Historical dialectology and the Poetry Classic

« previous post | next post »

[This is a guest post by John Carlyle written in response to the following comment by E. Bruce Brooks to "Similes for female pulchritude in an ancient Chinese poem" (7/1/20):

The formation of the Shr* corpus is currently under serious study, and it can be said with some certainty at this preliminary stage that this particular poem was added to the growing Shr collection at the end of the 05c. How much older it may be, in its own country (Wei) will depend on scrutiny of its dialect position: some of the poems from that area show traces of (original) local pronunciation; others do not. Stay tuned.

*Shījīng 詩經, aka Poetry Classic, Classic of Poetry, Shijing, Shih-ching, Book of Songs, Book of Odes, Odes, or Poetry.]

   There is justification that Wey's 衛 dialect position might suggest something about the age of some of the poems in the Wey airs. The dialect position of Wey is better understood for the later period. What that might suggest about earlier poetry is still not clear. I'll try to give a quick summary of what we know so far.

   At least by the time of Fangyan 《方言》, Wey belonged to an eastern group of Chinese dialects. The exact limits of this eastern group are not entirely settled nor are the phonological features shared by the group since studies of Fangyan are primarily lexical. Since the time of Lin Yutang's (1927) first approximation of Fangyan dialect boundaries, the dialects of Wey and Song have been grouped together. Later scholars also included neighboring states like Qi (but not "Eastern Qi") and Lu. More recently, Matsue Takashi (1999, 2006, 2013) argues that the eastern group's boundaries extent as far as Chen and that the dialect of Chen was a transitional dialect between the eastern and southern groups due to Chu incursion (2006, 2013).

   As far as phonology goes, the defining feature of the eastern dialect group is contacts between yin and yang Old Chinese (OC) rime groups, e.g. 歌 vs 元, 脂 vs 真, 脂(微) vs 文. However, for a long time, this phenomenon wasn't correlated with dialects but was rather characterized as an "alternation" (轉. By a curious stroke of fate, Yang Xiong himself used this exact term to explain phonetic variation in dialects, e.g. 語之轉也, though I suspect this has little bearing on the Qing philologists' conception of OC "alternations."). Karlgren (1954) reconstructed an -r coda to account for the rhyming alternations with -n coda words, i.e. -ar vs -an, -ǝr vs -ǝn, etc.

   In their study of Han rhymes, Zhou Zumo and Luo Changpei (1958) began to correlate this alternation with dialects and noted that the yinyang alternation seemed to correspond to what commentators called the speech of Qi, Song, and Chen.

   Coblin (1983) explored this alternation further in his study of Eastern Han sound glosses in the glosses of Du Zichun, Zheng Zhong, Xu Shen, Fu Qian, and Ying Shao. The hometowns of these scholars correspond to the eastern portion of the central dialect group, the eastern group, and the southern group of Serruy's (1959) Fangyan dialect groups. They all fall into the larger eastern group of Matsue's more recent study. Coblin explained the alternation in terms of final -r loss, nasalization of -n coda words, and then loss of nasalization, e.g. *ar > a, *an > ã > a. This explanation assumes a full merger of the alternate rime groups.

     Starostin (1989) observed that the yinyang alternations of the rime groups we are investigating tend to correspond to a subset of symphonic series and rarely correspond with rime table division II words. He posited that originally there were three coda, /-n, -r, -j/, increasing the total number of distinct rime groups.  In some dialects, the -r coda mergers with -j, in other dialects it merged with -n. He observed this merger must have been in progress during the times of the Odes, but he didn't take a stance on exactly when the merger was in progress of where. He did note that the merger of -r and -n seemed to be later, likely during the Warring States period. The rhyming practices of Qu Yuan, Song Yu, and Xun Zi all keep -r and -n separate, but the Laozi merges them. Starostin's proposal allows for a partial merger in the eastern dialect group. I'll try to illustrate this below using the 歌元 alternation.

