Effects of spectral degradation on contextually driven shifts in phonetic categorization.

Abstract

Under conditions of spectral degradation, listeners have been shown to experience deficits in perception of some consonant and vowel contrasts that are driven by spectral cues. In everyday listening situations, cues to phonetic contrasts are dynamic, owing to varying listening contexts both across‐ and within‐talkers. This adds an added layer of spectral information not typically captured in assessment of cochlear implant users or normal‐hearing listeners in simulated conditions. A series of experiments was designed to assess the ability of these listeners to use spectral cues to normalize and adjust phonetic boundaries across talkers and across various phonetic contexts. These tasks involved the adjustment of the alveolar/palatal fricative boundary, which has previously been shown to be sensitive to talker gender as well as vowel context. Preliminary results suggest that when spectral resolution is degraded, listeners show an impaired ability to adjust to these contexts. It is also suggested that visual information can substantially mitigate these deficits, at least for the phonetic contrasts tested here. These results help to explain some of the difficulties experienced by cochlear implant users in common conversational settings, and reinforce the importance of compensatory strategies (such as audio‐visual integration) to aid in perception.