Abstract

In evaluating the Linguistic Relativity Hypothesis, it is necessary to avoid using terms in such a way as to empty the Hypothesis of empirical content; it is also necessary to separate related but distinct issues. There is no reason to accept any strong version of the Hypothesis when this is understood to pertain to differences in cognition due to non-universal aspects of language structure. Generative grammarians have been led by their orientation and findings to reject the Hypothesis, but their ideas on the relation between language and thought have often been confused and burdened by gratuitous assumptions. It is argued that 'semantic representations', the semantic objects determined by linguistic principles, cannot be equated with the primary structures manipulated in cognition, termed 'conceptual structures'. It is further argued, with lexical and grammatical examples from various languages, that semantic representations are not universal, even granted essential uniformity of cognition for all speakers and the viability of an informal notion of semantic equivalence between sentences. Semantic representations are not neutral characterizations of conceived situations; rather they portray situations through various 'images', so that the study of semantic representations cannot be divorced from the general problem of imagery and figurative language. One kind of syntactic change involves reanalysis of semantic representations, shifting the image they embody from one type to another.