Effects of Masking Noise on the Representation of Vowel Spectra in the Auditory Nerve

Abstract

Steady-state vowels can be discriminated in background noise at signal to noise ratios of less than -15 dB (Dewson, 1968). Thus, the information relating to the spectrum of a vowel stimulus which is contained in the firing patterns of the ensemble of auditory-nerve fibers must remain when the vowels are presented in background noise. Spectral features may be encoded in auditory-nerve discharge patterns in at least two ways: One in terms of the distribution of average discharge rate as a function of fibers’ characteristic frequencies (CF) (rate-place representation, Sachs and Young, 1979), and the other in terms of temporal or phase-locked responses (temporal-place representation, Young and Sachs, 1979). We have shown previously (Young and Sachs, 1979) that the temporal-place code provides a more robust representation of vowel formant frequencies than does the rate-place code. In this paper we shall compare the effects of broad-band noise on these two representations of the spectrum of the vowel /ε/.