Abstract:
Many studies have examined the ability of cochlear implant recipients to perceive words and sentences. The findings have frequently been analyzed within the structure of traditional linguistic units such as phonemes and features. However, it is possible that these units do not accurately represent the type of acoustic information encoded and transmitted by cochlear implants. In an alternate system of representation, broad phonetic categories (BPCs) parse the speech stream into acoustically based units [D. Shippman and V. Zue, IEEE ICASSP, Paris (1982)]. Although BPCs were developed to aid computer speech recognition systems in the face of ambiguous phonetic information, they may also be valuable in modeling the perception of words in cochlear implant recipients. The present experiment investigated whether BPCs provided a better representation of the speech signal in cochlear implant recipients than phonemes and features. Subjects were adults who had been implanted with either the Nucleus-22 or Clarion cochlear implant. A version of the modified rhyme test was used to assess phoneme and BPC recognition in word context. These results were then compared to the listeners' scores on the CUNY (auditory-only) sentence test because of its closer relationship to everyday listening experiences. [Work supported by NIH.]