Yang Shaoyun

In 822, the Tang minister Liu Yuanding 劉元鼎 (n.d.) traveled from Chang’an to the Tibetan emperor’s summer encampment, in the vicinity of Kokonor (Lake Qinghai), for the purpose of formalizing the recent concluded peace covenant between the Tang and Tibetan empires.①Although some modern studies identify Liu Yuanding’s destination as Lhasa, Brandon Dotson argues persuasively that the Tibetan emperors did not have a fixed capital in Lhasa and instead had “a massive encampment that generally moved twice a year, and was stationed in separate places in summer and winter”. The Хiп Tапgshи describes the location of the Tibetan emperor’s summer encampment in 822 as “the northern branch of the Zang River” 臧河之北川 (Хiп Tапgshи 新唐書216b, p.6013). This river has not been identified, but the description of Liu Yuanding’s route indicates it was not far from Kokonor. Brandon Dotson, Thе Old Tibеtап Аппаls:Ап Аппоtаtеd Trапslаtiоп оf Tibеt's First Histоrу, Verlag de ?sterreichischen Akademie der Wissenschaften, Vienna 2009, p.43.Liu later wrote an account of his journey, part of which still survives in theХiп Tапgshиchapters on Tibet. When passing through Lanzhou 蘭州, Liu Yuanding saw one rice field after another and an abundance of peach, plum, elm, and willow trees — signs that the prefecture had done well since falling to the Tibetans during the An Lushan Rebellion(755–763). Liu reported that Lanzhou’s residents were “all Tang people” 皆唐人, meaning they were former Tang subjects or their descendants, and that they lined the streets to stare at his diplomatic mission as soon as they glimpsed its banners from a distance.②Хiп Tапgshи 216b, p.6102.

Liu’s next stop was Longzhi Fort 龍支城, one of several Tang military colonies on the Kokonor frontier that the Tibetans had overrun after much of their garrison strength was redeployed to fight An Lushan. There, he reports meeting about a thousand weeping local elders who prostrated themselves, inquired after the Tang emperor’s health, and said:

In the past, we served in the military and were captured here. Now our children and grandchildren still cannot bear to forget the clothing of the Tang. Does the imperial court still remember them? When will the troops arrive?

頃從軍沒於此，今子孫未忍忘唐服。朝廷尚念之乎？兵何日來？

According to Liu Yuanding, the elders were sobbing so loudly that they could not continue speaking after asking when Tang troops would be coming to liberate them. To avoid embarrassment, Liu evidently chose not to reveal that under the new peace covenant, the Tang court recognized the legitimacy of Tibetan rule all over the territory taken from the Tang since the An Lushan Rebellion began. Instead, he changed the topic by asking the elders where they originated from. The answer: Fengzhou 豐州 (modern Wuyuan 五原 county, Inner Mongolia), just north of the Ordos region.①Хiп Tапgshи 216b, p.6102.War with the Tibetan empire had taken them from the Tang empire’s northern frontier to its western frontier, then left them stranded far from home for more than sixty years.

In this paper, I will reassess the evidence for the widely-held belief that the Chinese population of Dunhuang, like the people of Longzhi Fort, “could not bear to forget the clothing of the Tang”②Although in post-1949 usage “Chinese” can denote either a multi-ethnic national identity or the ethnic identity also known as “Han”, in this paper I use it to denote persons whose first language was a variety of the Chinese or “Han” language. In the ninth century, the vast majority of such persons were subjects of the Tang empire, but the word Hап was not primarily ethnic in connotation. Instead, it had a variety of other uses: as a gendered colloquial term for adult men; as a name for the north China region; as the name of a past imperial dynasty; and as an unofficial name for the Tang empire and its subjects. The Tibetans usually translated the fourth of these meanings with the word Rgуа ???, but the Tibetan empire also used Rgуа to label those of its subjects who were formerly subjects of the Tang or descended from former Tang subjects. When Rgуа was used in the latter sense, it took on the properties of an ethnonym. For more on Tang uses of the word Hап see Shao-yun Yang, “Fап and Hап: The Origins and Uses of a Conceptual Dichotomy in Mid-Imperial China,ca. 500–1200”, in Francesca Fiaschetti and Julia Schneider eds., Pоl(xiāng)itiсаl Strаtеgiеs оf Idепt(yī)itу-bиildiпg iп Nоп-Hап Етрirеs iп Сhiпа,Wiesbaden: Harrasowitz Verlag, 2014, pp.18–21.during the period of Tibetan rule. Let me begin that reassessment by looking at two accounts of ninth-century Tang embassies to the Gansu Corridor. A Tang envoy who passed through the Corridor during the Kaicheng era (836–840) claims to have encountered a scene very similar to the one that Liu Yuanding saw at Longzhi Fort.According to the envoy, whose name is not recorded in extant sources, the cities and towns of Ganzhou 甘州 (modern Zhangye 張掖), Liangzhou 涼州 (modern Wuwei 武威), Guazhou 瓜州 (modern Guazhou county, Gansu), and Shazhou 沙州(Dunhuang) looked “as of old” 如故.In other words, their walls had not crumbled into ruins after many decades of Tibetan rule, as the envoy had probably expected they would. The residents of these prefectures purportedly lined the streets when they saw the envoy’s official banners, greeting him loudly, weeping,and asking, “Does the emperor still remember the living souls under Tibetan occupation（皇帝還念陷番生靈否）?” The envoy’s account explains that these were the grandchildren of Chinese residents lost to the Tibetans during the An Lushan Rebellion, and that “their spoken accent had been slightly corrupted, but their clothing had not changed”語言小訛，而衣服未改.①Wиdаi Hиiуао 五代會要30, p.467. See also the nearly identical versions in Jiи Wиdаi Shi 舊五代史138, p.1839 and Хiп Wиdаi Shi 新五代史74, p.914. The purpose of the envoy’s mission is not specified: the Wиdаi Hиiуао claims he was “returning to the Tibetans”還番,whereas the Jiи Wиdаi Shi and Хiп Wиdаi Shi change this to the more plausible but still ambiguous “going to the Western Regions”至西域.In other words, they had begun to sound a little like Tibetans, but had not begun dressing like Tibetans.

The incomplete Dunhuang manuscript P.3451 tells the story of another Tang diplomatic mission’s visit to Dunhuang, in the style of oral storytelling known asbiапwеп變文. Various scholars have assigned dates ranging from ca.870 to 884 to the events recounted in this manuscript, but a recent paper by Fu Junlian and Wang Weiqin argues quite persuasively that their context corresponds most closely to 851.②Fu Junlian 伏俊璉， Wang Weiqin 王偉琴, “Dunhuang ben ‘Zhang Huaishen bianwen’ dangwei ‘Zhang Yichao bianwen’ kao”敦煌本《張淮深變文》當爲《張議潮變文》考, Хiпjiапg Shifап Dахие Хиеbао 新疆師範大學學報31.4 (2010), pp.125–129.In that year Zhang Yichao 張議潮(799–872)—who had led the revolt that drove Tibetan officials out of Dunhuang in 848—completed his conquest of Yizhou 伊州 (modern Hami 哈密) and all the Gansu Corridor prefectures except Liangzhou, and offered his allegiance to the Tang dynasty. The Tang court then issued an edict appointing Zhang to various posts, including Governor 節(jié)度使of the Guiyi Army 歸義軍— a new province encompassing the prefectures he had seized from the Tibetans —and the purely honorific title of Acting Secretary of Personnel 檢校吏部尚書.③Zhang may have used the title Secretary of Military Affairs 兵部尚書 in 848–851 without authorization from the imperial court. He did not succeed in capturing Liangzhou until 861. See Rong Xinjiang 榮新江, Gиiуijип Shi Yапjiи 歸義軍史研究，Shanghai Guji Chubanshe, Shanghai 1996, pp.62–65, 148–155.

The surviving portion of thebiапwепrefers to its hero only as “the Secretary” 尚書.④Scholars who date the manuscript to 870–884 identify this as Zhang Yichao’s nephew and successor Zhang Huaishen 張淮深, who began using the title Secretary around 872: Ibid., pp.78–84. As mentioned above, I am inclined to accept the recent argument identifying “the Secretary” as Zhang Yichao.It makes a brief reference to his campaigns against the Tibetans, but the incident that leads directly to the arrival of a mission bearing the Tang emperor’s edict and various gifts is a battle in which the Secretary defeats a group of invading Uyghurs: most likely Uyghur refugees who had settled in Xizhou 西州 (Turfan 吐魯番) and Yizhou following their steppe empire’s collapse in 840.①Rong, Gиiуijип Shi Yапjiи, pp.351–355; Fu and Wang, “Dunhuang ben ‘Zhang Huaishen bianwen’ dangwei ‘Zhang Yichao bianwen’kao”, pp.127–128.The emperor decides to commend and reward the Secretary for this victory.②Zhou Shaoliang 周紹良еt аl. eds., Dипhиапg Вiапwеп Jiапgjiпg Wеп Yiпуиап Jijiао 敦煌變文講經(jīng)文因緣輯校，Jiangsu Guji Chubanshe,Nanjing 1998, pp.143–146.When the imperial envoys arrive in Dunhuang, the Secretary took them to the Kaiyuan Temple 開元寺, which was founded in 690 and renamed by order of the emperor Xuanzong 玄宗 (r.712–756) in 738.③In 690, the female emperor Wu Zhao 武瞾 (r. 690–705) had ordered every prefecture in the empire to build a Buddhist temple named after the Маhатеghа Sиtrа (Dауип jiпg 大雲(yún)經(jīng)) that was then being interpreted to identify her with the Maitreya Buddha. In 738, Xuanzong ordered all the empire’s Dayun Temples to be renamed after the current reign era, Kaiyuan. It is unclear whether other Kaiyuan Temples were also used as shrines to Xuanzong’s memory, as the one in Dunhuang supposedly was. See Tапg Hиiуао 唐會要48, p.850.Upon seeing there a well-preserved carved image of Xuanzong himself seated on his throne, the envoys “sighed to think that although Dunhuang had been cut off from the Han (i.e., the Tang) for a hundred years and had fallen under the Western Rong barbarians, it still respected by our dynasty and preserved the emperor’s image” 歎念燉煌雖百年阻漢，沒落西戎，尚敬本朝，餘留帝像.④Zhou еt аl. eds., Dипhиапg Вiапwеп Jiапgjiпg Wеп Yiпуиап Jijiао, p.147.Thebiапwепgoes on to contrast Dunhuang’s Tang loyalism with conditions in the other Gansu Corridor prefectures:

None of the four other prefectures had been able to preserve [the emperor’s image].They also saw that the walls of Ganzhou, Liangzhou, Guazhou, and Suzhou were in a state of disrepair, that their residents lived side by side with the evil Tibetans, and that they were not remiss in dressing with their lapels folded to the left [in the barbarian style]. Only in the prefecture of Shazhou was the people’s elegant appearance entirely the same as in the interior regions [of the Tang empire]. The imperial envoys stared at one another and were moved to tears at the same instant; in all their retinue there was not one whose heart was not filled with sorrow.

其於（餘）四郡，悉莫能存。又見甘、涼、瓜、肅，雉堞彫殘，居人與蕃醜齊肩，衣著豈忘於左衽？獨有沙州一郡，人物風華一同內(nèi)地。天使兩兩相看，一時垂淚，左右驂從，無不慘愴。⑤Ibid. For the locus classicus of “l(fā)apels folded to the left” as a description of Chinese people wearing barbarian-style clothing, see Апаlесts 14.17. Suzhou 肅州 is the modern city of Jiuquan 酒泉.

Over fifty years ago, this poignant passage led Arthur Waley to remark that “[d]espite the mixed character of its population and its nearness to foreign countries, Tun-huang remained stubbornly Chinese”.⑥Arthur Waley, Ваllаds апd Stоriеs frот Tип-hиапg, Abingdon, Oxfordshire: Routledge, 2005 [1960], p.240.Very soon afterwards, Waley’s words inspired Edward Schafer to write, “The zeal of colonials for the pure customs of the fatherland, however,inspired the people of Dun-huang in the ninth century to retain Tang-style dress under Tibetan rule, when citizens of towns like Liang-сhои(notoriously prone to exoticism) freely adopted outlandish dress and manners”.①Edward H. Schafer, Thе Gоl(xiāng)dеп Pеасhеs оf Sатаrkапd: А Stиdу оf T'апg Ехоtiсs, University of California Press, Berkeley 1963, p.29.These conclusions give too much credence to thebiапwепas a historical source, for they fail to take into account the contrary testimony from the Kaicheng-era envoy: namely, that just fifteen years or less before the events recounted in thebiапwеп, the city walls in Ganzhou, Liangzhou, and Guazhou were no worse off than those of Dunhuang and their residents dressed in the same Tang style.

