Attempts to date the composition of Sanguo Yanyi from textual evidence have largely been based on the intratextual notes preserved in the 1522 edition, in particular a number of geographical glosses on Three Kingdoms place-names.[1] Those scholars who believe that these notes came from the hand of the novels putative author, Luo Guanzhong, argue that they should reflect the geographical nomenclature and administrative regions of the time at which the author was writing. Thus, if the time period which these glosses reflect could be isolated, the approximate or possibly even the precise date during which they must have been written could be established, and this would have a direct bearing on the date of composition of Sanguo Yanyi. That is to say, if we accept that the notes were written by the textual author, then the date of textual composition would coincide with the date of composition of the geographical glosses; or if we believe that the notes were written by a later editor rather than the actual author, then the date of textual composition could not post-date the composition of the geographical glosses, and we would still be able to establish a terminus ante quem for the date of textual composition. In fact there is a serious flaw in this argument, which I will go into below, but as these glosses are of such potentially crucial importance, I will first list them in full. There are twenty-three different Three Kingdoms place-names glossed in the common text of Sanguo Yanyi (i.e. derived from the archetypal Sanguo Yanyi), and these are listed below with references to one edition of each of the AB and CD systems for each example.
There have been various interpretations of these "modern" glosses to Three Kingdoms place-names. On the one hand, Zhang Peiheng and Ma Meixin have argued that they largely reflect Yuan usage, and none are exclusively Ming place-names. They further point out that in three instances (see i, k, and r above) the notes refer to place-names that were made obsolete in Tianli 2 [1329], and thus Sanguo Yanyi must have been written prior to this date.[2] On the other hand, Ouyang Jian cites the glossing of Leiyang county as within Hengzhou (see q above) and Jia county as within Henan (see u above) as evidence that the work must have been composed during the Ming. He further fixes the date of textual composition of Sanguo Yanyi to circa Hongwu 3 [1370] in the early Ming.[3]
However, a close examination of the glosses throws doubt on both of these conclusions. Firstly, the two examples that Ouyang Jiang suggests can only be examples of Ming geographical nomenclature do not stand up to scrutiny. Although not a county during the Yuan period, Leiyang was a county under Hengzhou during the Song dynasty. Furthermore the CD text places Hengzhou within Jing-Hu southern circuit, which is definitely not a Ming term, but would rather seem to be a slight corruption of the name of a Song administrative region. Ouyang Jian takes the gloss of Jia county as within Henan as referring to the Ming Henan province, but the CD text expands Henan to Henan prefecture which is not consistent with Ming geography, but is consistent with Yuan geography as Jia county was indirectly under Henan prefectural route. There are thus no incontrovertible examples of geographic glosses that could not have been written by a pre-Ming author. This does not, however, mean that the glosses must reflect Yuan geography as argued by Zhang Peiheng and Ma Meixin. In fact, almost every geographical gloss is also consistent with Song geography, and of those glosses that are inconsistent with Ming geography (see c, f, g, i, k, l, n, q, r, t, and u above), at least five examples are only consistent with Song, Jin, or early Yuan (pre-1308) geography (see g, i, j, q, r, t, and possibly also b above). The only example of an exclusively Yuan dynasty place-name is the reference to Guiyang route (see l above), but as this name occurs in a note appended to the gloss on Guiyang as Chenzhou (consistent with Song, Yuan and Ming geography), it may well be a later comment on an original Song gloss.
Are we then to conclude that Sanguo Yanyi was written during the Song dynasty? No, such a conclusion would be unjustified. The glosses may well reflect pre-Ming or even pre-Yuan geographical nomenclature, but that does not necessarily disqualify a late Yuan or Ming date for their incorporation into the text of Sanguo Yanyi. Studies which have used these geographical glosses as evidence for the date of composition of Sanguo Yanyi all base themselves on the premise that the glosses were written by the author from his own personal knowledge of contemporary geography at the time he was writing, and that any inconsistencies (e.g. the glossing of Wuling commandery as the Song dynasty place-name of Dingzhou) were due to a haziness in his personal geographical knowledge. There are two counter-arguments to this somewhat dubious premise:
There are more obvious ways of giving the text of Sanguo Yanyi the veneer of a Yuan dynasty composition than using obscure and ambiguous Yuan dynasty geographic glosses (it would, for example, be far simpler and more effective to give a false Yuan dynasty preface signed by Luo Guanzhong), and so I think that we can discount the first possibility.
