“Literary reading” or “Literary misreading”: an “archaeological” study of the pronunciation of 車

UPDATE: I have written a paper on this topic with much more updated information. So you may read the paper on my academia.edu account here.

This weekend I attended a workshop on developing standard teaching materials for Chinese as a second language. One of the presenters from China used a common idiom “bì mén zào chē” (閉門造車) to describe how some teaching materials were created without consulting real makevehiclebehinddoorlanguage data. The idiom 閉門造車 literally means “make vehicles behind shut doors”, and it is now used to describe people who do things without really knowing the relevant facts. What drew my attention to the presenter’s use of this phrase was his pronunciation of 車 as “jū” instead of the standard “chē”. This presenter is from Mainland China. Interestingly, one of my colleagues from Taiwan also pronounced the 車 in 閉門造車 as “jū”, as can be confirmed by Taiwan’s 中華民國教育部重編國語辭典修訂本. But in contrast Mainland China’s 現代漢語詞典 only lists “chē” as the pronunciation. The character 車 is pronounced as “jū” in Chinese chess, as a convention. But the pronunciation of 車 as “jū” is considered non-standard in Mainland China now, although it seems that this pronunciation maintains a certain degree of “literary reading” status in Taiwan. What is worth noting here is that in both Mainland China and Taiwan, the standard pronunciation of 車 in most cases is still chē, and only in certain cases, such as idioms or conventional specialty pronunciations, does the pronunciation “jū” appear. So there is no doubt that in spoken language, the pronunciation of 車 is “chē”. The alternative pronunciation is strictly speaking just alternative. I remember having read an article on the pronunciation “jū” for 車 a long time ago, and I vaguely remember that the conclusion of that article was that the correct pronunciation of 車 should be “chē”. But since I could not really find that article now, I decided to venture out by myself and digging out some information in “old piles of paper” (故紙堆).

Basically, my missions are (1) what are the historical sources for the pronunciation of 車; (2) according to these historical records, which pronunciation is the correct pronunciation, chē or jū; (3) where did the other pronunciation come from and how.

Also broadly, my conclusions are (1) historical records show clearly that the pronunciation of 車 should be “chē”, which can also be corroborated by the reconstructions of Chinese historical phonology/phonetics; (2) the other pronunciation “jū” was a literary reading, or literary pronunciation, which had been a curious result of a reverse use of the 聲訓 “definition by pronunciation” practice in traditional Chinese commentaries.

Ok, now let me take you on my journey of this interesting discovery. As usual, our starting point is 說文解字, the earliest Chinese character dictionary dated to the 2nd century. The definition of 車 goes:

車: 輿輪之緫名。夏后時奚仲所造。象形。凡車之屬皆从車。尺遮切。

So what is relevant is the pronunciation as 尺遮切, which gives “chē” as the pronunciation in Modern Standard Chinese. Now let’s see what the “authoritative” Kangxi Dictionary 康熙字典 says:


This would give the pronunciation of 車 as “jū”, and that’s the only pronunciation! It did not even mention the possibility of pronouncing it as “chē”. Since this note on pronunciation of 車 cites 廣韻, now let’s look at 廣韻. In the 廣韻, the character 車 is listed under two different rimes: 麻 and 魚. The definition of 車 under the rime 魚 is very short, which is:


The pronunciation of the homophonic group to which 車 belongs under the rime 魚 is 居, and that’s where the pronunciation of 九魚切, i.e. “jū”, originated. Many rime dictionaries after the 廣韻, such as the 集韻 in the Song Dynasty, and the 古今韻會舉要 in the Yuan Dynasty, probably also listed this pronunciation. Kangxi Dictionary apparently took it and used it as the standard pronunciation. But even in the definition cited above from 廣韻 under the rime 魚, it still says that another pronunciation of 車 is 昌遮切, which gives “chē” as the standard pronunciation. Actually the character 車 is also listed under the rime 麻 in the 廣韻. The definition there is very extensive, and shows that that is the main entry for this character. For the pronunciation, it says:


So here the standard pronunciation of 車 is “chē”, and it can also be pronounced as “jū”, but apparently “jū” is not the norm here. Therefore taking all this into consideration, we can draw the following inferences:

  • In the 說文解字, the only pronunciation is “chē”.
  • In the 切韻/廣韻 system, the standard pronunciation is still “chē”, although “jū” is also listed as an alternative pronunciation.
  • In the 康熙字典, “jū” is listed as the standard pronunciation. There is no mention of “chē”.

Therefore the pronunciation of 車 gradually changed from “chē” to “jū”. But is this change a result of natural sound change, or an “artificial misreading”? That’s what we are going to figure out next. So apparently somewhere between the 說文解字 in the 2nd century and the 廣韻 in the 11th century, the pronunciation of 車 as “jū” appeared. According to the 說文解字註 by 段玉裁, he noted many commentaries about the meaning and pronunciations. E.g.:


In here, Duan explained in more detail the meaning of 車, and he used the word 居 (to stay) to explain that Xu Shen used the part where a person that rides a vehicle stands to refer to the concept of 車. Therefore there is a connection between the meanings of 車 and 居.