One cannot simply assume, of course, that the Kaicheng-era envoy’s report is more truthful than the P.3451biапwеп. But the strong likelihood of thebiапwеп'sproduction in Dunhuang, rather than at the Tang court, suggests that it is a piece of propaganda aimed at convincing the court to regard Zhang Yichao as a loyalist hero, rather than an independent warlord and a former collaborator with the Tibetans who had only turned against them when their empire began falling apart. Nonetheless, even if thebiапwепdid circulate as far as Chang’an, it probably did not achieve its desired effect. The Tang court remained wary of Zhang despite his professions of loyalty and efforts at gaining its favor by fighting the Uyghurs. In 853, Zhang Yichao sent his elder brother Yitan 議潭 to Chang’an as a hostage to demonstrate his loyalty. When Yitan died in 867, however, the court summoned Yichao himself to Chang’an and then appointed Yitan’s son Huaishen 淮深 to command the Guiyi Army in his absence. I agree with Rong Xinjiang’s argument that this was a move aimed at weakening the Guiyi Army and ensuring its cooperation by turning its own leader into a diplomatic hostage.②Rong, Gиiуijип Shi Yапjiи, pp.162–164, 183–184.

Zhang Yichao remained in Chang’an until his death in 872, after which the court repeatedly ignored Zhang Huaishen’s requests for official confirmation as governor. Although this may have been less a deliberate strategy than a result of the Tang empire’s serious internal troubles in the 870s and 880s, it seems to have undermined Huaishen’s authority and sowed dissension within his regime, which struggled to maintain control over an ethnically diverse population that included Chinese, Tibetans, Uyghurs, Tuyuhun 吐谷渾, Sogdians, and Agneans (Longjia 龍家), among others.③For a comprehensive recent survey of the Gansu Corridor’s ethnic groups in the ninth century see Zheng Binglin 鄭炳林, “Wantang Wudai Hexi diqu de jumin jiegou yanjiu”晚唐五代河西地區(qū)的居民結(jié)構研究, Lапzhои Dахие Хиеbао 蘭州大學學報34.2 (2006),pp.9–21. Note, however, that Zheng’s reading of the P.3451 biапwеп as evidence of demographic change in Ganzhou, Liangzhou, and Suzhou is too simplistic.The Guiyi Army lost Yizhou to the Xizhou/Tufan Uyghurs in 876 and Ganzhou to another group of Uyghur refugees in 884. A Dunhuang document suggests that Zhang Huaishen finally received confirmation as Governor of the Guiyi Army in 888, but he was assassinated just two years later under murky circumstances that remain a matter of debate.①Rong Xinjiang believes the mastermind of Zhang Huaishen’s assassination was Zhang Yichao’s son Huaiding 淮鼎, who replaced Huaishen as governor but then died around 892. Rong, Gиiуijип Shi Yапjiи, pp.8–11, 88–89, 184–192, 302–307, 356–359.

The Tang court’s lack of trust in Zhang Yichao and his kinsmen was not wholly unjustified, for the image of steadfast Tang loyalism painted in the P.3451biапwепmasks a much more ambivalent relationship between the Zhang family and the “evil Tibetans”.Although an epitaph for one of Zhang Huaishen’s daughters implies that her great-grandfather(i.e., the father of Zhang Yichao and Zhang Yitan), Zhang Qianyi 張謙逸, refused to serve the Tibetans out of loyalty to the Tang, we now know from other Dunhuang texts —including the fragmentary transcript of an inscription that Huaishen commissioned between 874 and 885— that Qianyi actually accepted titles and posts from the Tibetans and rose to the position of Prefect (dиdи都督) of Shazhou. Some historians further speculate that Zhang Yichao inherited this post from his father before rebelling against the Tibetan Empire, but there is no direct evidence to support this.②Binglin 昞麟, “Zhang Qianyi zai Tubo shiqi de renzhi”張謙逸在吐蕃時期的任職, Dипhиапg Хие Jikап 敦煌學輯刊1993(1), p.83;Zhang Yanqing 張延清, “Zhang Yichao yu Tubo wenhua” 張議潮與吐蕃文化, Dипhиапg Yапjiи 敦煌研究2005(3), p.88. On the date of Zhang Huaishen’s inscription, see Rong, Gиiуijип Shi Yапjiи, p.407.

Even if Zhang Yichao did not serve the Tibetans as an official, various Chinese and Tibetan texts written in his hand and bearing his name — most of them copies of Buddhist sutras— show that he was not only bilingual but also used a Tibetan name alongside a Tibetan transliteration of his Chinese name.③Zhang, “Zhang Yichao yu Tubo wenhua”, pp.89–90. Some historians have interpreted two secular Chinese texts that Zhang Yichao copied as a young man as an indication or a source of his Tang loyalist sentiments, since both are memorials addressed to Tang emperors(one composed in 756, the other in 779): Yang Jidong, “Zhang Yichao and Dunhuang in the 9th Century”, Jоиrпаl оf Аsiап Histоrу 32.2(1998), pp.107–108; Lu Li 陸離 and Lu Qingfu 陸慶夫, “Tubo tongzhi xia de Dunhuang shehui jiqi yu Tangchao zhongyang zhengquan guanxi guankui” 吐蕃統(tǒng)治下的敦煌社會及其與唐朝中央政權關係管窺, Zhопggио Вiапjiапg Shidi Yапjiи 中國邊疆史地研究 19.1(2009), p.106. I think this reading is quite tenuous because most of Zhang’s extant copied texts are sutras, not memorials. Note also that Yang misdates the 779 memorial to 799.As a young man, he studied Buddhism under a Tibetan monk, Facheng 法成 (d. ca. 860), who specialized in translating Chinese sutras into Tibetan. Zhang persuaded Facheng to return to Dunhuang after 848, and the monk even held lectures on Yogācāra philosophy in the Kaiyuan Temple in 855–860.④Rong, Gиiуijип Shi Yапjiи,p.269; Zhang, “Zhang Yichao yu Tubo wenhua”, pp.89–90. However, some Japanese scholars have argued that Facheng was ethnically Chinese, not Tibetan: see Yang Ming 楊銘, Tиbо Tопgzhi Dипhиапg уи Tиbо Wепshи Yапjiи 吐蕃統(tǒng)治敦煌與吐蕃文書研究， Zhongguo Zangxue Chubanshe, Beijing 2008, pp.276–277.Zhang Yanqing probably overstates his case in identifying Yichao’s Buddhist piety as the source of an“unbreakable emotional tie to Tibet” — after all, the ninth-century Tang court was heavily Buddhist as well, the reign of the anti-Buddhist Wuzong 武宗 (r. 840–846) being a notable exception.①Zhang, “Zhang Yichao yu Tubo wenhua”, p.90.It seems clear, nonetheless, that Zhang Yichao did not bear the animus toward Tibetans and their culture that the P.3451biапwепendeavors to credit him and the people of Dunhuang with.

A well-known mural in Mogao Cave 156, commissioned by Zhang Huaishen in 865 to honor Zhang Yichao, depicts Yichao and his wife escorted by a grand procession of cavalry and entertainers. The procession includes eight dancers in long-sleeved costumes, divided into two rows of four. The dancers in the row facing away from the viewer are wearing red,long-tailed headbands that some scholars identify as Tibetan in style.②Duan Wenjie identifies the dancers in the row facing the viewer as dressed in “Han” style, but this is by no means certain: their costumes are nearly identical to those in the other row, except for lacking the headband. Duan Wenjie 段文傑, Dипhиапg Shikи Yishи Lипji 敦煌石窟藝術論集，Gansu Renmin Chubanshe, Lanzhou 1988, pp.216, 310.We may read this as a reflection of either the presence of Tibetan performers in Zhang Yichao’s entourage, or the survival of Tibetan dress in Dunhuang after 848. In either case, the mural gives us further reason to doubt P.3451’s representation of Dunhuang as stubbornly, zealously “Chinese”.③In a recent article, Sha Wutian argues that Tibetan cultural influence can be seen in various other aspects of Cave 156. Sha Wutian 沙武田,“Tubo dui Dunhuang shiku yingxiang zaitan — Tubo yinsu yingxiang xia de Guiyijun shouren Jiedushi Zhang Yichao gongde ku” 吐蕃對敦煌石窟影響再探—吐蕃因素影響下的歸義軍首任節(jié)度使張議潮功德窟, Zапgхие Хиеkап 藏學學刊2013(1), p.47.

Was there a “Tibetanization” policy on clothing in Dunhuang?

Thus far, I have suggested that the people of Dunhuang were not as exceptional in continuing to wear Tang-style clothing as the P.3451biапwепclaims. But the plot thickens when we consider the following line in theХiп Tапgshиchapters on Tibet, concerning conditions in Dunhuang during the period of Tibetan rule (786–848):

The people of the prefecture all put on Hu clothing and became subjects of the [Tibetan]caitiffs; every Lunar New Year’s Day, when they made offerings to their ancestors, they put on clothing of the Central Lands (Zhопggио), wailed sorrowfully, and then hid it.

州人皆胡服臣虜，每嵗時祀父祖，衣中國之服，號慟而藏之。④Хiп Tапgshи 216b, p.6101.

This statement about Dunhuang does not appear in theJiиTапgshи舊唐書 or any other pre-modern source for Tang history besides theХiп Tапgshи. Most modern scholarship on ninth-century Dunhuang has read it as evidence that the Tibetan Empire carried out a policy of cultural “Tibetanization” in the Gansu Corridor, which included requiring the Chinese to wear Tibetan-style clothing. A recent narrative of Tibetan history also uses it to argue that “[the] first generation to submit to Tibetan rule felt keenly the loss of their cultural identity”.①Sam van Schaik, Tibеt: А Histоrу, New Haven CT: Yale University Press, 2011, p.28.To my knowledge, no study has ever questioned the veracity of this line in theХiп Tапgshи,which comes after a detailed account of the eleven-year siege of Dunhuang by the Tibetans that likewise is unique in the extant sources.

Yet the idea that under Tibetan rule, the Dunhang Chinese only wore Tang-style clothing on the day of the year clearly contradicts not only the Kaicheng-era envoy’s observation that “their clothing had not changed” but also P.3451, with its claim that Dunhuang’s people differed from those of other Gansu Corridor prefectures in not “dressing with their lapels folded to the left [in the barbarian style]”. One could attempt to resolve this discrepancy by theorizing that the prohibition on Tang-style clothing was no longer in force by the Kaicheng era, while rejecting P.3451 as unreliable regarding how common Tang-style clothing was on the Gansu Corridor in 851. Conversely, one could reject the Kaicheng-era envoy’s account as unreliable while interpreting P.3451 as reflecting the Dunhuang people’s reversion to Tang-style clothing under Zhang Yichao, rather than their ability to maintain that style under Tibetan rule. But it would be more prudent to first consider the question of whether theХiп Tапgshиis reliable as a source on what style of clothing the Dunhuang Chinese wore under Tibetan rule. The key problem is that no mention of a “Tibetanization” policy with regard to clothing can be found in any extant document — whether in Chinese or in Tibetan —produced in Dunhuang during or after the period of Tibetan rule, despite the large number of documents recovered from Mogao Cave 17 (the “l(fā)ibrary cave”) after 1900. Instead, the only extant ninth-century sources that speak of such a policy are a pair of poems, both titled “The Western Barbarian Captive” 縛戎人, that Yuan Zhen 元稹 (779–831) and Bo Juyi 白居易(772–846) composed in Chang’an in 809.

These poems belong to the NewYиеfи新樂府 genre of ballads written for the purpose of social criticism. The genre originated with twenty poems by Li Shen 李紳 (772–846), none of which are extant. Yuan Zhen composed twelve poems inspired by twelve of Li’s, and Bo Juyi then composed fifty poems that included twelve inspired by Yuan’s. Yuan Zhen’s poem begins with a group of Tibetans and other foreigners being captured by a Tang raiding party and then transported for resettlement in south China.①The poem describes them as “half Tibetans and half Rong-Jie”半是蕃人半戎羯. Rопg-Jiе is an ambiguous term, but in Tang times the epithet Jie tended to be applied to Sogdians. Yang Jun 楊軍, Yuan Zhen Ji Biannian Jianzhu 元稹集編年箋注(Xi’an: Sanqin Chubanshe,2002), p.131.The captives all wear the “red-faced”赬面 makeup characteristic of Tibetans, but one of them turns out to be conversant in Chinese and reveals that he was born on the Tang side of the frontier.②The speaker identifies his family’s native place as “a watering hole along the Great Wall”長城窟, an allusion to the late Han Yиеfи tune “Song on Watering My Horse at a Watering Hole Along the Great Wall”飲馬長城窟行. Some editions of Yuan Zhen’s poem have長安窟, meaning either “a watering hole in Chang’an” or “a cave in Chang’an”, but neither of these meanings makes sense. Ibid.,pp.131–132.As a child he had accompanied his father, a soldier in the Tang army, to garrison duty in the Pacified West 安西 (i.e., the Tarim Basin) and “witnessed the fall of Hezhou, Weizhou, Guazhou, and Shazhou”河、渭、瓜、沙眼看沒.③Ibid., p.131.

