Abstract
-
This paper identifies a case of textual dislocation in each of two chapters of the Shiji: the “Memoir of Li Shang” and the “Memoir of the Xiongnu.” Reconstructing these misarranged passages suggests that the Memoirs section of the Shiji was originally composed on bamboo slips containing approximately 21 to 23 characters each—a format consistent with other narrative texts of the period. This suggests that Sima Tan and Sima Qian may have directly excerpted or copied earlier source materials in their compilation of the Shiji.
The displaced slips in the “Memoir of Li Shang” were already present in the version known to Ban Gu, leading to longstanding misinterpretations and textual modifications beginning with the Han shu. Once restored, the passage shows that Li Shang and Fan Kuai were appointed Right and Left Chancellors, respectively, and took command in suppressing the rebellions of Zang Tu and Chen Xi after Gaozu’s withdrawal from the front. This restoration offers new insight into the structure of the chancellorship in the early Han dynasty.
The disruption in the “Memoir of the Xiongnu,” on the other hand, suggests that prior to the Eastern Han, there were at least two competing accounts of the final years of Emperor Wu’s reign. Eventually, only the version aligned with the Han shu narrative prevailed and was established in the received historical tradition.
-
Keywords: Shiji, Misordered slips, Historical writing, Li Shang, Li Guangli
Research on the Shiji 史記 (Records of the Grand Historian) has generally followed two main approaches: one focused on its historical discourse—particularly Sima Qian’s 司馬遷 (139-86 BCE) use of weiyan dayi 微言大義 “subtle words with weighty meaning” — and the other concerned with the Shiji’s material form as a written text. Since the twentieth century, the discovery of a wide range of excavated bamboo slip manuscripts from early China has transformed our understanding of early textual practices. Although no early copy of the Shiji has yet been discovered, the physical features of these manuscripts have helped shape modern perspectives on how the Shiji may have circulated in its initial transmission.
Nienhauser
1 and Boltz
2 have proposed that, in “composing” the
Shiji, Sima Tan and Sima Qian first excerpted the passages they wished to use from earlier sources, marked them, and only thereafter arranged and compiled those extracts into one
juan (chapter). Their hypothesis both suggest a editing process of transcribing text on single slip and then bind them together sequently, which shed light on the methodology this paper uses.
This paper begins with two instances of misordered bamboo slips in the Shiji, offereing preliminary reflections on its initial stage of compilation. It considers questions such as the number of characters per slip, the editorial and structural process of assembling the text, and the evolution of early versions. As part of this inquiry, it also seeks to clarify several specific historical events mentioned within the affected passages.
The Misplacement of a Strip in the “Memoir of Li Shang”
Li Shang 酈商 (d. 180 BCE) was active as both a military commander and administrator during the early Han dynasty. Having followed Liu Bang 劉邦 (256-195 BCE) from his early days as Magistrate of Pei, Li Shang participated in various campaigns and later held positions such as General, Commandant of the Guards, Right Chancellor, and Chancellor of Zhao. He played a major role in three important military operations during the reign of Gaozu: the suppression of Zang Tu 臧荼 (d. 202 BCE) in the fifth year (202 BCE), the pacification of Chen Xi 陳豨 (d. 196 BCE) in the eleventh year (196 BCE), and the defeat of Qing Bu 黥布 (d. 195 BCE) in the twelfth year (195 BCE). The Shiji, in the “Memoir of Li Shang,” records these events as follows – the logic of the sequencing (a)–(e) will be addressed later. Particular attention should be given to sentence division and punctuation in this section:
其秋，燕王臧荼反，商以將軍从击荼，戰龍脱，先登陷陣，破荼军易下，却敌，遷爲右丞相，賜爵列侯，與诸侯剖符，世世勿絕，食邑涿五千户，號曰涿侯。
That autumn, Zang Tu, the King of Yan, rebelled, and Shang, as a general, followed Gaozu to strike at Zang Tu, gave battle at Longtui, was first to climb the walls and cause the enemy’s ranks to fall, and defeated Tu’s army beneath the walls of Yi, driving back the enemy. He was promoted to be Chancellor of the Right, granted the rank of a Ranking Marquis, and together with the other feudal lords, he received a tally from the Emperor pledging that the succession would not be cut off for generation after generation. He also received the fief-town of Zhuo, with five thousand households, and was titled Marquis of Zhuo.
As Chancellor of the Right and operating separately, he pacified Shanggu (commandery), took advantage of the situation to attack the commandery of Dai,
(b) 受趙相國印。
以右丞相趙相國别4與絳侯等定代、雁門，
and received the seal of the Chancellor of the State of Zhao. As Chancellor of the Right of Han and Chancellor of State of Zhao, operating independently and together with the Marquis of Jiang and others, Li Shang pacified the commanderies of Dai and Yanmen,
(c) 得代丞相程縱、守相郭同、將軍已下至六百石十九人。
and captured Cheng Zong, the Chancellor of Dai, Guo Tong, the Acting Chancellor; and 19 men from general down to the rank of 600 shi.
(d) 還，以將軍
爲5太上皇衛一歲。
七月6，以右丞相擊陳豨，残東垣。
After he returned, as a general he became the Guard of Taishang Huang (His Most Honored Majesty) for a year. In the seventh month, as Chancellor of the Right, he struck at Chen Xi and destroyed Dongyuan.