The second possibility seems to me to be far more reasonable. That the glosses were not all written by the author or an editor of Sanguo Yanyi is evidenced by the gloss on Guandu in Item 59 (see d above), which is copied from a note by Hu Sanxing 胡三省 (1230-1302) to Sima Guang's historical chronicle Zizhi Tongjian 資治通鑒. A further six glosses (see g, h, j, k, l, m above) correspond exactly to glosses provided in the annotation to juan 2 (the section covering the Three Kingdoms period) of the historical chronicle of the Shu region (modern Sichuan) entitled Shu Jian 蜀鑒, which was compiled during the Southern Song by Guo Juren 郭居仁 (Guo Yundao 郭允道). I believe that the close correspondence between the Sanguo Yanyi glosses and the Shu Jian glosses is not mere coincidence, and that Shu Jian was indeed one of the sources used for the geographical glosses in Sanguo Yanyi. A thorough survey of Song and Yuan editions of historiographical texts will, I suspect, come up with sources for most or all of the remaining geographical glosses, and perhaps also for many of the other intratextual notes common to both the AB and the CD text.
Although it is not possible to prove whether the intratextual notes were written by the textual author or whether they were added by a later editor, the fact that the great majority of them can be shown to have been present in the archetype (by their common presence in editions of both the AB and the CD system) demonstrates that they were a component of the text at an early stage of its history. Moreover, as the notes seem to have been derived largely from just such historiographical sources as the author would have used for the composition of the narrative text, I am inclined to believe that the common intratextual notes to Sanguo Yanyi were copied into the text by the original author (putatively Luo Guanzhong) from Song or Yuan annotated editions of historiographical texts such as Sanguozhi, Zizhi Tongjian and Shu Jian.
If we reject the notion that the geographical glosses in Sanguo Yanyi were the product of the textual author's personal knowledge of contemporary geography, then we cannot use these glosses to support the precise dating that authors such as Zhang Peiheng and Ouyang Jian have attempted. Nevertheless, they do still potentially have some bearing on the dating of textual composition. If we could locate the specific editions of historiographical texts used as sources for intratextual annotation in Sanguo Yanyi, this could provide a terminus post quem for the date of composition of Sanguo Yanyi. Although we cannot use intratextual annotation to establish a terminus ante quem for the date of textual composition as an author or editor could quite conceivably copy geographical glosses from earlier source texts long after they had become anachronous, common sense would suggest that it is unlikely that an author would deliberately insert geographical glosses that would be incomprehensible to his readership. As there was far less change in geographical nomenclature between the Song and Yuan dynasties than there was between the Yuan and Ming dynasties, most of the geographical glosses in Sanguo Yanyi would have been familiar to a Yuan readership, but unfamiliar to a Ming readership. Although speculative, this does strongly suggest that such geographical glosses would not have been added to the text any later than the early Ming, and the absence of a single Ming gloss in the common text makes a Yuan date seem more plausible. That Ming readers did find some of the original glosses incomprehensible is demonstrated by changes made by an editor to the AB text who changed several of the Song and/or Yuan geographical terms to Ming terms (e.g. deletion of the term Jing-Hu Southern circuit in example q, changing Jiankang fu to Jinling in example r, and changing the Song-Yuan Huazhou to the Ming Huazhou in example c). Other changes to anachronous geographical glosses were also made by editors of the B text, B0a text and D text (see XXX, XXX and XXX).