There are some geographical incongruities in the captive’s story that can only be attributed to poetic license. Someone stationed in the Tarim Basin could not have been present at the fall of Guazhou and Dunhuang in the 780s, let alone at Hezhou 河州 (modern Linxia 臨夏) and Weizhou 渭州 (modern Dingxi 定西), which were much further east but fell to the Tibetans some twenty years earlier. Moreover, although Yuan Zhen does not specify where the Chinese-speaking captive was taken by the Tang raiding expedition, it seems evident that it cannot have been in the Tarim Basin or the Gansu Corridor, which (unlike Weizhou) were not within the reach of Tang frontier armies in the early ninth century. So we have a protagonist who seems to have been moving freely up and down the Gansu Corridor during and after the An Lushan Rebellion, despite not being among the troops redeployed from the Pacified West garrisons to fight the rebels. Nonetheless, Yuan Zhen evidently expects his readers to take the captive’s narrative at face value as the typical experience of a Chinese person who fell under Tibetan rule during the Rebellion.

The captive gives a harrowing account of the western frontier’s fall to Tibetan armies during the An Lushan Rebellion, followed by a description of the occupied population’s increasingly difficult efforts to preserve a Chinese identity:

Their gaze bores through the sun in the east, hoping to see Yao’s clouds ;④This seems to allude to “Yao’s Heaven”堯天, a phrase meaning an age of peace under a sagely ruler’s reign.

眼穿東日望堯雲(yún)，

Their heart breaks on Lunar New Year’s Day when they comb their hair up in the Han style.

腸斷正朝梳漢髮。

In recent years, of those who yearn for the Han (i.e., Tang) like this,

近年如此思漢者，

Half are old and sick, and half are buried bones.

半爲老病半埋骨。

They teach their grandsons to speak in their hometown accent,

常教孫子學鄉(xiāng)音，

And still tell stories of the capital’s grandeur in the old days of peace.

猶話平時好城闕。

When the old ones are gone and the youths have become men,

老者儻盡少者壯，

They will be like rebellious Tibetans, having been born and raised in Tibetan lands.

生長蕃中似蕃悖。

If they do not know that their grandfathers and fathers were all subjects of the Han,

不知祖父皆漢民，

I fear with all my heart that they will become Tibetans！

便恐爲蕃心矻矻?、資ang, Yиап Zhеп Ji Вiаппiап Jiапzhи, p.132.

The line about combing hair in the Tang style on Lunar New Year’s Day is accompanied by a note that explains the source of this information:

Li Ruxian of the Yanzhou garrison is the son of the Fleabane General. He was once captured by the Tibetans, and said after he returned that the law of the Tibetans only permits Tang people who have been captured by the Tibetans to wear [Tang-style] robes and caps on Lunar New Year’s Day. On that day, Ruxian was filled with unbearable sorrow because of this and therefore secretly made a plan with his Tibetan wife to return [to the Tang].

延州鎮(zhèn)李如暹，蓬子將軍之子也。嘗沒西蕃，及歸云，蕃法唯正嵗一日許唐人沒蕃者服衣冠。如暹當此日，由是悲不自勝，遂與蕃妻密定歸計。②Ibid.

We know nothing else about Li Ruxian’s family or career. The curious title or nickname“Fleabane General” does not appear in any extant Tang text apart from the poems by Yuan Zhen and Bo Juyi. Yanzhou (modern Yan’an 延安) could be either his hometown or his current home; it was not very far from the Tang-Tibetan border after the An Lushan Rebellion, but far enough to never have been the target of any recorded Tibetan raid.①Chen Yinke hypothesized that the text originally read “Tingzhou” 庭州, the name of a Tang prefecture in Beshbaliq that fell to the Tibetans in 790. However, there is no strong reason to accept this hypothesis since neither Yuan Zhen nor Bo Juyi states that Li Ruxian served in the Central Asian garrisons. Chen Yinke 陳寅恪, Yиап Ваi Shi Jiапzhепg Gао 元白詩箋證稿 ( Beijing: Sanlian Shudian, 2001 [1955]),p.221.So Li Ruxian was presumably captured when serving elsewhere, but it is implausible that he escaped back to the Tang from the Tarim Basin, so the poem’s scenario can only be very loosely based on his experience.

Bo Juyi’s version of “The Western Barbarian Captive”attempts to match the fictional captive’s story more closely with Li Ruxian’s (even though their fates still differ dramatically), but does so partly by altering Li’s story. In Bo’s poem, the captive tells his tale to the Tibetan captives in the group (presumably in Tibetan) during their miserable journey to south China, claiming that their anguish is nothing compared to his:

He said that his hometown was in Liangzhou and Yuanzhou,

自云鄉(xiāng)管本涼原，

That he fell into the Tibetans’ hands during the Dali era (766–779).

大歷年中沒落蕃。

Forty years passed after he was taken into Tibetan lands,

一落蕃中四十載，

During which he was made to wear coats and belts of fur.

遣著皮裘繋毛帶。

Only on Lunar New Year’s Day was he allowed to wear the Han style,

唯許正朝服漢儀，

Straightening his robe and kerchief while weeping secretly.

斂衣整巾潛淚垂。

He vowed in his heart to plan secretly to return to his hometown,

誓心密定歸鄉(xiāng)計，

Not letting his wife and sons in Tibetan lands know about it.

不使蕃中妻子知。②Zhu Jincheng 朱金城 ed., Во Jиуi Ji Jiапjiао 白居易集箋校，Shanghai Guji Chubanshe; Shanghai 1988, p.198.

The last line of this passage has a note about Li Ruxian that is essentially the same as Yuan Zhen’s, but with two key omissions: It does not mention Li’s origin in Yanzhou, and it omits his Tibetan wife’s participation in his plan. In this way, Bo Juyi distorts Li Ruxian’s story to produce the bitterly ironic tale of a Chinese man leaving his Tibetan family behind to return to his people, only to be rejected as a Tibetan.His protagonist escapes alone, evading Tibetan patrols, and crosses the Gobi Desert (Dато大漠) and the frozen Yellow River to return to Liangzhou and Yuanzhou 原州 (modern Guyuan 固原 in Ningxia), but is taken for a Tibetan and seized by the first Tang troops he encounters — men too careless or too eager for rewards to notice his ability to speak Chinese. The poem ends with the captive lamenting his ironic fate, as well as the mismatch between identity and appearance that caused it: “My heart is Han, my speech is Han, but my body is Tibetan！” 漢心漢語吐蕃身?、賈hu Jincheng 朱金城 ed., Во Jиуi Ji Jiапjiао 白居易集箋校，Shanghai Guji Chubanshe; Shanghai 1988, p.198.

Like Yuan Zhen, Bo Juyi takes significant liberties with geography. If the captive had to cross both the Gobi Desert and the Yellow River, the starting point of his journey can only have been the Mongolian steppe, not the Tibetan Empire;the poem even has him hiding beside Wang Zhaojun’s 王昭君 tomb, the Green Mound 青塚, which was located on the steppe.②Ibid.“Liangzhou and Yuanzhou” is also a highly imprecise destination, meant only to indicate the region between the southern end of the Gansu Corridor and the present Tang-Tibetan border, which runs past Yuanzhou.

Some commentators have argued that the poem is chronologically inaccurate, since before Liangzhou and Yuanzhou fell to the Tibetans, not during, the Dali era — in 763–764, to be exact.③Ibid., p.199.But this criticism is based on an unfounded assumption that the captive was in his hometown when the Tibetans took it, rather than being captured elsewhere later on.A much bigger problem lies in the conventional assumption that he is now trying to escape to Tang territory. This assumption is just as unfounded, since the poem depicts him leaving his wife and children behind in “Tibetan lands” 蕃中 or “Western (Hи) lands” 胡地 to return to a hometown that was itself still under Tibetan rule as of 809. In other words, the identification of “Liangzhou and Yuanzhou” as the captive’s hometown and destination means that his “returning to the Han”歸漢 involves moving from one part of the Tibetan Empire to another, not — as in Li Ruxian’s case — returning to the Tang empire.④Ibid., p.198. The captive is clearly not referring generically to the entire Tang empire as his hometown, as in the poem he bemoans the fact that because of his capture by Tang troops, “I cannot see my hometown in Liangzhou and Yuanzhou/And have abandoned my wife and children in Western (Hи) lands for nothing” 涼原鄉(xiāng)井不得見/胡地妻兒虛棄捐.The Tang troops he encounters are therefore a raiding party (as in Yuan Zhen’s poem), not a border patrol. The sentiment that drives him to leave his family, brave sandstorms, and walk on thin ice is homesickness, not Tang loyalism. This homesickness is somehow bound to an ethno cultural identity that is stirred and awakened whenever he puts on Tang-style clothing for Lunar New Year’s Day.Yet the Chinese people of Liangzhou and Yuanzhou, whom he longs to live among again,may well be no freer to wear Tang-style clothing than he has been in his “Tibetan lands”home of the last forty years. The captive’s perception of “Liangzhou and Yuanzhou”as “Han lands” 漢土①Ibid.is frozen in time, locked in a nostalgic memory of life as it was before these prefectures came under Tibetan rule. I would argue that Bo Juyi intended both the irony of this misplaced nostalgia and the irony of mistaken identity to lie at the heart of the poem’s message.

Let us further consider the motivations behind the “Western Barbarian Captive” poems.A note added to the title of Yuan Zhen’s version states that Li Shen’s version, which inspired his, was written to satirize (fепg諷) a recent imperial decree that “whenever Tibetan prisoners are taken on the western frontier, they should typically be transported to the south and not killed”西邊每擒蕃囚，例皆傳置南方，不加剿戮.②Yang ed., Yиап Zhеп Ji Вiаппiап Jiапzhи, p.131.It is unclear what aspect of this ostensibly humane policy Li Shen wished to satirize, but it was very likely the lack of long-term strategic ambition that Yuan Zhen criticizes in his version’s last four lines:

We have been feeding a hundred thousand troops along the border,

緣邊飽喂十萬衆(zhòng)，

So why not unleash them now, all at once?

何不齊驅(qū)一時發(fā)？

Taking just two or three captives a year

年年但捉兩三人，

Is like Jingwei picking reeds to fill up the ocean！

精衛(wèi)銜蘆塞溟渤?、跿he reference to Jingwei alludes to a tale in the Shапhаi jiпg 山海經(jīng): Yandi’s 炎帝daughter Nüwa 女娃drowns in the Eastern Sea, and her spirit transforms into a bird called Jingwei that continuously drops tree branches into the Eastern Sea in an attempt at filling it up.Ibid., pp.132, 134.

This poem is thus a call for an irredentist campaign to reclaim the Tibetan-occupied prefectures before their inhabitants lose their Chinese identity altogether.

Bo Juyi’s version of “The Western Barbarian Captive” does not express an irrendentist message explicitly. Instead, it tries to milk the captive’s ironic life story for pathos — or,as Bo’s subtitle to the poem reads, “to express the feelings of a desperate subject” 達窮民之情也.①Zhu, Во Jиуi Ji Jiапjiао, p.197.As I explained above, however, part of that pathos lies in the fact that the captive’s intended destination, “Liangzhou and Yuanzhou”, ought to be “Han lands”in the Tang empire but no longer are. There is no doubt that Bo espoused irredentist sentiments as well, since his version of “Liangzhou Music” 西涼伎, subtitled “A satire on border generals” 刺封疆之臣也, and Yuan Zhen’s earlier version (which is not subtitled) both lament Liangzhou’s loss to the Tibetans and lambaste Tang generals on the western border for indulging in performances of the Central Asian lion dance known as “Liangzhou Music”, even as Liangzhou languishes under Tibetan rule and the Tibetans now loom within striking range of Chang’an.②Ibid., 210; Yang ed., Yиап Zhеп Ji Вiаппiап Jiапzhи, pp.114–115. The “Liangzhou Music” poems’ vivid descriptions of Liangzhou’s heavily “western” musical culture are probably the source of Waley’s characterization of that prefecture as “notoriously prone to exoticism”.Bo’s version also echoes the ending of Yuan’s “The Western Barbarian Captive”, in the words of a seventy-year-old veteran brought to tears of exasperation by yet another presentation of Liangzhou Music for his commanding general’s pleasure:

We aimlessly station a hundred thousand troops on the border,

緣邊空屯十萬卒，

Feeding and clothing them well but letting them pass their days in idleness.

飽食溫衣閑過日。

At Liangzhou there are lost subjects whose hearts are breaking,

遺民腸斷在涼州，

While our generals and soldiers look at one another without any intent of recovering

them.

將卒相看無意收。

……

Even if you lack the intelligence and ability to recover them,

縱無智力未能收，

How can you bear to entertain yourselves with Liangzhou Music?

忍取西涼弄爲戲？③Zhu, Во Jиуi Ji Jiапjiао, p.210. See also Chen, Yиап Ваi Shi Jiапzhепg Gао, p.218.An original NewYиеfиpoem by Bo Juyi, “Fortifying Yanzhou”城鹽州, delivers an even harsher rebuke at generals on the western border, accusing them of cynically allowing Tibetan raids in order to justify their salaries and large armies.①Ibid., pp.179–180.

It thus seems clear that Li Ruxian’s claim about “the law of the Tibetans” with regard to clothing proved compelling to Yuan Zhen and Bo Juyi (and probably Li Shen as well)because of the emotional force and urgency that it could lend to irredentist arguments.It confirmed the Chang’an elite’s hopes that forty years of Tibetan rule had not made Tibetans of the Chinese in the occupied prefectures — even as it raised fears that a policy of“Tibetanization” would eventually erase their Chinese identity as time went on and the older generations passed away. And it provided rhetorical ammunition against border generals who claimed to be doing their job by launching mere raids into Tibetan-occupied territory, for it showed that in that territory were Chinese people longing to be liberated, not raided and occasionally taken captive by accident.

