Abstract:

Recent results [Hawks et al., in revision for Ear Hear.] suggest that disambiguation of neighboring lax vowels by cochlear implant (CI) recipients using the spectral peak (SPEAK) speech coding strategy may be hindered by the lack of perceptually significant differences in the coding of F1 and F2 as represented in the electrode activation patterns. Since multiple adjacent electrodes are activated when coding these spectral prominences, it is important to know how much activation is necessary to elicit the percept of a spectral peak in the presence of adjacent stimulation. To investigate this issue, CI recipients and normal hearing adults were compared on a detection task. Continua of three-tone complexes were constructed with tone frequencies equal to the centers of the frequency bands encoded by three adjacent electrodes in the F1 region and again for the F2 region. Stimuli were varied in intensity of the center tone, incremented both above and below a fixed level for the bounding tones. An adaptive procedure was used to estimate subjects' thresholds for detection of a change in center tone amplitude from the pedestal level. Results will be discussed in terms of their significance to CI processor coding strategies of spectral information for vowels.