Patterns of Induced Speech Errors and Their Implication for the Mora

Speech error data have been known to be informative of phonological units that are functional during speech processing. Using short-term memory tasks, we gathered speech error data from Japanese speakers to investigate how they combine segments into higher-level suprasegmental frames. The findings suggest that the phonological affiliations of segments are an important factor to Japanese speakers’ speech errors. Specifically, onsets and vowels are more likely to be kept intact as a unit than vowels and codas. We suggest that this finding supports the idea that Japanese speakers’ more permanent knowledge of the moraic structure contributes to the their short memory recall of segment sequences.