idiom, the theory would read as follows : matrilateral cross-cousin marriage produces social cohesion between larger social units and therefore optimizes the chances of satisfaction of individuals in these populations, at equilibrium point. How then are we to explain the occurrence of parallel cousin marriages ? We will [60] assume that the conditions for optimum satisfaction can vary ; for instance, one must also take into account the social cohesion of smaller family groups. Where joint ownership of land is practiced, we will then assume that the chances of satisfaction are optimized when the cohesion of smaller family groups is given priority since we find a connection between parallel cousin marriage and joint land ownership. The R-evolution mechanism would thus explain the correlation between the two phenomena. Once more, the so-called 'explanation' appears superfluous and redundant, for it adds nothing to what we already know. But it does suffer from more serious limitations. Let us quote van Parijs again :

Parallel cousin marriage will be adopted at equilibrium because of its local optimality in a society with joint ownership, while cross-cousin marriage will be adopted elsewhere for the same reason, [p. 169.]

How shall we then explain the existence of joint ownership (which, incidentally, is found in a myriad of societies without parallel cousin marriage) ? Certainly not by its 'local optimality in a society with parallel cousin marriage' ! We will then look for another 'control variable'. We might find that the Maghrebian type of joint ownership is found in a particular type of environment. Will we then invoke the same mechanism and argue that 'joint ownership will be adopted at equalibrium because of its local optimality in a given type of environment' ? And will the same mechanism explain every single causal connection in a complex model ? What does it then add to our understanding of joint ownership and parallel or cross-cousin marriage to burden every one of our statements with the incantation that 'it is adopted at equilibrium because of its local optimality in a society with...' ? It adds strictly nothing.

Interestingly enough, van Parijs shows a clear preference for anthropological explanations in terms of social cohesion. Would it matter to him that rivalries between brothers are probably at their fiercest in Maghrebian societies ? Probably not. But he might not have realized that we can always demonstrate that any social practice whatsoever produces social cohesion at one level of grouping or another. This is why younger generations of anthropologists use the argument only sparingly, and mostly in its negative form. If we find that a particular practice would engender deep conflicts which could hardly be solved in the polity under study, we might then infer that such-and-such a practice does not occur because of its negative consequences. But only a minority of the more philosophically-minded anthropologists now resort to the crude form of functional explanations in terms of social cohesion because the 'social cohesion' argument, like the R-evolutionary mechanism, can never be refuted.

Indeed, if one finds a correlation, it is explained by the R-evolutionary mechanism. If not, one may conjecture that other 'control variables' have been overlooked. But if all the relevant control variables have been tested and no correlation emerges between matrilateral cross-cousin marriage or parallel cousin marriage and some other social phenomena, for instance, there is no cause for alarm.

... for what matters in an evolutionary approach is local optimality, and not global optimality. In the space of possible marriage systems, the pure cross-cousin type and the parallel cousin type may well fail to constitute neighbouring alternatives.... Under such circumstances, the two extreme types of marriage system are locally optimal at the same time, and the result situation of attract or conflict can be represented as a catastrophe surface. [p. 171.]

With this conceptual armour anthropologists have nothing to fear. With or without causal connections between social phenomena there is always at hand a [61] universal mechanism to explain their success or failure. Beyond (or beneath) the sophistication of the evolutionary language, there is nothing more than outmoded models of thought. In this respect, van Parijs reminds me of St. Thomas Aquinas. Van Parijs would like to emerge as the Galileo or Darwin of social sciences but, sadly enough, he will only appear as their Aquinas. Where the medieval philosopher tried to rescue Aristotelian philosophy by giving it a Christian veneer, van Parijs is trying to salvage the tautological (and, interestingly enough, Aristotelian—see Verdon 1982) form of functionalism by giving it the pseudo-scientific lustre of three-dimensional geometrical representations and catastrophe theory. Admittedly, the comparison between the two philosophers should not be allowed to go further.

