Between the Great Walls: Northern Song Imaginations of the Northwest

Abstract

After the political fragmentation of the “Five Dynasties and Ten Kingdoms” period, the Northern Song consolidated much of the lands under these regional states into a larger polity in a process often described as “reunification.” But this “reunification,” judged against the domain of the Tang dynasty, was incomplete. The “Sixteen Prefectures” to the northeast were ceded to the Liao, and this became a vexing issue for Song emperors and officials. But the northeast was not the only region once under Tang rule that did not enter the Song domain. In this article, I discuss the area to the northwest of the Song, much of which was eventually governed by the Tangut Xia state. This area, roughly the modern provinces of Ningxia and Gansu, featured prominently in Northern Song political discussions, national geographical treatises, and national and regional maps. By analyzing the treatment of the northwest in these diverse genres of representation, I demonstrate a spectrum in the perceptions of the northwest. It was sometimes seen as little different from areas under Song rule; in other cases, it was treated as “beyond the sphere of civilization” (huawai). Such ambiguity is visualized in many Song cartographers who placed this area between two segments of the Great Wall. For Song emperors and officials, the northwest sat uncomfortably in their imaginations of the world, not easily dismissed and forgotten, yet irrecoverable.

Introduction

In the fall of 1001, Emperor Zhenzong (968–1022) of the Northern Song (960–1127) assembled a group of high-ranking officials in the palace in Kaifeng to view “a map of Twenty-Three Prefectures in Shaanxi” (Shaanxi ershisan zhou ditu 陝西二十三州地圖). The Emperor pointed out key military and strategic information contained in the map, such as the traversability of mountains and rivers and the location of non-Han fan polities. Then, looking at Qin Prefecture 秦州 (modern Tianshui in Gansu Province) on the map, the Emperor said that he had dispatched someone to guard this important prefecture. Having carefully examined this map, presumably spread out on a surface—a table or the floor—in the palace, the Emperor then turned his gaze toward the walls of the palace hall, which was evidently covered with maps, either in the form of banners or painted directly on the walls.

On the northern wall, there was a map of Ling Prefecture, about which the Emperor remarked: “This was drawn by Feng Ye and is very detailed. The mountains and rivers are so strategically significant—where can I find someone wise and brave who can guard it for me?” Then, turning to the southern wall, the Emperor pointed at the map of Gan Prefecture, Sha Prefecture, Yi Prefecture, Liang Prefecture and other prefectures, and said: “this map records the mountain from which the Yellow River originated. It is beyond (west of) the Jishi [mountain], which is different from the record in The Tribute of Yu” (Yugong 禹貢). Finally, a “map of the state borders of the Kitans to the north of You Prefecture” (Youzhou bei Qidan guojie tu 幽州北契丹國界圖) on the eastern wall shows that the land of Kitan measured 1500 li from the south to the north, and 900 li from the east to the west. Looking at this map, the Emperor lamented that “the territorial domain is not very extensive, it is a pity that Yan and Ji fell to this foreign domain” (yisu 異俗).¹

It is impossible to know the shape and making of the maps on the walls, or if the mural maps included other areas within or beyond the Song domain. What is clear is that the three maps mentioned were all about places (Ling Prefecture, the Hexi Corridor, and the Kitan domain) that were once part of the Tang domain but were either precariously held by the Song state—Ling Prefecture was under siege when Emperor Zhenzong made these remarks, and it would be conquered by the Xia state a few months later, in the spring of 1002—or had never entered the Song domain. Painting (or hanging) these maps of areas beyond the Song domain on walls of the imperial palace was a powerful reminder for the Song Emperor and officials of these “lost” lands. Yet, it is equally intriguing to discern the difference in the Emperor's attitude toward these three areas: while he expressed the desire to seek a capable general to hold on to Ling Prefecture and regretted that the Yan-Yun area was lost to the Kitan, he seems to have only showed antiquarian interest in the area of the Hexi Corridor.

In this article, I try to put Emperor Zhenzong's map viewing party in the context of the shifting views of the northwest in the Northern Song. The broadest background is familiar to students of Middle Period China: the Tang Empire controlled a much larger territory than the Song. Emperor Zhenzong's remark points to two areas where such “loss” is particularly poignant to him. In the northeast, the area around modern Beijing had been, since the Later Jin dynasty (936–947), ceded to the Kitan Liao state (916–1125).² In the northwest, the vast land from the bend of the Yellow River to the western end of modern Gansu was, from the late ninth to the early eleventh century, gradually taken over by the Dingnan Garrison (dingnan jun 定難軍) and its successor, the Tangut Xia state (1038–1227).³ By the mid-eleventh century, the areas represented on both the northern and the southern walls in Emperor Zhenzong's party were part of the Xia state.

The northeast—the Sixteen Prefectures—was a key target in the early Song military, and later became “a concrete symbol of the imperfect formation of the dynasty.”⁴ Partly because of Song people's attention to this area, much modern scholarship has been devoted to it too.⁵ The understanding of the Sixteen Prefectures, and its place in the broader Song–Liao relation, have implications for how the idea of “China” was conceptualized.⁶ The Song views on the northwestern region, in contrast, have been the subject of only a small number of serious scholarly works.⁷ But as the map-viewing party shows, the emperors and officials of the Northern Song by no means forgot about the northwest, as they envisioned the geographical prospect of their state. This article is an investigation of the ways that the northwest was discussed, mapped, and conceived in the Northern Song.

To do so, I need to first clarify what I mean by the “northwest.” The term “northwest” in my use is not strictly speaking a translation of the Chinese term xibei 西北. In Song sources, this term is used more in the sense of a northwestern direction, as in the case of phrases like xibei bian 西北邊, meaning the “border in the northwestern direction.”⁸ Instead, I employ this term to denote a more precise region that included modern Ningxia and Gansu provinces and northern Shaanxi Province, areas that were once Tang official prefectures (zhou) but that never entered the Song domain. This area can be further divided into two regions, “Henan” (south of the Yellow River), which had been the heart of the Dingnan Garrison since its inception in the late ninth century, and “Hexi” (west of the Yellow River), which had been conquered by the Xia state, the successor of the Dingnan Garrison, by the mid-eleventh century. These two regions were represented separately in emperor Zhenzong's palace hall: Henan, of which Ling Prefecture was the center, on the northern wall, and Hexi (“Gan Prefecture, Sha Prefecture, Yi Prefecture, Liang Prefecture”) on the southern wall. I discuss these two regions together because they eventually constituted the domain of the Xia state and were thus unified in some way in the minds of the Song people. For the same reason, I do not include the areas further to the west, including modern Xinjiang and Central Asia, in my discussion. These areas, further away from the Song both in distance and in imagination, occupied different places in Song geographic thinking, and should be treated separately.⁹ I will call this area “the far west” for the sake of clarity and distinction. It is no coincidence that emperor Zhenzong did not mention these areas as he examined the mural maps. In this sense, this article is also an investigation of how the Xia land was viewed in the Northern Song.

I try to access the Northern Song view of the northwest in three domains: political discourse, geographical treatises, and officially commissioned maps. With the rise of the Xia state, and its gradual encroachment on Song territory in the northwest, Song officials and emperors debated the merits and pitfalls of different strategies against the Xia. Their views on whether the Song should attack or retreat, as I show, often depended on different ways of conceiving the northwest in relation to the Song domain. Such difference is more subtly expressed in Song geographical treatises, such as the late-tenth century Universal Geography of the Taiping Reign (Taiping huanyu ji 太平寰宇記) compiled by a Song geographer Yue Shi 樂史 (930–1007).¹⁰ These treatises drew from earlier Tang geographical works, but also incorporated new information and new systems of geographical thinking. As they were often created, as in the case of the Universal Geography, in order to celebrate Song rule, their views reflect the official thinking about different parts of the Song world, including the northwest. Finally, the Song dynasty represents the first era in Chinese history from which large quantities of officially commissioned maps have been preserved and transmitted to us. The scale of these maps can be “ecumenical,” like the Hua-Yi tu 華夷圖 (Map of China and the Barbarians), “national,” like the various shidao tu 十道圖 (Map of the Ten Circuits) and the shanchuan xingshi tu 山川形勢圖 (Map of the Shape of Mountains and Rivers), or “regional” such as the map of Shaanxi presented to Emperor Zhenzong in 1001. A small number of these maps have been preserved, either as stone inscriptions or in printed books. The map makers, if the domain of their project included the northwest (as I define it), had to come up with a way of representing it. Did they draw the northwest the same way as they did the heartland of the Song? Or did they try to represent the northwest in visually distinct ways? Through my close reading of some of these maps, I show that the views revealed by these maps corroborated those found in political discourses and geographical treatises, while offering uniquely visual expressions.

These three groups of sources reveal different aspects of the Song geographic thinking about the northwest. Combining these sources, in the Conclusion I chart the changing landscape of the Song view of the northwest, and compare it to the most prevalent views of the northeast and of the far west. I suggest that a careful consideration of the place of the northwest in the Song geographic thinking is an integral part of a comprehensive understanding of how the Song people viewed their state and its place in the world.

Political Discourse

As the works of Li Huarui and Tsang Shui-lung show, the relation between the Song and the Xia had many twists and turns. This section will not be able to fully recount this long history and the political discussions that occurred in the process. Instead, I will single out a few representative cases that can illustrate the views of key Northern Song politicians about the place of the northwest in relation to the Song domain.

The origin of the Xia state can be traced back to the late Tang. Tuoba Sigong 拓跋思恭 (?–886), after helping the Tang state in suppressing the Huang Chao rebellion, was appointed the military commissioner (jiedushi 節度使) of the Dingnan Garrison and bestowed the imperial surname Li. His state governed five prefectures (Xia, Yin, Sui, Jing, and You) in the northwestern border of the Tang. Like in many of the late Tang warlord states, leadership of the Dingnan Garrison passed within the Tuoba/Li family, as they maintained nominal deference to the Tang. After the fall of the Tang, the Dingnan Garrison continued the vassal state status in their relation to the Later Liang (907–923) and the Later Tang (923–937), until 933, when Li Yichao 李彝超 (?–935), a grandson of Tuoba Sigong, repelled a Later Tang effort to encroach on its independence. Afterwards, the Dingnan Garrison essentially became an independent state from the “dynasties” in North China.¹¹

Little changed with the founding of the Song in 960, as the first Song emperor initially targeted southern states in his attempt at “reunification.”¹² This grand strategy was specifically designed to avoid engaging directly with warlords in the northwest, including the Dingnan Garrison. In a likely hagiographical conversation between emperor Taizu (927–976), the founder of the Song, and Zhao Pu 趙普 (922–992), the chief strategist, in 968, the emperor expressed the desire to attack Taiyuan 太原, the seat of the Northern Han state (951–979) in modern Shanxi Province. Zhao Pu objected to this proposal, alerting the emperor that “Taiyuan borders both the northern and the western borders. Suppose we conquer it, we alone will be exposed to border harassment (bianhuan 邊患). Why not leave it until we have eradicated all the other kingdoms? [Afterwards], how can this tiny state escape [our conquest]?”¹³

The sequence of events that played out afterwards famously confirmed Zhao Pu's proposal. Emperor Taizu and his brother Emperor Taizong (939–997) conquered several southern kingdoms in the 960s and 970s, before Emperor Taizong finally turned back north and conquered the Northern Han in 979. Zhao Pu's strategy proposal does not directly implicate the “northwest” that I talk about. But implicitly, his use of the phrase “the northern and the western borders” (xibei erbian 西北二邊) reveals an important feature in the thinking among imperial leadership about the northwest in the early Song. To the north, the Northern Han bordered the Liao state, the chief geopolitical rival of the Song. To the west, the Northern Han faced several warlords across the Yellow River, including the She 折 family in Fu Prefecture (modern Fugu in Yulin, Shaanxi) and the Dingnan Garrison under the rule of Li Yixing 李彝興 (r. 936–967) and his son Li Guangrui 李光睿 (r. 967–978). Therefore, underlying Zhao Pu's proposal of “reunification” is an assumption that these areas west of Taiyuan and south of the bend of the Yellow River, all of which were once prefectures of the Tang empire, were borders to be guarded, not lands to be conquered and “reunified” into the Song domain.