OC       Eastern trend (-r > -j)             QYS trend (-r > -n)
ar         aj (Coblin 1983: a)                  an
an        an (Coblin 1983: a)                 an
aj         aj  (Coblin 1983: a)                 aj (Coblin 1983: a)

   Baxter (1992) acknowledged Starostin's discoveries but chose to explain the contacts as an unconditional change of -n codas into -j (where -j is reconstructed for Karlgren's old -r, sorry this is rapidly becoming confusing), e.g. aj vs -an > -ãj > -aj. Again, like Coblin, this proposal only allows for a full merger. Schuessler's OC also explains the alternation in these terms and only reconstructs the codas -n and -j. Baxter

   Baxter and Sagart (1992) followed Starostin in reconstructing a third -r coda and explicitly pinned the merger of -j and -r to eastern China around Shandong based Luo and Zhou and Coblin's observations. Strangely, they draw the boundary between -n, -r, and -j coda rhyme groups different than Starostin, but not elaborate on the bases for this change. It seems to be tied to Han renderings of foreign names ending in -r first observed by Pulleyblank, but they don't set out strict criteria for how they interpret this data.

   Now, back to the issue of chronology. We are interested in a merger of OC -r and -j codas (though scholars that argue for the distinction disagree exactly where the -r coda should be reconstructed). To find explicit instances of this merger, we need to find instances of Qièyùn System (QYS, sometimes called “Middle Chinese” or “Ancient Chinese.” I give transcriptions in Baxter’s (1992) system.) open or -j ending rimes mixing with QYS -n ending rhymes. This assumes that fanqie readings in Qieyun 《切韻》(QY) largely follow the trend -r > -n, but not -r > -j. (Of course, there are exceptions and doublet readings reflecting both trends, but they are likely not statistically significant. I feel more or less comfortable making this assumption.) We could also look for cases of -n ending symphonic series graphs mixing with -j ending symphonic series graphs, but then we have to take a leap of faith that the orthography in the received editions of the Odes accurately reflects a distinction of -r and -j as it was during the composition of the Odes in dialects where they did not merge. (I am not comfortable with this assumption).  

   Here's all the cases of -j and -r merger I found using this standard. Cases of QYS open and -j ending rimes are underlined. I have marked cases where the rhyme is not obligatory with an asterisk:

Section

Subsection

Rhyme Sequences

 

國風

邶風

43.1a(沘彌鮮),43.2a(洒浼殄)

 

鄘風

*47.3a(展袢顏媛)

 

陳風

137.2a(麻娑)

小雅

彤弓之什

182.3a(晨輝旂)

 

桑扈之什

222.2a(芹)

 

都人士之什

228.1a()

大雅

文王之什

*241.6bc(原)

 

生民之什

*245.1(嫄)

魯頌

299.1a(芹)

           


   These are just cases I was able to identify on a single pass. It's probably not an exhaustive list.

   It's important to remember that not all of these are definitely cases of -r > -j merger. Some of them might be cases of QY reflecting the merger in fanqie readings that reflect the eastern dialect. I am most skeptical of 182.3a,222.2a, and 299.1a considering the phonetics involved in the graphs suggest -r codas that "should" have become -n in the QYS. However, the rest probably do reflect the merger. It is no coincidence that the cases that occur in the Airs all coincide with Matsue's eastern dialect continuum (or in other words Wey, Song, Qi, Lu, Chen). 

   So, the geographic distribution does seem to be around the area of Wey (though the Airs of Wey don't not have an immediately identifiable example, more on that in a second). But what about this distribution through time? Without reading up on the scholarship of the dating of the different parts of the Odes, I can't really make a sound hypothesis about what my sample tells us. However, Baxter (1992) points out that this type of merger also occurs in bronze inscriptions studied by Zhou Fagao et al. My intuition is the merger itself is quite old, perhaps even as old as the oldest pieces in the Odes. I am just unsure where the change originated and exactly when it would have diffused to Wey if it did not originate there.

Now, about Ode 57. We do have one curious line:

碩人其頎、衣錦褧衣。

   頎 has the reading gj+j in QY. However, Lu Deming gives a sound gloss as 懇 khonX in Jingdian shiwen 《經典釋文》 (JDSW). Likely before the QY, other -n coda readings of this graph, like gj+n, would have possible. As far as the OC goes, 斤 as a phonetic indicates an -r coda according to both Starostin and Baxter and Sagart. 衣 indicates a -j coda according to both. If we can establish that 頎 was not an orthographic innovation to write a -j coda word in the received text of the Odes, we can establish a case of -r > -j coda merger in this poem and more general in Wey during the time it was composed.