In here, Duan said that the ancient pronunciation is jū, and the pronunciation “today” was chē. This is useful because he confirmed that in colloquial speech of his time, people pronounce 車 as chē, but not jū.


釋名 is a book written probably a century later than the 說文解字. In 釋名, the method called 聲訓 (“phonetic gloss”) was first used and this method proved to be quite popular throughout later historical stages. 聲訓, literally meaning “definition by pronunciation”, uses a word that has basically a similar sound and meaning to define the meaning of another word, and also probably to approximate the pronunciation of the word that is defined. For example, 天 tiān (“sky”) can be defined by 巔 diān (“mountain top”), since they have both similar sounds and related meanings. Here we used the Modern Standard Chinese pronunciations of these two words, but in principle we should have used their pronunciations in late Old Chinese, but conveniently here the modern pronunciations of these two characters are still quite similar. So according to this passage quoted by Duan from 釋名, it is said that in ancient times the pronunciation was jū, and this would correspond well to the meaning of 居 jū as well.


Here the same “definition by pronunciation” method is used to explain the meaning of 車 with a word 舍 shè (“hut”) that sounds similar to its “contemporary” pronunciation of chē, and we can see the connection between their sounds and meanings, saying that when one travels on a vehicle it is comparable to staying at one’s house.


Here another commentator quoted by Duan said that actually the pronunciation of 車 should first be 尺遮切, which corresponds to the modern pronunciation of chē, while the other sound of 居 jū appeared since the Han dynasty.


Here Duan refuted the above claim. He said that the word 遮 was not pronounced as zhē in ancient times. This is definitely correct. He then used this to say that therefore the old pronunciation of 車 should not be chē, but not jū either, and he proposed qū as the ancient pronunciation using the same “definition by pronunciation”, since 祛 has the meaning of “expel, dispatch”, and it is also used in a line from the Book of Songs 詩經 in a context where 車 is also used. The original line is 以車祛祛 “these are all stout carriage horses” (詩經・魯頌・駉). But this proposal of Duan’s is pure conjecture without any concrete linguistic evidence.

Now let’s go back to the original claim in the 釋名 where it says the old pronunciation of 車 should be 居. Note that the 釋名 is a book that exploits the “definition by pronunciation” method, and the normal logic of the method is to try to find a word with a similar sound and meaning. But if one uses this principle sloppily or somewhat in a reverse way, it will be something like this: since character B has a similar meaning to character A, which we want to define, then for lack of a better word, character B seems to be the natural correct definition for character A. But then ideally, let’s emphasize the word “ideally” here, the pronunciation of B should be similar to A. So probably the “correct” “original” pronunciation of A should be the same as B. This kind of logic is totally sloppy in many different ways, but we can still see how it could be possible for one to believe firmly that 居 is the ideal or natural definition for 車, and the pronunciation of 車 should be similar to that of 居, i.e. jū, due to the “definition by pronunciation” method. Admittedly in many cases the “definition by pronunciation” is very forced, i.e. the definition is too vaguely related. Thus in order to find a good “definition” word that also shares the same or a similar pronunciation is very difficult in many cases. Let’s say for now that the purported “ancient” “original” pronunciation of 車 as jū was first fabricated in the book  釋名 in the Han Dynasty. Ever since then, many educated people tried to pronounce 車 as jū, since they believed that that was the more correct pronunciation, although in their spoken language, the pronunciation of 車 had never been even remotely similar to that of 居. Therefore we can see that in literary pronunciations, or literary readings, the purported “ancient” “original” pronunciation jū was favored, as can be seen from the Kangxi Dictionary and the argument put forth by Duan Yucai.

