* Kuhn, T. S., The Structure of Scientific Revolutions (Chicago 1970)Google Scholar; Hagstrom, W. O., The Scientific Community (New York 1965)Google Scholar; Fisher, C. S., The Death of a Mathematical Theory: a Study in the Sociology of Knowledge, Archives for the History of the Exact Sciences, III (1966), 137–159CrossRefGoogle Scholar; Mulkay, M., Some Aspects of Cultural Growth in the Natural Sciences, Social Research, XXXVI (1969), 22–52Google Scholar; Id., The Social Process of Innovation (London 1972); Dolby, R. G. A., The Sociology of Knowledge in Natural Science, Science Studies, I (1971), 3–21CrossRefGoogle Scholar; Barber, B., Resistance by Scientists to Scientific Discovery, Science, CXXXIV (1961), 596–602CrossRefGoogle Scholar.

(1) I have not attempted to relate the main arguments of this paper to episte-mological questions. They accord a position of skepticism in that they question the ways in which particular epistemologies are supposedly logically grounded.

For my own part I see no need to timeless, universal criteria of rationality, whether for sociological purposes or any other. Thus, whatever present arguments have, I would regard them as contingent, just as Wittgenstein (1964) characterises mathe-matical arguments and proofs.

(2) The problems of intelligibility and attribution of significance are among the most neglected, yet important, in the social sciences. I have simplified their treatment in this paper by largely ignoring the disquestion tinction between professed and inferred significance. Thus, although I frequently stress the importance of understanding the actors' viewpoint, I do not imply that this should derive solely from his verbal accounts. An actor who vomits when offered meat to eat, or who laughs uncon-trollably on hearing a ‘blue’ joke, is providing evidence of what is significant to him; indeed such evidence could well be regarded as more convincing than that of (conflicting) verbal accounts.

(3) Philosophers have sometines criticised Kuhn's account of science as ‘irrationalist’, a charge which, if in the last analysis incorrect, nonetheless makes sense in terms of the rules of their game. If Kuhn's work is viewed as a sociological enterprise, the claim becomes irrelevant. The difference between the disciplines most salient here is that, whereas many philosophers conceive their activity as prescriptive for the actors they study, such an objective would be fatal to sociology. Thus, recent work in the Popperian tradition by Lakatos (Lakatos and Musgrave 1970; Cohen and Buck 1971) differs from Kuhn's approach mainly in the author's more overtly prescriptive concerns, which link with his conception of what philosophy should be about (cf. Bloor 1971). These concerns account for an interest in standards of rationality of no relevance to the sociologist.

(4) McIntyre's (1966) exposition of the relationship between reasons and causes in the explanation of action is accepted as valid throughout this paper.

(5) The argument implies that “normal practice”, or some equivalent, must be an essential regulatory concept in sociology. Its use does not, however, prevent the identification of situations where there is no “normal practice”—any more than Newton's first law prevents an identification of bodies not moving in straight lines with uniform velocity! Use of a concept as a regulatory idea does not involve an assumption about what is necessarily the case. This is a point insufficiently appreciated, for example, in the controversy between conflict theories and functionalism. Regulatory functionalist ideas, such as those of Parsons' “Social System” are not always distinguished from those which constitutively involve a particular concrete vision of society, as in Merton's work on deviance.

(6) It is worth restating here that the argument does not imply that actors can be studied without the intrusion of any external standards or preconceptions. Hollis has produced a powerful argument against this kind of extreme idealism (1967). He shows that in the attribution of significance to actors' utterances the existence of a high level of conformity to standards of identity and contradiction must be pre supposed. That is, without assumptions of generally consistent linguistic usage, meaning cannot be derived from utterances (see also Lukes 1967).

It could be argued that Hollis has shown consistency to be a universal rationality criterion, and hence has produced an important limitation to my general argument. However, to show that cultures must be assumed to possess high levels of consistency, if they are to be regarded as cultures at all, is not to show that there is any general, spontaneous tendency to consistency within cultures; to show that water is very hot tells nothing about whether it is likely to grow hotter or cooler. To be sure, actors may be assumed universally to possess the capacity to detect inconsistency, and even universally to regard it as significant. It may also be admitted that actors often try to eliminate perceived inconsistency from their beliefs. But the anthropological literature is clear that, on occasion, they may be indifferent to it. And contradiction may actually be institutionalised and more or less consciously utilised by actors. It may gain power and significance precisely from its irrationality in this sense, just as excrement may gain power from its relationship to norms of cleanliness and order (Douglas 1966; cf. Wittgenstein 1964, pp. 130–131).

