讲座题目：Spokenword recognition across regional accent variation: I. Native and secondlanguage adults
Unfamiliar regional accents disrupt spoken word recognitionby L2 adults and young L1 learners, can also trip up L1 adults, and oftenconfuse ASR and smart systems. Little research, however, has addressed theaspects of non-native accents that hinder word recognition, or the processes involved.We used a Visual World task to assess how English regional accent differencesinfluence the time course of spoken word recognition by L1 and L2 adults. Basedon the principles of the Perceptual Assimilation Model (PAM: e.g., Best, 1995,and PAM-L2: Best & Tyler, 2007), we selected cross-accent differences inconsonant and vowel pronunciation to fall into two broad categories: CategoryShifting (CS) and Category Goodness (CG). CS differences refer topronunciations in a relatively unfamiliar regional accent that L1 listeners arelikely to perceptually assimilate to their native accent as a different phonemethan the speaker intended, e.g., Australian (AusE) listeners tend to hear TH inCockney-accented THIEVES as [f]. CG accent differences instead involvenon-native accented phoneme pronunciations that are likely to be assimilatedinstead as merely deviant pronunciations of the same phoneme in the listener'snative accent, e.g., AusE listeners hear the affricated T in Cockney-accentedTINY as a /t/ with a deviant or marked pronunciation. Two listener groups oflisteners, for whom AusE was the L1 or L2 (Chinese L1-Mandarin speakers), heardwords spoken in AusE and two unfamiliar accents, Cockney-(CknE) andJamaican-Mesolect-accented English (JaME), which display both CS and CGdifferences from AusE, primarily in their consonant (CknE) or vowelpronunciations (JaME). Listeners heard each word and identified it by clickingamong printed choices of the target word, word onset competitor, word offsetcompetitor, and phonologically unrelated distracter, or 3not there2.Proportions of fixations to onset and offset competitors during the decisionperiod indicate that cross-accent perceptual assimilation to the AusE accentplayed a key role in recognition of JaME and CknE pronunciations for both L1and L2 listeners, especially at word onsets but also at offsets. Vowel andconsonant variations affected lexical competition similarly in both groups,suggesting the L2 listeners had formed AusE-accented lexical representations,with one exception: for CknE words the L2 listeners failed to show the sameassimilation (CS>CG) difference for word offset competitors as the L1listeners. Thus, although the Mandarin listeners had formed AusE-likephonological categories for their L2-English, they paid less attention toconsonant coda information than did L1-AusE listeners, suggesting a persistentinfluence of their L1-Mandarin, in which only nasal consonants can occur incoda positions. Implications of these findings for current perspectives onspoken word recognition and cross-language speech perception will be discussed.