A key moment in the Song's relation with its western neighbors came in the early summer of 982, when Li Jipeng 李繼捧 (?–1004), the leader of the Dingnan Garrison, submitted himself to the Song court in Kaifeng. The son of Li Guangrui, Li Jipeng succeeded in the role of governor of Dingnan Garrison in 980 after the death of his older brother who only ruled for one year. Because Li Jipeng was a youngest son of Li Guangrui, his ascension was objected by many elders of the Li family, notably Jipeng's uncle Li Kewen 李克文 (?–1005). Unable to quell internal dissent, and no doubt encouraged, perhaps even coerced, by the Song, Li Jipeng decided to give up his status of a regional warlord, and entered the Song capital as a willing vassal. The Song court welcomed Li enthusiastically, showering him and his immediate family (including his grandmother) with gifts and titles, kept him in Kaifeng, and distributed the lands he once governed to pro-Song officials, many from the same Li family.¹⁴ Li Jipeng's submission to the Song was later framed as a key step to the reunification effort. In the geographical treatises included in the Songshi, for instance, we see the following characterization of this event:¹⁵

In the seventh year [of the Taiping xingguo reign] (982), Li Jipeng came to the court, and [our state] acquired Four Prefectures and Eight Counties. At that moment, all-under-heaven was already unified, and the [Song] domain almost recovered the old [domains] of the Han and the Tang dynasties. The only area that did not enter the map of the imperial domain were the Sixteen Prefectures including Yan and Yun.

七年，李繼捧來朝，得州四縣八。至是，天下既一，疆理幾復漢、唐之舊，其未入職方氏者，唯燕、雲十六州而已。

The four prefectures were Xia 夏, Yin 銀, Sui 綏, and You 宥, a relatively small area in modern northern Shaanxi and southern Inner Mongolia, and the territory ruled by Dingnan Garrison at the time.¹⁶

The claim made in Songshi is extraordinary in two ways. First, although the nominal integration of these four prefectures extended the Song border by only a few hundred kilometers to the west, the author of the passage felt confident enough to compare the Song to the Han and the Tang. The key word here is, of course, that it “almost recovered” (jifu 幾復) the domain of these two empires. Lurking behind this hesitant claim was no less than the relinquishing of the entire Hexi Region in modern Gansu and large parts of modern Ningxia, as well as eastern Xinjiang such as the area around modern Turfan, all of which were once part of the official prefecture-county system of the Tang, but beyond the Song border.¹⁷ Such a gargantuan omission in this geographic vision highlights a second feature of this claim: while being willing to simply forget about an enormous area in the northwest, the author was not willing to concede the Sixteen Prefectures, a comparatively smaller area in the northeast that, according to this claim, was the “only area that did not enter the map of the imperial domain.” The view on the northwest revealed in this claim is broadly similar to Zhao Pu's view: Song officials knew very well that their domain did not equal those of the Han and the Tang in the northwest, but they showed little irredentist interest in the northwest, and the old Tang domain of Hexi and beyond were not considered the core interest of the Song, in the same way that the Sixteen Prefectures were.

For all the rhetoric, Li Jipeng's action did not lead to the incorporation of the lands under Dingnan Garrison into the Song domain. Instead, it prompted strong reactions from dissenting factions within Li's state, foremost of which was his cousin Li Jiqian 李繼遷 (963–1004). By rallying local support, both from the Tangut people and various other non-Han groups in the region, and marrying a Liao princess (989), Li Jiqian consolidated his status as the new ruler of his state. It is from this period onward that, because of Li's hostility toward the Song and his acceptance of the title of the “King of the Xia State” (xia guowang 夏國王) in 990 from the Liao emperor, we can speak of an independent Xia state. The Xia state contended with the Song along its eastern border, but expanded further to the west, conquering Ling Prefecture in 1002, and the Hexi region in the 1020s and 30s, under the rule of Li Deming 李德明 (981–1032) and Li Yuanhao 李元昊 (1003–1048). The latter would, after the final conquest of the entire Hexi region, assume the title of the emperor (huangdi) and establish a new Xia dynasty in 1038.¹⁸ This half-century ascent of the Xia took place in the context of a Song state that was becoming less adventurous militarily. Faced with the possibility of losing Ling Prefecture to the Xia, there was an extensive debate at the court of emperors Taizong and Zhenzong about the merits and pitfalls of continuing to defend Ling Prefecture, now located further to the west of the rising Xia power, becoming almost an enclave. Many who espoused the position of essentially giving up Ling Prefecture to the Xia commented on the location of the area.

One official in 997 described Ling Prefecture as “located beyond the frontier fortifications [saiwai 塞外], although it is often called a strategic place on the western borderland, it in fact has become an area that drains [the resources; duqu 蠹區] of the Central Land of the Xia.”¹⁹ The term duqu is a hapax legomenon in the Song textual corpus. But judging from its components—du means a kind of insect that eats and corrupts, while qu means “region,” the meaning of the term is fairly clear: Ling Prefecture was a leech of a prefecture that could drain all the resources of the “Central Land of the Xia” (zhongxia 中夏), the Song domain, if military actions were taken to defend it.²⁰ While the phrase “beyond the frontier fortifications” (saiwai) had by this time become a common term for areas considered beyond the northern border, the need to coin the neologism duqu reveals a lack of standard vocabulary to describe the place of Ling Prefecture.

This view is more systematically expressed by Yang Yi 楊憶 (974–1020) in 1001, who joined the debate about the fate of Ling Prefecture. A young official in charge of drafting imperial edicts, Yang first invoked the history of the Han dynasty to show that the Ling Prefecture area, also called Shuofang, had always been difficult to govern, and was an area “unreachable by civilized teachings” (shengjiao buji 聲教不及).²¹ He then described Ling Prefecture as “the land of Helian Chang 赫連昌 (?–434),” the emperor of the Xia state in the Sixteen Kingdoms period, and the son of Helian Bobo 赫連勃勃 (381–425), and then the “old domain of the Xiongnu” (xiongnu zhi jiurang 匈奴之舊壤). By invoking these historical precedents, Yang argued for the alien status of Ling Prefecture that was “distant from the various Hua areas” (miaojue zhuhua 邈絕諸華).²² After describing the failures of the Song in holding onto Ling Prefecture, Yang continued with a broader claim about the changing domains of historical dynasties. He claims that the sage kings Yao, Shun, and Yu only ruled an area of a few thousand li, while the area under King Wuding of the Shang and King Cheng of the Zhou, both also sage kings of the past, was also fairly restricted. Importantly, he pointed out that the northern edge of both kings’ political domain “did not reach beyond Taiyuan” (bei buguo Taiyuan 北不過太原).²³ The further northwestward expansion only began in the Qin and the Han dynasties.

Yang's argument for giving up Ling Prefecture thus depended on his assessment of its geographical location and its relation to the Song heartland. He simply refused to make recovering the lost frontier territories of the Tang a priority. If someone objected to giving up Ling Prefecture on the basis that it would become a “lost land” to the Song, Yang argued, they should look at many other lost lands such as “the eight prefectures of You-Ji (northern Hebei) and five cities of Hehuang (southern Gansu and eastern Qinghai).” Since many of these areas were never parts of the Song domain, to characterize them as “lost lands” implies that Yang Yi was comparing the Song with the Tang. Therefore, Yang Yi was not unaware of the fact that the Tang domain extended much further to the north and the west than that of the Song. He was not even trying to avoid mentioning this comparison. Instead, in this long treatise, he mounted a two-fold argument. First, political domains change, and one of the reasons for such changes was the reckless adventurism of certain past rulers; there was no reason to blindly follow the examples of the most expansionist Han and Tang emperors and venture too far into the northwest. Second, in contrast to emperors of the Qin, the Han, and the Tang, the ancient sage kings ruled a much smaller area, centered in North China, where the Song capital was also located. This smaller area formed the core interest of the Song, while any regions beyond were dispensable. According to this argument, Ling Prefecture and the areas further to the west would all be readily dispensable without hurting the core interest of the Song state.

The battle over Ling Prefecture foreshadowed the increasing tension between the Song and the Xia in the eleventh century. After Li Yuanhao declared himself the emperor in 1038, the two sides engaged in three major battles in the early 1040s, all of which resulted in the defeat of the Song.²⁴ In 1044, both courts agreed upon a treaty in which the Xia emperor accepted a relatively low status vis-à-vis the Song emperor in exchange for Song material payment to the Xia in silk, tea, and silver. This treaty, known among historians as the “treaty of the Qingli reign” (Qingli heyi 慶曆和議), along with the internal chaos (the assassination of Li/Weiming Yuanhao in 1048 and the succession of Yuanhao's two-year-old son Liangzuo) that thwarted Xia military ambitions, resulted in relative peace between the two states. As Liangzuo (Emperor Yizong 1047–1068, r. 1048–1068) and his son Bingchang (Emperor Huizong, 1061–1086, r. 1068–1086) both began their rule as young children, their mothers held the real power for extended periods. In the process, the Song slowly gained an upper hand in military power against the Xia. Liangzuo died, for instance, from wounds suffered in a battle against the Song.

Major Song counteroffensives began again with the reign of the young and ambitious Emperor Shenzong (1048–1085, r. 1067–1085).²⁵ After Emperor Shenzong assumed the throne, Wang Shao 王韶 (1030–1081) submitted a policy proposal titled “Strategy to Pacify the Barbarians” (ping rong ce 平戎策), which laid out a plan to take over the Hehuang region (modern southern Gansu and eastern Qinghai, in the southwestern part of the Xia domain). Wang followed this policy proposal by leading the Song army into the Hehuang region in 1071 and in fact taking much of the Hehuang region. Wang Shao's proposal is a guiding document that in many ways influenced, even predicted, the expansionist policy against the Xia in Emperor Shenzong's reign.

The body of this proposal discusses primarily the need to conquer the Hehuang region to threaten the Hexi area of the Xia. To justify this position, Wang opened his treatise with a broad discussion of the relation of the northwest to the Song heartland.²⁶

Since antiquity, there has been no ideal strategy in defending against the barbarians (rong). It is said that the northern and western barbarians [rong-di] were capricious, and were located beyond the Yao domain.²⁷ Now the Li family of Hexi occupies two circuits, both old prefectures of the Han and the Tang times. They were within the boundaries of our domain. This is what is called sores and rashes that linger underneath one's side that threatens the heart and the belly [xinfu zhihuan]. How can our state treat [the Xia] as barbarians? Should we not worry that, generations after our sons and grand-sons, they might pose the threat of encroaching on our state and disrupting the interior?

