^{page 515 note 1} In this paper, ‘Chiehyunn’ refers to all the existing versions of the rime-book called Chiehyunn, and includes also the Tarngyunn and the Goangyunn. All quotations from these rime-books are from the Shyryunn hueybian Peking, 1936.Google Scholar

^{page 515 note 2} On the chorngneou, see especially Fa-Kao, Chou, ‘Studies in the fan-ts'ie doublets in Kuangyün’, Academia Sinica (Peking), XIII, 1948, 49–117,Google Scholar and Tung T‘ung-ho, ‘A preliminary study of the fan-ts‘ie doublets in Kuangyün ‘, ibid., 1–20.

^{page 515 note 3} By ‘rime-set’ is meant the set of two, three, or four rimes which differ only in tone; e.g. the rime-set comprises (In this connexion, it is convenient to take the ruhsheng simply as another tone, for which no special statements are required.) Although taken from the Sonq rime-tables, the rime-sets are probably deducible from the Goangyunn material alone.

^{page 515 note 4} It is convenient to describe the various kinds of finals with terms showing their distribution in the Sonq rime-tables, specifically, the Chi-in Liueh and Yunnjinq The meaning of the terms is self-evident, except that ‘mixed 3rd grade’ refers to the finals occurring with all initials in the 3rd grade, the 2nd grade chyyin (fricatives and affricates), and some 4th grade initials; i.e. Karlgren's ‘type a ’. The ‘pure 3rd grade ’ finals are Karlgren's ‘type β’. (See his Études de la phonologie chinoise, Stockholm, 1915, 625,Google Scholar 626.) My terms are merely loose translations of the terms used in many recent Chinese works on the subject, e.g. Luh Chih-wei's ‘The phonology of Ancient Chinese’ Peiping, 1947, 39.Google Scholar

^{page 515 note 5} Unless otherwise specified, all reconstructions of Ancient Chinese are taken from Karlgren's Grammata Serica (reprinted from BMFEA), Stockholm, 1940.Google Scholar

^{page 516 note 1} ‘Medial i’ here refers to any kind of medial palatal semivowel queleonque, or palatalization, occurring in the mixed 3rd grade rimes and pure 3rd grade rimes only. I follow Luh Chih-wei and others in taking the pure 4th grade rimes (such as and ) as lacking this ‘medial i’. See his Phonology, 6.

^{page 516 note 2} Throughout this paper I have used bold-face type to transcribe gross phonetic features, with IPA symbols, and italics for systematic transcription. I quote Cantonese and Moiyan Hakka (Meihsien from my own knowledge. The Pekingese readings are taken from the Chunghua Press, 1936, the Fuchow forms from Lan Ya-shiu's ‘Phonetic system of the Fuchow dialect’, Bull, of the College of Arts, National Taiwan Univ. , No. 5,1953. The Wenchow examples are from Montgomery, P.H.S., Introduction to the Wenchow dialect, 1893.Google Scholar

^{page 517 note 1} Note also that in two versions of the Chiehyunn (Chieh 3 and Wang 2) the syllable exemplified by also has a 3rd grade final chieh

^{page 518 note 1} See Chao, Y.R., ‘Distinctions within Ancient Chinese’, HJAS, v, 3–4, 1941,Google Scholar 207.

^{page 518 note 2} See his ‘The finals with medial i in Ancient Chinese’,Academia Sinica (Shanghai), XIX, 1948, 217,Google Scholar 218.

It is otherwise with the very daring reconstructions put forward by Luh Chih-wei in his Phonology (esp. pp. 29–46) where most of the wayjuann 2nd and mixed 3rd grade rimes have the same vowels, and differ only in presence or absence of ‘medial i’. While many of his theories are most attractive, it seems safer, for a purely Chiehyunn reconstruction, to keep to the rule’ one rime, one vowel’, at least for the wayjuann.

^{page 519 note 1} This transcription transposes their symbols into Karlgren's symbols. In effect, this change associates the vowel of with that of etc., and, in Karlgren's system, necessitates the change of See Luh, op. cit., 49, 54; Fa-kao, Chou, ‘A study of the phonemes of Ancient Chinese’ Bull. Acad. Sinica (Taipeh), XXV, 1954, 9.Google Scholar

^{page 520 note 1} With one variant graph in ṡăp, to make up six.

