Predictors of individual differences in accented speech perception

Speech perception relies on quick mapping of spectral-temporal cues onto linguistically meaningful units. This becomes difficult when there is a mismatch between the listener’s phonological representations and phonetic cues produced by a speaker with an unfamiliar accent. This difficulty is similar to other adverse listening conditions and the ability to resolve phonetic ambiguity has been linked to individual differences in phonetic category representations. In this study we measure phonetic category structure and accent perception accuracy for 85 adults (55 younger and 30 older) to explore the impact of bottom-up processing on accent perception. We use a discrete Visual Analog Scale (VAS) phonetic decision task requiring listeners to rate tokens on a continuum spanning canonical endpoints and ambiguous tokens. Phonetic category structure is modeled by using a four-parameter logistic regression (slope, boundary, minimum, and maximum) to estimate each individual’s unique behavioral function. Interestingly, slope (or gradience interpreted as increased phonetic sensitivity) did not predict accent perception accuracy, but the error of responses to that function (consistency) did. Age was not a significant predictor of accent performance in this sample. Implications of response consistency, slope, and the stability of speech perception processes in healthy aging will be discussed.