Effects of Text Structure on the Comprehension of Natural and Synthetic Speech

Author: Scott E. Lively, James V. Ralston, David B. Pisoni, and Susan M. Rivera

Abstract:
Subjects were required to listen to passages of connected speech using a self paced sentence-by-sentence listening time task. Passages of natural speech and poor quality synthetic speech were presented in either normal sentence order or in a scrambled order. Following passage presentation, postperceptual word and proposition recognition statements were presented. Results indicated that the voice and order variables interacted in the listening time data. Subjects who listened to passages of synthetic speech in random order displayed the longest sentence listening times. An interaction of passage sentence order and recognition memory sentence type was observed in postperceptual comprehension date. Subjects' performance decreased for proposition recognition memory items, but remained constant for word recognition items across sentence orderings. An attentional explanation similar to the one suggested by Ralston et al. (1991) is given for the present set of results. Subjects who listen to synthetic speech attend more closely to the acoustic-phonetic characteristics of the passages. Listeners who hear natural speech, in contrast, are able to selectively attend to higher levels of the text structure. This different focus of attention may cause listeners who hear synthetic speech to have slower on-line comprehension and to have poorer performance on postperceptual comprehension measures.