The latest dateable source text used for the compilation of Sanguo Yanyi that I have been able to identify is Hu Sanxing's 胡三省 (1230-1302) annotated version of the Zizhi Tongjian. This has a self-preface dated at Zhiyuan 22 [1285], and provides us with the only reliable terminus post quem for the composition of Sanguo Yanyi.
The use of Hu Sanxing's version of Zizhi Tongjian as a source for Sanguo Yanyi is evidenced by the presence of some of the Hu Sanxing notes in Sanguo Yanyi (see XXX), as well as by the fact that text copied into Sanguo Yanyi from Zizhi Tongjian sometimes includes Sima Guang's Kaoyi 考異 [Investigation of Textual Differences] notes, which were originally published as a separate work, and were only inserted into the appropriate place in the main text with the publication of Hu Sanxing's annotated edition. For example, in Item 38 the section of text which relates Lü Bu's conversation with Cao Cao and Liu Bei before his execution (e.g. 1522 4:68b-69a) copies one of Sima Guang's Kaoyi notes quoting from Xiandi Chunqiu 獻帝春秋 [The Annals of Emperor Xian] together with the surrounding main text (see Zizhi Tongjian 62/2007).
Ogawa Tamaki notes that the Taichang Yuan 太常院 [Academy of Imperial Sacrifices] is mentioned in Item 159 of Sanguo Yanyi (see 1522 16:71a; YFC 7:57a), and that this name was first used in Zhida 1 [1308] of the Yuan dynasty.[5] This would thus seem to extend the terminus post quem for the composition of Sanguo Yanyi to 1308. However, it is possible that the Taichang Yuan mentioned in Sanguo Yanyi is a corruption or abbreviation of the Tang-Song institution known as the Taichang Liyuan 太常禮院 [Ritual Academy of the Court of Imperial Sacrifices], and so this is not an entirely reliable example for dating composition.
The terminus ante quem for the composition of Sanguo Yanyi is given by the earliest extant reference to the novel. This is the "Vulgar and Fooloish" (Yongyuzi 庸愚子) preface dated at Hongzhi 7 [1494]. Thus, we can only claim with certainty that Sanguo Yanyi was composed sometime between the years 1308 and 1494, a window of opportunity of nearly two hundred years.
Although at present it is not possible to either extend forward the terminus post quem or extend backward the terminus ante quem, we can attempt to approximately locate the composition of Sanguo Yanyi within this two hundred year time range based on the evidence of textual corruption in Sanguo Yanyi. Textual corruption is a function of transmission, and transmission is itself a function of time: the longer a period of time, the more times that a text is transmitted; the more times that a text is transmitted, the more textual corruptions will be introduced. Thus, if two texts descended from a common ancestor have few exclusive textual corruptions, then we can assume that each text has undergone few transmissions since splitting off from the common ancestor a short time previously. On the other hand, if two texts descended from a common ancestor have very many exclusive textual corruptions, then we can assume that each text has undergone very many transmissions since splitting off from the common ancestor a long time ago.
We can use homœoteleutic omission as an index of textual corruption, and as a general principle state that the number of examples of homœoteleutic omission that separate stemmatic branches is a function of the number of transmissions that a text has undergone (the more occurrences of homœoteleutic omission, the more times that a text will have been transmitted), and this is further a function of time (the more times that a text is transmitted, the longer a period time will have elapsed). Therefore, the larger the number of homœoteleutic omissions between stemmatic branches, the longer the period of time between each stage in the evolution of the text.
Although it is not possible to quantify the rate of textual corruption (i.e. put an exact figure on the time or number of transmissions needed to produce a certain number of homœteleutic omissions), we can still get a general idea of the length of textual transmission involved between evolutionary stages from the empirical evidence of homœoteleutic omission.