古者禦戎無上策，謂戎狄荒忽在要服之外也。今河西李氏據兩路，皆漢唐舊郡，在邦域之中，所謂癰疽伏疹，留滯脅下，心腹之患也。 … … 國家奈何以戎狄待之？曾不念子孫百世之後，將有蹙國内侮之憂也？

Here, Wang first conceded that there had not been a good policy in dealing with the “barbarians” (rong, or rongdi) since antiquity, and they were often left to their own devices without interference from the state in North China. But such lack of strategy should not apply in this case because, according to Wang, the Xia state—“the two circuits under Li family rule”—were located “within the boundaries of our domain” (bangyu zhizhong). Therefore, they should not be left alone, but should be taken seriously as a threat to the very core existence (“the heart and the belly”) of the Song state. In this case, the assessment of the northwest's geographical relation with the Song heartland is a central pillar of Wang's argument for an expansionist policy.

This policy, espoused by Emperor Shenzong and his trusted advisor Wang Anshi, was initially successful, but the major campaigns the Song waged against the Xia in the early 1080s resulted in Song defeat.²⁸ After the death of Emperor Shenzong, such expansionist attitude was almost immediately reversed.

The view of the other side of the debate is expressed clearly by Sima Guang (1019–1086), the paragon of the anti-reform clique in Song politics. Sima's argument against Wang Anshi and Emperor Shenzong's border policy is summarized succinctly in a policy proposal he submitted after emperor Shenzong's death. Sima's proposal begins with a general assessment of the ills of the time: the depletion of public and private resources and the exhaustion of people near and far were all caused by unnecessary military actions. To support this central argument, Sima recounted a history of the development of the Song domain. The first two Song emperors, Sima argued, ended an era of division that lasted for two hundred years since the mid-Tang (i.e. the An Lushan rebellion). As a result of their effort, “the traces of the Great Yu [Da Yu zhiji 大禹之跡] are all possessed by the Song.”²⁹ Here, Sima Guang recycled a concept of the “traces of the Great Yu” that he had already used decades earlier in 1061 in his petitions to Emperor Renzong, where he claimed that the great accomplishment of the two early Song emperors was the reunification of “the traces of the Great Yu.”³⁰ Curiously, he also claimed that this was the first time the area was unified in the past 225 years, since the onset of the An Lushan Rebellion (755–763). This timeframe means that, unlike the writers of the preface of geography chapter in Songshi, Sima considered the conquest of the Northern Han in 979 (225 years after 755), rather than the submission of Li Jipeng in 982, as the event that marked the final reunification of the land under the Song.

After describing the reigns of the next three emperors as largely peaceful—a gross simplification—the 1086 petition focused on the reign of emperor Shenzong, and noted the geopolitics of the time:³¹

You Prefecture, Ji Prefecture, Yan Prefecture, and Shuo Prefecture were lost to Kitan (the Liao); Lingwu and Hexi regions were dominated by the Tuoba (the Xia), Jiaozhi and Rinan were governed by the Li family (the Ly dynasty in Vietnam). It was impossible to dispatch officials and clerks to these places to compile household registers and collect taxes. Compared to the domains of the Han and the Tang, [the Song domain] was still not complete. [The emperor was] deeply ashamed, and vehemently developed a desire to expand [the Song domain] through warfare.

以幽、薊、雲、朔淪於契丹，靈武、河西專於拓跋，交趾、日南制於李氏，不得悉張置官吏，收籍賦役，比於漢、唐之境，猶有未全，深用為恥，遂慨然有征伐、開拓之志。

The view about the nature of the Song domain quoted here differs very little, on the surface, from that of Wang Shao. Both Sima and Wang acknowledged that places under the Xia rule were once within the domains of the Han and the Tang. But Sima then continues to deride the “borderland military men” (bianbi wufu 邊鄙武夫), who engaged in recklessly adventurous behaviors against the Xia, in efforts to win favor with emperor Shenzong. In using this term, Sima was likely referring to military officials like Wang Shao, even though the latter also received a jinshi degree.³² To Sima, there was no shame in not matching the Han and the Tang in geographical expansion, because the two founding emperors had already established the core of the Song state that encompassed “the traces of the Great Yu” in its entirely. The state of King Yu, compared to the Han and the Tang, was a much more admirable example to emulate.

As with other aspects of the Song politics, the Xia policy also shifted again after the next Song Emperor Zhezong (1077–1100) assumed real power in 1093.³³ Emperor Zhezong vowed to resume the policy of his late father, Emperor Shenzong, and engaged in military actions with the Xia in both the Fuyan region in the north and the Hehuang region in the south. After initial success, the Xia was able to fight the Song to an impasse. The last bout of Song military ambition occurred in the mid-1110s, when Emperor Huizong (1082–1135, r. 1100–1126) again wished to continue the unfinished campaigns of Emperor Shenzong.³⁴ There were significant changes in the Song–Xia border throughout the four decades from the death of Emperor Shenzong to the fall of the Northern Song in 1127, but neither the Song nor the Xia were able to reach any conclusive victory. It was the rise of the Jin dynasty that conquered the Song and physically separated the Song and the Xia, bringing an end to the centuries-long military engagement between the two states.

For all the political and military vicissitudes in these last four decades, the rhetoric about the relation of Xia territory to the Song among Song officials continued along earlier lines. For instance, in 1099, when the Song conquered the Qingtang state, a Tibetan state located to the southwest of the Xia, the prime minister Zhang Dun 章淳 (1035–1106) stated that “the Longyou and Origin of the [Yellow] River [heyuan 河源] areas have long fallen into the domain of the remote borderland” and the result of the recent military success was that “the old domains of the Han and the Tang was entirely restored” (xifu Han-Tang zhi jiu jiang 悉復漢唐之舊疆), echoing the claim made by Wang Shao when he designed the expansionist policy toward the northwest.³⁵

To conclude, in the long history of Song–Xia relations, in times of negotiated peace and open warfare, a wide spectrum of opinions regarding how to treat the Xia state were voiced by generations of Song officials. Among the voluminous debates on border policy, I have selected a few influential examples. Their views are sometimes diametrically opposed, but they all acknowledge a basic historical fact: the northwest was once within the domain of the Han and the Tang dynasties. The views diverged on how to deal with this fact. Hawks like Wang Shao argued that because of this history, the northwest should be considered within the Song boundaries too and not be treated as rongdi. Doves like Sima Guang, on the other hand, made a distinction between the domain of the Han/Tang and the smaller domain of older regimes under ancient sage kings such as Great Yu, arguing that since the northwest was never under the rule of the sage kings, they should not be central to the concern of the Song emperors either.

These Song officials were eloquent rhetoricians and expert historians. Whatever views they held, they could almost always find precedents in the long imperial and pre-imperial history. They sometimes also manipulated their language to advance their argument. The unknown author of the passage from the chapter on geography in the Songshi, if this section indeed came from Song-era sources, placed enormous weight on the small character ji 幾 (almost) when they claimed that the submission of Li Jipeng meant that the domain of the Song almost equaled that of the Han-Tang empires. Sima Guang, on the other hand, created the idea of an area known as the “traces of the Great Yu,” which constituted the heartland of civilization. The genre of the policy recommendations allowed such selective use of history and manipulation of language. But the space of rhetorical maneuvering narrowed and transformed in other genres of geographical writing, where more concrete decisions had to be made about whether to include a place within a geography of the Song domain or not, which chapter to put it in, and where to draw the line on a map. In the next sections, we can further observe the Northern Song views on the northwest in geographical treatises and maps.

Geographic Treatises

The Song dynasty witnessed a new era of geographical knowledge and writing. Many older kinds of geographies, such as the “Illustrated Geographies” (tujing 圖經) and “Treatise on the Ten Circuits” (shidao tu 十道圖) continued to be produced, but new kinds of works, most notably the local gazetteer (difang zhi 地方志), also began to emerge. The genre of geographical writing that took shape in the Song would become the standard for the rest of the imperial period.³⁶

In this era of the explosion of geographical writing, specialized treatises and perhaps even local gazetteers were produced about the northwest. For example, a text titled An Illustrated Geography for the Accumulation of Rice in Shaanxi (Shaanxi jumi tujing 陝西聚米圖經) was said to have been written by Zhao Xun 趙珣, a military official active on the northwestern border, in 1041. According to a later catalogue, this text describes the “mountains and rivers, condition and distances on the road of the five circuits” along the western border of the Song.³⁷ Unfortunately, none of these specialized geographical works on the northwest have come down to us. Instead, of the vast number of geographical texts produced in the Northern Song, the best preserved is a particular genre that documents not a specific area, but the entire Song domain. In this section, I survey four such texts that were produced in the Northern Song dynasty and examine the ways they treat the geographical information of the northwest. These four texts are Record of the World during the Taiping [xingguo] Reign (Taiping huanyuji 太平寰宇記), completed in the late 980s,³⁸ Essentials of Military Classics (Wujing zongyao 武經總要), composed between 1043 and 1047 that includes a survey of the Song borders,³⁹ Treatise of the Nine Regions in the Yuanfeng Reign (Yuanfeng jiuyu zhi 元豐九域志), finished in 1080,⁴⁰ and Extensive Record of Geography (Yudi guangji 輿地廣記), produced in the 1110s.⁴¹ By tracing how the information about the northwest was organized in these four texts, I hope to illustrate the changing views of this area in geographical writing over the century from the early Song to the eve of its conquest by the Jin.

Yue Shi's work is a monumental compilation of geographical data. In the “Preface,” he acknowledged that there were precedents in the Tang dynasty to his work, such as Jia Dan's 賈耽 (730–805) Record of the Ten Circuits (Shidao lu 十道錄) and Li Jifu's 李吉甫 (758–814) Illustrated Gazetteer of the Prefectures and Counties in the Yuanhe Reign (Yuanhe junxian tuzhi 元和郡縣圖志).⁴² However, the brevity of these earlier works, the change of the dynasty, and the shifting local geographies in each of the counties and prefectures meant that it was necessary to produce an updated geography of the new realm of the Song in order to celebrate the deeds of the Song emperor, in Yue's words the “one sovereign of over [the world's] myriad countries” (wanguo zhi yijun 萬國之一君).⁴³

In 200 juan, The Record of the World describes the Song domain and the world around it. The first 171 juan of the book are devoted to the thirteen circuits (dao 道) of the Song, beginning with the Henan circuit, where the capital Kaifeng was located, and ending with the Lingnan circuit in the deep south. Under each circuit, entries are organized under different prefectures (zhou 州) and other equivalent administrative units (such as jun 軍 and jian 監). The structure of this part of the book largely follows that of Li Jifu's Illustrated Gazetteer.⁴⁴ Each entry includes a brief history of the prefecture in question, the counties under its jurisdiction, its location in relation to its immediate neighbors, the two capitals Kaifeng and Luoyang, and the old Tang capital of Chang'an, its population, unique cultural traits, and local products. In the last 29 juan, the text turns to what it calls “the Four Barbarians” (siyi 四夷), and lists states that lay to the east, south, west, and north of the Song domain. The kind of information included in this section devoted to the non-Song states is broadly similar to that included for the Song prefectures. Since the Illustrated Gazetteer does not have any discussion of the “Four Barbarians,” Yue Shi turned for a model to another famous Tang text, Du You's 杜佑 (735–821) Comprehensive Compendium of Institutions (Tongdian 通典).⁴⁵ The majority of the “barbarian” states included, and the sequence in which they appear in the text, are the same as those in the Comprehensive Compendium.⁴⁶

The overall reliance on Tang predecessors should not diminish the originality of Yue Shi's work. Yue consulted an enormous number of other sources and incorporated updated Song-era information on various aspects of the places he recorded. Take, for example, Hua Prefecture (Huazhou 華州), a key location on the route between Chang'an and Luoyang: according to Comprehensive Compendium, it was listed as being 180 li east of Chang'an, 670 li west of Luoyang.⁴⁷ In the Illustrated Gazetteer it was 180 li east of Chang'an, but 680 li west of Luoyang.⁴⁸ Evidently, the Illustrated Gazetteer, composed a decade after the Comprehensive Compendium, used an updated piece of information, perhaps because of the construction of a new road, or a modification of an existing road. As distances between places in the Tang and the Song were based on experiences of travelers on the road, when roads shifted, distances between two well-known places such as Hua Prefecture and Luoyang could also change.