Up to this point, I have argued (/) that van Parijs' class of function statements is heterogeneous, (2) that his claim to legitimate functional explanations in terms of an R-evolutionary mechanism is tautological and, (J) that his R-evolutionary mechanism is utterly superfluous and redundant for any of the social theories to which it is applied in the book (and, I might add, probably to any social theory). I now wish to question his more general assumption, namely that a (consensual) paradigm will emerge from an understanding of the morphology of explanatory models.

EXPLANATORY OR CONCEPTUAL MODELS ?

If van Parijs' understanding of social phenomena appeared naive, simplistic and crudely reductionist, his understanding of science and its evolution displays similar handicaps. By postulating that 'stigmatization breeds insecurity, and insecurity prevents the competence from displaying itself in a consistent, reliable fashion' (p. xii), van Parijs claims to account for the preparadigmatic nature of social sciences which, in his opinion, will be resolved by a clear understanding of the morphology of explanatory models. In this more general thesis van Parijs is once more resorting to the crudest form of psychological reductionism, and he is also putting the cart before the horses.

Let us go back to the question of marriage systems. Van Parijs manipulates casually various anthropological concepts, and gives the impression that the real problems lie at the explanatory level. Nothing would be further from the truth. This, he only achieves by ignoring the lengthy debates surrounding such concepts as 'joint ownership' or 'cross-cousin marriage'. For instance, anthropologists are still unable to define the class of 'societies with matrilateral cross-cousin marriage' because they are still disputing the very definition of their concepts. Does 'cross-cousin marriage' refer to prescriptive marriages only, or to preferential ones also ? If it includes preferences, does it have to be recognized as a preference by the actors themselves, or is it to be decided on a purely statistical basis ? If the latter, at what percentage do we start talking of a 'preference' ? Does the preference apply to 'real' cross-cousins only, or to 'classificatory' ones as well ? Recent evidence from South America (Kaplan 1975) and reanalyses of Australian material (Verdon and Jorion 1981) now suggest that the prescriptive form of matrilateral cross-cousin marriage occurs in societies without descent groups, when Levi-Strauss had associated it to exchange between descent groups. But anthropologists do not even agree on the definition of descent groups !

Does all this support van Parijs' contention that we are preparadigmatic because we are all muddled up, and that our confusion arises out of the insecurity bred by our social stigma ? But, as I have just demonstrated, the philosophy of social sciences suffers from the same conceptual confusion in the definition of its basic classes. Indeed, twenty grammarians of equal competence would presumably [62] write basically the same grammar of a given language, but twenty philosophers of the social sciences will write twenty different grammars of these sciences. The philosophy of social sciences is thus at least as preparadigmatic as the social sciences (can a philosophy be 'paradigmatic', anyway ?) and, to be quite honest, it is equally 'stigmatized' in the eyes of the general public and social scientists more particularly. It could not act as midwife, since it is itself in labour. Are we then to assume that its preparadigmatic nature results from the insecurity generated by its social stigma ? Surely, even a casual reading of van Parijs gives an impression of over-confidence originating from a feeling of superiority. We must certainly seek the answer elsewhere, and the history of science can give us a few clues.

Although paradigmatic shifts in evolutionary biology and geology may appear related to transformations in the explanatory models (and even this is quite debatable) the great paradigmatic shifts in the history of physics (associated with the names of Galileo or Einstein) or chemistry (associated with the name of Lavoisier) were essentially epistemological revolutions, deep transformations in the representation of the universe of phenomena to be studied, and in the concepts used in their description and analysis. These paradigmatic shifts led to radical reinterpretations, to the solution of old problems, to the formulation of new questions. It enabled scientists to derive new, unsuspected corollaries and did affect the manner in which causal linkages were to be made intelligible, but only after the paradigm had emerged at the conceptual level. I would thus contend that paradigmatic shifts do not usually occur as a result of clarifications or reshuffling of explanatory models, but as a result of deep transformations in our manner of seeing the world, and of describing it.

