The ‘‘noisy’’ speech chain

Abstract

Kenneth Stevens’ seminal contributions to the field of acoustic phonetics are far reaching and have important implications for modeling normal and disordered speech production and perception. His studies of the relations between underlying discrete linguistic features and the analog worlds of articulation, sound, and hearing have influenced a generation of speech researchers, including myself. Our studies of speech perception in noise are inspired by the work of Ken and his colleagues on uncovering the acoustic and perceptual correlates of linguistic features. I will report on experiments which examined acoustic and perceptual (in noise) cues for the place of articulation and voicing features for syllable-initial plosive and fricative consonants. Results are analyzed in terms of voicing, vowel context, manner of articulation, place of articulation, shape of the noise masker, and gender of the talker. One of the interesting findings is that certain attributes which cue a feature acoustically and perceptually in quiet conditions (such as the VOT for the feature voicing) do not correlate well with the perceptual robustness of the feature in noise.