Record of the World generally notes the distances from a certain place to Kaifeng, Luoyang, and also Chang'an. Because of the change of the capital to Kaifeng, the “eastern capital” in the Song meant Kaifeng, while the “western capital” meant Luoyang. In Record of the World, Hua Prefecture was “around 980 li” west of Kaifeng, “around 610 li” west of Luoyang, and 150 li east of Chang'an.⁴⁹ The road between Chang'an and Luoyang via Hua Prefecture had been well-established in the Tang dynasty. Yet this well-trodden road became significantly shorter—760 li versus 860 li according to the Illustrated Gazetteer—in the Song dynasty. Was the old road renovated or was a new road constructed? It is impossible to know. What is clear, however, is the fact that Yue Shi's work incorporated new information, likely reported by actual travelers, regarding the distances among different places within the Song domain. The same can be said to varying degrees about other aspects of the information included in this book.

If adjusting Tang sources to Song reality in places like Hua Prefecture at the heart of the Song domain demanded the revision of certain statistics, the northwest presented Yue with a much thornier issue. Yue must have been faced with a decision about what to do with an area that featured prominently in Tang geographies he relied on, but did not remain in the Song domain. At the first glance, Yue seems to have ignored or papered over the political change that occurred after the mid-Tang: the northwest as defined in this article is recorded in chapters of the Guanxi 關西 Circuit and the Longyou 隴右 Circuit along with other regular prefectures under the Song official jurisdiction with little distinction. If we read only the table of contents of this book, it would appear that Yue presented the Song domain as differing very little from the Tang by including places like Xia Prefecture, Gan Prefecture (Zhangye), and Sha Prefecture (Dunhuang) in this book.

The difference between these northwestern regions and regular Song prefectures is visible only when we get into specific entries themselves. For instance, Xia Prefecture, the center of the future Xia state, is placed within the Guanxi Circuit in juan 37 of this book along with other regular Song prefectures. Only two places hint at Xia Prefecture's status beyond the Song jurisdiction. At the end of the brief history of the place, it is acknowledged that Xia Prefecture once “succumbed to foreigners” (xianfan 陷蕃) in the Five Dynasties era, but returned to the Song in 983, a reference to Li Jipeng's symbolic submission in 982 discussed above. Afterwards, Yue Shi's book states, it is “Dingnan Military Garrison,” without specifying its relation with the Song.⁵⁰ The second place to note is the end of the entry, where Yue Shi lists eight “commanderies” (dudu fu 都督府), such as the Yunzhong 雲中 Commandery, that were “established by the Tang but are now abolished.”⁵¹ The inclusion of these Tang military establishments and the acknowledgement of their abolishment at Yue Shi's time reveal rather indirectly the lack of a Song presence in this area. Elsewhere, similar information is conveyed in a more straightforward manner. The entry for Gan Prefecture (Zhangye), for instance, clearly indicates that it was “abolished” (fei 廢).⁵² This is of course a judgment made from the Song perspective, as we know that the Uyghur state was alive and well in this region during the time that Yue's geographical treatise was composed.⁵³

Additionally, the Song presence and its relation with the northwest is also expressed in a more subtle manner in Record of the World. As in the case of Hua Prefecture, for all the prefectures in the northwest Yue Shi provides us with their distances to Kaifeng, Luoyang, and Chang'an. Table 1 lists 23 prefectures in the northwest, and their distances (in li) to these three cities according to the Record of the World. The numbers are followed by initials of texts they originate from. TD indicates that they were copied from the Comprehensive Compendium, while YH indicates that the data were from the Illustrated Gazetteer. Of course, these two Tang works only recorded the numbers relating to the two capitals during the Tang: Chang'an and Luoyang. The way that Yue Shi came up with the distance to Kaifeng was, in the vast majority of cases listed in the table, simply adding 420 li, the distance between Luoyang and Kaifeng, to the number of distance to Luoyang.⁵⁴ All of the numbers calculated in this manner are marked with d. to indicate that they were derivational of the distance to Luoyang.

Table 1. The Northwestern Prefectures and their Distances (in li) to Kaifeng, Luoyang, and Chang'an in the Record of the World during the Taiping [xingguo] Reign

A number of observations can be made about the data collected here. First, Yue Shi had no new data for the more distant places. For all the prefectures more the 2500 li from Luoyang (mostly places in the Hexi Corridor), Yue copied the data from old Tang geographies. Second, for certain places, Yue had only new data for their distances to Chang'an, but he copied old numbers for those to Luoyang and Kaifeng. An example is Liang Prefecture's distance to Chang'an, which is 1800 li according to Record of the World. In both Tang works, the distance between these two places were about 2000 li. Yue Shi copied the number from the Illustrated Gazetteer for Liang Prefecture's distance to Luoyang, but flagged his new data by specifying that “if one travels by way of the Pass of Qin Prefecture,” the distance from Liang Prefecture to Chang'an would be 1800 li. Third, for places closer to the Song domain, or those within the Song, Yue Shi often had new data. For Dang Prefecture, Qin Prefecture, Wei Prefecture, Yi Prefecture, and Sui Prefecture, he even had nonderivational distances to Kaifeng recorded, indicating that the Song government was diligent in collecting new data about the distance to these border regions. In fact, an approximate northwestern borderline for the Song can be drawn on the basis of the presence/absence of new distance information in the Record of the World.

A final feature of the Record of the World's treatment of the northwest is the clear distinction it made regarding the area further to the west of Xi Prefecture (Turfan) and Sha Prefecture (Dunhuang), the “far west” as I define it. While both of these prefectures, along with the area to their east, were placed in the Longyou Circuit chapter, the far west was actually listed twice in this book. The first reference is in the Longyou Circuit chapter, under the entry of the Great Anxi Protectorate (Anxi da duhufu 安西大都護府).⁶⁰ This entry first lists the Four Garrisons (Kucha, Khotan, Shule/Kashgar, and Yanqi/Agni) of the Anxi Protectorate in modern Southern Xinjiang, and then sixteen commanderies (dudu zhoufu 都督州府) further to the west that were located in Afghanistan and Iran, areas that were purely nominally affiliated with the Tang government. This list does not come from the Illustrated Gazetteer, the usual basis of this section of the book. A note at the end of this entry reveals Yue Shi's thinking:

The kingdoms in the Western Regions listed above … … were under the governance of the Anxi Protectorate. Before the fourteenth year of the Tianbao reign (755, the onset of the An Lushan rebellion) they paid tribute (to the Tang government) without interruption. Now I list them under the Anxi Protectorate in order to illustrate my intention of compiling a Record of the World [yi biao huanyu zhi zhi 以表寰宇之志].⁶¹

From this note, we can see that Yue included this region so that the content of the chapter lives up to its ambitious title, the Record of the World.

The second time the northwest is recorded is in the “Western Barbarian” (Xirong 西戎) chapter toward the end of the book. Here, Yue Shi offers a much more detailed account of the history and geography of many of these kingdoms. For instance, while the Longyou Circuit chapter makes reference to the kingdom of Khotan but merely discusses its status as host to a Tang military garrison, the “Western Barbarian” chapter goes into the details of Khotan's history from the Han dynasty onward, its distance to neighboring kingdoms, and local customs and products.⁶²

To summarize, Yue Shi's Record of the World describes the land purportedly of the Song (chapters 1–171) surrounded by the “four barbarians” (chapters 172–200). Together they constitute the world (huanyu) in Yue Shi's eyes. But the area discussed in the first 171 chapters under the headings of the Ten Circuits is modeled on the Tang geography (and the Tang domain) and is much larger than the land actually administered by the Song government. Specific to the northwest, this mismatch means that many places not under Song control were included in the first 171 chapters of book as if they were part of the Song. Only through close reading of the content of their history and geography can one detect their status beyond the Song administration. A clearer distinction is made between the northwest and the far west, and the latter is only tangentially mentioned in the first 171 chapters of the book. Yue Shi retained the structure of the Tang national geographical treatises, especially Li Jifu's Illustrated Gazetteer, but incorporated updated information about specific places in the Song. In this way, he presented a picture of the Song world that was broader than the actual Song polity, without distorting the historical and geographical information regarding specific places.

This vision of the Song domain and its relation to the northwest changed in the mid-Northern Song. In 1043, Emperor Renzong ordered the compilation of a work titled Essentials of Military Classics (Wujing zongyao 武經總要).⁶³ This imperial order came after a series of Song military defeats by the Xia, which necessitated, in the mind of the emperor, fresh gathering of military techniques and information. A prominent academician Zeng Gongliang 曾公亮 (999–1078) took control of the work. In 1047, when the work was finished, emperor Renzong had signed a treaty with Xia two years earlier. The first installment of this work, in 20 juan, contains a thorough examination of military institutions, technologies, and the geographies of the Song border. Chapter 18 deals with the northwestern border and is further divided into two halves. The first half, titled “Shaanxi Circuit” (Shaanxi lu 陝西路), is a detailed account of the different garrisons and other military establishments within the Song domain on its northwestern border.⁶⁴ The second half of this chapter, titled “the Geography of the Western Foreigners” (Xifan dili 西蕃地理), records information in the area to the northwest of the Song border.⁶⁵ This distinction itself is significant, because it reveals that in Song strategic thinking, a clear border had appeared between what was actually under the jurisdiction of the Song government and what was not. The ecumenical pretension of the Record of the World, which purported to depict the entire known world, was thus eliminated, and a more practical picture of the Song relation with the northwest emerged.

