Effects of Intonation Pattern of Synthesized and Natural Speech Upon Listener Resolution of Semantic Ambiguity

The semantically ambiguous phrase, "I thought so," was tape-recorded using three sources: (1) a professional radio announcer, (2) a computer speech synthesizer employing within-word intonation glides, and (3) a computer speech synthesizer without glides. Each subject heard four intonation patterns from one source and marked the "speaker's" apparent intended meaning for each utterance: (A) "and I was right," (B) "but I was wrong," or (C) "Can't tell which the speaker meant." Analyses indicated: (1) For the announcer and synthesized glide versions, listeners inferred confirmation from falling intonation on thought, and denial from rising thought. (2) Percentages of "Can't tell which the speaker meant" responses and intra-subject response agreements indicated that the announcer utterances were least ambiguous, while the synthesized non-glide utterances were most ambiguous.