(e) 又以右丞相從高帝擊黥布，攻其前拒，陷两陳，得以破布軍，更食曲周五千一百户，除前所食。
7
Again, as Chancellor of the Right, he followed Gaozu to strike Qing Bu, attacking his front lines of defense, causing his first two ranks to fall, and thus enabling the Emperor to defeat Bu’s army. The Emperor changed his sustenance fief to Quzhou with 5,100 households and abolished the fief from which he had formerly drawn sustenance.
8
Commentators throughout the history, including Takigawa Sukenobu 瀧川資言(1865–1946), have agreed on the reading and punctuation of this quoted passage as reflected in the Zhonghua edition (1959): The first paragraph refers to events from the battle of 202 BCE, while sentences (a) through (d) describe those of 196 BCE, and (e) refers to 195 BCE; the wording and sequence are consistent across all received editions of Shiji. However, this reading gives rise to several interpretive challenges.
The first question concerns the geographic plausibility of sentence (a) and whether it refers to the 196 BCE campaign. In the ninth month of Gaozu’s tenth year, Chen Xi—then in command of the military forces in Zhao and Dai—rebelled. In response, the Han court organized a two-pronged campaign: the western force, led by Zhou Bo周勃 (d. 169 BCE), advanced into Dai from Taiyuan 太原, while the eastern front was directed personally by Gaozu. Based on the “Basic Annals of the Emperor Gaozu” in the
Shiji and “Annals of the Emperor Gaozu” in the
Han shu, this eastern front proceeded entirely within Zhao territory prior to the recapture of Dongyuan; Gaozu first reached Handan 邯鄲, then moved northward to take Dongyuan 東垣, and subsequently returned to Luoyang 洛陽. Fan Kuai樊噲 (d. 189 BCE), according to his Memoir in the
Shji, was at this time a principal general on this front and also departed from Handan to advance northward, taking Xiangguo 襄國 and Bairen 柏人, as well as the commanderies of Qinghe 清河 and Changshan 常山—both in Zhao – before joining Gaozu at Dongyuan. Other generals, including Guan Ying 灌嬰 (d. 176 BCE), Xiahou Ying 夏侯嬰 (d. 172 BCE), and Liu Ze 劉澤 (d. 178 BCE), reclaimed further counties in Zhao such as Anping 安平, Anguo 安國, Lunu 廬奴, Quni 曲逆, and Shangquyang 上曲陽.
9
Given this context, it is intriguing that, aside from Shanggu as mentioned in the “Memoir of Li Shang,” no location within the Yan 燕state appears in either the Shiji or Han shu accounts of the 196 BCE campaign. It is therefore more plausible that Li Shang, as one of Gaozu’s generals advancing north from Handan, remained within Zhao territory rather than operating independently in the faraway northern region of Yan at the outset of the conflict. This interpretation is further supported by the “Memoir of Lu Wan” in the Shiji (p. 2638), which states: “When Chen Xi rebelled in Dai, Gaozu advanced to Handan to attack him, while the King of Yan, Lu Wan moved from the northeast” 陳豨反代地, 高祖如邯鄲擊稀兵, 燕王綰亦擊其東北. Since Shanggu was under Lu Wan’s jurisdiction and faced Zhao across the border, any rebel forces there would likely have been engaged by Lu Wan’s troops, not those of Li Shang. In short, Li Shang most likely entered Shanggu not during the 196 BCE campaign against Chen Xi, but during the earlier conflict with Zang Tu in 202 BCE.
Supporting evidence can be found in the “Basic Annals of Gaozu.” In the seventh month of the fifth year, Zang Tu, King of Yan, rebelled and seized control of Dai. After Gaozu defeated Zang Tu’s main force beneath Yi 易下 in Yan, he appointed Fan Kuai as Chancellor and sent him northwest to eliminate the remaining rebels in Dai.
10 Given that Shanggu Commandery lies north of Yi, and that Zang Tu had occupied both Yan and Dai, it is reasonable to conclude that Li Shang accompanied Gaozu in the attack at Yi and was subsequently appointed Chancellor of the Right. He then continued northward to
bie ding 別定 “independently pacify” Shanggu, before advancing on Dai together with Fan Kuai. The term
bie ding indicates that Gaozu had already departed from Yi to return to the capital – in the
Shiji, the character
bie 別 is typically used to describe military operations conducted independently of the emperor or highest commander.
11 At this stage, Fan Kuai likely held the position of Chancellor of the Left, while Li Shang, his counterpart in this operation, had been appointed as Chancellor of the Right.
12
The second problem is therefore about the punctuation between sentence (a) and the beginning part of sentence (b):
(a) As Chancellor of the Right and operating separately, he pacified Shanggu, took advantage of the situation to attack the state of Dai,
(b) and received the seal of the Chancellor of State of Zhao.
The question is whether the first part of sentence (b) should be read as continuing the account of the campaign against Zang Tu (202 BCE), or as part of the later campaign against Chen Xi (196 BCE).