Furthermore, it is worth noting what was included in the chapter on “the Geography of the Western Foreigners.” This chapter opens with Xia Prefecture, the center of the Xia state and its namesake. This entry begins with a discussion of its history starting from the Shuofang Commandery in the Han dynasty. After briefly touching on its history in the Northern Wei and the Tang, the narrative turns to the rule of Tuoba Sigong in the late Tang, who was credited with founding the polity that eventually became the Xia state. The entry then continues to describe the familiar story of Li Jipeng's submission and Li Jiqian's rebellion. Importantly, when it turns to the present, the entry tells its readers that “now [the Xia state] possesses twelve prefectures including Xia, Yin, Sui, You, Ling, Hui, Yan, Lan, Sheng, Liang, Gan, and Su.”⁶⁶ The entry includes distances between Xia Prefecture and its bordering regions. In contrast to the Record of the World, there was no attempt at recording the distance between these places to the Song capitals; only distances to its most immediate neighbors were included. After this summary on the history and present condition of Xia Prefecture, the text then turns to brief discussions of six military establishments within its jurisdiction. These sub-entries contain the location and history of these establishments. It is notable that the names used for some of these places were old Tang names; the Song writers often tell us, at the end of the entry, what their current and, to the Song, “illegitimate names” (weihao 偽號) were at this time.⁶⁷

After this entry, the narrative turns to other places, starting with Yin 銀 Prefecture. The places included in this chapter are, according to the sequence of their appearance, Xia Prefecture, Yin Prefecture, Sui Prefecture, You Prefecture, Ling Prefecture, Yan Prefecture, Sheng Prefecture, Liang Prefecture, Gan Prefecture, Su Prefecture, Gua Prefecture, Sha Prefecture, Yi Prefecture, Xi Prefecture, Anxi Pretectorate, Shule Kingdom, Yutian [Kingdom], Yanqi Kingdom, Shan Prefecture, the Old Wei Prefecture, Lan Prefecture, Hui Prefecture, He Prefecture, Chao Prefecture, Min Prefecture, Kuo Prefecture, Die Prefecture. The places in roman typeface are those mentioned in the first entry on Xia Prefecture as among the twelve prefectures of the Xia state, while those in italics are presumably areas beyond the Xia, at least according to the Song understanding at the time. One can immediately see, therefore, that this chapter is divided into two parts. Until Yanqi Kingdom, the narrative follows in a broadly westward progression, from modern Ningxia along the Hexi Corridor to Xinjiang, introducing the areas under the Xia rule and then those beyond; afterwards, the narrative turns to the southwest and discuss the areas in modern Qinghai and southern Gansu. The land of the “Western Foreigners” in this text thus includes not just the domain of the Xia, which by this time largely overlapped with what I call “the northwest,” but also part of the far west, and some areas to the southwest of the Xia.

The record of the northwest in the Essentials of Military Classics differs in several ways from that found in the Record of the World. There was no pretense about the status of the northwest beyond the Song domain. The line between the Song domain and the area to its northwest that remained blurred or invisible in the Record of the World was made strikingly transparent in the Essentials of Military Classics, by the chapter divisions—the first half of chapter eighteen is on the Shaanxi region in the Song domain, while the second half is on areas further to the northwest beyond the Song realm. The westmost places included in the Essentials of Military Classics are the Four Garrisons of the Tang (Anxi/Kucha, Shule/Kashgar, Yutian/Khotan, and Yanqi/Agni) in modern southern Xinjiang. There was no longer any attempt to record any place further to the west. Under each entry, unlike the Record of the World that focuses on customs and products, the Essentials of Military Classics unsurprisingly centers on military establishments. Overall, the Essentials of Military Classics offers a more pragmatic and realistic treatment of the northwest that largely put it beyond the Song domain.

Such division between the domain under the Song rule and those beyond the Song border persisted in the Treatise of the Nine Regions in the Yuanfeng Reign composed in 1080 in Emperor Shenzong's reign by an editorial team headed by academicians Wang Cun 王存, Zeng Zhao 曾肇 and Li Dechu 李德芻. This text, in ten juan, is much shorter than Yue Shi's geography. Unlike the Record of the World that distinguishes the “four barbarians” into different chapters, or the Essentials of Military Classics that contains chapters on both “the Geography of the Western Foreigners” and one on “Northern Foreigners,” the Treatise of the Nine Regions does not specify the directions of the areas beyond the Song border but lists all of these areas in the last chapter, which is divided into three sections.⁶⁸ The first, “Eliminated or Abolished Prefectures and Military Commandaries” (Shengfei zhoujun 省廢州軍), lists administrative units that once existed in the Song, but were subsequently combined with other units or abolished. The second section includes those Prefectures and Military Commandaries that were “beyond the sphere of civilization” (huawai 化外). Virtually all the areas in the northwest are placed in this category. The last section is that of the “loose rein prefectures” (jimi zhou 羈縻州), which consists exclusively of areas in the south and southwest.

Unlike the Record of the World, which details the history of prefectures in the northwest and their distances to the capitals, or the Essentials of Military Classics, which includes information regarding the military establishments under each prefecture, the Treatise of the Nine Regions is extremely laconic in the information it includes for the northwest. The entry for Xia Prefecture, for instance, only tells the reader its historical administrative level (middle commandery, zhong dudu[fu] 中都督[府]), its old names (Shuofang jun 朔方郡 and Dingnan jun jiedu 定難軍節度), and the three counties under its jurisdiction.⁶⁹ But the relative brevity of the Treatise of the Nine Regions is not the only reason for such a treatment. For areas within the Song domain in the first nine chapters of the book, the editors included much more detailed information, similar to that of the Record of the World. In contrast, the northwest and other areas beyond the Song border appear like an afterthought. The authors considered it necessary to include these places, but the editorial interest clearly focused on the Song domain itself.

Nevertheless, the Treatise of the Nine Regions still marks a transitional moment in the representation of the northwest (and the border regions in general) in Song national geographies. This transition is epitomized by the term huawai (化外). This term, which I translate as “beyond the sphere of civilization,” is a shortened form of the phrase wanghua zhiwai 王化之外, which literally means “beyond the transformative realm of the king.” The “transformative realm of the king” or wanghua, is a key concept in early China that posits a kind of transformative power of the king, “the hearth from which civilization shines forth.”⁷⁰ Therefore, categorizing somewhere as beyond such power is a rather definitive statement about that place's distinct nature from the areas within the wanghua. It is notable that this term huawai does not appear in any of the Tang geographies as a category for places, or the first two Song geographies I surveyed here. In the Treatise of the Nine Regions, however, this term not only appears but serves as an organizing category to which a number of places belonged. Treatise of the Nine Regions thus went one step further in distinguishing the Song domain from areas beyond, including the northwest, not only in geographic, but also in cultural, terms.

What explains such editorial choices? The clue is found in the very title of the book. The preface indicates that Treatise of the Nine Regions follows a long tradition when “ancient kings” collected information on the “domains of the Nine Provinces (jiuzhou fengyu 九州封域) and the numbers of its population.” This tradition continued, the preface tells us, in the form of the Maps of the Ten Circuits (shidao tu 十道圖) in the Tang and as the Maps of the Nine Regions (jiuyu tu 九域圖) in the Song.⁷¹ Therefore, the term jiuyu in the title of the book is a shortened form of jiuzhou fengyu, and the composition of this text followed the tradition that went back to the “Tribute of Yu” (Yugong 禹貢) that divided the civilized world into Nine Provinces.

This view that connects the Song domain (“the Nine Regions”) directly with the ancient “Nine Provinces” mirrors that of Sima Guang, who considered the unification of the Song a thorough recuperation of the “Traces of the Yu.”⁷² Wang Cun, the chief editors of this book, sympathized with Sima Guang's politics and was elevated as a prime minister after the death of emperor Shenzong.⁷³ Both Wang Cun and Zeng Zhao, the other editor of the text, ranked among the “Yuanyou partisans” and had their names listed in the notorious Stele for the Yuanyou Partisans (Yuanyou dangji bei 元祐黨籍碑).⁷⁴ It is not impossible that the Treatise of the Nine Regions reflected this generally less expansionist understanding of what the Song domain ought to be. Even though the text was composed during emperor Shenzong's reign, the more expansionist views held by hawks such as Wang Shao were not reflected in the Treatise of the Nine Regions.

The final geographical treatise I examine here is titled Extensive Record of Geography (Yudi guangji 輿地廣記) by a certain Ouyang Min 歐陽忞.⁷⁵ The text was composed, according to its preface, during the Zhenghe reign (1111–1118) of Emperor Huizong. Nothing is known about the author, and Southern Song cataloguers such as Chao Gongwu even questioned the existence of this Ouyang Min.⁷⁶ The obscurity of the author of the text indicates, at the very least, that Extensive Record was not a geography commissioned by the government (like Record of the World and Essentials of Military Classics) or compiled by prominent Kaifeng officials (like Treatise of the Nine Regions).

Unlike these three geographies that describe, if often in anachronistic terms, the domains of the Song, the Extensive Record is explicitly a historical geography. The first three juan survey the changing political domains from the “tribute of Yu” to the end of the Five Dynasties periods. The book opens with a discussion of the Nine Provinces (jiuzhou 九州) according to the “Tribute of Yu,” where it places the entire Shaanxi region, the northwest, and the far west until the Anxi Protectorate—an obviously anachronistic use of the Tang institutional name—in the region of the Yong Province (Yongzhou 雍州).⁷⁷ The third juan on the Tang follows the fifteen investigative circuits (caifang shi 採訪使) devised in 733 during the reign of Emperor Xuanzong of the Tang.⁷⁸ The prefectures that constituted the “northwest” as defined here are placed under the Guannei circuit and the Longyou Circuit.⁷⁹ There is thus no question that the northwest was considered within both the bounds of the Nine Provinces of King Yu and the domain of the Tang dynasty.

The Song section of Extensive Record, which constitutes the majority of the book, is divided into the four capitals, the twenty-three circuits, and eight areas of “prefectures beyond the sphere of civilization” (huawai zhou 化外州). It incorporated the information of areas outside of the Song domain by attaching them after the chapters on places within the Song that were geographically adjacent. For instance, the northwest is recorded in juan 17, which is titled “the prefectures beyond the sphere of civilization in the Shaanxi Circuit.” This section begins with the Great Anxi Protectorate and the Great Beiting Protectorate and continues with Ling Prefecture, Xia Prefecture, and other prefectures in the area. So the organizing principle is primarily one of rank: the higher administrative units are listed before lower ones.

For each entry, the Extensive Record gives the name and history of the prefecture, and then the name and history of the counties under the jurisdiction of this prefecture. No additional data on distance and products are included. The record of Xia Prefecture, for instance, first recounts its history since antiquity. Unlike the Record of the World, which begins the history of Xia Prefecture by stating that it belonged to the “Yongzhou region in The Tribute of Yu, and the Wei Kingdom in the Spring and Autumn period and the Warring States period,”⁸⁰ or the Essentials of Military Classics that begins this history with Xia Prefecture's status as the Shuofang Commandery in the Han dynasty,⁸¹ the Extensive Record offers a new beginning where Xia Prefecture “was a barbarian land in antiquity” (guwei rongdi 古為戎狄). This phrase guwei rongdi, with some slight variation, appears at the beginning of every prefecture listed in this chapter. It is unclear when this “antiquity” was, but since the entry on Xia Prefecture continues afterwards to record, chronologically, the name of the area under the Qin dynasty, it is safe to assume that this “antiquity” meant the pre-Qin era, most likely the time of the imagined sage kings. In so doing, the Extensive Record creates a non-Han origin for all the areas in the northwest—something that earlier geographies did not claim—and justifies their inclusion into the chapter on areas “beyond the sphere of civilization.”