The “Western Barbarian Captive” poems did not succeed in effecting a shift toward irredentism in Tang policy toward the Tibetan empire, but they did succeed in turning Li Ruxian’s story into a powerful and enduring image of former Tang subjects doggedly maintaining their Chineseness under Tibetan rule. That image resurfaced nearly forty years later in an irredentist-themed essay by Chen An 陳黯(ca. 805–871), a Quanzhou 泉州 literatus who failed thejiпshiexaminations about eighteen times between 845 and 864 before finally giving up and retiring into reclusion.②Our main source for biographical information on Chen Anisa preface to his collected works,written by his nephew Huang Tao 黃滔 (ca.840–911). This is supplemented by short biographical notes in the Хiп Tапgshи “Treatise on Literature”藝文志 and the Qиап Tапgwеп 全唐文.Хiп Tапgshи 60.1609; Qиап Tапgwеп 767.7983; Huang Tao 黃滔, Pиуапg Hиапg Yиshi Ji 莆陽黃御史集，Shangwu Yinshuguan,Shanghai 1936, pp.178–182. The version of Huang Tao’s preface at Qиап Tапgwеп 824, p.8684–8685 contains a few errors.Of the nine essays by Chen that have survived, most if not all would have been written for the literary portfolios (хiпgjиап行卷)that he presented to examiners and potential patrons in Chang’an during the examination season. The essay that concerns us here, titled “A Memorial on Behalf of the Elders of He-Huang”代河湟父老奏,was almost certainly composed in 846–848, when the Tang court was holding discussions on the possibility of taking advantage of the collapse of the Uyghur and Tibetan empires by retaking the He-Huang region (i.e., the Gansu Corridor, the Kokonor frontier, and the Longyou 隴右 region of southeast Gansu) from the Tibetans.③These discussions had already begun in 844: see Zizhi Tопgjiап 資治通鑒 247, p.7999. Tang military operations across the western border began in the winter of 848–849; Zizhi Tопgjiап 248, p.8036–8037.Chen An may not have been a fervent irredentist, but given the mood of the time, he certainly hoped to benefit from appealing to the irredentism of potential patrons at the imperial court.

“A Memorial on Behalf of the Elders of He-Huang” is written in the form of a message from the people of He-Huang to a newly-enthroned Tang emperor, this being Xuanzong 宣宗(r. 846–859), urging him to launch a military campaign to liberate them from Tibetan rule and recover “the former territories west of Qinzhou and Weizhou that our dynasty has lost” 國朝已來所沒秦渭之西故地. Chen An imagines the He-Huang people beginning their memorial with these words:

Your subjects and their ancestors were people of the Han (i.e., the Tang). Although our land was conquered by the western barbarian caitiffs, we have long harbored the desire to return [to Tang rule]. Back then, we could not find a way to do so, and therefore had no choice but to drag out an ignoble existence. But now that an age of peace and prosperity has begun, how can we remain silent?

臣等世籍漢民也，雖地沒戎虜，而常蓄歸心。時未可謀，則俛僶偷生，既遭休運，詎可緘默？

Further down, Chen uses the Lunar New Year’s Day motif as an illustration of the He-Huang population’s unwavering Tang loyalism:

Nonetheless, although our strength is inadequate, our hearts have not abandoned [the Tang].That is why we who live in the He-Huang region admonish one generation after another [to be loyal to the Tang] and to this day have handed down and preserved the caps and robes of the Han.Every Lunar New Year’s Day, when we make offerings to our ancestors, we always put them on to show that we have not forgotten the Han style…. Your Majesty has newly acceded to rule over the empire, transforming and nurturing it with benevolence and humaneness. When you hear of this, can you not be moved to sorrow and longing for us?

然雖力不支而心不離，故居河湟間，世相爲訓，今尚傳留漢之冠裳。每歲時祭享，則必服之，示不忘漢儀。…… 陛下新統(tǒng)寰區(qū)，以慈仁化育，聞之得不惻然而軫念乎！①Q(mào)иап Tапgwеп 767, p.7983–7984.

Although this passage does not explain why the He-Huang Chinese only put on Tangstyle clothing on Lunar New Year’s Day, it evidently assumes that the reader is familiar with the supposed “Tibetanization” policy from having read the poems by Yuan Zhen and Bo Juyi.Instead, Chen An fills in a detail that was only implicit in the poems: the symbolism of Lunar New Year’s Day resides in the customary ancestral sacrifice, which — like the wearing of Tang-style clothing— reminds participants of their Chinese ancestry and, presumably,thereby inoculates them from assimilation by the Tibetans during the rest of the year.

It seems to me that Chen An failed to recognize that even if Li Ruxian’s claim about the “l(fā)aw of the Tibetans” was true and not a self-justifying fiction to explain why he took so long to decide to escape (long enough to take a Tibetan wife), this was not necessarily a law or policy that applied to the entire He-Huang region. Zhao Lin’s 趙璘 (fl. 834–853)account of Tan Keze 譚可則, a Tang military officer who was a prisoner of the Tibetans from 820 to 826, does not mention any law on clothing — only that Tan received a tattoo on one arm to mark him as the Tibetan emperor’s subject.①Li Zhao 李肇、Zhao Lin 趙璘, Tапg Gиоshi Ви 唐國史補/Yiпhиа Lи 因話錄， Shanghai: Shanghai Guji Chubanshe,1957, 4, pp.96–97.Likewise, no Tang envoy who passed through Tibetan-occupied territory reported the existence of such a “Tibetanization” law.Liu Yuanding’s account of the elders of Longzhi Fort only quotes them as saying that their“children and grandchildren still cannot bear to forget the clothing of the Tang” — an indication of a persistent link between clothing and identity, no doubt, but not of any Tibetan policy aimed at severing that link. Liu does not elaborate on what the elders were wearing,nor does he comment on the clothing of the “Tang people” in Lanzhou, even though he would surely have found it worthy of noting if either group was dressed like Tibetans. Indeed, since the people of Lanzhou seem to have stared at him silently, one wonders how he would have recognized them as “Tang people” at all if they were dressed in Tibetan style.

Lü Wen 呂溫 (772–811), who was an envoy to the Tibetans in 804–805, comes somewhat closer to corroborating the existence of a “Tibetanization” policy in two irredentist poems inspired by the Chinese people whom he saw in Hezhou and Heyuan 河源 (modern Xining 西寧). The poem on Hezhou depicts local elders, who had experienced the Tibetan conquest,standing on a bridge across the Yellow River, “with hair untied, crying over the east-flowing waters” 被髮哭東流. This implies that they had at least adopted the Tibetan hairstyle and had not done so willingly. The poem on Heyuan contains the line, “Their Chinese customs,too, are still secretly preserved at the autumn and winter sacrifices” 伏臘華風亦暗存.②Qиап Tапgshi 全唐詩371, p.4165.Yu Shucheng and Zheng Chuanrui argued that this line is based on direct observation and validates the Lunar New Year’s Day motif in the “Western Barbarian Captive” poems by Yuan Zhen and Bo Juyi.①Yu Shucheng 余恕誠、Zheng Chuanrui 鄭傳銳, “Tangren chushi Tubo de shishi — lun Lü Wen shifan shi” 唐人出使吐蕃的詩史—論呂溫使蕃詩, Мiпzи Wепхие Yапjiи 民族文學研究2012(4), p.165.On the contrary, it is quite ambiguous which Chinese custom Lü Wen is referring to, and if he means the wearing of Tang-style clothing, this still contradicts the notion that the He-Huang Chinese only did so on Lunar New Year’s Day. It is certainly possible that the Chinese of Heyuan wore Tang-style clothing legally for the Lunar New Year’s Day sacrifice (although Lü Wen does not say so) and illegally for the autumn and winter sacrifices, but I find it very doubtful that a Tang envoy passing through could witness or hear about the latter practice while the Tibetan authorities knew nothing of it. If the Tibetans did know of it, then it was an open secret at best.

Even if Lü Wen’s poems count as evidence for the imposition of Tibetan-style dress in Hezhou and Heyuan, the same was not necessarily true of the Gansu Corridor, which Lü did not pass through during his mission. Likewise, since we have no evidence that Li Ruxian ever set foot on the Gansu Corridor during his time among the Tibetans, we cannot assume that his story reflects the situation there, let alone the situation in Dunhuang specifically. TheХiп Tапgshиeditors seem to have made that assumption when they associated Dunhuang in particular with the Lunar New Year’s Day motif — which essentially originated with Li Ruxian— and ignored the Kaicheng-era envoy’s report that all along the Gansu Corridor, people were still dressed in the Tang style after fifty or more years of Tibetan rule.②I do not think the editors of the Хiп Tапgshи can plead ignorance of the Kaicheng-era envoy’s account, since one of the editors was Ouyang Xiu 歐陽修(1007–1072), who had earlier copied the Jiи Wиdаi Shi summary of that account into his Хiп Wиdаi Shi.Moreover, theХiп Tапgshиintensifies the severity of the “l(fā)aw of the Tibetans” by implying that Tibetan-style clothing was mandatory throughout the year, including Lunar New Year’s Day — such that the Dunhuang Chinese had to keep their Tang-style clothing hidden away,only daring to wear it secretly during the Lunar New Year’s Day ancestral sacrifice.

I would argue that the most logical way to ascertain which source — theХiп Tапgshиor the Kaicheng-era envoy’s report — is closer to the truth is by looking to visual and textual evidence from Dunhuang itself. As we have seen, P.3451 contradicts both sources to some extent: unlike theХiп Tапgshи, it asserts unbroken continuity in the use of Tangstyle clothing in Tibetan-ruled Dunhuang; unlike the Kaicheng-era envoy, it claims that such continuity was unique to Dunhuang among all the Gansu Corridor prefectures. Unfortunately,the leading Chinese experts on Dunhuang art seem to have started their analyses of the visual evidence by reading both the Kaicheng-era envoy’s report and P.3451 as fully compatible with theХiп Tапgshиaccount. They have thus assumed that the people of Dunhuang were indeed ordered, against their will, to adopt Tibetan-style clothing.①E.g., Duan, Dипhиапg Shikи Yishи Lипji, p.261; Sha Wutian 沙武田, “Tubo tongzhi shiqi Dunhuang shiku gongyangren huaxiang kaocha” 吐蕃統(tǒng)治時期敦煌石窟供養(yǎng)人畫像考察, Zапgхие Yапjiи 藏學研究2003(2),88; idет., Tиbо Tопgzhi Shiqi Dипhиапg Shikи Yапjiи 吐蕃統(tǒng)治時期敦煌石窟研究，Beijing: Zhongguo Shehui Kexue Chubanshe, 2013, pp.62–63.In the next section of this paper, I will critique these experts’ analyses and present a new interpretation of clothing styles in Tibetan-ruled Dunhuang that does not begin with this unfounded assumption.

Visual evidence for clothing styles from Dunhuang cave murals

The Buddhist murals in Dunhuang’s Mogao caves include numerous images of local donors, both male and female. Efforts to interpret these images in terms of the “Tibetanization”policy claimed in theХiп Tапgshиseem to have begun in the 1980s. Twentieth-century China’s foremost authorities on Dunhuang art, Shi Weixiang and Duan Wenjie, recognized that both Chinese and Tibetan clothing styles can be seen in the murals dating from the period of Tibetan rule, but they came to very different conclusions about this.②Among the various items of clothing that art historians generally agree upon as constituting ninth-century Tibetan style, the most distinctive are a red, brimless cylindrical hat, a high collar, a capelet, and a leather belt with seven hanging straps. For images and analysis see Wang Jingyi 王婧怡, “Dunhuang Mogao ku bihua Tubo zanpu fushi kao — fanling yu yunjian, xue ji diexie dai” 敦煌莫高窟壁畫吐蕃贊普服飾考—翻領與云肩、靴及蹀躞帶, Zhеjiапg Fапgzhi Fиzhиапg Zhiуе Jishи Хиеуиап Хиеbао 浙江紡織服裝職業(yè)技術學院學報2009(4), pp.38–43.Shi argued that the Tibetans enforced the “Tibetanization” policy rigorously, requiring all Dunhuang Chinese to adopt Tibetan clothing and hairstyles. Therefore, all living persons depicted in the murals were dressed in Tibetan style, whereas their deceased parents could be depicted dressed in Tang style.③Shi Weixiang 史葦湘, Dипhиапg Lishi уи Моgао Kи Yishи Yапjiи 敦煌歷史與莫高窟藝術研究，Lanzhou: Gansu Jiaoyu Chubanshe,2002, p.76. This reprints an article that Shi wrote in 1979 and published in 1982.Duan, however, argued that only a minority of donor images from this period are dressed in Tibetan style, and inferred from this that except for a small number of Dunhuang officials who collaborated with the Tibetans, the people of Dunhuang mostly refused to adopt Tibetan clothing. Duan cited Liu Yuanding’s account of Longzhi Fort, the Kaichengera envoy’s report, and P.3451 as evidence of a defiant adherence to Tang-style clothing in Dunhuang.④Duan, Dипhиапg Shikи Yishи Lипji, p.261 (see also 207).But he also recognized that a servant in a mural in Cave 159, which depicts the Tibetan emperor and his entourage, has a Tibetan-style robe but a Chinese hairstyle.Duan explained this, rather paradoxically, as a hybrid style resulting from a long period of coexistence and mutual acculturation between the Chinese and Tibetans in Dunhuang, even claiming that the two ethnic groups became “l(fā)ike one family”.①Ibid., 262 (see also 305).

