ABSTRACT
A slots-and-fillers model of the speech production planning process, based on patterns and constraints in sublexical speech error data, proposed the separate representation of (a) single phonemic segments and (b) the utterance-specific framework that guides their processing, as well as (c) a segment-by-segment process to associate segments with their target locations in the framework. In this model, all sublexical errors occur during the operation of this serial ordering mechanism. Further analysis of sublexical errors, however, shows that the error data are concentrated largely in word-onset consonants. Evidence from both spontaneous and experimentally elicited speech supports the claim that word-onset consonants in particular (rather than syllable onsets or prestressed consonants in general) undergo separate processing that renders them particularly liable to confusions. An expanded slots-and-fillers model is proposed to take account of these findings; it includes a processing mechanism that deals with word-onset consonant information separately during the elaboration of metrical structure.
