社会学方法論の反省

通語構造の分析を手懸りとして

抄録

Sociological studies of language can be divided into two main classes. One is the study of the language in general especially taking its social functions into consideration and the other is that of a language in relation to a particular society. Since these studies, however, are mainly interested in its phenomenal descriptions, they can not explain the causes bringing about various types of languages and their changes, but only understand or interpret them. If they try to pursue its causes, they must have recourse to some assumptions which are often of psychological nature. Nevertheless a language must be treated as a social fact which has its own raison d'etre. So it has to be explained by itself, that is, in its own term without resorting to such an assumption.
1) We must be careful to reseach a language as a whole, because dividing it into units must not lead to the vanishment of the nature of totality. It must keep the nature even when it is operationally divided into its constructive units. What is this nature ? It consists in the fact that it is a pattern, which means an organization, for example, an arrangement in such a form as a sequence of time-series. In short, it may be taken for granted that a language is a patterned form of representation, which is possible only by an arrangement of its representation elements. So we must bivide a language into its representation elements and (re-) integrate them into a pattern. We, sociologists, must be interested in this way of (re-) integration.
2) But there are three levels of the study of language corresponding to the levels distinguished by Weaver, which will presumably correspond to the levels of integration ; that is. (1) technical level of communication, (2) semantic level, (3) effectiveness level. Therefore, we must choose the technique appropriate to each level. Now, while level (1) is included in the statistical information theory, (2) must be treated by the semantic information theory, because the probability theory which corresponds to these two levels respectively must be distinguished ; the former is the statistical one and the latter the inductive. By the statistical description of a language, we can explain the objective aspect apart from its meaning and by the description of its inductive probability the subjective one, that is, the aspect of its meaning. On the basis of these descrption, we shall concern with level (3), that is, the behaviorial or effective aspect of a language. However, in this paper we shell concentrate our attention only to the problems of the former two, because only by the establishment of the logical foundation of description, we can proceed on the problem of causality empirically.