LEXICALISATION IN SPEECH PRODUCTION: EVIDENCE FROM FORM RELATED WORD SUBSTITUTIONS IN APHASIA

One of the major empirical sources of theories of speech production are speech errors in normal speakers. Speech errors occurring during lexicalisation of a concept to be expressed can result in whole-word substitutions that are target related in form and/ or meaning or can appear as nonword productions (neologism). Similar error phenomena have been reported for aphasic patients. The present study describes the aphasic, HZ, who produced mainly form-related word substitutions and neologisms in several single-word processing tasks (picture naming, repetition, and reading aloud). In picture naming, meaning-related substitutions also occurred as well as substitutions that were related both in meaning and form (mixed errors). Three hypotheses of the origin of formal errors were tested: the full interactive activation hypothesis postulating meaning-form interactions, the lexical (form) retrieval hypothesis, and the post-lexical phonological encoding hypothesis. HZ's performance on repetition and reading aloud (tests showed no mixed errors and no effects of imageability and of target frequency on formal errors) failed to support the predictions of the first two hypotheses. However, the phonological encoding hypotheses (enriched by a comprehensionbased editor) could also not account for the data (e.g. for some task-specific asymmetries in the error pattern). Instead, an attempt is made to account for formal and mixed errors by construing word form encoding as an interactively organised component within a two-stage model of lexicalisation. Formal errors are traced back to interactions between lexical forms and sublexical phonological information during the second stage of lexicalisation. Mixed errors occur when lexical forms of the target's semantic competitors are involved in word form encoding.