Thus, although Hollis's point must be conceded, indeed regarded, in the same way as Lévi-Strauss's work, as a contribution to the general study of culture, it does not establish contradictory beliefs as a separate class with reference to sociological explanation, nor does it show that, within a given existing culture, particular consequences are potentially entailed by their inconsistency.

(7) Briefly and loosely, poison oracles provide ‘yes’ or ‘no’ answers to questions of interest to Azande. They involve administering a substance, benge, to a chicken, which may die as a result. As the chicken lives or dies, so the oracle is held to have answered one way or the other.

Azande, however, are careful to ask the question twice, and sometimes find themselves receiving contradictory answers. On such occasions they are apt to find their benge at fault, and suspect inadequacies in the way it was prepared. Western critics are apt to find this explanation inferior to one resting upon variability in the concentration of poison in the ‘benge’ and individual differences in the constitutions of chickens.

Perhaps if Azande had produced a philosopher, who had shown that their views were “in principle” testable using several thousand chickens, western academics would have been more impressed with Azande rationality? Indeed, a few may have been persuaded, by such a philosopher, that Azande explanations were more rational or scientific than our own!

(8) Compare the approach of Peel (1969) I have, of course, deliberately selected examples of ad hoc scientific work which may be characterised respectively as leading ‘towards’ and ‘away from’ current scientific views.

(9) The crucial parts of Popper's theory of knowledge seem compatible with a more extended conception of the sociology of knowledge than he himself has been disposed to favour. Particularly relevant here is Popper's discussion of the “empirical base” (1959, ch. V, and see also 1969, ch. xi).

(10) Cf. Jarvie and Agassi (1967).

(11) There is, in addition, the point that beliefs do not have to be untrue in order to have latent functions. It is, for example, not necessary to await a ‘final’ verdict on the claims made in Jensen's (1969) on the connection between intelligence and ethnicity in the United States, before attempting to analyse their latent functions in that society.

(12) For simplicity and clarity in developing the argument, I have taken Azande as a paradigm of a highly stable society, an assumption encouraged by much of Evans-Pritchard's work. As Ernest Gellner has pointed out to me, however, the full range of Evans-Pritchard's writings on Azande reveals them as a society undergoing considerable social change. The same points can still be made in such a context, but with more difficulty; hence my decision to stay with the idealised and oversimplified description of Azande conventionally accepted in current theoretoping ical debates.

(13) Cf. Gellner (1962, p. 14). It could be argued that Azande don't wish to falsify their oracular systems whereas scientists try to falsify their systems. However, it is on the whole more accurate to say that scientists sometimes try to falsify other scientists' systems. It is the highly differentiated structure of science, and of the societies which contain it, that largely explain its more dynamic culture.

(14) For another discussion of the particular difficulties in Barber's analysis see Barnes (1972).

(15) It is interesting to note that Kuhn's philosopher-critics have often interpreted his work, albeit in caricature form, along the lines of Mulkay's interpretation (long periods of dreary conformity, short periods of mysterious deviance). They have pointed out that there is more to “normal science” than dreary conformity, but have been less than helpful about what more there is (see, for example, Lakatos and Musgrave 1970).

Kuhn himself, on the other hand, has always explicitly claimed that paradigms are more than mere sets of rules (or, presumably, norms), feeling it necessary, from the first, to discourse at length on “The Priority of Paradigms” (1970, ch. v).

(16) The role of interpretation is, of course, much easier to see with legal rules than with scientific ones. Characteristically, Wittgenstein chooses to make this sort of point in the most difficult context of all—mathematics.

—But am I not compelled, then, to go the way I do in a chain of inferences ?—Compelled ? After all I can presumably go as I choose!—But if you want to remain in accord with the rules you must go this way.—Not at all, I call this ‘accord’.—Then you have changed the meaning of the word ‘accord’, or the meaning of the rule.—No;—Who says what ‘change’ and ‘remaining the same’ mean here ?

However many rules you give me, I give a rule which justifies my employment of your rules.

We might also say: when we follow the laws of inference (inference-rules) then following always involves interpretation too.

[…] Nevertheless, the laws of inference can be said to compel us; in the same sense, that is to say, as other laws in human society. (Wittgenstein 1964, pp. 33e–34e).

