Unit definition in the mental lexicon

Abstract

Two experiments were conducted to investigate the way in which derived words are stored in the mental lexicon. The experiments were designed to reveal the criterial feature or features responsible for lexical definition.Four word classes and a range of storage modes were considered. The chosen word classes involved contrasts in pronunciation and stress between the base word and the stem of the derivative. The storage modes could be distinguished on the following features: (a) use of common or separate storage for base and derivational forms, (b) whether, in models which include separate entries for base and derivational forms, derivational access is achieved via the base, or independently, and (c) whether, in models which involve separate entries and independent access, morphologically related items form linked sets, distinct from a general semantic network.Predictions concerning the relationship between the word class features and storage modes were tested by comparing the word classes in two experiments which involved priming effects in word recognition and incidental memory for derivational status respectively. The results of the experiments were complementary. There were no significant priming or attribute retention differences among the word classes. The priming effects observed under derivation-to-base conditions were, in each case, less than, but not significantly different from those observed under base-to-base conditions. Similarly, attribute retention was in each case poor, as if derivational status was a surface attribute, like case or color.It was concluded, firstly, that phonology and stress are not diagnostic features governing the lexical organization of morphologically related items and. secondly, that the results were inconsistent with a model in which derivational representation involves: (a) separate lexical entries and (b) base-independent access and (c) no special linkage among morphologically related items. Difficulties in distinguishing among models which relaxed one or more of these constraints are discussed.