If it belongs to the former, then Li Shang must have been appointed Chancellor of Zhao as early as Gaozu’s fifth year (202 BCE). Yet in the tenth year, he is again recorded – now also as Chancellor of Zhao – leading troops against Chen Xi, who was himself serving as Chancellor of Zhao at the time. This would imply that Zhao had two Chancellors simultaneously, which is highly implausible. Moreover, in 202 BCE, the state of Zhao already had its own Chancellor,
13 and the campaign against Zang Tu did not directly involve Zhao (apart from the possibility of passing through it). There would have been no clear reason for the Han emperor to appoint Li Shang as Chancellor of Zhao in that context.
If, however, we depart from the usual reading and instead understand the beginning of sentence (b) as part of the campaign against Chen Xi – which unfolded across both Zhao and Dai – then the appointment becomes far more logical. As Gaozu pacified Zhao and ordered his generals to advance northward into Dai, it would have made strategic sense for him to install a new Chancellor of Zhao
14 – Li Shang – in order to reclaim military authority over Zhao and continue the advance northward.
The third problem involves the chronological inconsistencies introduced by sentence (d). On either interpretation, Li Shang is said to have joined Zhou Bo in subduing Dai during the 196 BCE campaign, only to then return to the capital and serve as Guard General to the Taishang Huang for a full year, before once again heading north to participate in the recapture of Dongyuan alongside Gaozu. However, the Taishang Huang had already passed away before Chen Xi’s rebellion, making it impossible for Li Shang to have served him at that time.
15 Moreover, if Li Shang had in fact returned to the capital for a year before rejoining the campaign, the rebellion would likely have been resolved in his absence.
16 Compounding the problem, the capture of Dongyuan occurred in the early phase of the campaign, not after such a prolonged interruption. In short, sentence (d) introduces a number of chronological contradictions, making the transposition of bamboo slips the key for restoring narrative coherence.
In light of these problems, sentences (a) and (b) should be repunctuated: the comma at the end of sentence (a) may be replaced with a full stop, thereby clarifying that sentence (a) refers to the 202 BCE campaign, while sentence (b) pertains to that of 196 BCE.
(a) 以右丞相别定上谷，因攻代。
As Chancellor of the Right and operating separately, he pacified Shanggu (commandery) and took advantage of the situation to attack the commandery of Dai.
(b) 受趙相國印, 以右丞相趙相國别與絳侯等定代、雁门，
He received the seal of the Chancellor of the State of Zhao, and as Chancellor of the Right of Han and Chancellor of State of Zhao, operating independently and together with the Marquis of Jiang and others, Li Shang pacified the commanderies of Dai and Yanmen,
As for the chronological inconsistency in sentence (d), since Taishang Huang passed away between the rebellion of Zang Tu and Chen Xi, the part of sentence (d) concerning Taishang Huang should at least be replaced between sentence (a) and (b). It is also clear that Gaozu granted Li Shang the seal of Chancellor of Zhao to transfer command over the Zhao forces previously led by Chen Xi. Sentence (b) should therefore come after the defeat of Chen Xi at Dongyuan, as recorded in latter part of sentence (d). This interpretation is further supported by Fan Kuai’s action of hitting Dai as Chancellor of the Left following the victory at Dongyuan – Gaozu appears to have appointed two new Chancellors of Right and Left at the same time – Li Shang and Fan Kuai – who together led the army northward from the eastern front to join Zhou Bo in Dai. Placing sentence (b) after the sentence (d) not only straighten up the logical sequence of all the events but also addresses the chronological inconsistency on Li Shang’s service as Guard General to the Taishang Huang: after pacifying Zang Tu in the fifth year, Li Shang could plausibly have returned to the capital to become guard of the Taishang Huang, who received the honorific title in the sixth year.
As for the chronological inconsistency in sentence (d), since the Taishang Huang passed away between the rebellions of Zang Tu and Chen Xi, the reference to him in sentence (d) should, at the very least, be repositioned between sentences (a) and (b). It is also clear that Gaozu granted Li Shang the seal of Chancellor of Zhao in order to transfer military command over Zhao forces that had previously been led by Chen Xi. Sentence (b), then, should follow the defeat of Chen Xi at Dongyuan, as described in the latter part of sentence (d). This interpretation is further supported by Fan Kuai’s advance into Dai after Dongyuan as Chancellor of the Left. Gaozu likely appointed Fan Kuai and Li Shang as Left and Right Chancellors, placing them in charge of the eastern front as it moved north to support Zhou Bo in Dai. Placing sentence (b) after sentence (d) not only clarifies the logical sequence of events but also resolves the chronological problem concerning Li Shang’s service as Guard General to the Taishang Huang. After pacifying Zang Tu in the fifth year, Li Shang could plausibly have returned to the capital and assumed this post, as the Taishang Huang received the honorific title in the sixth year.
