Perceptual normalization for speaking rate II: Effects of signal discontinuities

Abstract

In a series of experiments, we examined how rate normalization in speech perception is influenced by segments that occur after the target. Perception of the syllable-initial target was influenced by the durations of both the adjacent vowel and the segment after the vowel, even when the identity of the talker was changed during the syllable. These results, together with earlier findings of a temporal window that follows a target phoneme within which segment duration influences perception of the target, help to resolve apparently conflicting results that have been reported previously. Overall, the results fit within a theoretical framework in which the rate at which events take place is extracted early in processing, prior to segregating voices, and the use of this information is obligatory in subsequent processing.The process of perceiving speech involves achieving perceptual constancy in the face of substantial stimulus variation. In order to recognize a word, the listener must deal with variation between talkers and variation that occurs within the speech of a single talker. Between-talker differences include the length of the vocal tract, vocal fold mass and tension, articulatory habits, intrinsic speaking rate, and dialect. Within-talker effects include the emotional tone of the speech signal, the context in which segments occur, and the rate of speech.Our focus here is on how listeners achieve constancy in the face of variation in the rate of speech.

Speaking rate from proximal and distal contexts is used during word segmentation.

“…An attenuated effect of proximal context relative to Experiment 2 would support accounts that argue for the use of rate information averaged over a time window of approximately 300 ms around the target sound (e.g., Newman & Sawusch, 1996;Sawusch & Newman, 2000). Parts of the distal contexts in our Experiment 3 fall into this time window even for the slow rate proximal contexts (which had a durations of maximally 249 ms) and thus allow for a potential attenuation of the proximal rate effect by distal rate.…”

“…Additional results from the present experiments suggest that time-window accounts of rate effects (e.g., that rate is evaluated only within a 300 ms window; Newman & Sawusch, 1996;Sawusch & Newman, 2000) may be too simplistic. In the categorization tasks, a longer distal context led to stronger rate effects than a short distal context (Experiment 4 vs.…”

“…In comparison to the effect of the long distal context in Experiment 4, however, the short distal context had a much smaller effect. This suggests that either the time window over which rate information is averaged is larger than 300 ms (Newman & Sawusch, 1996;Sawusch & Newman, 2000), or that there is some other mechanism that allows a longer distal context to have a larger effect (e.g., Wayland et al, 1994).…”

“…For the interpretation of pho-neme categories, speaking rate information has been shown to exhibit a stronger effect the closer it is to the target phoneme. This research suggests that speaking rate is calculated within a small time window of about 250 to 300 ms from the target in both directions (Newman & Sawusch, 1996;Sawusch & Newman, 2000;Summerfield, 1981). Rate context within this time window of approximately one syllable affects the perception of a target sound even if the rate-manipulated material is not adjacent to the target sound (Newman & Sawusch, 1996;Sawusch & Newman, 2000).…”

Listeners recover /t/s that speakers reduce: Evidence from /t/-lenition in Dutch

“…In order to explain the context effects and their interaction with phonetic detail within models making hard decisions on a prelexical level, we have to assume that these decisions are postponed until all segments that may affect the identification of a given segment have been perceived. This requires that decisions about individual segments be deferred for at least 200 ms (e.g., Nooteboom, 1979;Nooteboom & Doodeman, 1980;Sawusch & Newman, 2000). However, Phillips (2001) presented electrophysiological evidence that at least preliminary phonetic categorizations for stops are available around 150 ms after stimulus onset.…”