   The graph 頎 probably isn't an innovation. This "long" word seems to have been known among Han commentators like Mao Heng, Zheng Xuan, and Xu Shen. Xu Shen gives a 讀若 reading of 頎 as 鬢 and does the same for 芹. Likely he pronounced them the same (though he may have had the -r > -j merger himself). I think this really is a case of an -r > -j merger.

   In summary, the Wey dialect seems to have belonged to the eastern group and Ode 57 contains a rhyme that exhibits a well-known eastern dialect feature. When the -r > -j merger completed in Wey is difficult to date. My intuition is that the merger, in general, is quite old, but that doesn't necessarily mean it was a feature of the Wey dialect as far back. One final musing: Since Wey is located around the old Shang capital, it might be worth inspecting OBI for cases of -r > -j. If there are none, we might be able to at least put a cap on how far back we can project the merger. Presumably, there would be none, unless unique -r coda phonetics are an innovation of the Zhou script.

Sources

Baxter, William H. 1992. A Handbook of Old Chinese Phonology. Trends in Linguistics. Studies and Monographs [TiLSM] 64. De Gruyter Mouton.

Baxter, William H. and Laurent Sagart. 2014. Old Chinese: A New Reconstruction. Oxford: Oxford University Press.

Coblin, W. South. 1983. A Handbook of Eastern Han Sound Glosses. Hong Kong: The Chinese University Press.

Karlgren, Bernhard. 1954. “Compendium of Phonetics in Ancient And.” Archaic Chinese  Reprinted from Bulletin of the Museum of Far Eastern Antiquities, Stockholm 22.

Lín Yǔtáng 林語堂. 1927. “Qían Hàn fāngyīn qūyù kǎo” 前漢方音區域考. Yǔyánxué lùncóng, 16–44.

Luó Chángpéi 羅常培, and Zhōu Zǔmó 周祖謨. 2007. Hàn Wèijìn Nánběicháo yùnbù yǎnbiàn yánjiū 漢魏晉南北朝韻部演變研究. Vol. 1. 2 vols. Běijīng: Zhōnghuá shūjú.

Matsue Takashi 松江崇. 2013. "Old Chinese Dialects According to Fangyan." Dialectologia (Special Issue IV): 181-197.

—. 2006. "Hàndài fāngyánzhōng de tóngyánxiànshù yě tán gēnjù Fāngyán de fāngyán qūhuàlùn 漢代方言的同言線束–也談根據《方言》的方言區劃論." Huáxuéchéng 1509-1533.

—. 1999. Yōyū Hōgen chikujōchizushū 揚雄《方言》逐條地圖集. Vol. Project 09301022 Vol. 4. Tōkyō: Ministry of Education Science and Culture Japan.

Serruys, Paul L.-M. 1959. The Chinese Dialects of the Han Time According to Fang Yen. Berkeley: University of California Press.

Starostin, Sergej. 1989 [2010]. Gǔdàihànyù yīnxì de gòunǐ 古代漢語音系的構擬 (Chinese Translation of Rekonstruckcija drevnekitajskoj fonologic̆eskoj sistemy, Moscow: Nauka, Glavanaja Redakcija vostoc̆noj Literatury 1989). Translated by Sūn Shūxiá 孫淑霞. Shànghǎi: Shànghǎi jiàoyù chūbǎnshè.



12 Comments

  1. Chris Button said,

    July 9, 2020 @ 1:22 pm

    it might be worth inspecting OBI for cases of -r > -j

    The problem, as I see it, is the inconvenient truth about the lateral -l coda. Pulleyblank reconstructed it, but it throws off the symmetry of many other reconstructions, and so is unfortunately often ignored.

    And so we end up with things like -ij/-i (should we really assume a surface distinction of [i] versus [iː] ?) being used for -əj such that -əj is then used for -əl !

