자립 자립분절음 음운론

Abstract

This study is an attempt toward a revised theory of Autosegmental Phonology the typical models of which are put forward in Goldsmith(1976) and Haraguchi(1977). In the approach taken here the autosegmental tier is represented as a sequence of slots associated in a one-to-one fashion with the P-bearing units on the segmental tier. As for tonal autosegments, a slot filled with a tone is represented by H or L, and an empty one by OH. A single Tone-bearing unit (TBU) may be associated with more than one slot, but not vice versa. A derivation within this model is illustrated with an example from the Tokyo dialect of Japanese(cf. Haraguchi 1977) :

(1) a t a m a - made 'head + even' ｜ ｜ ｜ ｜｜ OH OH H H OH Underlying Representation OH H-Erasure L Polarity one Assignment (6.a.ⅰ) H H L Autosegment Spreading(8) L Initial Lowering Rule(3) L H H LL Surface Representation

The structure of phonological representation presented here is supported by abundant evidence drawn from a number of languages, especially from the Kosung dialect of Korean and the 15th-century Korean; and it is shown that it entails the literal independence of the autosegmental tier. For instance, in the model which allows one single tone to be linked with more than one TBU, Initial Lowering Rule which applies in deriving surface forms like the one in (1) should be stated as in (2), as is the case with Haraguchi (1977)

(2) VC_OV VC_OV ＼／→ ｜ ｜ ／ ＃＃ C_0 ___ H L H

Within the framework of A(uto)-A(utosegmental) Phonology, however, it is formulated as in :

(3) H → L ／＃＃ ___ H

As is clearly seen, this rule refers only to the tonal tier independently of the segmetal one. Tone rules may be stated in as simple and independent a manner as this, and even in case they must refer to the segmental tier, the reference will be remarkably simpler. Tone languages are classfied typologically as core-tone languages (formerly pitch-accent languages) and non-core-tone languages (formerly true tone languages). The former need universal Polarity Tone Assignment stated as in:

(4) OH → -αT % αT ___

The environment αT is the core tone, which is either lexical or assigned by rule. This rule is expanded into the following subrules:

(5) a. OH → -αT / αT ___ b. OH → -αT / ___ αT

These in turn are expanded into the following:

(6) a.ⅰ. OH → L / H ___ ⅱ. OH → H / L ___ b.ⅰ. OH → L / ___ H ⅱ. OH → H / ___ L

The Kosung dialect described in §2 requires rules (6.a.ⅰ) and (6. b.ⅰ). As is exemplified in (1), Japanese dialects employ either rule (6.a.ⅰ) alone or both rule (6.a.ⅰ) and rule (6.b.ⅰ) (cf. Haraguchi 1977). And rule (6.b.ⅱ) is operative in Tonga (cf. Goldsminth 1981a). On the other hand, among non-core-tone languages are Chinese, the 15th-century Korean described in §3, Mende (cf. Leben 1973 and Goldsmith 1976), Margi (cf. Williams 1976), and Kikuyu (cf. Clements and Ford 1979). In place of Polarity Tone Assignment (4), non-core-tone languages adopt universal Tone Mapping Rule (7), to which language-particular condition may be added.

(7) Assign a tone from the Bisic Tone Melody to each OH in a one-to-one fashion from left to right.

Polarity Tone Assignment (4) and Tone Mapping Rule (7) alike feed universal Autosegment Spreading (8), which may also be qualified by language-particular condition.

(8) Spread α, where α=the feature(s) of autosegment

As in Hyman (1982 b,c), Poser(1983) and Pulleyblank(1983), the diacritic use of the abstract accent marker is abandoned in favor of the lexically linked core tone. And the device of default rules proposed in Pulleyblank(1983) is shown to be untenable. Lastly, the tone systems of the Kosung dialect and the 15th-century Korean are extensively analyzed to serve as sources which provide evidence in support of the revised model of A-A Phonology and as its testing grounds.