The phonetics and phonology of shiki-accent systems in mid-western Japanese dialects

For the mid-western region of Japan, a lexical contrast in word prosody shiki has been documented, defined as "the direction of pitch within a word" in Japanese accentology ("dynamic view"). Four experiments on f 0 production were conducted on the speech data from 22 speakers in 7 cities in the north-central Shikoku and Kinki areas. The first experiment revealed that word-level fundamental frequency (f 0) contour systematically varies in its degree of downtrend depending on the number and structure of shiki types. This shows that shiki downtrend cannot simply be reduced to the f0 declination that defines the backdrop global f0 baseline, but a part of phonetic knowledge of the speakers to maximize perceptual distinctness. The result also shows the cross-dialectal variation in the peak timing within a word. Combined with the first result, we can identify the 3 shiki-accent systems: System-A: Central Kinki; Level vs. Rising-shiki: System-B: Ibukijima; Level vs. Falling vs. Rising-shiki: System-C: Western Kagawa & Tokushima; Falling vs. Non-falling-shiki. The second and third experiments demonstrated that the word-level f 0 contour patterns are largely unmodified under f0 range rescaling across words governed by the same principle in the 3 systems above: downstep: compression of f0 range after falling f0 contours; upstep: expansion of after rising f0 contour. The results lend further support to the typology of shiki systems and to introducing the 3-way contrast of tones: H(igh), M(id) and L(ow). The last experiment showed that f0 downtrend in Falling-shiki is different from that of Level-shiki in system-A in that both the degree and the size of f0 fall are correlated with duration of the f 0 fall, consistent with the "tone target model" and the "constant rate model" of f0 production, respectively, latter of which is better understood in light of the dynamic view. Overall, the findings calls for a reconsideration of some of the fundamental tenets of the f0 production model called the AM theory by showing that (i) f0 production patterns of shiki-accent systems cannot be completely captured by binary tone level and (ii) the concept of tone as the sole universal phonological unit for f0 production may not be maintained