Prosody Based Speech Segmentation

Two experiments were conducted to verify whether prosody can be a unit of phonological segmentation. In experiment 1, 24 participants were asked to rate meaningless speech imitating 40 meaningful sound sequences produced by one male speaker. It was found that 94.7% of the selected combinations conformed to Japanese accent rules. Similarly, in experiment 2, 19 participants were asked to rate meaningless speech imitating 76 meaningful sound sequences produced by a different male speaker. 92.8% of the combinations selected conformed to Japanese accent rules. These experiments suggest that native speakers of Japanese can also recognize segment boundaries based on prosody.