Pharyngealization in early Chinese.

• Author: Norman, Jerry

In the past three or four decades there has been a great deal of ferment in the field of Chinese historical phonology. Many new and exciting proposals have been put forth and new reconstructions of early stages of the language have been attempted.(1) Yet several vexing problems continue to challenge people working in the field; one of these is the origin of the pervasive palatalization found in our chief documentary source for the study of Chinese historical phonology, the Chiehyunn rimebook compiled by Luh Faayan at the junction of the sixth and seventh centuries A.D. The chief purpose of this paper is to suggest a new model to account for this feature of the Chiehyunn.

The Chiehyunn and its subsequent redactions in themselves reveal only a very simple analysis of Chinese phonology. The syllable is the basic unit of analysis; this is undoubtedly because the Chiehyunn was a dictionary of graphs, each representing a single syllable. In the Chiehyunn, all syllables were first analyzed into four tones: pyng, shanq, chiuh, and ruh; each tone was then broken down into yunn "rimes." Under each rime, the syllables were divided into sheauyunn "small rimes," which were groups of homophonic characters. The first graph of each homophonic group was provided with a faanchieh formula which in essence "spelled" the graphs of each homophonic group. Looked at in this way, the Chiehyunn is a syllabic inventory arranged according to tone, rime, and small-rime or homophonic group. Faanchieh is a device whereby the pronunciation of a graph is indicated by the use of two different graphs, the first of which represents the initial consonant and the second the rime and the tone. Faanchieh was not entirely systematic because the same initial could be (and normally was) indicated by several different graphs and the same rime likewise by more than one character. While such a device worked in a practical sense, it did not represent a very high level of phonological sophistication.

With the coming of Buddhism to China, concepts from Indian phonology were also introduced. It is very likely that knowledge of Indian alphabetic writing was the impetus that led to the discovery of the thirty-six initials of medieval Chinese phonology. Not only were the initials themselves identified, information concerning their places and manners of articulation was also provided. With such detailed knowledge of the initial consonants (represented in the faanchieh formulae by the upper graph), it was possible to study the co-occurrence patterns of initials and rimes. These patterns provided the most important element that produced the notion of deeng "rank" which is the single most important concept underlying the Sonq rimetables.(2)

The oldest extant timetable is the Yunnjinq or "Speculum of Rimes" which was composed during the Northern Sonq period (Lii Shinkwei 1983: 164). The phonological theory on which this table is based, however, is clearly older. We know from the fragmentary Yunnshyue tsarnjiuann of the monk Shoou'uen, which was discovered at Duenhwang, that the concept of deeng existed already at the end of the Tarng.(3)

Medieval Chinese phonologists, in their study of the co-occurrence relationships among initials and rimes, developed a system in which all of the rimes of the Chiehyunn could be classified into one of four deeng or ranks. In doing this they were inevitably influenced by their own vernacular, which clearly differed from the Chiehyunn in numerous ways. The results of their analysis were eventually presented in the form of a tabular matrix, with initials listed horizontally across the top of the table, and the rimes arranged vertically according to tonal category and rank. The notion of deeng or rank seems to represent a valid insight into the nature of the Chiehyunn system. In effect, this analysis divides all syllables into four basic categories characterized chiefly by the classes of initial consonant permitted in each syllable class.

In order to discuss the origin of the palatalized syllables of the Chiehyunn more clearly, I will make use of a reconstruction based on the faanchieh of the Chiehyunn, the rimetable categories and the phonetic values of these categories in modern Chinese dialects as well as the school pronunciation of Chinese characters employed in Korea, Japan, and Vietnam when reciting Chinese texts. The main outlines of this reconstruction were worked out by Bernhard Karlgren; for the sake of simplicity I have incorporated certain revisions to Karlgren's system suggested by Li Fang-kuei (1971) and W. South Coblin (1986). In addition I will add a medial *j in all third-division type times even where the above scholars have omitted it.(4)

Before proceeding further, I would like to make a distinction between what I will call "rank" and "division." I will use rank to refer to the number of the line in a table to which a given graph is assigned. "Division" will be used to refer to the classes of times established on the basis of co-occurrence patterns. There is a simple one-to-one correlation between division and rank for divisions one, two, and four: division-one words are all placed in the first rank; division-two words are all placed in the second rank; division-four words are all placed in the fourth rank. Since certain initials that occur in division-three words were restricted to the second and fourth ranks, words in this division that have the initials in question may be moved (so to speak) from the third rank into the second or fourth rank. For example, all words with retroflex sibilant initials - the initials juang (*ts), chu (*tsh), chorng (*dz), and sheng (*s) in current Chinese terminology - are placed in the second rank; therefore syllables like tzou *tsjeu and shuang *sjang are placed in the second rank, although they are in a division-three rime and in the reconstruction employed here have the division-three medial *j. The plain sibilants as well as the so-called yuhsyh (*ji) initial are placed in the fourth rank when they occur in division-three rimes: shiou *sjeu,jang *tsjang and yang *jiang will all be found in the fourth rank even though they are division-three type words. In addition, chorngneou words, that is, words with labial and guttural initials, which in certain rimes have double ocgurrences, are divided between the third and fourth ranks despite the fact that they are all division-three type words; thus mean *mjan will be found in the third rank while mean *mjian will be found in the fourth rank. My use of "division" corresponds to the usage of contemporary Chinese linguists; it is found, for example, in the widely used Fangyan diawchar tzyhbeau.(5) Hereafter I will employ the term in the sense defined above.(6)

There is nothing sacrosanct about the four divisions. Other analyses of the syllabic inventory of the Chiehyunn are possible without abandoning the fundamental insights of the medieval phonologists. Lii Rong, for example, in his well-known 1956 work, Chiehyunn inshih, gives a somewhat different interpretation of the divisions. In his scheme, the Chiehyunn finals are divided into six categories; three of these correspond to the traditional divisions one, two, and four of the rimetables, but the third-division finals are subdivided into three new categories based on the co-occurrence patterns of initials and finals. Below I would like to propose a simpler system of divisions using basically the same criteria as those employed by the framers of the early rimetables.

In examining the system of divisions, it seems at first sight surprising that the set of initials occurring with the first and fourth divisions are identical, while the second and third divisions are nearly characterized by different sets of initials. This raises the important question of why the first and fourth divisions were differentiated at all. To understand this we must remember that the notion of deeng (division) appears at least two centuries after the completion of the Chiehyunn; hence...