Perceptual weighting of acoustic cues for word stress in normal hearing listeners and cochlear implant users

Abstract

Accurate perception of word stress is key to segmenting and understanding speech. We examined perception of four acoustic cues to word stress in English—vowel quality (VQ), pitch, syllable duration, and intensity—and how perception of these cues is affected by specific kinds of signal degradation. Cochlear implants (CIs) severely degrade VQ and pitch, which could demand different relative reliance on these cues. We tested NH listeners in unprocessed, vocoded, and simulated bimodal (unprocessed low frequencies, vocoded high frequencies) conditions as well as CI users. Stimuli included stress-contrastive word pairs (DESert versus dessERT, SUBject versus subJECT) with orthogonal manipulations of each cue. For both NH and CI listeners, there were strikingly different patterns of cue weighting between these word pairs, highlighting how stress cue weighting is flexible across words rather than being absolute. Vocoding the stimuli led to a downweighting of VQ and pitch cues and a compensatory upweighting of duration and intensity; perception of pitch and VQ were largely restored in the bimodal conditions with 250 and 800 Hz bands preserved, respectively. CI users relied on duration more than the NH controls in any condition, while also showing a surprising ability to use VQ and pitch cues.