The representation of normally-voiced and whispered speech sounds in the temporal aspects of auditory nerve responses

The acoustics of spoken phonemes can be varied by phonation, speaker, word-position and co-articulation. Whispers lack many of the frequency and voicing cues of normally-voiced speech. Both forms of speech are accurately perceived and placed in the same linguistic categories. This perception may rely upon the same invariant cues or other compensatory cues. The auditory nerve (AN) was hypothesized to produce common representations for acoustically varied phonemes.; Individual AN fiber responses were pooled to create global average peri-stimulus time (GAPST) histograms. GAPSTs for six different tokens of /da/ contained a two-peak pattern within the first 38 ms. All tokens of /ta/ produced a single peak within the first 38 ms. These patterns were consistent across normal and whispered voicing and intensity changes. Locally averaged PSTs reflected varied acoustics. Individual fibers did not demonstrate phoneme-specific response patterns. Acoustically varied word-final /d/ and /t/ also produced distinct ensemble AN patterns. These common representations for voiced and voiceless alveolar consonants were also shown in AN responses to voiced and voiceless labial consonants. Averaged patterns of auditory nerve responses across seven frequency bands were distinct for labial and alveolar consonants, regardless of normal or whispered voicing. In these AN patterns that were specific to consonant features, the contribution of all fibers was essential, particularly those with high characteristic frequencies. These findings suggest that temporal aspects of ensemble AN processing contribute to common representations of the voiced or voiceless feature of normally-voiced and whispered stop consonants; aspects of locally-averaged responses contribute to common representations of place of articulation.; The fine-timing of AN responses to multiple repetitions of stimuli represented normally-voiced and whispered vowel spectra. Again, the inclusion of all fibers' responses, especially those with high characteristic frequencies, was important for recovering high frequency third formant information. Details of vowel frequency information were represented in the fine-timing of AN responses but gross formant patterns were represented by rate-place patterns across 30 dB of intensity changes.; Whispered and normally-voiced speech sounds were encoded by distinct patterns in the AN, when responses of all fibers were included. These common representations support and describe auditory mechanisms for speech perception.