Effects of practice on speeded classification of natural and synthetic speech

Abstract

At the last meeting of the Society in Miami, one of us (Pisoni) presented results from a speeded classification task demonstrating increased processing time for recognition of synthetic speech. In a lexical decision task, we showed an overall increase in response time of 145 ms for synthetically produced words and nonwords over naturally produced control items. Moreover, there was no interaction between signal type and classification response suggesting that the observed differences were due to early stages of perceptual analysis in which the segmental representation is developed from the acoustic‐phonetic input. In the present investigation, we were interested in determining whether the observed differences in processing time between natural and synthetic speech could be attenuated or possibly eliminated by practice with the experimental materials and task over several days. We ran ten undergraduate subjects for five days in the lexical decision task. As expected, accuracy improved and latency decreased over the test sessions. However, the relative differences in response latency between natural and synthetic items remained roughly constant. Thus, the earlier differences observed in this speeded classification task are not due to short‐term familiarity effects. Rather they seem to reflect real differences in the perceptual and cognitive processes used to extract segmental information from the speech signal. [Supported by a research grant from NINCDS to Indiana University.]