If this revision is accepted, the difficulties in the original passage are effectively resolved, and the narrative aligns more closely with other accounts in the Shiji. While the inconsistencies may have arisen from combining divergent sources, the fact that the inconsistency appears only in this passage – and that the overall coherence can be restored simply by adjusting sentence order without altering the original wording – makes such a reinterpretation the more reasonable explanation. The original sequence of sentences (a) through (d) can thus be reconstructed as follows:
(a) ...... 以右丞相别定上谷，因攻代。
(d) 還以將軍為太上皇衛一歲。七月以右丞相擊陳豨殘東垣，
(b) 受趙相國印，以右丞相趙相國别與絳侯等定代、雁门，
(c) 得代丞相程縱、守相郭同、将军已下至六百石十九人。
Based on the revised sequence and the nature of bamboo-slip manuscripts, it is not difficult to observe that sentences (b) and (c) contain a total of forty-two characters, while sentence (d) contains only twenty-three – roughly half as many. This supports the conjecture that (b) and (c) were originally written on two separate slips, and (d) on a single one (see reconstruction above), suggesting that sentence (d) was likely misplaced after sentence (c).
17
A close comparison of this passage with the
Han shu version gives rise to further intriguing hypotheses. The first is that the disordered sequence of slips likely predates Ban Gu. In the
Han shu “Memoir of Li Shang,” (p. 2075) Ban Gu introduced two key modifications to the
Shiji account: the alternation of a character and the deletion of a clause. First, he changed the
Shiji phrase “he became Guard General of the Taishang Huang” 以將軍爲太上皇衛 to “he commanded [
jiang] the guard of the Taishang Huang”以將軍將太上皇衛. Ban Gu evidently recognized that it would have been implausible for Li Shang to “become” the Taishang Huang’s Guard General at that point in time. To make the sentence more defensible, he replaced
wei 爲 “became” with
jiang將 “commanded,” allowing for the possibility that even if the Taishang Huang had already died, Li Shang might still have commanded the imperial guard unit.
18 In addition, Ban Gu omitted the phrase 以右丞相 from sentence (a), and the longer phrase 以右丞相趙相國别 from sentence (b). Such deletions would only have been necessary if sentences (a) and (b) had already appeared in this sequence and were being read together as part of the Chen Xi campaign. Ban Gu’s editorial approach always involved removing redundant phrasing when adapting
Shiji material.
19 The fact that Ban Gu chose to revise characters and omit certain clauses – yet left the sequence of sentences unchanged – suggests that he, too, found this passage problematic but was unable to resolve its internal inconsistencies. This strongly implies that the disordered structure found in the
Shiji “Memoir of Li Shang” reflects its original form, prior to Ban Gu’s editorial intervention.
That this slip-misplacement could have persisted into Ban Gu’s time and remained in received editions suggests that the textual misorder likely originated during the
Shiji’s early compilation.
20 If Ban Gu’s copy of the
Shiji was indeed transcribed with these slips containing 21–23 characters each, then given the estimated length of 520,000 characters (excluding extra space needed for tables), the full text of
Shiji would have required nearly 30,000 slips.
In sum, the disordered structure in the “Memoir of Li Shang” may have taken shape before or during the initial compilation of the Shiji draft, resulting in later misunderstanding and editorial misreading. Recovering this misplacement not only helps resolve longstanding interpretive challenges, but also offers new insight into the material and editorial practices underlying both the Shiji and the Han shu, as well as into the interpretation of related historical events.
The Misplacement of Slips in the “Memoir of Xiongnu”
The full passage containing the proposed slip-misplacement is transcribed below.
後二歲，復使貳師將軍將六萬騎，步兵十萬，出朔方。彊弩都尉路博德將萬餘人，與貳師會。遊擊將軍說將步騎三萬人，出五原。因杅將軍敖將萬騎步兵三萬人，出雁門。匈奴聞，悉遠其累重於餘吾水北，而單于以十萬騎待水南，與貳師將軍接戰。貳師乃解而引歸，與單于連戰十餘日。
Two years later [97 BCE], the Ershi General [Li Guangli] was again sent to lead sixty thousand horsemen and hundred thousand infantry troops and go out through Shuofang. Lu Bode, the Chief Commandant of the Strong Crossbowmen, led more than ten thousand men and met with the Ershi [General]. [Han] Yue, the Youji General, led thirty thousand men of infantry and cavalry and went out through Wuyuan. [Gongsun] Ao, the Yinyu General, led ten thousand horsemen and thirty thousand infantry troops and went out through Yanmen. When the Xiongnu heard about this, they removed all their baggage to far away in the north of the Yuwu River, and the Shanyu with hundred thousand horsemen waited south of the River to take up the battle with the Ershi General. The Ershi General then withdrew and led [his troops] back, battling with the Shanyu successively for more than ten days.
(a) 貳師聞其家以巫蠱族滅，因並眾降匈奴，得來還千人一兩人耳。遊擊說無所得。因杅敖與左賢王戰，不利，引歸。
When the Ershi [General] heard that his family on account of voodoo sorcery had [suffered the punishment of the whole] clan being wiped out, he took the opportunity and together with his troops surrendered to the Xiongnu, only one or two out of thousand men being able to come back home. [Han] Yue, the Yuqi [General], achieved nothing. [Gongsun] Ao, the Yinyu [General], battled with the Worthy King to the Left, did not gain the upper hand and led his troops returning home.
(c) 是歲漢兵之出击匈奴者不得言功多少，功不得御。
That year, it was not allowed to talk about how much merit had been acquired by the Han troops’ setting out to attack the Xiongnu; merit was not to be mentioned [at all].
