The Duhem-Quine Thesis

As we noted in the preceding chapter, W. V. Quine rejects every construction of the nature of Carnap’s concept of an observational language (either strict or extended) as substantively incorrect since—in his opinion—it is based on the false assumption that the vocabulary serving to verbalize our knowledge about the world also includes predicates, elsewhere referred to as primitively observational, each of which is equipped, independently of all other predicates and therefore in a ‘natural’ way, with a denotation in the form of a definite observable relation. This assumption is false because first of all, there are no such terms in our language which denote objects (in the wider meaning of the word) of one kind or another in a manner totally independent of that which all the remaining defined terms denote. Secondly, the denotations of these terms, in particular of the terms regarded by Carnap as primitive observational predicates, are not assigned to the terms directly and ‘naturally’, but rather on the basis of a specified set of ontological-semantic assumptions. It is this very set of assumptions that designates the complex distribution, so to speak, of references of individual elements of the lexical system. A distribution of this kind can be carried out in a variety of ways—corresponding to a variety of systems of ontological-semantic assumptions, each of which is capable of accounting for the purely empirical data represented by ‘stimulus meanings’, that is by types of physically characterized situations determining positive or negative replies to appropriate occasion sentences. Thus, the choice of one of the possible ontological-semantic systems is empirically arbitrary. It is dictated by considerations of a formal-technical nature. This is true both when we acquire our own mother tongue and when we conduct linguistic research into some foreign language.