Let’s look at the modern pronunciations of 車 in various Chinese dialects. According to Yuan Jiahua, et al. (1956) 漢語方音字彙, among the 20 dialects listed, the pronunciation of 車 include tʂʰɤ in Beijing, Ji’nan and Xi’an;  tsʰɤ in Taiyuan, Wuhan and Changsha;  tsʰe in Chengdu;  tʂʰe in Hefei;  tsʰo in Suzhou; tsʰei in Wenzhou;  tʰo in Shuangfeng; tsʰa in Nanchang and Meixian; tʃʰɛ in Guangzhou and Yangjiang; tsʰia in Xiamen, Chaozhou, Fuzhou, and Jian’ou, and tɕʰiɪ in Yangzhou. It is clear that the most common type of initials for the pronunciation of 車 in modern dialects is some kind of affricates: retroflex affricates and alveolar affricates mostly, since in Guangzhou and Yangjiang, the palatal affricates are also interchangeable with the alveolar affricates. The initial tʰ in Shuangfeng might have been the original initial. The only exception is the tɕʰ in Yangzhou, which is the most close to the pronunciation of jū. But this might have been a result of palatalization of some kind of affricates. As for the vowels, it is quite varied, including a, ia, o, e, ei, ɛ, ɤ, iɪ, but it seems low or back vowels are more common, and it also seems that there is a certain degree of a high front vowel such as i, or e. Then we could temporarily posit tʰja as the Old Chinese form of 車, and in Middle Chinese, tʰj>tɕʰ, and consequently tʰja>tɕʰja, and in Early Modern and Modern Chinese, tɕʰja>tʂʰɤ. This seems to be a very reasonable explanation. On the other hand, the modern pronunciations of 居 mostly starts with tɕ or k, and it can be assumed that the original initial of 居 is k. Thus it is not very straightforward how k could change into t, if we take jū as the correct pronunciation.  Our temporary reconstruction of the pronunciation of 車 as tʰja in Old Chinese and tɕʰja in Middle Chinese is very close to Wang Li’s reconstructions of these pronunciations as ȶʰiɑ in Old Chinese (昌 initial and 魚 rime) and tɕʰĭa in Middle Chinese (昌 initial, 麻 rime, 3rd grade 三等, open rime 開). Moreover the reconstructions of the pronunciation of 居 in Wang Li’s system is kĭɑ in Old Chinese (見魚), and kĭo in Middle Chinese (見魚開三). Thus by this line of reasoning, it seems that the pronunciation of 車 has never been the same as that of 居, although they belonged to the same rime class in Old Chinese.

However, there is indeed some evidence for the other view. Since the 廣韻 did list 居 as a possible pronunciation for 車, and if we believed this was a result of natural sound change, we need to account for this. One way of accounting for this is to resort to a more complicated process of sound change.  We might just as well say that the original initial of 車 in Old Chinese or earlier was kl-, instead of t-. Thus a possible reconstruction of 車 in Old Chinese is kʰlja, and then we have to posit kʰlj>tɕʰj in Middle Chinese. Phonetically speaking kʰ is [-continuant; +velar], while l is [+continuant; +alveolar], thus it is possible to get a sound change which mix these features up, i.e. [-continuant; +alveolar], which corresponds to “tʰ”, and the “t” can further palatalized into tɕʰ. Note that if we posit kʰl as the Old Chinese initial of 車, it is now very close to the pronunciation of 居, which had a k initial in Old Chinese. In fact many reconstructions, such as Zhengzhang Shangfang’s, and Fang-kuei Li’s, posit kʰl-as the original initial. In Zhengzhang Shangfang’s system, the Old Chinese pronunciation of 車 is kʰla, and that of 居 is kas. Also they posit a k- initial as another pronunciation of 車 in Old Chinese, which would eventually develop into the 居 pronunciation of 車 in Middle Chinese.

Thus it seems that there is also a possibility that the pronunciations of 車 and 居 were quite close in Old Chinese. However, because k>t is not a very common sound change process, and also that there is no trace of the original k in any modern dialects, I am reluctant to agree with the second point of view. As our textual archeological findings above show, the pronunciation of 車 as 居 started in 釋名, and it is very likely that saying that the pronunciation of 車 should have been more like 居 could have been a sloppy reverse use of the “definition by pronunciation” method. All in all, there is very little real evidence, except for the mention in the 廣韻, for the pronunciation jū for 車. It might just have been a literary misreading, due to the distinction of colloquial and literary pronunciation systems, and also the attitude of traditional Chinese scholars to distrust the spoken pronunciations of characters and therefore the tendency to make distinctions, but in many cases, there might not be any real argument for such distinctions.

PS: I just did some google search, and found out that the original paper on this topic that I referred to at the beginning of this blog is 时建国 (1997) 说车字的”居“音,in 语文研究 65-4, and there is a newer article by 孟蓬生 (2002) “车”字古音考——兼与时建国先生商榷, in 古籍整理研究学刊, which argues that the pronunciation of 車 as jū indeed is reasonable and gives a lot of textual evidence. However I am still not totally convinced by this, mainly because the uncertainty of the value of the textual evidence. Also the nature of the character 車 as a pictogram is also relevant, since unlike the phono-semantic compounds, you can basically attach any pronunciation to a pictogram. Therefore we have to be more clear what we are talking about. Is it how many pronunciations that the character 車 can have? Or is it the original pronunciation of the WORD 車 was more like that of 居?I am pretty sure it is the second question that we are more concerned about. So imagine there was a dialect in which they say kia when they refer to the same thing as 車, but this might not be a cognate word of tja at all, although we can use the same character 車 to represent these two different words.

Also on the point that I raised about the rarity of the sound change kl->t, I found another useful paper by Juliette Blevins and Sven Grawunder (2009) about a very rare sound change in Germanic: *kl>tl, and this seems very similar. So it does lend some support to the reconstruction of the pronunciation of 車 as kʰlja.