Introduction

Comparative syntactic studies of the world's languages are greatly facilitated by the use of glosses, which allow for a closer analysis than is possible through conventional translation. However, the creation of glosses can be misleading if the elements given as lexical analogues differ from language to language, to the extent that the meanings of words determine possible syntactic environments. In fact, it is extremely difficult to find true matches in the open-class lexicons of any two languages, and this is certainly the case for closed-class items, whose syntactic properties are inevitably language-particular (Emonds, 1985: 165–170, 2000: 115). Such differences in lexicalization across languages appear to be so ubiquitous as to be indicative of a fundamental organizing principle in the mental lexicon. I draw on observations by Saussure (1983 [1916]) to argue for a principle of lexical relativity, according to which lexical concepts are determined relative to other items in the lexicon, with the result that no two languages lexicalize concepts in the same way. This does not necessarily lead to a radical Whorfian approach, as the semantic machinery used to build lexical items contains elements of meaning that are plausibly universal, and mental construals of the world are not only constructed through language; however, it does support an approach in which the particular language we speak determines linguistically mediated construals of events, states and objects, in line with Slobin's (1996a) proposal of Thinking for Speaking. For the present purposes, discussion is restricted to encoded aspects of lexical meaning, rather than aspects of meaning that may change according to context, on the assumption that encoding and inference are independent systems of interpretation (Sperber & Wilson, 1995 [1986]). For example, the English verb spray, regardless of context, encodes the interpretation of the moving object in the event as a three-dimensional (3D) aggregate of psychologically dimensionless points; smear requires that the moving object be a 3D semisolid substance; and the verb wrap specifies that the moving object be a two-dimensional (2D) flexible solid (see Pinker, 1989: Chapter 5 for discussion).