(b)
有詔捕太醫令隨但，言貳師將軍家室族滅，使廣利得降匈奴。21
There was an edict to arrest Sui Dan, the Prefect Grand Physician, for telling the Ershi General that his family, house, and clan had being wiped out, letting [Li] Guangli be able to surrender to the Xiongnu.22
This passage concerns an event in the fourth year of the Tianhan period (97 BCE), when Emperor Wu “mobilized the seven classes of reprobated persons in the empire together with resolute and courageous gentlemen”
23 發天下七科謫及勇敢士, and dispatched the Ershi General and other commanders on a large-scale expedition against the Xiongnu, which ended in a fruitless withdrawal.
24 On the other hand, the two underlined sentences (a) and (b) refer to the Ershi General Li Guangli’s 李廣利 surrender to the Xiongnu in the third year of Zhenghe (90 BCE), an event that occurred seven or eight years later and thus cannot belong to the same context. Notably, as early as the Eastern Jin, Xu Guang 徐廣 (352–425) annotated Li Guangli’s surrender with the following remark: “According to the ‘Table of Generals and Ministers’ and the
Han shu, the witchcraft affair began in the second year of Zhenghe (89 BCE); in the year after, Guangli and Shangqiu Cheng marched against the Xiongnu, were defeated, and then surrendered” 《將相表》及《漢書》, 征和二年巫蠱始起. 三年, 廣利與商丘成出擊胡軍, 敗, 乃降.
25 This indicates that the erroneous textual arrangement was already in place before Xu Guang’s time.
A comparison with the
Han shu “Xiongnu” chapter shows that, apart from the numbered sentences (a), (b), and (c), the rest of the wording is virtually identical. During the Tianhan fourth-year campaign (97 BCE), the Ershi General and Chief Commandant of the Strong Crossbowmen 彊弩校尉 Lu Bode 路博德 engaged the Chanyu for more than ten days without result; the General of Roving Cavalry 游擊將軍 Han Yue韓說 (d. 91 BCE) achieved nothing; and the Yinwu General 因杅將軍 Gongsun 公孫敖 (d. 96 BCE) having fared poorly, withdrew.
26 This outcome matches precisely with sentence (c), which states that “merit was not to be mentioned” – since in fact there was no merit to report. By contrast, in the campaign of the third year of Zhenghe (90 BCE), although Li Guangli ultimately surrendered, the other two lines of operation were successful: the Marquis of Chonghe, Mang Tong 莽通 (d. 88 BCE), “returned with the king and populace [of Jushi] in their entirety” 盡得（車師）其王民眾而還, and the Grandee Secretary “broke enemy lines, drove them back, and inflicted heavy casualties” 陷陣却敵, 殺傷虜甚眾.
27 Such outcomes cannot be described with the phrase “no merit was to be mentioned.” Sentence (c) must therefore be part of the original
Shiji account and rightly remains in its current position, even though it was omitted in the
Han shu. By contrast, sentences (a) and (b) are evidently later insertions that have been mistakenly placed in this context.
Interestingly, sentences (a) and (b) contain 25 and 24 characters, respectively. Their lengths strongly suggest that they correspond to two misplaced bamboo slips, each written on strips with a capacity of approximately 24–25 characters. The “Memoir of the Xiongnu” could be divided into sections which are headed by the name of a new Chanyu and begins on a new slip.
28 The quoted passage belongs to the section of Qiedihou Chanyu, which, if considered as an independent textual block, would fit precisely into the calculation of slip capacity. The text extending from
Qiedihou chanyu ji li 且鞮侯單于既立 “After Qiedihou ascended the throne” to
lian zhan shi yu ri 連戰十餘日 “fought continuously for more than ten days”– which immediately precedes sentence (a) – contains 339 characters. Distributed across 14 slips, this yields an average of 24.2 characters per slip (or 22.6 characters per slip if written on 15 slips, which would match the average capacity previously deduced for the “Memoir of Li Shang”). This average is consistent with the 24–25-character capacity of the two displaced slips represented by sentences (a) and (b).
The section from Youji Yue wu suo de 游擊說無所得 “[Han] Yue, the Youji [General], achieved nothing” to gong bu de yu 功不得御 “merit was not to be mentioned,” just before sentence (b), totals only thirty-eight characters. However, gong bu de yu clearly marks the end of this section – after which Sima Qian turned to his remarks, “Taishi gong yue” 太史公曰 – signaling the beginning of a new section and, in material terms, would have been written on a new slip. Sentences (a) and (b), therefore, are best understood as two bamboo slips that were misplaced during transmission and erroneously inserted into this position.
