62
J. MCCOY
contrasts in meaning when compared with all other such sound groups in the given language, ie, hat as contrasted with bar, cat, rat, etc. By convention, phonetic notations are enclosed in brackets, as [ylt2] ‘leaf’, while phonemic notations are enclosed in slant lines, /it2/ 'leaf'. I will follow this convention whenever it is necessary to record the distinction.5 For typographical reasons ad hoc symbolization will be used in this paper to express phonetic and phonemic notation represented elsewhere by special type. These are:
a. [ng] will be used for the velar nasal. As with the aspirate stops, two symbols here represent a unit phoneme.
b. [*], the apostrophe will be used to represent the glottal stop.
c. (ê), a circumflex 'e' will represent the mid central vowel elsewhere written with the inverted 'e' or schwa.
d. [ô] a circumflex 'o' will represent the low back rounded vowel elsewhere written with the reversed 'c'.
* For good descriptions of SC consonants see Chao (1947, pp.18-21) and Wong (1963, Part I, pp. xi-xii),
7 These and other examples may not all be minimal pairs in the strictest sense because of tones differences. However, I found no instances of change in the segmental phonemic structure of a syllable which was correlatable with tone change and I have ignored tone in order to select more familiar examples.
8 The chief reason for setting up the phoneme /kw/ in SC seems to be the fact that this permits a neater distribution pattern when all possible syllable types are recorded. If only /k/ is postulated, the total number of syllable types beginning with /k/ will be about double the average for other initials. If both /k/ and /kw/ are set up, the syllable types for these two initials are about equal in number to each other and to those for other initials. Here again, the arguments seem equally strong for either interpretation but I personally opt in favor of dropping the /kw,kwh/ from the SC analysis. My reasons are to some extent arbitrary and stem first from a desire to make the original phonemic selections on purely phonemic grounds and second from a desire to simplify comparative work with other subdialects which do not have /kw/ under any phonemic approach.
9 In spite of a general preference for postulating a phoneme of length in analyses of SC, there is equally good argument for eliminating length and adding one segmental phoneme. For my work I prefer the second alternative and include a mid central vowel /ê/; again my reasons for choosing this method are based on the resulting convenience in terms of comparing SC with other Kwangtung Province dialects which do not have length phonemes. If we dismiss the interpretations of Wong and Yuan, assuming the former to be purposely overdone for practical or pedagogical reasons and the latter to be more phonetic than phonemic, we find no real economy in a choice between Chao's five vowels plus length or my proposed six vowels without length. In either of these two latter systems roughly the same amount of explanation will do to fit the phonetic facts to the phonemicization. In any case SC length is significant only in the contrasts which Chao writes -aai versus ai, aau versus au. In other occurrences -aa- is described as differing from a in vowel quality, a very clear [a] as opposed to [ê]. When using /ê/ throughout instead of short /a/ the description must read that /a/ and /e/ have their cardinal values in all occurrences except /-au, -ai/ versus /-êu, -ei/ where the difference is essentially one of length; thus /-au/ would be [-a:u], /-êu/ would be [-au], etc.