Alternative solutions in componential analyses

Abstract

In this paper we shall discuss one of the central problems of ethnoscience, namely the indeterminacy of the analysis.1 As the discussion of this topic centers on componential analyses of kinship terminologies, we shall illustrate our arguments with the example of a formal kinship analysis. The problem of alternative solutions and indeterminacy in formal semantic analysis is most succinctly formulated by Burling (1962, 1963, 1964, 1965, 1969) in a number of articles, of which Cognition and componential analysis: God's Truth or Hocus Pocus? (Burling 1964) is the best-known. We shall concentrate on the arguments put forward in that paper. a componential analysis of a lexical system is to have some kind of psychological validity, so Burling argues, it must deal with the problem of alternative solutions. alternatives are found in each analysis, the idea that one particular formal analysis is the actual representation of informants' cognitions becomes improbable. Burling's main point concerns the astonishing number of theoretically possible alternative interpretations of any given set of terms: If there are three items in the set (call the items, a, and c) ... one has three obvious choices: use a component which separates a from and c; one which separates from a and c; or one which separates c from a and b (Burling 1964: 20). The possibility of using components which are relevant for only a part of the set doubles the number of possibilities. Whereas for a three term set the possibilities number 6, for any given four term set they number 124. Technical problems of the analysis increase the indeterminacy of an analysis even further; these include: homonymy, empty semantic spaces, non-binary components, parallel com ponents and redundancy. In a paper on Burmese kinship, Burling again