The identification of morphologically complex spoken words: continuous processing or decomposition?

Models of word recognition differ in the claims they make about the processing and representation of morphological complex words. The research reported here on the processing of spoken prefixed words contrasts predictions made by continuous left-to-right processing models and by models adhering to the notion of decomposition and discontinuous access to the mental lexicon. In left-to-right models, the speech input is mapped onto full-form representations in the lexicon in a continuous manner. In discontinuous models, the particular stretch of input corresponding to the stem of a complex word is used as the access unit, irrespective of where this information is located in the speech signal. Three sets of experiments, using gating and phoneme monitoring tasks, investigate whether the processing of prefixed (and psuedo-prefixed) words is determined by the complete, full-word form, or by the stem of the word. The results clearly disconfirm the decomposition model of word-recognition. None of the experiments provide evidence for the stem as the crucial unit in lexical access. Moreover, the results are not compatible with continuous processing models which claim that the lexicon contains only full-word forms. The implications of these findings for the processing and representation of morphologically complex spoken words are discussed.