Assuming an average of 24–25 characters per slip, the layout of the passage on bamboo slips may be tentatively reconstructed as follows, with attention to physical arrangement rather than translated content. Notably, in the reconstruction, sentences (a) and (b) would correspond to slips 15 and 18:
1 且鞮侯單于既立盡歸漢使之不降者路充國等得歸單于初立
2 恐漢襲之乃自謂我兒子安敢望漢天子？漢天子我丈人行也
3 漢遣中郎將蘇武厚幣賂遺單于單于益驕禮甚倨非漢所望也
4 其明年，浞野侯破奴得亡歸漢其明年漢使貳師將軍廣利以
5 三萬騎出酒泉擊右賢王於天山得胡首虜萬餘級而還匈奴
6 大圍貳師將軍幾不脫漢兵物故什六七漢復使因杅將軍敖
7 出西河與彊弩都尉會涿塗山毋所得又使騎都尉李陵將步騎
8 五千人出居延北千餘里與單于會合戰陵所殺傷萬餘人兵
9 及食盡欲解歸匈奴圍陵陵降匈奴其兵遂沒，得還者四百人
10 單于乃貴陵以其女妻之後二歲復使貳師將軍將六萬騎
11 兵十萬出朔方彊弩都尉路博德將萬餘人與貳師會遊擊
12 軍說將步騎三萬人出五原因杅將軍敖將萬騎步兵三萬
13 出雁門匈奴聞悉遠其累重於餘吾水北而單于以十萬騎待
14 水南與貳師將軍接戰貳師乃解而引歸與單于連戰十餘日
15
貳師聞其家以巫蠱族滅因並眾降匈奴得來還千人一兩人耳29
16 遊擊說無所得因杅敖與左賢王戰不利引歸是歲漢兵之出擊
17 匈奴者不得言功多少功不得御
18 有詔捕太醫令隨但言貳師將軍家室族滅使廣利得降匈奴
When, then, did these sentences enter the text in the form of mis-ordered slips, and by what mechanism were they preserved? Yi Ping 易平 has argued that the misplacement must have occurred before the time of Xu Guang in the Eastern Jin,
30 but his reasoning differs somewhat from the approach taken in this paper. More importantly, he does not address three critical questions: (a) why the rest of the accreted material concerning the 89 BCE campaign was later excised, (b) when that excision took place, and (c) how the misplaced slips themselves managed to survive the editorial process intact.
The Zhonghua edition Shiji included Tang commentator Sima Zhen 司馬貞 (679–732) Suoyin note immediately after this passage:
《漢書》云：明年且鞮死，長子狐鹿姑單于立。張晏云：
自狐鹿姑單于已下，皆劉向、褚先生所錄，班彪又撰而次之，所以《漢書·匈奴傳》有上下兩卷。
31
The Han shu states: “In the following year, Qiedi [Chanyu] died, and his eldest son, Hulugu Chanyu, succeeded him.” Zhang Yan (from late Eastern Han to early Three Kingdoms period) comments: “From Hulugu Chanyu onward, all entries were recorded by Liu Xiang and Master Chu and later compiled and arranged by Ban Biao; hence the Xiongnu chapter in the Han shu is divided into two scrolls.”
Yu Jiaxi 余嘉錫 (1884–1956) has pointed out that according to the Shiji suoyin dankeben 史記索隱單刻本, this annotation should be under the beginning rather than the end of this passage, that is to be addressing “since Qiedihou Chanyu succeded” 且鞮侯單于既立. Furthermore, Yu argues that the annotation should be corrected as:
張晏云：自且鞮單于已下，皆劉向、褚先生所錄，班彪又撰而次之，所以《漢書·匈奴傳》有上下兩卷。
32
Zhang Yan comments: “From Qiedi Chanyu onward, all entries were recorded by Liu Xiang and Master Chu and later compiled and arranged by Ban Biao; hence the “Xiongnu” chapter in the Han shu is divided into two volumes (juan).”
Yu Jiaxi emends the phrase “From Hulugu Chanyu onward” 自狐鹿姑單于已下 to read “From Qiedihou Chanyu onward” 且鞮侯單于已下. Yet Hulugu Chanyu, the son of Qiedihou Chanyu, did not ascend the throne until the first year of Taishi (96 BCE). What is striking is that both the Ershi General’s futile expedition and the end of the extant Shiji “Memoir of Xiongnu” off fall in the very year of Qiedihou Chanyu’s death – one year before Hulugu succeeded him. If the Shiji contained no material on Hulugu at all, why did Sima Zhen, in his commentary, attach an explanatory note on the son precisely at the point where the account of Qiedihou Chanyu ends?
Zhang Wenhu 張文虎 (1808-1885), drawing on textual variants in the
Suoyin, argues that the copy of the “Memoir of Xiongnu” available to Sima Zhen still contained a continuation down to the first year of the Taishi era (96 BCE). According to Zhang, this continuation was later excised by Song editors when compiling the printed edition of the
Three Commentaries三家注, and it was this omission, he contends, that prompted Sima Zhen to insert his note on Hulugu Chanyu.
33 Yu Jiaxi, by contrast, maintains that the
Suoyin edition does not include the sixteen-character line, “
Han shu says: ‘In the following year Qiedihou died, and his eldest son Hulugu became Chanyu.’” He regards those words as a later insertion added when the
Three Commentaries were printed, and he further insists that Zhang Yan’s note must have read “Qiedihou Chanyu,” the form “Hulugu Chanyu” in the received version being a subsequent corruption.
34 Yi Ping, however, argues that the sixteen characters beginning “
Han shu says…” and the name “Hulugu Chanyu” are original to Sima Zhen. Noting that the two misplaced slips discussed above concern events under Hulugu, Sima Zhen cited Zhang Yan precisely to show that these later notices did not belong to Sima Qian’s text.