(17) For an excellent introduction to this subject see Knight (1968). Also Russell (1971).

(18) This is merely a special instance of a completely general problem in the social sciences. Conservative approaches to it are particularly favoured by anthropologists; a notable illustration here is Leach's (1964) treatment of the Kachin concept of a ‘nat’.

Nats are apparently “magnified non-natural men” (p. 172). But, by noting the contexts in which they are referred to, and the phenomena they make intelligible, it becomes apparent that they are “nothing more than ways of describing the formal relationships that exist between real persons and real groups in ordinary human Kachin society” (p. 182). “[…] it is nonsense to ask such questions as: ‘Do nats have legs ? Do they eat flesh? Do they live in the sky?’” (p. 14).

(19) As a scientific philosophy, of course, the most extreme forms of positivism rejected the use of any terms, like atoms and molecules, which “went beyond the data”. This kind of position was very common among chemists in the mid-nineteenth century.

(20) See for example Black (1962), Hesse (1953, 1964), McMullin (1968), Schon (1963). Only a few of the far-reach-ing implications of this view of scientific theory will be discussed here.

(21) But see the fascinating discussion by Duhem (1914).

(22) A recent paper of outstanding interest by Mullins (1971) provides a valuable discussion of pre-paradigmatic research guided by common interest in a single problem; it also indicates important ways in which the framework used for discussion here is too simple. This framework is not entirely congruent with Kuhn's, or at least not obviously so; cf. Kuhn (1970, IV and V) and the slightly different analysis in Postscript (1 to 4); it does offer a means of relating the ‘broad’ and ‘narrow’ definitions of ‘paradigm’ used by Kuhn.

(23) It is not as easy as it might seem to decide when a model is replaced and when it is merely radically extended. See for example, Brush (1970).

(24) De Broglie's innovation is curious in simulthat wave properties are not associated with corpuscles in the primary system of the metaphor.

(25) It is interesting to speculate as to which kinds of scientist, and scientific role Alternativestructure, are best suited to this type of ‘creative’ normal science. Since it consists in the metaphorical transfer of beliefs and action from one system to another it can be regarded as “cross-fertilisation”, “an inter-play of divergent cognitive-normative frame-works” (Mulkay 1969). However “cross- performfertilisation” has generally been used to describe innovations made by transferring concepts and practices from one field or sub-field to another; the initiator of such innovation generally had special familiarity with two fields or types of practice, which explains why he, of all scientists, made the advance in question. He may have simultaneously occupied two scientific roles, or have transferred roles in such a way that the expectations of his original role were not abandoned (Ben-David 1960, Thackray 1966, Ben-David and Collins 1966). Alternatively the scientist may have found himself in a favourable position for more particular reasons (see the description of Bohr's early work in Heilbron and Kuhn 1969, and the description of Hilbert's work on Invariant Theory in Fisher 1966). However many acts of ‘creative’ normal science are performed by one of many actors familiar with both sides of the metaphor in theory and practice. Such scientists, with similarly structured concrete experiences, will tend to evaluate the innovative act alike. But the reason why a particular innovator was theoinvolved will generally not have a social-structural answer.

(26) For a summary of Arrhenius' appreciation by van't Hoff and Ostwald, and opposition to him from other scientists, see Dolby (1969).

The reception of Avogadro's hypothesis is also an interesting case. Favourable evaluation of this important principle was undoubtedly delayed by the difficulty with which it was reconciled with dualistic theories of chemical combination; these dualistic theories were dominant at the time when the principle was first formulated and later declined in influence. However, wide spread general interest in the principle only followed work which showed how it could be built into new, powerful and highly general procedures for determining atomic weights—constants of immense importance to all chemists (see Nash 1957, pp. 292–319).

(27) There is, in addition, the point that innovative ideas are often accompanied by descriptions of new concrete practice readily taken over by some of the scientific audience. Again, such ideas often suggest new concrete procedures to scientists, in the light of their own personal concrete concerns. New ideas disturb the context within which they are evaluated.

(28) I should emphasise that this paper is constiin no way whatsoever a criticism of the natural sciences. Depressing experience has convinced me of the necessity of making practithis point.

The beliefs of the natural sciences constitute the central element in my own world view. Nor is it surprising, socialised as I have been, that I regard them as practically reliable and aesthetically delightful.