Lexically Guided Phonetic Retuning of Foreign-Accented Speech and Its Generalization

摘要

Listeners use lexical knowledge to retune phoneme categories. When hearing an ambiguous sound between /s/ and /f/ in lexically unambiguous contexts such as gira[s/f], listeners learn to interpret the sound as /f/ because gira[f] is a real word and gira[s] is not. Later, they apply this learning even in lexically ambiguous contexts (perceiving knife rather than nice). Although such retuning could help listeners adapt to foreign-accented speech, research has focused on single phonetic contrasts artificially manipulated to create ambiguous sounds; however, accented speech varies along many dimensions. It is therefore unclear whether analogies to adaptation to accented speech are warranted. In the present studies, the to-be-adapted ambiguous sound was embedded in a global foreign accent. In addition, conditions of cross-speaker generalization were tested with focus on the extent to which perceptual similarity between 2 speakers’ fricatives is a condition for generalization to occur. Results showed that listeners retune phoneme categories manipulated within the context of a global foreign accent, and that they generalize this short-term learning to the perception of phonemes from previously unheard speakers. However, generalization was observed only when exposure and test speakers’ fricatives were sampled across a similar perceptual space.