Effects of rate and talker variability on the recall of spoken words.

Abstract

Changes in speaking rate and talker characteristics have profound effects on the acoustic structure of the speech signal. Current models of speech perception hypothesize that listeners must compensate for changes in rate and talker with a normalization process in which phonetic segments are evaluated relative to the prevailing rate of articulation and relative to specific talker characteristics. The present study was conducted to investigate how listeners adjust or normalize for changes in rate of speech and talker characteristics. A serial recall task was used to evaluate the consequences of rate and talker variability on the early encoding and rehearsal of the spoken words. Listeners were asked to recall words from single versus multiple rate lists; from single versus multiple talker lists; and finally, from lists with words produced by multiple talkers at multiple speaking rates. The results showed that words from lists produced at a single articulation rate were recalled more accurately in early serial positions than words from lists produced at multiple articulation rates; words from lists produced by a single talker were recalled more accurately in early serial positions than words from lists produced by multiple talkers; and finally, the combination of variability in talker and rate did not impair recall relative to lists with variation in rate or talker alone. These findings replicate and extend the results of previous research on talker variability [C. S. Martin etal., J. Exp. Psych:Learn. Mem. Cog. 15, 676–684 (1989); S. D. Goldinger etal., J. Exp. Psych:Learn. Mem. Cog. 17, 152–162 (1991)] suggesting that variations in speaking rate and in talker characteristics incur a processing cost that affects the initial encoding and subsequent rehearsal of spoken words. [Research supported by NIH.]