ABSTRACT
When we see, hear, or produce a familiar word (simple or complex) we reveal knowledge of an assortment of information about that word. For example, we know about the word want that it has a phonological form (/wantl) and an orthographic form ({want}). We also know how to use it and understand it in its syntactic context as being a verb with certain argumentrtaking requirements (John wants a house, * John wants to his friends)1 and certain inflection-taking requirements (wanted, wants, ·wantest). We also know exactly what want means and, of specific interest here, we know that the basic meaning and other lexical properties (subcategory requirements) remain constant when want occurs in its various inflectional forms (want, wants, wanted). Because inflectionally related words (want-wanted) share basic lexical information (components of form, morpho syntax, and meaning), a question arises as to how this 'sharing' is represented in the lexical processing system. Do wanted and want converge onto the very same representation in the lexical processing system? If so, at which level(s) do they converge: form, lexical-syntax, or meaning? Following the terminology of current lexical processing theories, we will refer to the level where a word's written or spoken form is encoded as the lexeme level, and to the level where its lexical-syntax and meaning are encoded as the lemma level. 2
The Internal Structure of Words 213
evidence for repeated access to a common decomposed stem for sized and size, might also be explained on a whole-word-based model where, for example, size and sized-like buy and boughtshare frequency-derived activation through a common lemma representation, but share no (decomposed) representation at the lexeme level.
