Informational masking in vowel sequences
Diane Kewley-Port, C. S. Watson, Mary P. Czerwinski · The Journal of the Acoustical Society of America · 1986
In a number of earlier studies, Watson and his colleagues have described various stimulus and training variables which determine the detectability and discriminability of single components within complex sequences of tones [e.g., J. Acoust. Soc. Am. 76, 1037–1044 (1984)]. A series of studies has begun to investigate whether, under the same conditions, similar discriminability will be observed for speech and nonspeech waveforms. Modeling a female talker, ten English vowels were synthesized at five durations (20–160 ms). Detectability of the vowels was studied using an adaptive tracking task. Large (20 dB) differences in thresholds were obtained among different vowels presented in isolation. Thresholds were then obtained for 40-ms vowels presented in ten-vowel sequences. In a high-uncertainty task, parameters previously shown to affect detection in tonal sequences similarly affected vowel detection. Vowels in initial or final position were detected most easily. Informational masking for vowels was observed to vary from 0 to 45 dB across subjects. At least for these preliminary experiments, the same mechanisms shown to underlie the perception of tonal sequences also appear to determine the processing of complex speech stimuli. [Supported by NINCDS and AFOSR.]