Homophone representation in monolingual and bilingual impaired and unimpaired speakers

Previous research has suggested three possible architectures for homophone production. Homophones may be stored either with shared representations, producing a processing advantage relative to non-homophones, or with independent representations and feedforward activation, producing no benefit for homophones, or with independent representations and interactive activation flow, where homophones may or may not have a production advantage depending on the relative balance between the influence of processing stages in retrieval. The purpose of this thesis was to attempt to reconcile the previously diverse findings in the literature and investigate which account was the most plausible theory regarding homophone production in the phonological and orthographic output lexicons.The first experimental chapter investigated whether previous conflicting evidence was due to differences in participants (monolingual vs bilingual), tasks (picture naming vs translation) or spelling (heterographs vs homographs). It did this by examining picture naming and translation of homophones and controls matched to their individual or summed frequency with monolingual and bilingual speakers.The second experimental chapter aimed to replicate phonological homophone treatment generalisation in the treatment of spelling with a participant with acquired dysgraphia. This treatment methodology examined whether improved lexical retrieval following treatment of one homophone resulted in generalisation to written naming of its untreated partner. Another aim of this experiment was to investigate if generalisation (if any) was due to homophones (homographs) sharing an orthographic word form or due to feedback from treated graphemes to independent representations. Therefore generalisation to items with varying degrees of orthographic overlap was also investigated (heterographs and direct neighbours).The final experimental chapter examined the effect of homograph priming compared to repetition, heterograph and direct neighbour priming in written picture naming.The results across these experiments provided no support for shared representations. Instead, the pattern of results favoured an account with independent representations and interactive activation including feedback from segments to the lexicon, however further research and computational modelling is needed to fully support this hypothesis.