Electrophysiological correlates of phonological processing during spoken word recognition.

摘要

The aim of this thesis was to clarify functional interpretations of the Phonological Mismatch Negativity (PMN), a negative-going event-related brain potential (ERP) elicited at 250--350 ms in response to deviations in phonological expectations during spoken word recognition. A phoneme deletion task, which necessitates phonological processing independently from lexical/semantic influences, was employed in three experiments. Experiment 1 examined whether the PMN is sensitive to whole-item phonological similarity or whether a mismatch in one phoneme is sufficient to augment the response. Participants were instructed to omit the initial phoneme from a prompt word (e.g., clap, /k/) and to anticipate the correct answer. They then heard a correct (e.g., lap) or incorrect answer that was phonologically similar (e.g., cap) or dissimilar to the anticipated answer (e.g., nose). The PMN was largest to incorrect items and did not differentiate between phonologically similar and dissimilar items. It was concluded that the PMN is not modulated by phonological similarity. Experiment 2 examined top-down influences on the PMN by manipulating the lexicality of the prompt. A word/nonword prompt (e.g., snoth, / s/) was followed by a word/nonword response that matched/mismatched expectations (e.g., noth/toth). The PMN was largest to mismatching responses and did not differentiate words from nonwords. Results indicate that the PMN is not dependent on lexical mechanisms. Experiment 3 measured whether the PMN is sensitive to mismatches beyond the initial phoneme. A prompt (e.g., snoob) was followed by a response that matched (e.g., moob) or mismatched expectations at either the initial (e.g., foob) or final phoneme (e.g., moop) position. As expected, the PMN was elicited to initial mismatch items. In addition, a negativity elicited at 500 ms was augmented in the Final Mismatch condition and was interpreted as a PMN-like response involved in monitoring the position of phoneme mismatch. Findings presented in this thesis indicate that the PMN serves as a neural marker for a pre-lexical stage of spoken word recognition involving the analysis of speech input merging with phonemic expectations developed by the linguistic context. Implications for models of spoken word recognition, as well as clinical applications are discussed.