Ever since that time we have witnessed disturbances and we cannot but look forward with dread to even greater ones that we shall have to suffer for these and other reasons. An outbreak of violence is to be expected between our Empire and Hungary […] From Poland come news of lamentable discord between three brothers, the territorial princes. We hear of continuous battle among powerful lords in Lorraine. In our own country [i.e., Bavaria] the confusion of minds has become so abominable that robbery and arson throw everything into disorder not only on the ordinary days of the year but even on days of fasting and penitence, in utter disregard of divine and human law. So heavily are we burdened by the memory of past, the onslaught of present and the fear of future vicissitudes that we are fain to yield to the sentence of death which is our lot from the beginning and to tire of life.

When Durkheim's The Division of Labor in Society first appeared in English translation in 1933, it was greeted by a terse and scathing critique by Ellsworth Faris, who suggested that there was no empirical foundation to Durkheim's thesis, and that subsequent anthropological research—in the forty years since the publication of the book in French—had rendered the work redundant. In particular, Faris called attention to the fact that in primitive societies—in stark contradistinction to Durkheim's basic theme—there was much division of labor and very little repressive law.

One of the major areas of dispute among social scientists working with South Asian material has been over what criteria enable one to isolate and define those characteristics attaching to the phenomenon of caste by which it can be described and analysed. Some scholars posit a set of fundamental attributes pertaining to caste which are so constituted as to limit caste solely to a Hindu context, while other scholars postulate a series of primary principles which permit comparative undertakings. One of the most vigorous and sophisticated proponents advocating the limiting of caste to the Hindu social system is Dumont. Dumont denies the validity of applying the term caste to non-Hindu social systems because for him the essential structural principle underlying caste is the polar opposition of purity and impurity. Hierarchy arises because of and is structured by this polarity. In turn, the requisite condition permitting the development of pure hierarchy is that ritual status and secular power must be conceived as completely separate domains and that status be made superior to power, the priest taking precedence over the ruler (Dumont 1972, p. 114). Caste exists only where this necessary disjunction between status and power is present and, furthermore, this disjunction is only found within the Hindu social system.

In this paper I am concerned with the lack of consistency in the criteria embodied in the concepts used to identify social relations. There are two main elements: what we can call the indigenous ideology and the observer's characterisation. In the former case we are dealing with role definition, more or less formalised in different cases and with cultural norms which define the scope of a relationship; in the latter, regularities of observed behaviour and their politico-economic significance are the basis for classification. Examples of concepts based on the former are relations of kinship; of the latter, pairs of terms like leader and follower, patron and client. It is to the last of these that this paper is devoted, for it is precisely in those areas where the ideas of the actors are given least weight in our analysis that our ethnocentric perceptions cause analytical confusion.