Pitch Accent in Japanese: Implementation by the C/D Model

In Tokyo Japanese, lexical accent is implemented by pitch pattern control, while phrasal stress patterns, along with pitch variation, convey non-lexical information in discourse. The C/D model represents pitch control by the tonal melody and stress control by the skeletal organization of the utterance. Phonetic implementation of pitch contours is exemplified here for different lexical accent patterns, including interphrasal interaction (catathesis). An example of phonetic variability is discussed. The issue of accentbearing units (syllable vs. mora) is revisited. A principle of nuclear prominence (PNP) is proposed explaining why pitch drops in the middle of an accented long syllable. 1. Pitch accent in Japanese Lexical accent in Japanese is generally implemented by pitch patterns, which vary greatly from dialect to dialect. While some dialects do not use any prosodic contrast lexically, Tokyo dialect, among many others, does have contrastive lexical patterns, which, as time functions of F0, have been widely discussed with acoustic data (Pierrehumbert [1980], Poser [1984], Pierrehumbert & Beckman [1988], Kubozono [1995], Kawakami [1995], Sugito [1996]). The phonological and (abstract) phonetic descriptions are typically couched in terms of a high vs. low binary status, using each mora as the accent-bearing unit. Hattori [1961] (and others following his work) used the accent kernel designated on a mora of a word to indicate a characteristic pitch change, in most dialects a sudden pitch drop, from the pertinent mora to the next. The kernel is given once, at most, for an accentual phrase (Fujimura [1966], McCawley [1968], Hayata [1997]). The pitch change due to the phonological accent kernel can also be a sudden rise, rather than the usual drop in F0, as in the Sizukuisi dialect, according to Uwano (Uwano [2007]). Some southern dialects of Korean exhibit pitch accent systems similar to Japanese dialects, in that an accent kernel is specified for a word (Hayata, [1997]). In addition to the pitch change due to the accent kernel, there may be lexically specified phrase-initial syllabic pitch pattern contrasts. Kansai dialects, spoken in Kyoto, Osaka, and elsewhere, exhibit this additional prosodic lexical feature (see for discussion based on data in Sugito [1996] with a report on phonetic variability referring to diachronic changes, Kinsui [1999]). In these dialects, according to the mora-based account, either high or low word-initial pitch is specified for lexical items independently from the designation of the kernel. There are other pitch changes that are not lexically distinctive, such as the rise from the initial mora to the next in Tokyo dialect, which can be described as a general phrasal characteristic, noting that even words spoken in isolation must be framed in a phrasal environment in order to be uttered. Thus in Tokyo dialect, a phrase typically starts with a mora in low pitch and the pitch goes up for the next mora, unless the first syllable is accented (Fujimura [1966, 1972], also see infra for long syllables). Depending on the discourse function of the utterance, a phonetic phrasal specification evokes different step response functions at phrase boundaries. A case of this is commonly observed in most languages in yes-no questions at the end of the utterance. Such stepresponse functions are selected according to phrase boundary features and their phonetic realization is senseitive to the given discourse situation as well as dialect. Fig. 1: CD diagram comparing three accent types of the key word: from top, /hasi/ “edge” (no accent kernel), /hasi’/ “bridge” (kernel on the second syllable), and /ha’si/ “chopsticks” (kernel on the first syllable), respectively, in ‘Hasi ga aru.’ (There is/are hasi). The last word (copula) ‘aru’ has an inherent accent kernel on the first syllable. In the case of the last two sentences, the pitch dropping effect (b) of this kernel of /aru/ is subject to a substantive reduction (traditionally considered elimination) by catathesis [Poser 1984], as shown by the curve (b’), due to the accent kernel of the word /hasi’/ and /ha’si/, respectively, that is contained in the preceding phrase. Also, the phrase-initial ascending pitch, (a), and its reduced form, (a’) in the case of catathesis, represent the property of (the edge of) the phrase. For the first sentence with /hasi/ with no kernel specified (and therefore no catathesis), no pitch resetting occurs at the small phrase boundary before ‘aru’. The F0 time functions, shown here by solid curves, are implementations of the underlying step functions, broken lines, of the phonetic status contours. The bottom panel shows syllable (solid) and boundary (broken) triangles, the size of each representing the syllable or boundary magnitude. In intonation theories (see McCawley, $hasi'#ga%aru$ $ha'si#ga%aru$ $hasi#ga%aru$