Word-tone dialects in Kyushu, where each lexical item in the speaker's mental lexicon has one of two tones, i.e., melodies, are found throughout western and southern Kyushu. Every lexical item is thought to be redundancy-free since the grammar is designed to minimize the amount of information that must be stored. Accordingly, each lexical item of Kyushu word-tone dialects must carry only one bit of tonal information. Predictable phonetic information is supplied by rules or constraints.
The "Yotsugana" Phonemic System disappeared from the Kyoto and East Japanese dialects in the late 17th century, yet the system has remained in the Kyushu dialect. At the present time, the "Yotsugana" Phonemic System is distributed in certain districts in Kyushu, and only old aged persons now use this system. These districts are Saga Prefecture, the southern part of Fukuoka Prefecture, Hita and Kusu counties of Oita Prefecture, the central and southern parts of Miyazaki Prefecture, and Satsuma Peninsula and Osumi Peninsula of Kagoshima Prefecture. The merging process of "Yotsugana" Phonemic System varies according to the phonetic environment. (1) First, the distinction between /di/;/zi/ disappears , and then the distinction between /du/;/zu/ disappears. (2) The medial affricate consonant changes into a fricative, and the initial affricate consonant remains.
The te-form verbs in the (West-)Kyushu dialects have a characteristic behavior in the verb conjugation system, which we call the "te-form phenomenon". This phonological phenomenon is that so-called SOKU-ON or HATSU-ON appears, corresponding to te in the standard dialect. This phenomenon depends on the difference of the stem-final segment. The te-form phenomenon is different across dialects. We assume that the e-deletion rule is the core of four rules governing this phenomenon. Differences of dialects are reflected in the applicational environment of the e-deletion rule. In addition, we discovered a "pseudo te-form phenomenon", which is partially similar to te-form phenomenon.
In the Fukuoka dialect, flat high pitch spreads between a WH-word and a [+WH] COMP which binds it. Two assumptions are made to explain this phenomenon: (i) there is a rule which forms a phonological phrase between a WH-word and a [+WH] COMP, (ii) only the underlying accent of lexical head will surface. These assumptions also hold for the Pusan dialect of Korean, which shows striking similarity to Fukuoka Japanese.
One characteristic of the Nagasaki dialect is that it has a "Nikei-accent system" that is widely distributed in the south-western part of Kyushu. This accentual system has two contrastive tonal patterns called "A" and "B", irrespective of the number of moras in the word. The first pattern involves a fall in pitch within the word, while the other pattern does not. Another characteristic of the two-pattern system is that the accentual pattern of compound words is determined by that of their leftmost element. The Nagasaki dialect shares these two characteristics with the Kagoshima dialect, but crucially differs in that it is a mora-counting rather than syllable-counting dialect.
In this paper I report two kinds of sound changes in Kagoshima dialect. First, I report close vowel's changes, such as [a?] (<aki, 'autumn'), [a?] (<aʒi, 'taste'), [a?] (<abu, 'horsefly'), [a?] (<aku, 'harshness'). These changes, commonly called 'nisshoka', are caused by devoicing final [i] or [u]. Second, I report alveolar's changes in Segami dialect, spoken in Kami-Koshiki island, Kagoshima prefecture, such as [arama] (<atama, 'head'), [abuja] (<abura, 'oil'), [neːzii] (neːzu, 'temperature'), [kuːnu] (<kudzu, 'trash'). In this dialect, these changes occurred very systematically, and so the distinction between /t/, /r/, /c/ and /d/ was not lost after the changes.