35
In any case, when Zhang Yan annotated the Han shu, he explicitly stated that all material following Qiedihou (or Hulugu) Chanyu was a supplement by Liu Xiang and others – implying that everything preceding this point was the work of Sima Qian. Although Zhang Yan was commenting on the Han shu, the text he had in mind was the Shiji; and the version of the Shiji he knew, like the received version today, ended with Qiedihou Chanyu and included nothing about Hulugu.
Yet a puzzle remains. Zhang Yan states that the Han shu “Memoir of Xiongnu” was divided into two parts (juan) at the reign of Hulugu Chanyu. However, the extant Han shu places this division much later – at the reign of Huhanye Chanyu, who ascended in the fourth year of the Shenjue era (58 BCE) under Emperor Xuan, several generations after Hulugu. Two explanations are possible. First, Zhang Yan may have consulted a version of the Han shu whose juan divisions differed from those of the received text. Or the Shiji “Memoir of Xiongnu” available to Zhang Yan already contained entries on Hulugu – and perhaps on later rulers as well – that are now lost. Such post-Hulugu material would likely have been marked, like the later notices in the “Memoir of Guice” with phrases such as “Master Chu says” 褚先生曰 or “Liu Xiang says” 劉向曰, signaling its status as later additions. It was on the basis of those textual markers that Zhang Yan commented as such in the Han shu “Xiongnu zhuan” 匈奴傳.In other words, these later insertions had already been woven into the transmitted text and were circulating together with the main narrative before Zhang Yan’s time, i.e. late Eastern Jin to early Three Kingdoms period but were excised again at some point prior to Xu Guang in the Eastern Jin.
Drawing together the evidence, the most plausible scenario is this: in the
Shiji “Memoir of Xiongnu,” the two sentences that refer to Li Guangli’s surrender are later interoplations, whereas the rest of the passage is Sima Qian’s original text. These supplementary sentences were probably added after the
Shiji notice “Guangli surrendered to the Xiongnu” 廣利得降匈奴, but were excised in a subsequent redaction. During that excision – when bamboo slips were dismantled and re-compiled
36 – two strips were mistakenly preserved and reintegrated into the main narrative, eventually making their way into the received text by chance.
Why were they removed? The later compiler may have recognized that the stories were not by Sima Qian
37 and that they diverged sharply from the version of events recorded in the
Han shu.
Indeed, the later additions concerning the arrest of Li Guangli’s family for cursing the throne and Li’s own surrender to the Xiongnu adopt a narrative and tone entirely at odds with the Han shu. The Han shu “Memoir of Liu Qumao” (p. 2883) states: “The Ershi General’s wife and children were likewise arrested. When the Ershi General heard of this, he surrendered to the Xiongnu, and his entire clan was exterminated” 貳師將軍妻子亦收. 貳師聞之, 降匈奴, 宗族遂滅. But the Han shu “Memoir of Li Guangli” (p. 2703) records only: “His army was defeated, and he surrendered to the Xiongnu” 兵敗, 降匈奴. A more detailed narrative appeared in the Han shu “Memoir of Xiongnu” (pp. 3779–3780): “It happened that the Ershi General’s wife and household were arrested on charges of witchcraft. When he heard the news, he became anxious and afraid… ‘My wife and household are all in the hands of the authorities. If I return without fulfilling the emperor’s expectations, I will meet them only in prison—will I ever be able to see them again north of Zhi?’ Because of this doubt, he resolved to press deeper in search of military merit… The army then fell into disorder and was defeated, and the Ershi General surrendered” 會貳師妻子坐巫蠱收, 聞之憂懼…‘夫人室家皆在吏, 若還不稱意, 適與獄會, 鄒居以北可復得見乎?’ 貳師由是狐疑, 欲深入要功…軍大亂敗, 貳師降. Therefore, it can be inferred that in the Han shu’s historical perspective, Emperor Wu first arrested Li Guangli’s family; upon hearing this, Li attempted to redeem himself through military success but, after defeat, surrendered when all hope of return was lost. Only after news of his surrender reached the court did Emperor Wu order the extermination of Li’s clan.
The two misplaced slips in the “Memoir of Xiongnu” in the
Shiji record instead that “When the Ershi [General] heard that his family on account of voodoo sorcery had [suffered the punishment of the whole] clan being wiped out, he took the opportunity and together with his troops surrendered to the Xiongnu, only one or two out of thousand men being able to come back home” 貳師聞其家以巫蠱族滅, 因並眾降匈奴,得來還千人一兩人耳, and that “There was an edict to arrest Sui Dan, the Prefect Grand Physician, for telling the Ershi General that his family, house, and clan had being wiped out, letting [Li] Guangli be able to surrender to the Xiongnu” 有詔捕太醫令隨但, 言貳師將軍家室族滅, 使廣利得降匈奴. In other words, here Emperor Wu exterminates Li Guangli’s family first and then orders strict secrecy – otherwise there would have been no need to seize the physician who leaked the news. Such an explicit censure of the emperor’s severity would have sat uneasily with the Eastern-Han historiographical view, which sought to fashion a late-reign image of Emperor Wu as one who “deeply repented of past excesses.”