After this start, the entry on Xia Prefecture continues to describe its history from the Qin until the late Tang, when it was renamed Dingnan Garrison.⁸² This account does not reach the Song, and it is unclear what this place's relation with the Song was from this text. A more direct telling of the Xia Prefecture (and the Xia state)'s history is buried in the subentry on Shuofang County, where it is stated that “in the late Tang, Tuoba Sigong became the commissioner of Dingnan Garrison. Generations of his descendants established themselves illegitimately [as rulers of the area] and continued to act as a thorn [to the Song].” This note, dismissive as it might be, acknowledges that Xia Prefecture was not only beyond the Song domain, but served as a center of a rival regime that posed problems for the Song.

In this way, Extensive Record's treatment of the northwest is inherently self-contradictory. It lists places like Xia Prefecture as a part of the domain attributable to the “Tribute of Yu” in the first juan; but when discussing the specific Xia Prefecture in the Song dynasty section, it claims that this place had been part of a “barbarian land” in antiquity. Such conflicting views reveal an ambiguous attitude toward places like the northwest that once belonged to the Tang domain but were no longer part of the Song. The Extensive Record inherited the stark contrast between the Song domain and those areas “beyond the sphere of civilization” first introduced in Treatise of the Nine Regions. But because it is a historical geographical treatise, it also conveys the idea that the northwest had been a part of previous dynasties. This idea is, historically speaking, incorrect in the case of the Yu and largely accurate for the high Tang. Perhaps because of their prior affiliations with states based in north China, the Extensive Record shows much greater interest in the history of these areas than Treatise of the Nine Regions, even while acknowledging that they were no longer part of the Song.

A broad trend regarding their treatment of the northwest is visible from my survey of these four national geographies that date to the Northern Song. While the earliest geography, Record of the World, maintained the pretense of a Tang dynasty work and placed prefectures in the northwest among other Song prefectures as if they were no different from areas within the Song domain, later works acknowledged the fact that the northwest was no longer part of the Song. Essentials of Military Classics categorized the northwest and the far west as part of the geography of western foreign lands, while Treatise of the Nine Regions and Extensive Record used the term “beyond the sphere of civilization” to describe this region. The records about the Xia Prefecture, for instance, became less detailed and less connected to the Song proper over the course of these four treatises.

This trend continued in Southern Song geographical works. Both Yudi jisheng 輿地紀勝 by Wang Xiangzhi 王象之,⁸³ and Fangyu shenglan 方輿勝覽 by Zhu Mu 祝穆⁸⁴ do not contain any information about the northwest. This omission might be the result of the different functions of these Southern Song geographical works. As Ruth Mostern observed, these works were composed as travel guides for elite Song travelers.⁸⁵ It is thus no surprise that they do not discuss the northwest, an area largely inaccessible to Southern Song elites. But the geography chapters of Songshi compiled in the Yuan dynasty after the fall of the Southern Song, which are not travel guides, are also silent about the northwest. Key places like Xia Prefecture are not mentioned in these works, while Sha Prefecture (Dunhuang), is placed within the chapters on “foreign countries” (waiguo 外國).⁸⁶ The geographical knowledge of the northwest, by the Southern Song, became a historical relic recoverable only from existing Northern Song and pre-Song sources.

Maps

One notable innovation in Song-dynasty geographical knowledge production is the proliferation of maps.⁸⁷ As the opening account of Emperor Renzong's viewing party shows, maps of potentially massive scale of parts of the northwest were produced by the Song court. As Ren Shumin pointed out in an exhaustive examination of the map-making activities sponsored by the Song government on the northwest, foreign envoys from the west who visited the Song court, Song envoys who traveled to the west, and Song military men all collected data for map-making. On the basis of these data, the Song government produced, according to Ren's counting, eleven types of maps, including border maps such as Sheng Du's Map of the Hexi and Longyou region (Hexi Longyou tu 河西隴右圖), military maps such as the one emperor Shenzong gave Wang Shao before his campaign in 1072, maps of conquered land such as the one Wang Shao commissioned after his conquest of the Xihe (熙河, modern southern Gansu) region in the same year, and maps of agricultural colonies (yingtian tu 營田圖).⁸⁸ However, with one exception that I will discuss in detail below, none of these regional maps centering on the northwest has come down to us. Instead, just as in the case of geographical treatises, an abundance of representation is found in national-scale works—thus, they are the focus of my discussion.

The first work I examine is an atlas titled Handbook of the Geography of the Successive Dynasties (Lidai dili zhizhang tu 歷代地理指掌圖) composed between 1098 and 1100 by Shui Anli 税安禮, an otherwise obscure figure who lived in Sichuan.⁸⁹ This atlas includes forty-four maps, arranged mostly in chronological fashion, starting with a map (Map 3) of the domain under King Ku 帝嚳, the son of the Yellow Emperor 黃帝, and ending with that of (Map 35) the Later Zhou dynasty (951–960). The first two maps chart the broad contour of the land of the Hua and the Yi throughout history, while the final nine maps include thematic treatments as well as chronological maps of different eras of the Northern Song. Many of these historical maps were accompanied by textual descriptions and elucidations (bian 辨) that further explain the cartographic choices of the mapmakers.

Guo Shengbo pointed out that much of these texts were copied from earlier sources. The texts that accompanied maps from the Eastern Han (25–220) to the Sui (581–618), for example, are mostly excerpts from the Tongdian. In the Southern Song, a critic characterized this atlas as erudite, but described the accompanying texts as “patchworks created by examination candidates,” in refutation to the widely discredited claim that this atlas was made by Su Shi 蘇軾 (1037–1101).⁹⁰ According to a Southern Song catalogue, Shui Anli planned to submit this atlas to the Song court in the Yuanfu reign (1098–1100), but died before he could do so.⁹¹ This failure to acquire imperial sponsorship might have obscured Shui's name, but it did little damage to the popularity of this atlas. By the Southern Song, the atlas was printed widely and in several revised versions. One of the Southern Song editions, now in the Tōyō Bunko, is preserved largely intact.⁹²

If we ignore the chronology and look at the shape of the maps, it is clear that forty-three of the forty-four maps were modeled on the same base map. Take map 41 (Figure 1) as an example: this map is titled “Map of the Unified Realm of Emperor Taizong (Taizong huangdi tongyi zhitu 太宗皇帝統一之圖).⁹³ A brief passage was carved on both sides of the map that explains the expansion of the Song domain under Emperor Taizong. It cites four events: Chen Hongjin's 陳洪進 (914–985) surrender of Quan Prefecture and Zhang Prefecture and Qian Chu's 錢俶 (929–988) surrender of the Wuyue Kingdom, both in 978, Emperor Taizong's conquest of the Northern Han in 979, and Li Jipeng's offering of Yin Prefecture and Xia Prefecture to the Song in 982. In this sense, the map aligns with the record of the Songshi cited above in that it sees Li Jipeng's submission as the final step in the Song “reunification” project.

Figure 1. Map of the Unified Realm of Emperor Taizong.

The shape of the map that accompanied this text gives expression to such an understanding of the Song domain. A long line begins from the northeast near the Bohai Gulf just south of the land marked by “Kitan” (qidan 契丹); it continues westwards just south of the Great Wall, then crosses the Great Wall toward the northwest before it turns south. After reaching modern western Sichuan, the line swerves sharply to the east and then back again to the west, before it finally connects with the sea at the lower bottom of the map. Unlike other lines in the map—shorter, thinner, and discontinuous—that signify waterways, this long and thick line represents a late Northern Song imagination of the Song border in Emperor Taizong's time.

The land within the Song border is described in more detail, and its rendering on the map is proportionally closer to modern maps. The land beyond the border, however, is squeezed into the edges of the map. Liao territory, for instance, is only a slim belt north of the Song border, are the extensive areas in modern Mongolia and Manchuria under Liao rule were excluded from the map. The treatment of the northwest is more complex. While the Song border was placed to the north of prefectures like Xia and Yin, a result of treating Li Jipeng's symbolic offering of land as tantamount to Song conquest, these areas were clearly separated from the Song mainland with a visually starker boundary, the elongated strip that meandered from the Bohai Gulf to southern Gansu that was meant to represent the Great Wall. It is notable that an additional, discontinuous stretch of the Great Wall was placed north of the Hexi Corridor. On this map, this stretch is represented by two segments, whose shape is identical to the continuous Great Wall to its south. With such symbology, the mapmakers intended to show these northern stretches (which we may call a “northern” Great Wall) were, although geographically distinct, conceptually identical, or at least connected to the continuous “southern” Great Wall.

As Nicolas Tackett has shown, the Great Wall in the Song dynasty was already seen, in cartographic works such as the Handbook of Geography, as “a timeless natural boundary between the Chinese and the northern barbarians.”⁹⁴ Hilde De Weerdt also pointed out that “the Great Wall was in Song times imagined as a marker of empire, an imperialist claim to the territories that fell south of it.”⁹⁵ Thus, in political and military realms, remnants of the old Great Walls also served to mark the border between the Song and the Liao and the Xia. Tackett points out that in at least three locations, the Warring States era Qin Wall served as markers of the Song–Xia border.⁹⁶ Tackett's insight can be applied to the stretches of the Great Walls north of the Hexi region. Unlike in North China, where successive dynasties from the Warring States to the Northern Zhou and Northern Qi continuously constructed defensive walls, the stretch of the wall in the Hexi Corridor was constructed in the Western Han dynasty, especially during the territorial expansion under emperor Wudi (156–87 BCE). Emperor Wudi's army conquered the Hexi Corridor in 121 BCE, and new administrative seats from Wuwei to Dunhuang were established in the following decade.⁹⁷ Along with the expansion of the Han administration to the Hexi region, a long stretch of defensive structures was built to the north of the Hexi corridor. Traditional written sources only vaguely allude to this structure; but archaeological work of the past century, starting from the investigation of Aurel Stein, has confirmed the existence of a roughly continuous defensive structure that dates to the Han dynasty, much of which is still visible today.⁹⁸ On the basis of his extensive onsite investigation, Li Bingcheng drew a map of this stretch of walls (see Figure 2). Because the Song dynasty never controlled the Hexi Corridor, there is little written record about this structure in Song sources. But the location of this northern Great Wall in the Handbook of the Geography leaves little doubt that the cartographers had at least some vague memory of the wall north of Hexi constructed in the Han dynasty. Importantly, they considered this structure culturally significant enough to be represented in these maps, and also similar enough to the southern Great Wall that they were represented by the same symbology.

Figure 2. Han dynasty Great Wall north of the western part of the Hexi Corridor according to Li Bingcheng.

