The phonetic and phonological triggers for downstep in Japanese: Evidence from perception

Abstract

In Japanese the F0 peak of a word is downstepped after a lexically accented word. Lexical accent can be characterized by its phonological specification in the lexicon or by its phonetic F0 pattern. A perceptual experiment was conducted in order to observe how these two properties interact with each other as downstep triggers. The experiment tested whether listeners compensate for downstep exclusively based on the phonological information of lexical accent or on the phonetic cues available. Participants were asked to judge the relative prominence between two F0 peaks. Results suggest that in order for two words to be perceptually equal in prominence, the second F0 peak must be lower than a given first F0 peak. This difference between the two F0 peaks was significantly greater when the first word was phonetically accented than when it was phonetically unaccented, which suggests that the downstepped F0 peak is perceptually compensated. A similar effect was observed for the phonologically accented first peak, but the effect size was much smaller than the effect observed for the phonetically accented first peak. These results suggest that the primary trigger of downstep in Japanese is the F0 pattern of accented words. Lexical specification plays, at best, a secondary role.