Sublexical representations in auditory word recognition: evidence from lexical learning
The main question this thesis addresses is whether auditory word recognition proceeds directly from the input to the lexicon or whether there is a prelexical level of processing where segmental units are recognised. In the first part, I situate the question in a wider context of representational issues, and show that it is a crucial question because it allows us to distinguish two broad types of word recognition models: what may be called direct- and mediated-access models. A review of the research literature addressing this question shows that existing experimental results are inconclusive. The second, experimental, part of the thesis addresses the research question with a lexical learning paradigm. English-speaking subjects are first trained to recognise novel words that contain a non-native speech sound (a voiceless bilabial fricative); they then perform two tasks designed to determine whether they have acquired prelexical representations for the nonnative segment. The tests used are repetition priming and phonetic categorisation. The results of the repetition priming task are consistent with direct-access models; but for methodological reasons they have to be regarded as inconclusive. The results of the phonetic categorisation task favour mediated-access models. They also suggest that the representations used at the prelexical level of processing are more likely to be position-specific segmental representations rather than syllable rhymes. These results are compared with those of other studies. They are consistent with a growing body of evidence that auditory word recognition involves a prelexical level of processing where segmental representations are recognised.