The northwest I discuss in this article, an area that eventually became the domain of the Xia, was largely placed in the area between two parallel stretches of the Great Walls. Within this area, key administrative units from the Feng Prefecture in the east to the Sha Prefecture (Shazhou, another name for Dunhuang) in the west are recorded. Places to the west of Sha Prefecture/Dunhuang were excluded from this map. In addition, names of certain mountains (such as the Sanwei mountain 三危山 and the Jishi mountain 積石山), rivers (the Ruo River 弱水 and the Ze River 澤水), and lakes (the Qinghai 青海 Lake and the Juyan 居延 Lake) are also placed on the map. Compared to the area within the Song border, the scale of this northwestern region was constricted. For instance, the distance on the map between Luoyang (xijing 西京) and Kaifeng (dongjing 東京), in the heart of the Song domain, is about twice as long as the distance between Sha Prefecture/Dunhuang and Su Prefecture/Jiuquan, even though the actual distance in the latter case (about 400 kilometers) is about twice as long as the former (less than 200 kilometers). The reduced scale allowed the mapmakers to include the entire Hexi Corridor into the map. While the administrative units are placed in a manner that is directionally accurate—Sha Prefecture is to the west of Gua Prefecture, which is to the west of Su Prefecture etc.—the locations of the mountains and waterbodies are often far from accurate. The Sanwei mountain, which was located near Dunhuang, was placed far from it in the map, on the edge of the Yellow River. When it comes to the depiction of the northwest, then, the mapmakers were more concerned with inclusion than accuracy. The mapmakers made an effort to include the northwest—an area bounded by two stretches of the Great Wall—in this map that purportedly is about the Song domain.

I describe this map in some detail because its shape is similar to, and in many cases identical with, all but one of the maps in this atlas. Cheng Yinong argued that these maps were created on the template of map 41.⁹⁹ For instance, if we look at the map of the Shang dynasty, the northwest is presented in an identical manner as in map 41, a wildly anachronistic depiction.¹⁰⁰ The inclusion of the northwest, the area “between the Great Walls,” reflects neither the Song border nor most other periods described in this book. Then what period does it represent?

If we examine the relation between the northwest and north-China-based regimes, it is easy to realize that the depiction was based on an understanding of the domain of the Han dynasty. The westernmost place, Dunhuang/Sha Prefecture included in this map marked the western border of the Han Empire.¹⁰¹ The idea that the Han border ends at Dunhuang/Sha Prefecture—the Han thus included the entire northwest as defined in this article—is canonized in the Hanshu 漢書: the northwest was placed in the Geography (dili) chapters, while the far west is discussed in the Western Regions (xiyu) chapters.¹⁰² By including the former and excluding the latter, The Handbook evidently adopted this Han-era border and projected it onto not just the Song dynasty, but the entirety of known history. As a result, in the two administrative maps (junguo tu 郡國圖) of the Han (western Han and eastern Han) included in the atlas, the lines that represented the borders were largely congruous with the frame of the maps themselves.¹⁰³

This adherence to the Song understanding of the Han domain is most transparently visible in the atlas's treatment of the Tang dynasty, the only polity whose control reached further northwestward than the Han. Map 27 in the atlas is titled “Map of the Ten Circuits of the Tang” (Tang shidao tu 唐十道圖) (Figure 3), and it depicts the domain of the early and high Tang.¹⁰⁴ The Song mapmakers were clearly aware of the extent of the Tang western expansion. However, in this map of the Tang domain that represents the early Tang at the height of its military expansion, they did not fundamentally change the shape of the map in the northwest inherited from their understanding of the Han map, but squeezed many place names in an already crowded space. The result is that places over a thousand kilometers away, as in the case of Ting Prefecture (Jimsar) and Anxi (Kucha), are placed right next to each other. Such disregard for scale allowed the mapmakers to represent the Tang domain in a manner that is broadly accurate in the places it includes, but visually similar to that of the Han-dynasty domain.

Figure 3. Map of the Ten Circuits of the Tang.

Against such uniformity that erases historical change, the first map of the atlas stands out. This map is grandly titled “a summary map of the domains of the Chinese and the barbarians in the past and the present” (Gujin huayi quyu zongyao tu 古今華夷區域總要圖) (Figure 4).¹⁰⁵ Unlike all the other maps that depict a specific historical period, this map is purportedly ahistorical (“the past and the present”) and designed to represent the entire known world, the lands of the Hua and the Yi. I call this an ecumenical map. Like all the historical maps, this map includes two stretches of the Great Wall in the northwest, and the places recorded in this area and their spatial arrangement are also largely the same. What changed was the area further to the west. Unlike the early Tang map that squeezed places from modern Xinjiang and beyond into the left edge of the map, this map is more faithful to the enormous scale of the far west, depicting places like Turfan (represented by the character Xi 西 as in Xi Prefecture 西州), Kucha, and Khotan in a much more accurate manner.

Figure 4. A summary map of the domains of the Hua and the Yi in the past and the present.

Why did the mapmakers feel the need to include this unique map in the atlas? And what explains their different treatment of the northwest (and other border regions) in this map in contrast to all other maps in the atlas? I think the answer lies in the term huayi. Unlike all subsequent maps in the atlas that were designed to represent the change of the domain of the Hua, this first map reveals a much broader world in which the domain of the Hua was surrounded by that of the many different Yi. This distinction is clear from the map itself, which includes three passages of texts in rectangular cartouches. The two on the upper left corner explain that emperor Taizong of the Tang established the Xi (Turfan), Yi (Hami) and Ting (Jimsar) prefectures, while emperor Gaozong further stationed the “Four Garrisons”: Kucha, Yanqi/Agni, Khotan, and Kashgar. The cartouche in the lower left contains a much longer passage that explains the meaning of the term xiyu 西域. According to this text, largely abbreviated from the xiyu chapter in Hanshu, this region included the land west of the Yumen Pass and east of the Pamirs (Congling). In effect, this passage explains that the two areas noted by the two shorter passages on the upper left corner can be grouped together into a broader region known as the xiyu. The important distinction between this area (the “far west” as I define it), which largely maps onto modern Xinjiang, and the northwest is that the latter is not explained with additional textual notes in cartouches. Put in the context of all the other maps of the atlas, I think it is reasonable to assume that the area east of the northern Great Wall was considered Hua, while the area further west, the xiyu region, was considered Yi. The mapmakers, we might surmise, anticipated that only the Yi area needs additional textual explanation, because the Hua area is well-represented and well-explained in the other maps in the atlas.

My analysis reveals a number of features of its mapmakers’ understanding of the northwest in this atlas. First, they were aware of, and did not refrain from showing the fact that the Song border did not extend to the west to the extent of the Han and the Tang dynasties. They nonetheless inflated the Song domain by drawing the borderline according to the moment—a fleeting moment in the grand scheme of things—when several of the prefectures in the Xia were nominally part of the Song. Second, even though the Hexi region was never part of the Song, it was consistently included in all the Song maps, and all the historical maps before them. This decision was evidently the result of using the domain of the Han dynasty as a default base map for the entire atlas; the use of the northern Great Wall visually underscored this decision. Third, the Song mapmakers were aware of the extent of the Tang domain but refrained from depicting it with the same drive for accuracy they exhibited for the Han domain. They evidently considered the area west of Dunhuang and east of the Pamirs as the land of the Yi, an area that is not included in most historical maps and requires explanation when it is included in the ahistorical, ecumenical map. The area further west, such as the lands of Persia, although well known in the Tang and the Song, was excluded from this atlas. These areas did not belong even to the Yi but lay beyond the realm of the known world to these Song mapmakers.

After this atlas, the earliest preserved map of the Song domain is the “Map of the Governance of the Nine Regions” (Jiuyu shouling tu 九域守令圖) (Figure 5). This map was carved on stone in 1121, and was placed in the Confucius Temple in Rong County 榮縣 in southern Sichuan. Based on the direction of the Yellow River and the names of prefectures, Zheng Xihuang argued that this map was modeled on another map made between 1094 and 1099, but underwent a final tweaking in 1121.¹⁰⁶ The colophon to the map is now almost completely destroyed. But its text is recoverable from Qing dynasty local gazetteers. The colophon first cites the chapter on geography in the Hanshu and charts a history of the all-under-heaven from the Yellow Emperor to the Han dynasty. Then it describes the division at the end of the Han, the unification of the Sui and the Tang, the political fragmentation of the Five Dynasties, and the reunification of all-under-heaven in the Song dynasty. After this account about history, the colophon then explains that there were several extant versions of the “Map of Governance,” but none was able to correctly mark the prefectures that had undergone many changes since the founding of the Song more than a century earlier. To correct these mistakes, this map was made on the basis of the “Treatise of the Nine Domains” (jiuyu zhi 九域志), but with corrections from other texts.¹⁰⁷ Zheng Xihuang suggested that this “Treatise of the Nine Domains” must have been the “Treatise of the Nine Domains in the Yuanfeng Reign” finished in 1080. This assessment is in line with a passage from the colophon, which specifically points out that the areas “recovered in the Shaosheng reign (1094–198)” were not recorded in the “Treatise of the Nine Domains,” and therefore not included in the map. Therefore, although carved in the early twelfth century, this map reflected the official Song thinking about its own domain in the Yuanfeng reign, before the conquest in the northwest in the 1090s.

Figure 5. Map of the governance of the Nine Regions.

It is thus no surprise that this map's depiction of the northwest is largely in line with the description of the “Treatise of the Nine Domains in the Yuanfeng Reign” discussed above. As I mentioned, this “Treatise” treats much of the northwest, the areas within the Xia domain, as “beyond the sphere of civilization.” In a similar fashion, the “Map of the Governance” does not include any places considered “beyond the sphere of civilization” in the Treatise. Remarkably, it draws a thin long borderline that runs from the southwest in what is modern Sichuan to the northeast, reaching the Bohai Gulf. On the original map, this line can easily be confused with the mountain lines and rivers that were also drawn in the form of long lines. But the cartographers made an effort to avoid such confusions. For every river, they copied the name of the river and placed it at the origin of the river, in a rectangular box. So there is no doubt that this lone line does not denote any river. The many mountains depicted in this map consisted of lines that might look a bit similar, but these lines are uniformly disjointed and modeled on how mountains are drawn in Chinese ink paintings. In contrast, it is clear that this line that runs through the upper edge of the map is not meant to depict any mountain, but the long and largely continuous borderline between the Song and its northern and western neighbors. North of this long line, the areas not under the Song rule are not marked by any characters for any specific locations, but simply short, twisting strokes that signify mountain ranges. The intention of the mapmakers is rather evident: they were not interested in representing anything concretely beyond the Song domain.

This overall principle of carefully describing areas under Song domain but ignoring anything beyond resulted in several curious features in the depiction of the northwest on the map. First, as Zheng Xihuang pointed out, this map is remarkably accurate in delineating the courses of the major rivers such as the Yellow River, the Yangzi River, the Huai River, the Pearl River, and their main tributaries. The only part that seems wildly inaccurate is “the part of the Yellow River that runs through and turns in modern Inner Mongolia and Ningxia.”¹⁰⁸ As anyone who has access to a modern map can immediately detect, the first third of the Yellow River is stretched further to the north and the west in this map, significantly expanding the territory south of the Yellow River. Why was this singular inaccuracy introduced? It is possible that there was a lack of knowledge about the course of the river at the time when this map was made. Indeed, as we have seen, geographical knowledge about the northwest seems to have stagnated after the early Song. But this is unlikely, in my opinion, because, as we will see momentarily, other Song dynasty national maps do not make this mistake. It is more likely that the inaccuracy was an issue of aesthetics rather than knowledge.

