Discourse comprehension of synthetic speech across three Augmentative and Alternative Communication (AAC) output methods

The purpose of this investigation was to determine the relative effects of three different Augmentative and Alternative Communication (AAC) speech output methods (word, sentence, mixed words and letters) on a listener's ability to summarize paragraph-length texts. Based on previous work on the discourse processing of synthetic speech, a monotonic trend in a listener's ability to accurately summarize texts produced by different output methods was hypothesized (word > sentence > mixed). Thirty able-bodied adults were individually tested over a 2-day period, listening to four texts produced by a DECtalk speech synthesizer at a speech presentation rate of 7.5 wpm. Using a discourse summarization taxonomy developed by Higginbotham, Lundy, and Scally (1993), the experimental hypothesis wee confirmed. Word-method listeners produced significantly more accurate renditions of the original texts than sentence-method listeners, who, in turn, did better than mixed- method listeners. Statistically significant differences also were found for the topic familiarity of the texts. The role of the above variables on AAC device comprehension and technology design is discussed.