Hearing acoustic details of speech. I. Level discrimination of stop‐consonant noise bursts

Abstract

In this series of experiments, listeners' abilities to detect and discriminate acoustic details of syllables and words are measured. This paper examines the discrimination of increments in the rms level of either the /k/ or the /t/ noise‐burst of the word “cat” (/kæt/). A token of /kæt/ produced by a female talker was digitized at 20 kHz, output via a D/A converter, and presented on each interval of a 2IFC task. For one of the observation intervals, the noise‐burst sample points were multiplied by (1 + k), 0.05⩽k⩽1.25 [Forrest and Green, J. Acoust. Soc. Am. 82, 1933–1943 (1987)] to obtain rms‐level increments ranging from 0.4–7.0 dB. Normal‐hearing listeners were instructed to select the interval that contained the louder burst. Level discrimination was studied: (a) for noise‐bursts excerpted from the word and presented in isolation (“non‐speech” condition); and (b) for the same bursts presented in word context (“speech” condition). Overall noise‐burst power was ∼60 dB in both conditions. Discrimination thresholds [P(c) = 75] were interpolated from maximum‐likelihood ratio probits fitted to 9‐point psychometric functions. Thresholds in isolation expressed as 20 log (1 + k) were about 1.0 dB for both the /k/ and /t/ bursts. Thresholds in word context were higher than in isolation. Worst performance was obtained with the /t/ burst in word context. [Work supported by OCAST Grant HSO‐005 and Presbyterian Health Foundation.]