38 Whether through oversight or deliberate choice, however, these sentences remained in the text.
Whatever the initial context, the fragments remind us that the events recorded in the
Han shu represents only one version of “history.” The post-Tang textual history of both the
Shiji and the
Han shu has been studied in detail, but the history of their transmission between the Han and the Tang – several centuries in which competing versions evidently circulated – remains largely unexplored.
39 Behind those versions lay a constant contest of selection and interpretation. The version we have today is what remains after the dust has settled, accompanied by a few accidental survivals from earlier textual layers.
Conclusion
The two cases of mis-ordered slips in the “Memoir of Li Shang” and “Memoir of Xiongnu” together indicate that the “Memoir” part of the
Shiji was first written on bamboo slips that carried roughly 21–23 characters each. The misplaced slips in the “Memoir of Xiongnu” hold 24–25 characters, but those sentences are later accretions; the
Shiji’s original text – like the “Memoir of Li Shang” – would have averaged about 21–23 characters per slip (e.g., 339 characters divided among 15 slips yields 22.6 characters per slip). That range matches the slips of the
Chunqiu shiyu 春秋事語excavated from the Western-Han tomb at Fuyang – during Emperor Wendi’s reign – which also average 23 characters.
40
As aforementioned, Nienhauser
41 and Boltz have both proposed a practice of Sima Tan and Sima Qian excerpting the passages they wished to use from earlier sources to compile the
Shiji, they copied the text with single slip with mark and thereafter arranged them into
juan. The reconstructed slip order in the
Shiji memoirs offers plausible support to their hypothesis.
42
This observation prompts a further distinction that future studies of Shiji historiography must attend to: which portions of the text convey the Simas’ own views, and which preserve the meaning already inherent in the materials they selected? What alterations, if any, did they impose on those sources? Did they review and evaluate the textual sources they worked from? These questions remain to be explored.
Translator: Jianan Shen, Nanjing University
Notes
REFERENCES
- Ban Gu 班固. Han shu 漢書. Beijing: Zhonghua shuju, 1962.
- Boltz, William G. “Myth and the Structure of the Shyy Jih.” Asiatische Studien Études Asiatiques 56, no.1 (2002): 573-585.
- Dubs, Homer H. The History of the Former Han Dynasty, Vol. 2. Baltimore: Waverly Press, 1944.
- Han Ziqiang 韓自強. Fuyang han jian Zhouyi yanjiu 阜陽漢簡周易研究. Shanghai: Shanghai guji chubanshe, 2004.
- Kuang Yantao 鄺彥陶. “Dunhuang chaoben Shiji shijia de tihang wenti” 敦煌鈔本《史記》世家的提行問題. Chinese Classics and Culture 中國典籍與文化, forthcoming.
- Li Feng 李峰. “Wu gu zhi huo shiyu xia Han Wudi yu Li taizi jiuge tanxi” 巫蠱之禍視域下漢武帝與戾太子糾葛探析. Qinghua daxue xuebao 清華大學學報, no. 2 (2020): 60-75.
- Li Ling李零. Jianbo gushu yu xueshu yuanliu 簡帛古書與學術源流. Shanghai: Sanlian shudian, 2004.
- Liang Yusheng 梁玉繩. Shiji zhiyi 史記志疑. Beijing: Zhonghua shuju, 1981.
- Nienhauser, William H., ed. The Grand Scribe’s Records, Vol. 8: The Memoirs of Han China, Part I. Bloomington: Indiana University Press, 2008.
- ____, ed. The Grand Scribe’s Records, Vol. 9: The Memoirs of Han China, Part II. Bloomington: Indiana University Press, 2010.
- Sima Qian 司馬遷. Shiji 史記. Beijing: Zhonghua shuju, 1959.
- Sun Jiazhou 孫家洲. “Han chu yi chengxiang, xiangguo tongbing kao” 漢初以丞相、相國統兵考. Junshi lishi 軍事歷史, no. 6 (1998): 21-23.
- Wang Shumin 王叔岷. Shiji jiaozheng 史記斠證. Beijing: Zhonghua shuju, 2007.
- Xin Deyong 辛德勇. “Han Wudi wannian zhengzhi quxiang yu Sima Guang de chonggou” 漢武帝晚年政治取向與司馬光的重構. Qinghua daxue xuebao 清華大學學報, no. 6 (2014): 5-50.
- Yi Ping 易平. “Shiji ‘Xiongnu liezhuan’ mo duan cuo jian kao bian” 史記·匈奴列傳末段錯簡考辯. Zhongguo dianji yu wenhua luncong 中國典籍與文化論叢, no. 1 (1995): 316-324.
- Yuan Zuliang 袁祖亮. “Zhanguo Qin Han Weijin Nanbeichao shiqi de xiangguo yu chengxiang” 戰國秦漢魏晉南北朝時期的相國與丞相. Zhengzhou daxue xuebao 鄭州大學學報, no. 6 (1988): 58-65.
- Zhang Wenhu 張文虎. Jiao kan Shiji Jijie Suoyin Zhengyi zha ji 校勘史記集解索隱正義札記. Beijing: Zhonghua shuju, 2012.
News