We would read these conclusions in van Parijs' own assumptions. According to him, an explanation presupposes the definition of a class (the class of phenomena contained in the theory) and a contrast class. Let us take one of his own examples. Social unrest in some industrialized countries at the end of the nineteenth-century would be explained by the gap between these countries GNP and the money supply. To transform this causal connection into an explanation we do need a mechanism but, first and above all, we must be able to define the class of countries where such a gap is found, and the contrast class of countries where it is not found. If we find social unrest systematically present in one class and absent in the second, we can then surmise a causal association and then look for an explanation.

I will go further and claim that any theory in the social sciences presupposes the definition of at least two classes, for which we need clear and unambiguous definitions. For instance, we must be able to define clearly what GNP and money supply are. This does not appear too difficult, but what about the notion of 'social unrest' ? When we come to 'marriage' or 'descent group', to name but two of the most notorious concepts in social anthropology, we cannot create clear classes for want of precise definitions. Before being concerned about patterns of intelligibility, we must first solve the basic conceptual problems which are plaguing us, a task to which I have already devoted a number of publications (Verdon 1980«, 19806, 1980r, 1981). When this is achieved, the problem of intelligibility will quickly and easily be solved.

The solution to our conceptual problems, however, will not come to us because we begin 'sharing common views' after having elucidated the 'deep structure' of our discipline. A change of paradigm is not a search for a common denominator, a simple set of definitions about which everyone would agree. Again, the history of science points to the contrary ; since Kuhn, many have come to realize that paradigmatic changes are like Gestalt shifts ; they do not [63] emerge slowly from common agreement about the 'deep structure' of the morphology of explanatory models. They rather explode, marking radical discontinuities with the past, changing our representation of the world and encountering stiff resistance among contemporaries. Their spread is not collateral, but pyramidal, with the new and younger generations more open to a new paradigm when older practitioners often cling to old world-views.

Van Parijs addresses problems central to any social science, he hopes to solve them, and offer us a paradigm but his work epitomizes to me what is wrong with many works in the philosophy of social sciences. By focusing obsessively on the logic or morphology of explanations, on the definition of true or false (or falsifiable) statements and other problems in logic, philosophers of the social sciences often miss the most important dimension, that of time. The greatest paradox of this book is that it is dedicated to 'evolutionary explanations' but is itself completely a-historical. I believe that if philosophers of the social sciences want to contribute positively to the advancement of any of the social sciences, they will have to spend more time acquainting themselves with the evolution of its concepts, with its underlying 'cosmology'. As long as they remain preoccupied with what constitutes a scientific statement or a scientific explanation they will go on writing for other philosophers about social sciences but their discourse will remain irrelevant to the development of the discipline.

REFERENCES

Goody, Jack. (1976) Production and Reproduction. Cambridge.

Kaplan, J. O. (1975) The Piaroa. London.

Levi-Strauss, Claude. (1949) Les structures élémentaires de la parenté. Paris.

Verdon, Michel. (1980a) 'Shaking Off the Domestic Yoke, or the Sociological Significance of Residence', Comparative Studies in Society and History, 22, 109-22.

_____, (1980b) 'Descent : an Operational View', Man, 15, 129-150.

_____, (1980c) 'From the Social to the Symbolic Equation...', The Canadian Review of Sociology and Anthropology, 17, 315-329.

_____, (1981) 'Kinship, Marriage and the Family : An Operational Approach', American Journal of Sociology, 86, 796-818.

_____, (1982) 'Durkheim and Aristotle : Of Some Incongruous Congruences', Studies in the History and Philosophy of Science, 13, 333-352.

Verdon, Michel and Paul Jorion. (1981) 'The Hordes of Discord : Australian Aboriginal Social Organization Reconsidered", Man, 16, 90-107.