Ning Qiang’s 2004 study of Cave 220, which belonged to many generations of the Zhai 翟family, essentially follows Shi Weixiang when interpreting an image of three kneeling figures dating from the Tibetan period: a woman dressed in Tang style on the left, and two men dressed in Tibetan styles on the right. Since an inscription accompanying the image speaks of the donor’s deceased mother, Ning argues that the woman could be depicted in Tang style because she was deceased, whereas her two sons (the donors) were alive and had to be shown in Tibetan style.②Ning Qiang, Аrt, Rеligiоп, апd Pоl(xiāng)itiсs iп Меdiеvаl Сhiпа: Thе Dипhиапg Саvе оf thе Zhаi Fатilу, University of Hawaii Press, Honolulu 2004, pp.70–72. On Cave 220 see also Sha, “Tubo tongzhi shiqi Dunhuang shiku gongyangren huaxiang kaocha”,pp.90–91; idет., Tиbо Tопgshi Shiqi Dипhиапg Shikи Yапjiи, pp.67–68.But Ning’s interpretation (and by extension, Shi’s) ignores the fact that every woman depicted in Dunhuang murals securely dated to the Tibetan period is dressed in Tang-style clothing. Clearly,these cannot all have been representations of deceased persons.

In several articles published in 2003–2010, Sha Wutian presented a number of new arguments about the “Tibetanization” policy’s impact on Dunhuang murals.③Note that these articles, in revised form, later became chapters in a book by Sha titled Tиbо Tопgshi Shiqi Dипhиапg Shikи Yапjiи 吐蕃統(tǒng)治時期敦煌石窟研究(2013). For the reader’s convenience, I have cited both the original articles and the corresponding pages in the book.First, Sha argued that during the early period of Tibetan rule (786–800), the local elite reacted to the policy by ceasing to include donor images in their caves, thus avoiding having to depict themselves in Tibetan-style clothing. In the later period, local resistance led to a relaxation of the “Tibetanization” policy, which in turn caused a revival in the number of donor images.④Sha, “Tubo tongzhi shiqi Dunhuang shiku gongyangren huaxiang kaocha”，pp.88, 90; idет.,Tиbо Tопgshi Shiqi Dипhиапg Shikи Yапjiи,pp.63, 66.Duan Wenjie had already argued that the number of donor images decreased sharply in the early period of Tibetan rule and rose in the later period, although he did not link this to the supposed “Tibetanization” policy.⑤Duan, Dипhиапg Shikи Yishи Lипji, p.206.According to Sha Wutian’s count, about 15 of the 56 caves dated to the Tibetan period have donor images; of these, Sha — following a dating schema established in 2000 — identifies four caves (191, 197, 200 and 201) as products of the early period. Yet at least three of these four caves (caves 197, 200 and 201) have murals in which the donors are dressed entirely in Tang style.⑥Sha, “Tubo tongzhi shiqi Dunhuang shiku gongyangren huaxiang kaocha”, pp.84, 86–87; idет., Tиbо Tопgshi Shiqi Dипhиапg Shikи Yапjiи, pp.8, 57–58; Liu Ying 劉穎, “Mogao ku Tubo houqi dongku jianzao niandai panding — Yi bozhuang renwu wei biaozhun” 莫高窟吐蕃後期洞窟建造年代判定—以蕃裝人物爲準, Zhопghиа Wепhиа Lипt(yī)ап 中華文化論壇2013(5), p.69. I have not found any analysis of clothing styles in Cave 191, suggesting that the images are too worn for such analysis to be possible.Sha Wutian himself recognizes the predominance of Tang-style clothing in murals from this period, and interprets it as a sign that the Dunhuang Chinese were using the murals to express their Tang loyalism and anti-Tibetan sentiment, despite the risk involved.①Sha, “Tubo tongzhi shiqi Dunhuang shiku gongyangren huaxiang kaocha”, p.91; idет., Tиbо Tопgshi Shiqi Dипhиапg Shikи Yапjiи, p.69.But if the Tibetans could not impose their clothing style by force on even such a small number of murals during the first fifteen years of their rule, one must question whether they were trying to impose it at all. It is very likely that the reason for the decline in donor images during this period lies elsewhere.

Sha Wutian also argued that Dunhuang’s elite families eventually developed a way of flouting the “Tibetanization” policy by depicting deceased ancestors in Chinese clothing above the cave’s eastern entrance — an elevated position previously reserved for Buddha images. In this way, the ancestors did not have to be depicted in Tibetan dress, unlike the living donors whose images were typically arranged in rows along the base of the cave walls.②Sha, “Tubo tongzhi shiqi Dunhuang shiku gongyangren huaxiang kaocha”, p.89; idem., Tиbо Tопgshi Shiqi Dипhиапg Shikи Yапjiи,pp.79–80; Bai Tianyou 白天佑 and Sha Wutian 沙武田, “Mogao ku di 231 ku Yin Bolun fufu gongyang xiang jiexi” 莫高窟第231 窟陰伯倫夫婦供養(yǎng)像解析, Dипhиапg Yапjiи 敦煌研究 2006(2), pp.8–10.Sha developed this theory through an analysis of Cave 231, which contains the most extensively studied example of such ancestor portraits.The cave was commissioned in 839 by Yin Jiazheng 陰嘉政 and his brothers, whose father Yin Bolun 陰伯倫 and mother Lady Suo 索氏 are depicted in Tang-style dress above the entrance. We know the donors’identities from an inscription in highly ornate parallel prose, titled “The Merit Record of the Retired Gentleman Master Yin at the Mogao Caves in the Former Dunhuang Commandery of Great Tibet” 大番故敦煌郡莫高窟陰處士公修功德記, that was carved on a stele outside the cave. The stele is no longer extant, but the inscription text is preserved in two Dunhuang manuscripts, P.4638 and P.4640.③Bai Tianyou and Sha Wutian misread the gu 故 in “Former Dunhuang Commandery” as indicating that Yin Jiazheng was deceased at the time of the cave’s opening. It actually indicates an administrative division of the Tang empire that had been abolished under Tibetan rule.Bai and Sha, “Mogao ku di 231 ku Yin Bolun fufu gongyang xiang jiexi”, p.9.Although Sha Wutian only uses the text cursorily, a close examination of it shows that an interpretation of Yin Jiazheng’s family as Tang loyalists would be quite unpersuasive.

The inscription mentions Yin Bolun’s service to the Tang as a military commander at Dunhuang, then describes his response to the Tibetan conquest in the following terms:

His beloved sons, brave as bears, were stripped of their swaddling cloths and tattooed;husband and wife, loving as mandarin ducks, removed their golden hairpins and braided their hair. How could he have planned to shift his gratitude from that of former days and bid a long farewell to his ruler of ten thousand generations? Matters having come to that year,he bent his knee to the lord of a second dynasty. From the time that the Bstan-po (i.e., the Tibetan emperor) entered the pass, he dressed with his lapel folded to the left and received promotions; beginning with his being conferred a seal by the chief minister, he let his sleeves hang down and was appointed to official posts. The Tibetan dynasty reappointed him as Great Commissioner for the Community of Daoist Clergy and their Families in Former Shazhou. Inheriting a foundation, he was overjoyed; his generation came into complete security. His kinsfolk received an abundance of relief grain; his family was exempted from the tithing tax. Relieved of their former land tax, their joy was such that they forgot they had been conquered; profiting from their newly added gardens and ponds, the years flew by.

熊羆愛子，拆襁褓以文身；鴛鴦夫妻，解鬟鈿而辮髮。豈圖恩移舊日，長辭萬代之君？事遇此年，屈膝兩朝之主。自贊普啓關之後，左衽遷階；及宰輔給印之初，垂袪補職。蕃朝改授，得前沙州道門親表部落大使。承基振豫，代及全安。六親當五秉之饒，一家蠲十一之稅。復舊來之井賦，樂已忘亡；利新益之園池，光流竟嵗。①Zheng Binglin 鄭炳林, Dипhиапg Веi Мiпg Zап Jishi 敦煌碑銘贊輯釋, Gansu Jiaoyu Chubanshe, Lanzhou 1992, p.239.

The first line of the passage is frequently cited as evidence for the forced “Tibetanization”of clothing in Dunhuang.Shi Weixiang also used it as evidence that the Tibetans made the Dunhuang Chinese tattoo their faces and bodies; Sha Wutian accepts this reading.②Shi, Dипhиапg Lishi уи Моgао Kи Yishи Yапjiи, p.76; Sha, “Tubo tongzhi shiqi Dunhuang shiku gongyangren huaxiang kaocha”,p.87.Some caution is in order, however, the Tibetans are not known to have tattooed themselves, unlike the ancient Wu and Yue peoples of south China, so the image of babies getting tattooed is most likely just a hyperbolic allusion to a stock image of foreign fashions. Ning Qiang readswепshеп文身as “printing a [slave] mark on their bodies”, presumably meaning branding, but there is no basis for this apart from Zhao Lin’s report (based on testimony from Tan Keze)that the Tibetans branded the faces of illiterate Chinese prisoners of war with a tattoo and used them as slaves, while tattooing Tibetan words meaning “the emperor’s servant” on the arms of those who were literate and designated for employment as scribes. Since the Tibetans did not turn the majority of their conquered Chinese subjects into slaves and certainly did not enslave Yin Bolun’s sons, these tattooing practices are irrelevant to interpreting the Yin family’s stele inscription.③Ning, Аrt, Rеligiоп, апd Pоl(xiāng)itiсs iп Меdiеvаl Сhiпа, p.71; Li and Zhao, Tапg Gиоshi Ви /Yiпhиа Lи, 4.96. For a refutation of the notion(once standard in Chinese Marxist historiography) that the Tibetans enslaved all their Chinese subjects, see Yang, Tиbо Tопgzhi Dипhиапg уи Tиbо Wепshи Yапjiи, pp.79–84.The idea of Yin Bolun’s wife untying her chignon and braiding her hair into pigtails is also contradicted by the absence of any depiction of women with this hairstyle in the Dunhuang murals.①According to Хiп Tапgshи 216a.6072, Tibetan women typically braided their hair and coiled it into a bun.Nor does the passage claim that these actions were taken under orders from the Tibetan authorities.

If Yin Bolun himself did adopt Tibetan dress, the line “he dressed with his lapel folded to the left and received promotions” suggests, tactfully, that he did so voluntarily to ingratiate himself with his new rulers. Moreover, we have no reason to assume that he and his family resented this to the point of being determined not to depict him wearing the same garb in a posthumous portrait.②Sha argues that the donor images in Cave 231, which are all in Chinese dress, were repainted during the Guiyi Army period and were originally divided into men in Tibetan dress and women in Chinese dress. However, this is a purely speculative inference based on the notion of a “Tibetanization” policy. Bai and Sha, “Mogao ku di 231 ku Yin Bolun fufu gongyang xiang jiexi”, pp.8–9.When read in its proper context, this passage in the inscription instead emphasizes and justifies (in a somewhat defensive tone) the many rewards and benefits that Yin Bolun and his family enjoyed in return for accommodating themselves readily to Tibetan rule and “[forgetting] they had been conquered”. As Zhang Yong’an recently argued,the Yin family did very well serving the Tibetans, only to reinvent themselves as Tang loyalists after 848.③Note that Zhang misreads the above-quoted passage on Yin Bolun as referring to Yin Jiazheng and his brothers. Zhang effectively echoes Shi Weixiang’s assessment of the Yin family, which is heavily influenced by a Maoist anti-elite bias but nonetheless largely correct. Zhang Yong’an 張永安, “Dunhuang Yinshi diwei yanjiu” 敦煌陰氏地位研究, Dипhиапg Yапjiи 敦煌研究2007(2), p.62; Shi, Dипhиапg Lishi уи Моgао Kи Yishи Yапjiи, pp.76–77.In this, they were very similar to Zhang Yichao’s family. In fact, the inscription commissioned by Zhang Huaishen in 874–885 is often also used to substantiate the “Tibetanization” policy, as it claims (directly contradicting P.3451) that after receiving posts from the Tibetans, Zhang Qianyi “adhered in physical form to the braiding of hair and beautified his body with fur coats; folded his lapel to the left and submitted, with hanging arms and bended knees” 形遵辮髮，體美織皮，左衽束身，垂肱跪膝.④Rong, Gиiуijип Shi Yапjiи, p.400. Ning Qiang misidentifies this inscription as the Zhai Family Stele and misreads “fur coats” (zhiрi)as “‘beauty mark’ (i.e., the slave mark)”, again reflecting his assumption that Dunhuang’s people were reduced to slavery. Ning, Аrt,Rеligiоп, апd Pоl(xiāng)itiсs iп Меdiеvаl Сhiпа, pp.71–72.The inscription goes on to claim that the Zhang family had no choice but to hide its resentment toward the Tibetans, but the circumstances of its composition, at a time when Zhang Huaishen was desperately trying to convince the Tang court of his loyalty, should incline us to skepticism.In any case, it is only possible to read these inscriptions commissioned by Zhang Huaishen and Yin Jiazheng as evidence of a “Tibetanization” policy regarding clothing if we assume theХiп Tапgshиaccount of such a policy to be true. If we do not, all we have here are two cases of collaborators choosing to adopt Tibetan dress as a gesture of loyalty to the Tibetan empire.

