Available from UMI in association with The British Library.; In recent years, the interactive activation framework has been used to model a number of cognitive processes including speech production. This thesis adopts such a theoretical framework and uses it to predict and explain a number of effects in both naturally occurring speech errors and experimental investigations of speech.; Naturally occurring word blend errors are analyzed in an attempt to determine whether such errors are more likely to result in real words than one would expect by chance. No strong lexical bias was found even when a large sample of blends was considered. It is argued that this finding creates difficulties for a particular model of speech production which maintains that the integrity of speech is ensured by the operation of a prearticulatory editor. The finding can be interpreted by an interactive activation account of speech production however.; Further analysis of naturally occurring word blends and word substitution errors looks for predicted influences on lexical access and lexical representation. It shows that frequency, certain aspects of phonology and the communicative context all have an effect on the form of such errors. The findings emphasize that for most blends and substitutions both phonology and context are influential; it is argued that pure phonological errors in particular are very rare.; The interactive activation framework suggests that it should be possible to obtain phonological priming in speech production. This was investigated in a series of experiments. The existence of such priming was established and attributed to a balance of facilitatory and inhibitory effects. The facilitation was attributed to repeated access of the same phonemic units and the inhibition to competition between higher level units. The priming technique was also used to investigate the nature of phonemic representations and their access.
展开▼