Masking-like phenomena in speech perception

Abstract

To study masking-like effects in the perception of continuous speech, we presented two-formant syllables /b ε b/ or /b ε g/ followed, at intervals from 0 to 150 msec, by /d ε/; listeners were instructed to identify the syllable-final consonant. An 80-msec intersyllable interval was required for recognition of syllable-final consonant to reach asymptote. To determine the level at which this “masking” occurred, we repeated the experiment, but with the second-formant transitions of the first syllable presented alone for judgment as rising or falling chirps. Recognition of the isolated transitions (chirps) was essentially unaffected, no matter what the interval between them and the second syllable. These data suggest that the “masking” in the initial study was due to the elimination of a necessary cue—in this case, a silent interval, corresponding to the stop closure, between the syllables—and not to backward masking of the auditory information. In a third study, we found that changing voice between the syllables from male to female also eliminated almost all the “masking.” This reinforces the conclusion of the first two studies, indicating in this case that the effect is not to be attributed to interruption of processing, but raises other questions that we have only begun to investigate.