    Coda -r quite simply merged with coda -l before Old Chinese. Coda -l shows a partial merger with coda -j during Old Chinese. Unsurprisingly coda -l also shows contacts with coda -n.

    Personally, I would reconstruct 衣 in Old Chinese with a lateral coda as *ʔə̀l. There is an overlap in semantic fields of its derivative 依 *ʔə̀l "reliant" with 因 ʔə̀ɲ < ʔə̀n (depicting an outstretched person on a mat in the OBI and in a ə/a ablaut relationship with the resting person in 安 ʔán). I think the overlap is coincidental and possibly fostered by the phonologically similarity, but a more primordial connection could be argued for.

  2. David Marjanović said,

    July 9, 2020 @ 2:13 pm

    an unconditional change of -n codes into -j

    That would be very odd if it happened in one step. But in two steps, [n] > [l] > [j] would be easy, and the interesting option of [n] > [r] > [j] doesn't seem outright impossible either.

    should we really assume a surface distinction of [i] versus [iː] ?)

    A surface distinction of [i] vs. [ij] isn't unheard of; it has a role in Russian grammar for instance. But this is not very common and should not be reconstructed before any alternatives have been exhausted.

  3. Michael Watts said,

    July 9, 2020 @ 5:47 pm

    and the interesting option of [n] > [r] > [j] doesn't seem outright impossible either.

    I wouldn't think so, since there are modern Chinese today who do not distinguish Mandarin "r-" from /j-/. I had a tutor who was endlessly amused that the radical in 脸/脂/腿 etc. historically derives from 肉, since to her "rou" and "yue" sounded so similar.

  4. Paul R. Goldin said,

    July 9, 2020 @ 8:50 pm

    FWIW (relevant to 碩人其頎、衣錦褧衣): 衣 (presumably *-j) = 殷 (*-r) is an attested jiajie. The most famous example is probably 壹戎衣 in Zhongyong, for which the parallel in Kanggao reads 殪戎殷. Also, the King of Shang is called 衣王 in Zhouyuan OBI.

  5. Chris Button said,

    July 9, 2020 @ 11:10 pm

    But wouldn't *-l and *-n, as in 衣 *ʔə̀l and 殷 *ʔə̀n, make a far more phonologically likely pairing than *-j and *-r?

  6. Chris Button said,

    July 10, 2020 @ 5:03 pm

    @ David Marjanović

    A surface distinction of [i] vs. [ij] isn't unheard of …

    I was actually being a little facetious. A phonological distinction between -i and -ij is not possible in Old Chinese. Even if something like "happy tensing" or the like were a phenomenon at the phonetic level, it's not a viable phonological distinction. The situation arises because the "six vowel" approach is fundamentally flawed.

    Another problematic area occurs with -e(j). The -j works better for Middle Chinese reflexes, but the open vowel is preferred for the "sacrosanct" -i, -u, -e, -o, -ə, -a vowel triangle. In fact, -e(j) is far better treated as Pulleyblank's -aj, and -aj is better treated as his -al.

  7. Victor Mair said,

    July 10, 2020 @ 9:30 pm

    When I remarked to Tsu-Lin Mei how surprised I was to see the name of Lin Yutang (whom I usually think of as a humanist and a humorist, and as someone who virtually bankrupted himself in a quixotic quest to invent a user-friendly Chinese typewriter, though I knew that he had a doctorate from Leipzig) among John Carlyle's list of serious sources on historical linguistics, Tsu-Lin Mei sent me this message:

    =====

    Malmquist, Goran 2011 Bernhard Karlgren: Portrait of a Scholar. Lin Yutang (1924) A Survey of the Phonetics of Ancient Chinese. Asia Major 1:134-146 is the English version of Lin Yutang’s German dissertation at Leipzig. 林語堂1924 《古有複輔音說》is the first account, written in Chinese, of the existence of consonant clusters in Old Chinese. Karlgren 1923 Analytic Dictionary of Chinese and Sino-Japanese, has a chapter on the principles of phonetic compound. This chapter was translated by Y.R. Chao as <高本漢的諧聲說>. Karlgren 1923 and Lin Yutang 1924 both tried to demonstrate the thesis that there were consonant clusters (e.g. gr-, gl- ) in Old Chinese and both were inspired by Conrady 1896 Eine indochinesische Causativ-Denominativ-Bildung und ihr Zusammenhang mit den Tonaccenten. In Conrady 1896, Conrady said that the causative prefix *s- is common to Sino & Tibetan, and that since Tibetan has consonant clusters, Old Chinese should also have consonant cluster and he cited Joseph Edkins 1874” The State of the Chinese Language at the Time of the Invention of Writing” as further support.

