Some quotes concerning historical Sinology from the correspondence of Jerry Norman (1936–2012) with W. South Coblin [source]:
I think already in the 1950s Lǐ Róng saw that there was not really much more to say about the QY [Qièyùn] system and that dialects were the logical next place to go to study the development of Chinese. You might say that Karlgren and his followers were like Western classicists - Greek and Latin scholars. What is needed now is something more akin to Romance linguistics. [2008]
I would say that 1) most of the material used for [Old Chinese] is of Hàn vintage and very little of it goes back to the period that people ordinarily claim for OC, 2) one cannot demonstrate that the xiéshēng characters form a sufficiently coherent body of evidence to base a reconstruction on, 3) OC is basically a backward projection of the Qièyùn system; where other kinds of data exist it is generally ignored. This is shown by the fact that OC practitioners do not try to derive modern forms from their OC reconstructions, 4) the fact that OC projects always lead to too much speculation makes the endeavor suspicious (…) [2008]
QY type Chinese is something like Latin or Classical Greek. When a classical tradition is preserved and elevated to a revered cultural icon, several things are required. You must have a substantial body of written material; you have to understand the material's meaning; you have to have a way to pronounce it. In the case of Latin, from an early period, the spoken and written languages diverged in many ways. Eventually you had a situation of diglossia; people treated the written language as a high or prestigious form of their spoken language. At this stage they carried over many vernacular elements when reading Latin: femina was pronounced femna, vetulus became veclus, etc. If this doesn't seem plausible, think about English; don't we do something like this even today? We see “night”, originally nɪxt, and we say nait. We in effect substitute a more evolved form of the word for a Middle English form. So what were the Chinese to do? When someone in the 6th c AD wanted to read a text aloud, he probably, as much as possible, substituted current vernacular readings for the old forms, much as we do today. We see ⽔ and say shuǐ, even though this is probably not how the word was pronounced in the Lúnyǔ when it was compiled. But then what about all those words for which there is no vernacular equivalent? Students had to rely on what their teachers told them; to some extent these scholastic pronunciations were handed down from generation to generation; in other cases the teachers depended on fǎnqiè glosses handed down from the Late Hàn. But even before fǎnqiè were employed, there must have been oral traditions about how to read texts. The Qièyùn, then, was a compilation of traditional sound glosses and traditional oral lore. Moreover, the QY was an aide to reading texts aloud and made no pretense to represent contemporary vernacular pronunciation. So the vernacular tradition and the reading tradition were really quite different, although undoubtedly there were places where the two intersected. To base a vernacular chronology on things like the QY and its descendants is a big mistake. For this I think we must work from the modern spoken dialects (19th and 20th century records). [2011]
I think he [Zev Handel] is right to say that OC is not properly speaking a reconstruction and certainly it is not a comparative reconstruction; it is based on a unique (and probably faulty) methodology that can only be used in the case of Chinese. (And it most probably is misguided.) [2011]
(…) it reminds me of a set of questions I have about the OC project. The first of these is just what is the corpus of OC? In the book I mentioned by Chén Fùhuá and Hé Jiǔyíng they say their corpus is approximately ten thousand graphs taken from pre-Qín texts. Now my question is, since a character generally represents a morpheme, do natural languages have this many morphemes? I am not sure anyone has ever studied this. In the case of modern Chinese few literate people seem to know more than about 4000 characters and some of these represent archaic morphemes. I recall that Qiú Xíguī [ 裘錫圭 ] somewhere says that an early study found only about 6000 different characters in the entire Shísānjīng [ 十三經 ]. The Shuowen has almost 10000 characters and the later rime dictionaries have many more, but many of these are not attested in actual texts. One thing that always struck me about Pulleyblank's work was that he seemed willing to use even the most obscure graphs in his phonological speculations. My suspicion is that a natural language has around 2000 separate morphemes. Reconstructed IE has many fewer than that. Modern languages like English have enormous collections of morphemes, but if one limited himself only to those words of Germanic extraction, there really wouldn't be very many at all. So why do almost all OC schemes claim to have on the order of 10000 reconstructable morphemes? One different approach would be to limit the number of relevant forms for reconstruction. Characters without textual occurrences should be discarded. Hapax legomena should also be eliminated. We should try somehow to establish a basic vocabulary for pre-Qín Chinese, which I would guess would be on the order of 1000 forms. [2012]
If we look at [Karlgren’s] own practice of Chinese dialectology, we see immediately that it had a peculiar character; he was not interested in the actual lexicon of local dialects but only in the way people read characters. This is a reflection of the tendency of both Chinese and foreign scholars to view Chinese as identical to the characters. As stated above, he had dismissed the relevance of genuine dialect data early on. I think in a way this explains why he put so much value on the Sino-Xenic systems: they were purely reading systems and there were no vernacular languages that they matched. Grootaers tried to point out this deficiency but was by and large ignored. The essence of the Karlgrenian approach then is to take the Chinese writing system as the only "real" Chinese, the only type amenable to scientific study. [2012]
(The full version also including several thoughts from Norman about his speciality, Min Chinese.)
