Masking and the perception of stop consonants: psychoacoustical and electrophysiological experiments.

Naturally spoken German words Stahl [Sta:1] and Spott [Spot] were prepared with gaps (silent intervals) of different lengths between the fricative (S) and the corresponding stop consonant [t] or [p], and with various fricative sound pressure levels. Reduction of the normally occurring pause between the fricative and the stop consonant [t] or [p] to less than 10 ms or 20 ms, respectively, leads to "suppression" of the acoustically present stop, e.g. Spott [Spot] is heard as Schott [Sot]. We compare the results of single neuron recordings from animals to results from psychophysical tests on humans, since animal experiments exclude the effects of possible speech-specific perceptual processes. The animal experiments were done on anesthetized cats, from which we recorded single unit activity in the medial geniculate body (MGB). For neurons which represented the stop consonant by a discharge increase, we varied the gap between the fricative and its following stop from 0 up to 120 ms. For all neurons the responses to the stop consonant were positively correlated with gap length and always showed a minimum for gaps of 30 ms or less. This critical interval for minimum responses matches that found in the human psychophysical measurements of stop consonant "suppression". The time course of the interval dependence of neural responses follows the pattern of the psychoacoustical measurements in humans. Suppression of neural responses to the stop did not depend on whether the neuron responded to either the preceding fricative or the following vowel. Neural (animal experiment) and perceptual (human experiment) responses to the stop consonant show the same dependence on the fricative sound pressure level. The results are interpreted as forward masking. The neurophysiological data correspond well to experiments in which discharge reduction is elicited with simple signals ("forward masking"). Our experiments are evidence that auditory masking plays a role in the suppression of stop consonants and that this phenomenon can be explained without assuming phonetic processes in speech perception at a higher level or specific for humans.