The effect of signal degradation on the status of cues to voicing in utterance-final stop consonants

In several recent studies, spectral acoustic cues (e.g., voicing during closure, preclosure transition) have been found to have more influence than temporal acoustic cues (e.g., vowel or syllable duration) on perception of the voicing distinction in final stops. Earlier studies, mostly using synthetic speech, had found vowel duration to be the primary cue to this distinction. The present stimuli were taken from a study in which spectral cues had dominated judgments in quiet listening conditions. These stimuli were presented against a high level of background noise (S/N=0) to seven normal adult subjects in a forced choice (e.g., BED–BET) format. In quiet, the correlation of the percentage of the original vowel remaining after deletion with judgments of the final stop as voiced had been 0.18. In the noise listening condition the correlation was 0.94. Syllables with voiced final stops from which 76% of the vowel (all but the final transition) had been deleted were judged voiced (82%) in quiet but voiceless (87%) in noise. These results indicate that the temporal cue of vowel duration is more resistant to degradation by noise than are spectral cues, and that vowel duration plays a dominant role only when other cues are unavailable.