Far more examples of homœoteleutic omission were introduced into the text of Sanguo Yanyi between the archetype and √A/B (31 examples) than were introduced between √A/B and the extant AB editions (e.g. 11 examples between √A/B and 1522, and 15 examples between √A/B and LZW). Likewise, more examples of homœoteleutic omission were introduced into the text of Sanguo Yanyi between the archetype and √C/D (between 22 and 34 examples) than were introduced between √C/D and the extant CD editions (e.g. 10 examples between √C/D and YFC, and 28 examples between √C/D and YMZ). This suggests that there must have been a longer of period of textual transmission between the archetype and √A/B and √C/D than there was post-√A/B and post-√C/D.
On the basis of the "Long Whiskers" (Xiuranzi 修髯子) preface common to both A branch and B branch editions we can assign a terminus post quem for √A/B of Jiajing 1 [1522]. Assuming that the AB text split into the A and B branches soon after the production of the Xiuranzi edition of 1522, most or all of the 31 examples of homœoteleutic omission common to the AB text would have occurred pre-1522. In other words, the AB branch split off from the archetype considerably before 1522.
On the basis of the Lin Han preface to Sui-Tang we can assign a terminus ante quem for √C/D of Zhengde 3 [1508]. Thus all of the 22-34 examples of homœoteleutic omission which are common to the CD text would have occurred pre-1508. In other words, the CD branch split off from the archetype considerably before 1508.
As the urtext would have pre-dated the archetypal stage still further (although, as the archetypal Sanguo Yanyi text probably closely represents the urtext, there was probably not a great time gap between these two stages), the evidence of homœoteleutic omission certainly would not seem to preclude a late Yuan or early Ming date for the original composition of Sanguo Yanyi. On the other hand one would be hard-pressed to argue for a mid-Ming date for the composition of Sanguo Yanyi (such as the 1494 terminus ante quem postulated above) as this would allow very little time for the necessary number of textual transmissions to account for the large number of examples of homœoteleutic omission found in the common text of each system.
I am therefore inclined to accept the traditional ascription of authorship for Sanguo Yanyi to Luo Guanzhong. As the evidence of homœoteleutic omission does point to a long history of textual transmission prior to the Jiajing period when the first extant editions of each textual system came to be published, it is plausible that Luo Guanzhong wrote Sanguo Yanyi in his later years during the early Ming dynasty, and that the missing period of 130 years between the last known date for Luo Guanzhong (1364) and the earliest recorded mention of Sanguo Yanyi (1494) can largely be accounted for by textual transmission in manuscript form. Hence the work would have been relatively unknown until it was first published in the mid-Ming.
1. For discussions on the use of the 1522 notes for dating the composition of Sanguo Yanyi see Liu Youzhu, "Sanguozhi Tongsu Yanyi shi Yuandai zuopin"; Ouyang Jian, "Shilun Sanguozhi Tongsu Yanyi de chengshu niandai"; Wang Changyou, "Jiajingben Sanguozhi Tongsu Yanyi xiaozizhu shi zuozhe shoubi ma?"; Zhang Guoguang, "Sanguozhi Tongsu Yanyi chengshu yu Ming zhongye bian"; Zhang Peiheng, "Guanyu Jiajing ben Sanguozhi Tongsu Yanyi xiaozhu de zuozhe"; and Zhou Cun, "Sanguo Yanyi fei Ming-Qing Xiaoshuo".
2. See pp.8-10 of their introduction to the Shanghai Guji Chubanshe 1980 edition of Sanguozhi Tongsu Yanyi.
3. See Ouyang Jian, "Shilun Sanguozhi Tongsu Yanyi de chengshu niandai".
4. See Zhang Guoguang, "Sanguozhi Tongsu Yanyi chengshu yu Ming zhongye bian" pp.270-272 for these two counter-arguments. That some of the intratextual notes were derived from notes in the historical source materials for Sanguo Yanyi has also been pointed out by Sun Kaidi ("Sanguozhi Pinghua yu Sanguo Zhizhuan Tongsu Yanyi p.117).
5. See Ogawa's Japanese translation of the Mao Zonggang recension of Sanguo Yanyi, Sangokushi, vol.7 p.223 note 12.