Let us return to the ancestor portraits of Yin Bolun and his wife in Cave 231. Sha Wutian argues that the Yin family was highly unusual in being powerful enough to defy the“Tibetanization” policy in this way, and that the practice of placing ancestor portraits above the cave entrance did not become popular until the Guiyi Army period.①Sha, “Tubo tongzhi shiqi Dunhuang shiku gongyangren huaxiang kaocha”, p.89; Bai and Sha, “Mogao ku di 231 ku Yin Bolun fufu gongyang xiang jiexi”, p.10.If the primary purpose of this placement was to avoid depicting ancestors in Tibetan dress, however, it seems strange that it would be more widely imitated onlyаftеrTibetan rule came to an end.Moreover, one of the two other caves with ancestor portraits of this kind that can be securely dated to the Tibetan period actually depicts the male ancestor in Tibetan-style dress and the female ancestor in Tang-style dress, according to Luo Shiping’s recent reading. This is Cave 144, commissioned by the Suo 索 family, Yin Jiazheng’s maternal kinsmen.②Luo Shiping 羅世平，“Shenfen rentong: Dunhuang Tubo zhuang renwu jinru dongku de tiaojian, celue yu shijian”身份認同：敦煌吐蕃裝人物進入洞窟的條件、策略與時間, Меishи Yапjiи 美術研究 2011(4), pp.62, 64. On Cave 144 see also Winston Kyan, “Family Space: Buddhist Materiality and Ancestral Fashioning in Mogao Cave 231”, Thе Аrt Виllеtiп 92.1/2 (2010), p.73. Kyan incorrectly dates the cave to the late ninth century; although the inscriptions in the image are now too damaged to read, transcripts have been preserved and show that they were produced under Tibetan rule.It would seem,then, that the appearance of ancestral portraits above cave entrances cannot be explained as a reaction to a “Tibetanization” policy. Winston Kyan has argued, more persuasively, that this was an innovation aimed at combining filial piety and ancestor worship with Buddhist faith in the same site.③Kyan, “Family Space”, pp.61–82. See also a paper by Madeleine Boucher that builds on Kyan’s argument: “Ancestors, Politicians,and Patrons: Portraits of the Dead in Ninth and Tenth Century Dunhuang Mogao Caves”, online at https://lucian.uchicago.edu/blogs/colloquium/2013/09/05/ancestors-politicians-and-patrons-portraits-of-the-dead-in-ninth-and-tenth-century-dunhuang-mogao-caves/#_edn14 (accessed July 18, 2016).That the innovation began under Tibetan rule is incidental, not crucial to its origins.

In 2010, Sha Wutian published an analysis of Cave 359, the third Tibetan-period cave with ancestor portraits above the entrance. In this cave, which is currently dated to the 840s,the ancestors are depicted in Tang-style dress, whereas the donor images are dressed in a mix of styles, with the men mostly in Tibetan style and the women all in Tang style. The men and women are in separate rows, each preceded by seven Buddhist monks. Duan Wenjie was unsure whether the male donors are Tibetan or Chinese. Sha, after initially identifying them as Chinese, later argued from the fragmentary inscriptions accompanying the images, as well as the male ancestor’s facial features, that they were neither Tibetan nor Chinese but rather Sogdians with the Chinese surname Shi 石, which typically denoted an origin from Tashkent.①Duan, Dипhиапg shikи уishи lипji, 207; Sha, “Tubo tongzhi shiqi Dunhuang shiku gongyangren huaxiang kaocha”, 89; idem., “Mogao ku Tubo qi dongku di 359 ku gongyangren huaxiang yanjiu — jiantan Sute jiuxing huren dui Tubo tongzhi Dunhuang de taidu” 莫高窟吐蕃期洞窟第359 窟供養(yǎng)人畫像研究—兼談粟特九姓胡人對吐蕃統(tǒng)治敦煌的態(tài)度, Dипhиапg Yапjiи 敦煌研究2010(5), 16–17;idет., Tиbо Tопgshi Shiqi Dипhиапg Shikи Yапjiи, pp.249–265.Sha argues that the striking absence of Tibetan dress in female donor images from this and all other Tibetan-period murals implies that Tibetan authorities did not enforce their “Tibetanization” policy as strictly toward women, or possibly did not apply it to women at all.②Sha, “Tubo tongzhi shiqi Dunhuang shiku gongyangren huaxiang kaocha”, p.89; idem., “Mogao ku Tubo qi dongku di 359 ku gongyangren huaxiang yanjiu”, p.23; idет., Tиbо Tопgshi Shiqi Dипhиапg Shikи Yапjiи, pp.68–69, 280.

In the specific case of Cave 359, Sha further argues that this Sogdian family’s choice of Tang-style women’s clothing reflects both its “complete sinicization” (wапqиап Hапhиа完全漢化) and its “sentiment of resistance” toward Tibetan rule. This interpretation clearly does not cohere well with the cave’s depiction of only two of its male donors (out of a total of at least 31) in unambiguously Tang-style dress, as well as its inclusion of the Tibetan emperor’s image in a mural of theViтаlаkīrti Sиtrа. Sha argues that the Shi, being Central Asians, were less attached to “traditional Han culture” than the Dunhuang Chinese and could “grudgingly accept” Tibetan dress because the Tibetan style itself bore Central Asian influences.③Sha argues that one of the two donor figures that he identifies as dressed in the Tang style has been repainted and may originally have been dressed in Tibetan style as well. He suggests that the repainting represents a repudiation of Tibetan rule after 848, but if so, why were the other figures in Tibetan dress not repainted? Sha, “Mogao ku Tubo qi dongku di 359 ku gongyangren huaxiang yanjiu”, pp.17–18, 21–23; idет., Tиbо Tопgshi Shiqi Dипhиапg Shikи Yапjiи, pp.275–276, 278–280.Obviously, this interpretation undermines Sha’s earlier point about the Shi family’s “complete sinicization”.

Sha identifies the first figure in the row of male donors, who is dressed in Tang style,as the head of the Shi family; suggests that he was able to be depicted in Tang-style dress as a “special case” because of his past service to the Tang empire; and credits him with using a “flexible” and “ingenious” method to evade the “Tibetanization” policy. But supposing a“Tibetanization” policy did exist and the Tibetans did allow such exceptions for former Tang officials in Dunhuang, surely the credit for flexibility should go to them, and such flexibility would render Sha’s argument regarding the origins of the ancestor portraits in caves 231 and 359 completely redundant. As for the choice to depict the Tibetan emperor in a mural, Sha reasons that the Shi family had no choice but to do so because earlier caves withViтаlаkīrti Sиtrаmurals (including Cave 231) had already set this precedent, but this only begs the question of why the owners of earlier caves chose to do it if they were so opposed to Tibetan rule. He also suggests that depicting the Tibetan emperor was a superficial gesture of loyalty for the purpose of “covering up” the depiction of Tang-style dress in donor figures, but the notion that such a tactic would have worked under the circumstances of a strict “Tibetanization”policy strikes me as grossly underestimating the Tibetan authorities’ intelligence.①Sha, “Mogao ku Tubo qi dongku di 359 ku gongyangren huaxiang yanjiu”, pp.22–23; idет., Tиbо Tопgshi Shiqi Dипhиапg Shikи Yапjiи, pp.278–279.

In reality, not all of the male donor figures that Sha Wutian identifies as dressed in Tibetan style can be neatly classified as such. Sha himself recognizes that whereas the men dressed in white are in “typical” Tibetan style, several others are in red or grey robes that closely resemble Tang-style male costume except for the Tibetan-style high collar. Sha interprets these robes as “Tang-influenced Tibetan dress” and traces it to Chinese cultural influence on the Tibetans, beginning with Princess Wencheng’s 文成公主 marriage to the Tibetan emperor in 641. He also argues that male members of the Shi family who were more opposed to the “Tibetanization” policy adopted this kind of “Tang-influenced Tibetan dress”as a compromise solution.②Sha, “Mogao ku Tubo qi dongku di 359 ku gongyangren huaxiang yanjiu”, p.18, 23; idет., Tиbо Tопgshi Shiqi Dипhиапg Shikи Yапjiи,pp.267–268, 279, 280.But I think it would be more plausible to interpret these hybrid robes as Tang-style dress with some Tibetan influence, and to conclude from them that the Shi family men were quite receptive to Tibetan fashions, albeit to different degrees — some adopting Tibetan dress wholesale, others blending it with the Tang style. Indeed, Cave 359 supplies intriguing evidence that such sartorial hybridization — which no extant textual source even hints at — also occurred among the Shi women, albeit to a smaller degree.Sha Wutian informs us that one of the female donors wears a distinctly Tibetan hairstyle along with Tang-style clothing— the exact opposite of the hybrid style worn by the servant whose image Duan Wenjie noticed in Cave 159.③Sha, “Mogao ku Tubo qi dongku di 359 ku gongyangren huaxiang yanjiu”, p.18; idет., Tиbо Tопgshi Shiqi Dипhиапg Shikи Yапjiи,pp.268–269.

Liu Ying has recently identified similar Sino-Tibetan hybrid styles in a mural of serving girls in Mogao Cave 147 and a wedding scene mural in Yulin 榆林 Cave 25 (near Guazhou).④Liu, “Mogao ku Tubo houqi dongku jianzao niandai panding”, pp.69–71. Liu argues that the serving girls have Tibetan-style high collars,and that a hybrid style can be seen in a male figure in the wedding scene who is conventionally identified as the father of the bride. On the wedding scene mural, which depicts the groom in Tibetan dress and the bride in Tang-style dress, see most recently Ma Junfeng 馬俊鋒and Sha Wutian 沙武田, “Tang-bo Qingshui huimeng zai Dunhuang shiku zhong de lishi yiji — Guazhou Yulin ku di 25 ku gongdezhu xinjie” 唐蕃清水會盟在敦煌石窟中的歷史遺跡—瓜州榆林窟第25 窟功德主新解, Хizапg Yапjiи 西藏研究2015(3), pp.15–22;idет., “Tang-bo Qingshui huimeng zai Dunhuang shiku zhong de tuxiang yicun—Guazhou Yulin ku di 25 ku hunjia tu huizhi niandai zaitan” 唐蕃清水會盟在敦煌石窟中的圖像遺存—瓜州榆林窟第25 窟婚嫁圖繪製年代再探, Shihеzi Dахие Хиеbао 石河子大學學報29.5 (2015), pp.31–38.These articles supersede Sha Wutian’s earlier analysis of Cave 25 in Sha, Tиbо Tопgshi Shiqi Dипhиапg Shikи Yапjiи, pp.355–382. Ma and Sha now date Cave 25 to 783, revising the date range of 776–781 proposed by Duan Wenjie and Shi Weixiang and the date range of 776–786 previously proposed by Sha Wutian. Other scholars favored a date of 823 or later.Like Sha Wutian and earlier scholars, Liu assumes the existence of a “Tibetanization”policy, but he argues (modifying Duan Wenjie’s position) that it was abandoned very quickly in Dunhuang due to strong resistance from the Chinese population. Therefore,the Dunhuang Chinese continued to dress freely in the Tang style. Liu also argues that the Dunhuang Chinese, being fiercely hostile toward their Tibetan conquerors, could not possibly have depicted themselves in Tibetan dress or depicted the Tibetan emperor in their murals before two things happened: first, the Tibetans’ adoption of more conciliatory policies toward the Dunhuang Chinese; second, the Tang-Tibetan peace covenant of 821. After 821, according to Liu, the Dunhuang Chinese began to depict male donors in Tibetan dress, the Tibetan emperor, and even hybrid clothing styles in their murals as symbolic gestures of friendship toward the Tibetans. On that basis alone, Liu argues that all caves containing such images should be dated to the period between 821 and 848. Luo Shiping has made essentially the same argument as Liu’s, with the additional point that the Dunhuang Chinese continued to depict women in Tang-style clothing in order to symbolize the union of the Tibetan and Chinese peoples into one family.①Liu, “Mogao ku Tubo houqi dongku jianzao niandai panding”; Luo, “Shenfen rentong”. Luo argues that the Tibetan emperor won the Dunhuang Chinese over by patronizing Buddhism, but also identifies the 821 covenant as a turning point in Dunhuang Chinese attitudes toward the Tibetan empire. Note also that Luo is more ambiguous than Liu as to how long the “Tibetanization” policy was in effect. He merely follows Sha in arguing that the Dunhuang Chinese reacted to the policy by ceasing to produce donor images for a time.

