Retuning speech sound categories: An eyetracking study.

Abstract

Speech categories are defined by multiple probabilistic acoustic cues. Fundamental frequency (F0) and voice onset time (VOT) are correlated in the English stop voicing contrast, for example. However, such correlations are often imperfect—especially in cases of non‐native or disordered speech. The present experiments investigate listeners’ ability to adjust perceptual cue weighting in online perception in response to changes in the cue correlations experienced across time. Native‐English listeners heard minimal‐pair words beginning with stop consonants varying along a VOT series. The F0 of the words was gradually shifted over the course of the experiment from the canonical English correlation (higher F0 for voiceless stops) to the opposite pattern (lower F0 for voiceless stops). Categorization was assessed via explicit responses while eye gaze data were simultaneously recorded using the visual world paradigm. Both data types revealed that the influence of F0 on voicing categorization changed in response to changes in F0/VOT correlation. Some listeners use of F0 reversed such that higher F0 led to more voiced responses; other listeners discontinued use of F0 in voicing categorization. These patterns suggest that listeners are continually monitoring the input for regularity and retuning acoustic cue use in an online manner to accommodate these regularities.