Perception of synthetic speech by children

Abstract

Many educational and commercial products that produce synthetic speech output are becoming widely available. However, there is little, if any, research on how adults or young children process (i.e., perceive, encode, retrieve) synthetic speech signals. Results of two experiments that studied children's perception of natural and synthetic speech indicated that subjects consistently showed a higher error rate when listening to synthetic speech than natural speech. In one experiment, children 5–8 years old listened to natural and synthetic speech tokens in a task that required them to point to a picture, from among four alternatives, that corresponded to the word heard. In a second experiment, children 8–10 years old listened to natural and synthetic digit strings and repeated the digit strings to the experimenter exactly as heard. Performance deficits were observed for the synthetic speech relative to the natural speech even for highly familiar material such as digits. These results have implications for the design, selection, and use of voice-response systems to be used in teaching and learning environments with young children and adults. Synthetic speech may require additional processing capacity and attentional demands that may well produce large decrements in other complex cognitive tasks that children are expected to perform routinely in educational settings. [Supported by NINCDS.]