These interpretations by Liu and Luo suffer from four weaknesses:

(1)Despite challenging Sha Wutian’s arguments, they follow past scholarship (including Sha’s) in taking a “Tibetanization” policy’s existence as a given, despite there being no strong supporting evidence for such a policy in the visual evidence from Dunhuang.

(2)Their emphasis on visual symbolism skirts the question of what exactly the Chinese men of Dunhuang were wearing between 821 and 848. Did they continue wearing only Tangstyle dress but have themselves painted wearing Tibetan or semi-Tibetan dress, or did they begin wearing Tibetan and hybrid styles too?

(3)Liu and Luo greatly overestimate the impact that the peace covenant of 821 would have had on attitudes toward Tibetan rule in Dunhuang. If the Dunhuang’s Chinese had a burning desire to return to Tang rule (and one should not assume they did), surely a peace agreement recognizing the current borders would not have erased that desire instantaneously. Indeed, both Liu Yuanding’s account and the Kaicheng-era envoy’s report appear to show that at least some Chinese people under Tibetan rule had a deeply emotional yearning for the Tang empire that could hardly be assuaged by a diplomatic agreement signifying abandonment by the Tang.

(4)They assume that before 821, the Dunhuang Chinese were uniformly hostile to Tibetan rule and loyal to the Tang, ignoring the evidence indicating that elite families like the Yin, the Suo, and the Zhang began serving the Tibetan empire not long after 786. The image of Dunhuang as a hotbed of anti-Tibetan sentiment is partly based on Dunhuang manuscripts containing information about a violent uprising that killed several Tibetan officials in 795, but it is noteworthy that the rebels were a very small group of couriers (уihи驛戶), not members of the local elite, and that a Prefect from the Suo family was responsible for their suppression.①See most recently Lu Li 陸離, “Dunhuang xieben S.1438 bei ‘Shuyi’ canjuan yu Tubo zhanling Shazhou de jige wenti”敦煌寫本S.1438背《書儀》殘卷與吐蕃占領沙州的幾個問題, Zhопggио Shi Yапjiи 中國史研究 2010(1), pp.87–100.

If one accords greater weight to the visual evidence than to theХiп Tапgshи, the most plausible interpretation is that there was no forced Tibetanization of clothing in Dunhuang at any time during the period of Tibetan rule. Instead, the people of Dunhuang were free to dress as they wished, but many members of the Chinese (and quite likely Sogdian) elite families adopted Tibetan dress or elements of Tibetan fashion voluntarily as symbols of prestige and loyalty to the Tibetan empire — a practice acknowledged subtly in the Yin Jiazheng and Zhang Huaishen inscriptions but elided in the P.3451biапwеп. The Tang-Tibetan peace covenant of 821 would have had no significant effect on the frequency of this practice, making it useless as a basis for dating cave murals. Unfortunately, the murals give us little indication of how the common people of Dunhuang were dressed. If the Kaichengera envoy’s report is reliable on this, it suggests that they had less motivation to dress in Tibetan style, so their clothing still “had not changed” fifteen years or more after the 821 peace covenant, with the likely exception of servants in elite households.

The question of why Dunhuang murals do not depict elite women in Tibetan dress —the serving girls in Cave 147 most likely being non-elite—is difficult to answer with the evidence available. The simplest explanation is that they generally did not wear Tibetanstyle clothes, but even texts from Dunhuang do not mention this. A recent study by Kate Lingley suggests, however, that this was not the first time male dress in Dunhuang changed to a foreign style while female dress did not. Using as a key example a mural in Mogao Cave 285, dated to 538 and in part patronized by the Yin family, Lingley argues that after ca. 500,the men in Dunhuang dressed in tunics, trousers, and boots—a style derived from the originally nomadic Xianbei or Xianbi 鮮卑who founded the Northern Wei (398–534) — whereas the women still dressed in wide-sleeved robes derived from Han-period prototypes. The difference was not confined to Dunhuang, as it became typical of Tang-period art, in which (Lingley argues)Xianbei-style male clothing lost its ethnic association with the Xianbei and became merely standard male dress.①Kate A. Lingley, “Naturalizing the Exotic: On the Changing Meanings of Ethnic Dress in Medieval China”， Аrs Oriепt(yī)аlis 38 (2010),pp.51–80.This, in fact, is the style of male clothing that art historians identify as“Chinese,” “Han,” or “Tang” when analyzing Dunhuang murals, and is presumably also the style that Tang texts describe as that of “Tang” or “Han” and theХiп Tапgshиdescribes as“clothing of the Central Lands”.

Although it remains unclear why the equivalent Xianbei female style of a tunic over a skirt did not become the norm for Chinese women in the Tang, Lingley hypothesizes that nomadic-style dress — like nomadic culture in general — had become associated with masculinity, making it less appealing to women than the Central Asian “Western” (Hи胡)styles that are known to have been highly popular in Chang’an during the eighth and early ninth centuries.②Ibid., pp.73–74.Since no extant Tang source represents Tibetan men as hyper-masculine, I find it unlikely that in Tibetan-ruled Dunhuang, Tibetan dress came to be seen as even more masculine than Tang-style (i.e., Xianbei-style) male clothing. However, it may well have had a strong association with political and social power that Chinese elite women — or more likely, their husbands — did not feel they needed.

Also worth considering is the strong possibility that there were very few Tibetan elite women in Dunhuang to provide examples of elite Tibetan-style female dress. As Xu Xiaoli and Zheng Binglin have shown, there is no trace of Tibetan women in the visual evidence from Dunhuang, and conclusive textual evidence is limited to the wives of Tibetan governors.③Xu Xiaoli 徐曉麗 and Zheng Binglin 鄭炳林, “Wantang Wudai Dunhuang Tuyuhun yu Tubo yimin funü yanjiu” 晚唐五代敦煌吐谷渾與吐蕃移民婦女研究, Dипhиапg Хие Jikап 敦煌學輯刊2002(2), pp.3–8.This suggests that Tibetan women had a very limited presence in ninth-century Dunhuang society. If, as I have argued, the Tibetans allowed the Dunhuang elite to choose how they dressed, the lack of Tibetan styles of elite women’s fashion to imitate would have made it very natural for Dunhuang Chinese women to continue dressing as before.Moreover, if under normal circumstances the only elite Tibetan woman in the prefecture was the governor’s wife, then Chinese women imitating her clothing could risk being seen as competing with her for prestige.④Gender differences in the use of fashion may also be relevant here: in modern contexts, men frequently dress in identical clothes to express group solidarity and identity, whereas two women who unintentionally wear the same outfit to a social event are committing a fashion faux pas by failing to express their individuality and setting themselves up for comparisons. Unfortunately, we do not know if such norms existed among Tang and Tibetan elites in the ninth century.

The ethnogenesis of the Wenmo and the limits of Chineseness

I have argued that elite clothing in Dunhuang during the period of Tibetan rule was probably characterized by levels of diversity, flexibility, and hybridity that belie the simple narrative of forced “Tibetanization” found in some of our textual sources. Instead of being a result of coercion, the wearing of Tibetan-style male clothing in Dunhuang was a voluntarily adopted symbol of collaboration with the Tibetans and the power and prestige that came from that collaboration. But what should we make of the emotional expressions of Tang loyalism that Liu Yuanding and the Kaicheng-era envoy claimed to have witnessed among non-elite Chinese at Longzhi Fort and all along the Gansu Corridor? I do not think we have any strong reason to doubt these accounts, but they clearly do not tell the full story, because subsequent events suggest that not all the Chinese in these lost regions had maintained their attachment to the Tang empire throughout the period of Tibetan rule.

In the autumn of 849, the Tang court celebrated its peaceful recovery of three Tibetanoccupied prefectures — Yuanzhou, Qinzhou 秦州 (modern Tianshui 天水), and Anlezhou 安樂州(modern Zhongning 中寧 county, Ningxia) — by having over a thousand local inhabitants brought to the imperial palace. As the emperor watched from a gate tower, his new subjects cheered, danced, and vied with one another to be the first to unbraid their hair and change into new Tang-style caps and robes. Clearly, this self-congratulatory display was choreographed by court officials acting on the assumption that the people of the recovered prefectures had hitherto been forced to dress in Tibetan-style clothes, but it also reflects how symbolically potent the choice between Tang and Tibetan clothing styles had become as an expression of allegiance and identity. The emperor issued an edict recruiting peasants to migrate to the reconquered prefectures in return for five years of tax exemption, and ordering these peasants to be supplemented by exiled convicts from the Chang’an metropolitan region.①Zizhi Tопgjiап 248, p.8037–8040; Хiп Tапgshи 216b, p.6107. The Xin Tangshu identifies the thousand or more people of Yuanzhou, Qinzhou,and Anlezhou as “elderly folk” 高年, whereas the Zizhi Tопgjiап claims that they included both “old and young” 老幼. On the recovery and administration of these three prefectures, see also Li Jun 李 軍, “Wantang zhengfu dui He-Long diqu de shoufu yu jingying—yi Xuan、Yi erchao wei zhongxin” 晚唐政府對河隴地區(qū)的收復與經(jīng)營—以宣、懿二朝爲中心, Zhопggио Shi Yапjiи 中國史研究2012(3), pp.115–121.

This project to push the frontier back westward to its original boundaries ultimately failed, as there was always a geographical area stretching between Qinzhou and Ganzhou that neither the Tang court nor the Guiyi Army was able to control effectively, even after the court deployed a small army to Liangzhou around 862–863 (more on which below). Words attributed to a local Tibetan chief in theZizhi Tопgjiапsuggest that the eastern end of this no-man’s-land, around Hezhou and Weizhou, was depopulated by the late 850s: the “Tang people” had left to settle around Yuanzhou and Qinzhou, while most of the Tibetans had “fled far to the west of Diezhou and Dangzhou”遠遁疊、宕之西 — that is, across the Yellow River and into the foothills of the Amne Machin Mountains. The Tibetan chief, Zhang Yan Sum (Chinese: Shang Yanxin 尚延心), blames this out-migration on famine and epidemics,but the constant warfare between rival Tibetan warlords in the Kokonor region during the 840s and early 850s must have had much to do with it as well.①Zizhi Tопgjiап 249.8064. On the conflict between the Tibetan warlords, see Zizhi Tопgjiап 246.7970, 247.7986–7987, 7992, 8000,248.8021, 8037, 249.8043–8044, 8047; Хiп Tапgshи 216b.6105–6107.

Nonetheless, there were still people living in this no-man’s-land, otherwise the Tang court could conceivably have reclaimed it using the same resettlement policies it had implemented in Yuanzhou, Qinzhou, and Anlezhou.②Li Jun claims that Zhang Yan Sum held Hezhou and Weizhou for the Tang throughout the 860s, but this ignores the sources stating that he and his followers were resettled in Qinzhou after submitting to the Tang. Li, “Wantang zhengfu dui He-Long diqu de shoufu yi jingying”,pp.124–127.The real problem was that the area’s remaining inhabitants did not see themselves as Chinese and had no desire to be under Tang rule. In a poem composed in the late ninth century, Sikong Tu 司空圖 (837–908) implied that there were people who had lost their Chinese identity under Tibetan rule:

Ever since the dust of battle rose up beyond the Xiao Pass,

一從蕭關起戰(zhàn)塵，

He-Huang has, for many springs, been cut off as a foreign land.

河湟隔斷異鄉(xiāng)春。

Han lads, all using the language ofHиlads,

漢兒盡作胡兒語，Now shout curses at we people of Han on our city walls！

卻向城頭罵漢人?、跶иап Tапgshi 633, p.7261. The Xiao Pass is one of several strategic passes around Yuanzhou that were recovered by the Tang in 849:Zizhi Tопgjiап 248, p.8037–8039.

The “l(fā)anguage ofHиlads” to which Sikong refers is presumably Tibetan, but it seems that the people who inspired this poem did not actually see themselves as Tibetans. Instead, they had coalesced into a new ethnic group, known to the Tang as the Wenmo 嗢末 or Hunmo 渾末.

TheZizhi Tопgjiап, which records that the Wenmo sent their first tribute mission to the Tang court in 862, explains that their name means “slave” in Tibetan, and that many wealthy Tibetan households had ten or more slaves who would accompany their owners on military campaigns.Many Tibetan slave owners were killed or fled during the warlord battles in the Kokonor region;their slaves then formed armed bands (bиlио部落) that also attracted non-elite Tibetans seeking security in numbers. TheZizhi tопgjiапclaims that such bands existed in Hezhou, Weizhou,areas further south and west, and all along the Gansu Corridor — indicating that the Guiyi Army’s campaigns had also indirectly liberated slaves by expelling their Tibetan masters.①Zizhi tопgjiап 250, p.8101–8102.TheХiп Tапgshиcontains a description of the Wenmo that is largely identical to the one in theZizhi Tопgjiап, but it includes the alternative name Hunmo and a few extra pieces of information:the Wenmo/Hunmo had previously served their masters by farming and herding in dispersed groups; they numbered in the thousands when they began forming bands; those “close to the Tibetan headquarters” 近蕃牙者 were the bravest fighters and had excellent horses —perhaps referring to slaves who had formerly served Tibetan generals as bodyguards.②Хiп Tапgshи 216b, p.6108.

