Abstract
Coarticulatory “noise” has long been presumed to benefit the speaker at the expense of the listener. However, recent work has found that listeners make use of that variation in real time to aid speech processing, immediately integrating coarticulatory cues as soon as they become available. Yet sibilants, sounds notable for their high degree of context-dependent variability, have been presumed to be unavailable for immediate integration, requiring that listeners hold all cues in a buffer until all relevant cues are available. The present study examines the cue integration strategies that listeners employ in the perception of prevocalic and pre-consonantal sibilants. In particular, this study examines the perception of /s/-retraction, an ongoing sound change whereby /s/ is realized approaching /ʃ/ as a result of long distance coarticulation from /r/. The study uses eye tracking in the Visual World Paradigm in order to determine precisely when listeners are able to utilize the spectral cues in sibilants in different phonological environments. Results demonstrate that while in most instances listeners wait until more cues are available before considering the correct candidate, fixation accuracy increases significantly throughout the sibilant interval alone. In the pre-consonantal environment, immediate integration strategies were strengthened when the coarticulatory cues of retraction were stronger and when they were more predictable. These findings provide further evidence that context-dependent variation can be helpful to listeners, even on the most variable of sounds.