^{page 520 note 2} e.g. Tables 13, 14, 23, 24, and 39 of the Yunnjinq.

^{page 520 note 3} It might be pointed out here that this paucity of contrasting syllables probably goes back to the original Chiehyunn. In the older versions and fragments which we possess, although the number of characters appearing twice is much reduced, the number of contrasting syllables remains very small. Especially suggestive are the rime-sets and ; in all extant versions, the former entirely lacks syllables with the retroflex initials. Moreover, the problem is not confined to the wayjuann; certain mixed 3rd grade neyjuann rimes also tend to have an -a- like vowel in many dialects in syllables with these initials, e.g. the rime-set; cf. also below, p. 524. The whole problem of these rimes needs going over again, but for the present it need not affect our transcription of the Chiehyunn rime-system; it is rather a problem for a comparative study.

^{page 521 note 1} See his Compendium of phonetics, Stockholm, 1954, p. 230.Google Scholar

^{page 522 note 1} All examples of the ḥPhags-pa transcription are taken from the Mengguu tzyhyunn (compiled by Ju Tzongwen preface dated (A.D. 1308). I have transcribed the digraph -hi- as -ə-, and -iu- as -ü-. See Dragunov, A., ‘The ḥPhags-pa script and Ancient Mandarin’, Bull. Acad. des Sciences de l'UBSS, 1930, 779.Google Scholar

^{page 522 note 2} For instance, Mackenzie, M.C., A Chinese-English dictionary (Hakka dialect), Shanghai, 1926;Google ScholarRey, Ch., Dictionnaire chinois-français (dialecte hac-ka), Hongkong, 1901.Google Scholar

^{page 523 note 1} Chap. 18 of his Études.

^{page 523 note 2} Seeming exceptions are some rimes in the Korean and Japanese transcriptions. These, however, agree with the Chinese dialects in treating the characters with these initials in both 2nd grade rimes and mixed 3rd grade rimes in the same way. The problem in this case is quite different. Here all the characters in the 2nd grade, of whatever initials, show traces of palatalization.

^{page 523 note 3} See his Études, 657, and Grammata Serica, 56.

^{page 524 note 1} I transcribe the ḥPhags-pa medial graph -h- by -ɔ-, for convenience.

It is interesting to note that in the rime-set, the characters with initial l- and n- had, in medieval Mandarin, the same herkoou development as the characters with the retroflex and plosive dental initials. Thus ḥPhags-pa Iwan and Iwav (Chiehyunn lȧng and lȧk). The former reading seems to have lasted until quite recent times (it appears in the early eighteenth century Inyunn chaanuei ed., Shanghai, 1936), p. 33;Google Scholar see also Chih-wei, Luh, Yenching Journal of Chinese Studies, No. 32, 1947, 187,Google Scholar on a seventeenth century rime-table, where it seems to be taken as ‘liuang’), but has died out in modern Mandarin.

^{page 524 note 2} Cantonese has some irregular forms with the -θ:- vowel, which usually points to a ‘medial i’ in Chiehyunn language. The remarkable thing is that not only the characters with Chiehyunn mixed 3rd grade final ( rime-set) exhibit this vowel, e.g. (Chiehyunn) ṡang (Cant.) sθŋ, but also characters of the pure 2nd grade rime-set e.g. above and (Chiehyunn) ṡȧng, (Cant.) sθ:ŋ) (sub-dialects sɔ:ŋ) and (Chiehyunn) (Cant.) tθ:k.

^{page 524 note 3} supra, p. 520.

^{page 524 note 4} Grammata Serica, No. 355a. Nos. 498a and 499a repeat this reading, but do not mention the ‘irregularity’.

^{page 525 note 1} I agree with Luh Chih-wei (Phonology, 49 ff.) in his discussion of the rime-set but have not changed Karlgren's reconstruction. For the rime-set, see supra, p. 519.

^{page 525 note 2} But and are Pek. Cant, si: Is this a case of a literary reading as against colloquial pronunciation ?