Another description of the Wenmo is preserved in the Dunhuang manuscript S.6342.③Images and a transcript of this manuscript can be found in Tang Geng’ou 唐耕藕 and Lu Hongji 陸宏基 eds., Dипhиапg Shеhиi Jiпgji Wепхiап Zhепji Shilи 敦煌社會經(jīng)濟文獻真跡釋錄, Vol. 4, Beijing: Quanguo Tushuguan Wenxian Suowei Fuzhi Zhongxin, 1990,pp.363–364.The text is a memorial from Zhang Yichao to the Tang court that begins with the following statements:

Zhang Yichao memorializes: “In the second year of Xiantong (861), I recovered Liangzhou. Now I do not know if it has been abandoned; moreover, it has a mixed population including Tibetans and Tuyuhun. Recently there were rumors that the Wenmo are obstructing travel to and from [Liangzhou]; I repeatedly inquired about this with witnesses,and they all said this is true. I humbly submit that Liangzhou is the empire’s frontier,and that the common people among the Wenmo are the descendants of Hexi (i.e., Gansu Corridor) and Longyou people who fell under [Tibetan] occupation. The empire cast them off instead of recovering them, so they turned into bands. They unbraided their hair only yesterday, so we have no choice but to mollify them gently.

張議潮奏：咸通二年收涼州，今不知卻廢，又雜蕃、渾。近傳嗢末隔勒往來，累詢狀人，皆云不謬。伏以涼州是國家邊界，嗢末百姓本是河西、隴右陷沒子孫④The original text reads 子將. I have followed the emendation in Lu Qingfu 陸慶夫, “Tang-Song zhiji de Liangzhou Wenmo” 唐宋之際的涼州嗢末, in Zheng Binglin 鄭炳林 ed., Dипhиапg Gиiуijип Shi Zhиапt(yī)i Yапjiи Хиbiап 敦煌歸義軍史專題研究續(xù)編，Lanzhou Daxue Chubanshe, Lanzhou 2003, pp.506, 509.，國家棄擲不收，變成部落。昨方解辮①The original text reads 解辨. I have followed Lu Qingfu’s emendation (see above).，只得撫柔。

This memorial has been variously dated, but I am persuaded by Zheng Binglin’s argument that it was composed in 870–872, when Zhang Yichao was in Chang’an, and was aimed at dissuading the Tang court from pulling out the 2,500 imperial troops from Yunzhou鄆州 (modern Dongping 東平 county, Shandong) that it had sent to garrison Liangzhou shortly after that prefecture’s conquest by the Guiyi Army in 861.②See also Li Jun 李軍, “Wantang (861–907) Liangzhou xiangguan wenti kaocha — yi Liangzhou kongzhiquan de zhuanyi wei zhongxin”晚唐 (861–907) 涼州相關問題考察—以涼州控制權的轉(zhuǎn)移爲中心, Zhопggиоshi Yапjiи 中國史研究 2006(4), pp.78–79, 82–83.The memorial and the appended response from the court show that the Liangzhou garrison was running low on supplies, in part because the Wenmo were “obstructing travel to and from [Liangzhou]” by force.③Two Dunhuang manuscripts also suggest that around 870, Zhang Yichao’s assistants were having problems getting through both the Wenmo of Longyou and the Liangzhou garrison’s officials when travelling between Chang’an and Dunhuang. Li Jun argues that this reflects a conflict between the Guiyi Army and the Tang court over control in Liangzhou. Ibid., p.81.

The Tang court seems to have subsequently dropped its plan to pull the Yunzhou troops out, but it also sent no reinforcements to their aid, effectively leaving them to fend for themselves. Various Song-period texts state that the Yunzhou troops’ descendants were the only “Han households” 漢戶 left in Liangzhou in the early tenth century.④Li Jun has argued that the Guiyi Army regained control of Liangzhou from the Wenmo sometime between 870 and 879, only to lose it to the Wenmo again in the 880s. I am not entirely convinced by Li’s evidence for this, including his reading of p.3451. Ibid., pp.83–87. On the Yunzhou troops’descendants see e.g., Wиdаi Hиiуао 30, p.467; Хiп Wиdаi Shi 74, p.914–915; Sопgshi 宋史492, P.14152.By 884, the Wenmo had established an independent state in Liangzhou, which was soon separated from the Guiyi Army by the new Uyghur kaghanate in Ganzhou.⑤The Dunhuang manuscript S.389 contains a report from the Guiyi Army’s commander in Suzhou regarding the Uyghur conquest of Ganzhou in 884. The same manuscript mentions that the Longjia (Agnean) community in Ganzhou had tried, without success, to obtain a contingent of reinforcements from the Wenmo of Liangzhou to fight the Uyghurs. The Longjia later fled to Suzhou. Li Jun points to evidence that after 884, the Tang court continued to appoint Guiyi Army personnel as officials in Liangzhou, but it is unclear how meaningful or effective such appointments were in the absence of direct access to Liangzhou from Dunhuang. See Lu, “Tang-Song zhiji de Liangzhou Wenmo”, pp.510–511; Rong, Guiyijun Shi Yanjiu,pp.303–306; Li Jun 李軍, “Wantang Guiyijun renyuan renzhi Liangzhou kao” 晚唐歸義軍人員任職涼州考, Dипhиапg Yапjiи 敦煌研究2010(4), pp.80–87.By 908, the Wenmo of Liangzhou merged with local Tibetan groups to form a joint polity that absorbed the Tang garrison’s descendants around 949 and lasted up to the beginning of the eleventh century.⑥I agree with Tang Kaijian’s argument that Zhebu (Chen-Po?) Jiashi’s 折逋嘉施 rise to leadership of the Tang garrison’s descendants in 948–949 marks their absorption by the joint Tibetan-Wenmo regime. Tang’s argument that Zhebu Jiashi was a Wenmo seems plausible to me, as does his argument that the Tibetan-Wenmo alliance finally fractured in 1004 when the leader of the Wenmo, Zhebu (Chen-Po?)Youlongbo 折逋遊龍缽, surrendered to the invading Tanguts. Tang Kaijian 湯開建, Sопg Jiп Shiqi Апdио Tиbо Виlио Shi Yапjiи 宋金時期安多吐蕃部落史研究(Shanghai: Shanghai Guji Chubanshe, 2007), pp.122–161; Lu, “Tang-Song zhiji de Liangzhou Wenmo”,pp.510–515; Li, “Wantang (861–907) Liangzhou xiangguan wenti kaocha”，pp.87–89. Note that this chapter of Tang Kaijian’s book was originally published as an article in 1988.

Although there remains some dispute over the ethnic composition of the Wenmo, I am inclined to agree with Tang Kaijian, Lu Qingfu, and Lu Li that Zhang Yichao’s claim about their origins is true for at least a majority of the Wenmo in the Liangzhou and Longyou areas.The Wenmo bands very likely included people of various ethnic origins, including Tuyuhun and Tibetans, but most of them would have been descended from “Tang people” who were enslaved during the period of Tibetan rule.①Lu, “Tang-Song zhiji de Liangzhou Wenmo”, pp.505–508; Tang, Sопg Jiп Shiqi Апdио Tиbо Виlио Shi Yапjiи, pp.124–126; Lu Li 陸離,“Wenmo yinyi kao” 嗢末音義考, Dипhиапg Yапjiи 敦煌研究 2009(4), pp.98–99.In that case, one could argue that the common experience of slavery under Tibetan masters had produced a new identity that gave the Wenmo greater affinity to one another and to the non-elite free Tibetans who joined them than to the free “Tang people”—including both locals and soldiers from far-off Yunzhou—whom they encountered. A recent history of bondage and servitude under the Tibetans thus led to a process of ethnogenesis somewhat similar to those seen later on in “maroon”communities in the Americas.②The analogy is not perfect, since the maroon communities formed through the merging of indigenous peoples and imported African slaves who had escaped from colonist masters, whereas the Wenmo were indigenous slaves joined by low-status colonists after the colonizing power’s withdrawal from Gansu. On ethnogenesis in the maroon communities see e.g., Kevin Mulroy, “Ethnogenesis and Ethnohistory of the Seminole Maroons”, Jоиrпаl оf Wоrld Histоrу 4.2 (1993), pp.287–305.With hindsight, Zhang Yichao’s statement that the Wenmo of Liangzhou had “unbraided their hair” and could be rehabilitated back to Chineseness by a conciliatory imperial government seems too optimistic. It was they who had “cast off” the Tang empire, not the other way around.

Conclusion

In this paper, I have argued that the Chinese elite families of Dunhuang accommodated themselves to Tibetan rule and adopted bicultural identities until the breakdown of Tibetan administration put their interests in jeopardy, upon which they drove the Tibetans out and sought to cover up their past collaboration by inventing a new tradition of “stubbornly Chinese” Tang loyalism. Despite this effort at reinvention, the Tang court was never fully convinced of the Dunhuang elite’s loyalty — with good reason, for the Chinese of Dunhuang seem to have posed as mortal enemies of the Tibetans only when they stood to gain better relations with the Tibetans’ other enemies. This tendency continued after the fall of the Tang dynasty, as seen from the manuscript P.3633, which contains a letter from 10,000 residents of Dunhuang to the Uyghur kaghan in Ganzhou.①An image and transcript of the letter can be found in Tang and Lu eds., Dипhиапg Shеhиi Jiпgji Wепхiап Zhепji Shilи, pp.377–380; for its context, see Rong, Gиiуijип Shi Yапjiи, pp.223–228. Fu Junlian and Wang Weiqin recently dated the letter to 851, but this is certainly erroneous: Fu and Wang, “Dunhuang ben ‘Zhang Huaishen bianwen’ dangwei ‘Zhang Yichao bianwen’ kao”, p.126.

The letter, dated 911, claims that after the revolt of 848, Zhang Yichao swore an oath with the people of Dunhuang to “wear the robes and caps of the House of Han”著漢家衣冠and “not to wear Tibetan [clothing]” 不著吐蕃. This letter, too, is frequently cited as evidence for the Dunhuang people’s resentment toward the supposed “Tibetanization” policy. The letter mentions no such policy, however, and its rhetorical context is usually overlooked. At the time of its writing, Dunhuang had recently been attacked by the Ganzhou Uyghurs and had sent a diplomatic mission to seek military aid from the “southern Tibetans” 南番 (probably the Tibetans of the Kokonor region). The mission failed, and the ruler of Dunhuang, Zhang Chengfeng 張承奉(r. 894–914), then submitted to the Uyghurs as a vassal under a “father-son”relationship.

In the letter, Zhang Chengfeng effectively uses his subjects’ voices to appeal for clemency from the Uyghur kaghan, seeking to reassure the kaghan that he has no further intention of allying with the Tibetans against the Uyghurs.Such an alliance would be impossible, the letter implies, because of the Dunhuang people’s longstanding enmity toward the Tibetans; Chengfeng had only contemplated it in a fit of anger. Clearly, no declaration of anti-Tibetan sentiment made under such circumstances of duress should be taken at face value. In the absence of other supporting evidence, I doubt we can even assume from the letter that Zhang Yichao and his followers did swear an oath never to wear Tibetan clothing again. In any case, if such an oath did exist, the very need for it would belie both the total persistence of Tang-style clothing claimed by the P.3451biапwепand the deep-seated rejection of Tibetan-style clothing claimed by theХiп Tапgshи. There is no way that P.3451,theХiп Tапgshи, and the letter in P.3633 are telling the same story about Dunhuang, and it is time for historians to stop assuming that they are.

In contrast to the Dunhuang Chinese elite’s attempt at presenting an unambiguously Chinese image to the Tang court, the Wenmo reimagined themselves as a new people created by slavery rather than as descendants of “Tang people”. This new identity prevented the Tang court from reasserting authority over the Gansu Corridor’s eastern end, left an opening for Uyghur refugees from the steppe to establish a new state in Ganzhou, and finally resulted in much of Liangzhou being controlled by an alliance of Wenmo and Tibetans for about a century. Although there is much that we still do not know about the reasons for these differing outcomes on opposite ends of the Gansu Corridor, due to a lack of sources, what we do know suggests that as in any colonial situation, there was no uniform experience of Tibetan rule on the part of either colonizing or colonized populations along the Corridor.①Tang Kaijian has argued that Dunhuang was not culturally “Tibetanized” like Liangzhou because it fell to the Tibetans twenty years later;because it was the cultural center of the Gansu Corridor; and because the elite Chinese families in Dunhuang acted as a countervailing force to the “Tibetanization” policy. I do not find these arguments convincing, partly because they arise from an uncritical attitude towards P.3451 and the Хiп Tапgshи. Tang, Sопg Jiп Shiqi Апdио Tиbо Виlио Shi Yапjiи, p.7 (note that this chapter was originally published in 1983).Instead, one’s experience varied according to one’s social status, social networks, and access to wealth,power, and information, among other factors, and the relative importance of different modes of identity — ethnic, class, gender, family, religious, political — therefore varied as well. I hope that by debunking the mythic image of Tang loyalism constructed by Chang’an irredentists and Guiyi Army leaders, and by contrasting that mythic construction with the origin of the Wenmo, this paper has convinced historians and art historians to reconsider the visual and textual evidence from Dunhuang with eyes more sensitive to the complexity and fluidity of identity in a place like ninth-century Gansu.