  8. Paul R. Goldin said,

    July 10, 2020 @ 10:52 pm

    @ Chris Button

    "But wouldn't *-l and *-n, as in 衣 *ʔə̀l and 殷 *ʔə̀n, make a far more phonologically likely pairing than *-j and *-r?"

    It doesn't really matter whether you call it *-l or *-r. We're reconstructing phonemes. If you want to call it *-l, call it *-l. But if you want *-n for 殷, you're going to have to find something else for words like 巾, because the point is that those two didn't end in the same phoneme. That's why we now say *-r for 殷. What's more plausible, if you want an *-l in the language, is *-j for 衣 and *-l for 殷. But that amounts to phonemically the same thing as what I wrote.

  9. Chris Button said,

    July 11, 2020 @ 6:54 am

    @ Paul R. Goldin

    Thanks for the further comment.

    It doesn't really matter whether you call it *-l or *-r. We're reconstructing phonemes.

    I actually quite strongly disagree here. That seems to be an excuse to reconstruct anything so long as it is distinct. In that case, it seems to me that we might as well call them *-X and *-Y and approach the whole thing like an algebra problem with no grounding in any linguistic reality.

    But if you want *-n for 殷, you're going to have to find something else for words like 巾

    I'm afraid I don't see any compelling evidence for that. In this case, even Baxter & Sagart (p.265) admit that: "The rhyme evidence for reconstructing *-r in 殷 … is open to more than one interpretation." They then talk about the overlap of 衣 and 殷, but as I said before *-l and *-n is far more plausible than -j and *-r on linguistic grounds. And more generally, it's no wonder that *-l and *-n have rhyming and xiesheng contacts.

  10. David Marjanović said,

    July 11, 2020 @ 1:19 pm

    there are modern Chinese today who do not distinguish Mandarin "r-" from /j-/

    Yes, but what's being postulated for Old Chinese here is an actual alveolar trill.

    Still, [r] can change into [j], and has done so in certain Cree dialects. (In others it has ended up as [l], [n], [ð] or remained [r]. The first three developments involve mergers with existing /j/, /l/, /n/, so a multi-step process like [r] &gt: [l] > [j] is not an option for how this could have happened.)

    if something like "happy tensing" or the like were a phenomenon at the phonetic level

    Sorry for going on an unannounced spin-off. I wasn't trying to discuss the OC vowel system, I was trying to pick a phonetic nit: phonetic [ij] is not the same as phonetic [iː] even though lots of languages allow at most one of these. (Similarly, bans on [ji] are common.)

    There is no [j] in any pronunciation of happy; the FLEECE vowel varies between [iː] and a diphthong along the lines of [ɪi̯] – it has often been interpreted as /ij/, but regardless of that, it's not [ij].

  11. Paul R. Goldin said,

    July 11, 2020 @ 6:13 pm

    @ Chris Button

    Even B&S currently reconstruct *-r for 殷.

    Over and out.

  12. Chris Button said,

    July 12, 2020 @ 7:40 am

    Even B&S currently reconstruct *-r for 殷

    And clearly that's a mistake stemming from a chain of problems ultimately going back to a faulty (and contextually inappropriate) approach to Old Chinese rhyming based on a confusion of underlying phonology with surface phonetics. Don't get me wrong, I have tremendous respect for B&S, but a little critical analysis from the rest of the admittedly minuscule field wouldn't go amiss.

    As for 殷 *ʔə̀n, a more interesting discussion is its graphic composition. It appears to contain 身 *ɬə̀ɲ or possibly an abbreviated 孕 *ɣə̀ŋs as its left component. A phonetic role is tempting, but it's problematic in either case.

RSS feed for comments on this post