By “aesthetics” I mean the general balance of the shape of the map. If we refocus our gaze on the northwestern section of the Song border, we can see that it did not follow the course of the Yellow River. Instead, from east to west, it first intersects with the river in modern Inner Mongolia, after which it lies south of the river for much of the Ordos region; then north of Lan Prefecture, the border crosses the Yellow River again, this time to the north, and encompasses a small region north of the river into the Song domain; afterwards, the border runs to the southeast of a mountain (the Jishi mountain), before it crosses the Yellow River for a third time, to the southwest of Xining. Here we can observe a closely intertwined relationship between the Yellow River and the Song's northwestern border. When we compare the Map of the Governance with a modern map of the Song–Xia border, it is clear that the Song mapmakers artificially extended the border—and along with it, the Yellow River—to the northwest, visually enlarging the Song domain and marginalizing the areas under Xia rule.¹⁰⁹ This need for visual marginalization, I think, explains the strange shape of the Yellow River more than technological incompetence.

Such manipulation of the shape of the Yellow River makes the map more Song-centric than most other Northern Song maps. This vision is in line with the Treatise of the Nine Regions, its source text, in that neither were interested in anywhere beyond the political borders of the Song. But unlike the Treatise of the Nine Regions that at least includes the names of places outside of the Song, the Map of the Governance goes one step further by eliminating any identifiable place names beyond the Song border while representing these areas with only mountains and rivers. It is notable that the mapmakers did not feel the need to include the Great Wall in this Song-centric map. The northwest, like other areas beyond the Song border, almost vanishes entirely from this map.

The last two national maps I investigate are perhaps the most famous maps in the history of Chinese cartography. A stone stele from 1136 includes one map on either side. The recto is the Yuji Tu (禹跡圖 Map of the Traces of King Yu), which was most likely created in 1080s, and the verso is the Huayi Tu (華夷圖 Map of China and the Barbarians), which was created between 1117 and 1125.¹¹⁰ Their names reveal the basic nature of and difference between the two maps: the first one, recording the traces of the legendary King Yu, charts what the Song cartographers considered to be part of the Hua/Chinese domain; whereas the second one outlines both the Chinese (Hua) and the foreign (Yi) lands.¹¹¹ This difference mirrors the distinction between the historical maps that describe the changing shape of the land of the Hua and the ecumenical map that combines both Hua and Yi areas in the atlas discussed above.

As Hilde De Weerdt pointed out, the Yuji tu follows a long tradition of mapping the Chinese empire that relies partly on the textual references from “The Tribute of Yu” (Yugong 禹貢) chapter of the Shangshu 尚書. The earliest example in this genre was produced by Pei Xiu 裴秀 (224–271) in the Jin dynasty (265–316). Pei's work, titled “A Map of the Regions of the Tribute of Yu” (Yugong diyu tu 禹貢地域圖) is, according to De Weerdt, “the presentation of a new and up-to-date interpretation of the geography of empire represented in the classical text.”¹¹² The same can be said about the Yuji tu produced in the Song dynasty (see Figure 6)—the cartographers also had to grapple with the relation between the classical text, The Tribute of Yu and geographical and political realities in the Song. The term “traces of Great Yu” can be a reference to a passage in the Zuo Commentary, as De Weerdt demonstrated. But it is also possible that the use of this term is influenced by, or at least produced in the same intellectual milieux as, Sima Guang's notion of the “traces of Great Yu” discussed earlier.

Figure 6. Line drawing reproduction of the Yuji tu.

Specific to the northwest, the mapmakers’ task was to decide where the traces of King Yu should end. At the very upper left edge of the map that denotes the northwest, we see three place names: Sha 沙 that refers to Sha Prefecture/Dunhuang, Flowing Sands (Liusha 流沙), and the Sanwei Mountain (Sanwei shan 三危山) (See Figure 7). While Sha denotes a Tang–Song era administrative unit, both “Flowing Sands” and “Sanwei Mountain” are names from The Tribute of Yu. The term sanwei first appears in the context of the description of the Yongzhou 雍州 area; it is identified as the area to which a group of people called the Sanmiao 三苗 was banished by King Shun 舜.¹¹³ Sanwei appears again in the account of the Black River (heishui 黑水), which was said to originate in sanwei and reached the Southern Sea.¹¹⁴ Although earlier references to sanwei often referred to it as a general area, by the Tang dynasty, it was increasingly identified with a specific mountain near Dunhuang.¹¹⁵ The term liusha appears in the penultimate sentence of The Tribute of Yu, which describes the work of King Yu as reaching the ocean to the east and the “running sands” to the west (xibei yu liusha 西被於流沙). As the Qing scholar Hu Wei shows in his erudite Yugong zhuizhi, which collects generations of commentaries on this brief text, the term liusha was generally read as meaning either somewhere in Zhangye (Gan Prefecture) or in Dunhuang (Sha Prefecture).¹¹⁶ For instance, Record of the World during the Taiping [xingguo] Reign records a place named liusha in both Gan Prefecture and Sha Prefecture.¹¹⁷ But there are other interpretations of this term. Shen Gua, the renowned Song polymath, suggests that a desert he once traveled to in northern Shaanxi near the Wuding River might have been the liusha.¹¹⁸ The creators of the Yuji tu made the decision to locate both sanwei mountain and liusha near Sha Prefecture, and placed these three names right next to each other, thus rejecting alternative understandings of the locations of liusha. They then used this cluster of places as the northwestern edge of this map, and end of the “traces of Great Yu.” Thus, the Yuji tu's assessment of the domain of the “traces of Great Yu” is more expansive than that of Sima Guang, who claimed that the “traces of Great Yu” stopped at the Song borders under Emperor Taizong.

Figure 7. Upper left section of the Yuji tu.

Unlike the atlas, the Yuji tu carved on stone is renowned for its remarkable accuracy, partly achieved by the use of the scaled grid. The decision to use Sha Prefecture/Dunhuang as the northwestern border of the map, combined with the attention to proportionality and accuracy, meant that there is a large empty space in the southwest (see the lower left corner of Figure 6). Such disregard for a balanced aesthetic, usually central in Song national maps, underscores the necessity felt by the mapmakers of including the northwest, from the Song border to Dunhuang, into the map that describe the traces of King Yu.

The Huayi tu carved on the other side of the Yuji tu depicts the heartland of the Chinese civilization (the Hua) surrounded by various areas of the Yi in all four directions (Figure 8). Here, in contrast with the Yuji tu, the proportion was greatly distorted to fit more places onto the map, especially in the northwestern direction. Cartouches containing texts were placed next to the lands of the Yi, explaining their history and their relations to the land of the Hua. The Hua land, on the other hand, is only marked by toponyms without such textual description. Unlike the Yuji tu that treats the northwest as a part of the land of King Yu, the Huayi tu gives the northwest—divided into the land of the Xia state (Xiaguo 夏國) and the “five prefectures of Liang-Gan” (Liang-Gan wuzhou 涼甘五州)—the text-in-cartouche treatment, implying that it was considered the land of the Yi rather than the Hua. This practice is identical with the “Summary Map” in the atlas. In the latter case, space constraints did not allow the mapmakers to include as many cartouches, but the same information is contained in the atlas, in the form of “elucidations” (bian 辨), on the following page.¹¹⁹

Figure 8. Line drawing reproduction of the Huayi Tu.

Cao Wanru compared the texts in cartouches in the Huayi tu and the “elucidations” in the “Summary Map,” and concluded that they are largely similar, and in many cases identical.¹²⁰ One reason for such similarity is that both maps were based on a now lost map created by the Tang dynasty cartographer Jia Dan 賈耽 (730–805) called “Map of China and the Barbarians within the Ocean” (Hainei huayi tu 海內華夷圖).¹²¹ We know that both Song maps simplified Jia's map, because both tell us that Jia's map records “hundreds of states” (shubai yu guo 數百餘國), from which they only retained the “renowned and well-known ones” (zhuwen zhe 著聞者). On the other hand, the Huayi tu and the “Summary Map” must have also added things. The notes on the northwest must have been added only in these Song maps for two reasons: First, they concern an area that was considered a part of the Tang, so one can imagine that, if Jia's map followed the same editorial principle of not remarking on lands under the Tang rule, there would not have been any textual notes on the northwest. A second and more straightforward reason is that the notes in the Song maps refer to events happening after the Tang dynasty.

The note on the Xia state mentions three incidents in the early history of the Xia state: the late Tang bestowal to Tuoba Sigong of the Tang imperial surname Li, the early Song bestowal of the Song imperial surname Zhao, and Yuanhao's “usurpation” of imperial pretentions in 1038.¹²² This account of the history of the Xia frames it entirely according to its relation with the major dynasty at the time (Tang and Song) in China proper.

In contrast, the note on “the five prefectures of Liang and Gan,” the western half of the northwest I discuss in this paper, recounts a much longer history:¹²³

The five prefectures (beginning with) Liang and Gan were (the area where) Emperor Wu of Han took the land of the Hunye and Xiutu kings and established the Four Commanderies of Hexi. To the south it shields against various Qiang peoples; holding the two passes, the right arm of the Xiongnu is cut off and the Western Regions are connected. Tribute came in uninterruptedly since the early Song.

凉、甘五州即漢武時取渾邪休屠王地置河西四郡，南隔諸羌，據二關，斷匈奴右臂，以通西域。宋初以來朝貢不絕。

The story told here about the history of the “five prefectures” is very similar to that told by modern historians.¹²⁴ But there is one very significant divergence: the most immediate predecessor of the Song, the Tang, was not mentioned at all; and the transition from being part of the Han to being beyond the control of the Song is indirectly expressed: whereas during the Han the area constituted the “Four Commanderies of Hexi,” an official part of the Han Empire, in the Song the relation became a tributary one (“tribute came in uninterruptedly since the early Song”).

The juxtaposition of the Yuji tu and the Huayi tu, carved on the same stone, conveys the ambiguity of the transitional status of the northwest clearly: whereas in the Yuji tu, the mapmakers made an effort to include the northwest in the domain that recreated the “traces of Great Yu,” in the Huayi tu it was clearly distinguished from the more properly Song regions and marked as belonging to the lands of the Yi. The two notes that accompanied the northwest explain how they once were parts of a north-China-based regime (the Han and the Tang), but had by Song times transitioned into a non-Hua territory. If we use the terminology introduced by Song geographies surveyed above, we might say that the mapmakers of the Huayi tu saw the northwest as “beyond the sphere of civilization.” Yet, like the maps in the atlas, a second segment of northern Great Wall, which runs in almost perfect parallel to the western half of the southern (and main) Great Wall, was also included in the Huayi tu. If the Great Wall in the Huayi tu, as Peter Bol remarked, serve to signify the “division between ‘us’ and ‘them,’”¹²⁵ then the northwest represented a transitional space where such division was blurred.

Finally, there is one Song-era map specifically about the northwest that survived (Figure 9). This map is titled “A Topographical Map of the Western Xia” (Xixia dixingtu 西夏地形圖). It has traditionally been included in a version of the collected writing of Fan Zhongyan 范仲淹. The earliest extant version of Fan's collected writing that includes this map dates to 1608. Huang Shengzhang and Wang Qianjin's meticulous study shows that this map included the new place names created all the way up to 1108, but did not incorporate any new names after that. On this basis, they date the map to 1108.¹²⁶

Figure 9. The shape of the land of the Western Xia.

This map centers on the Xia state, which borders four other states in four directions. The map notes the Xia borders with the Uyghurs (huihu jie 回鶻界) and the Tibetans (xifan jie 西蕃界) to the west, the Tatars (dada jie 達靼界) to the north, and the Kitans (qidan jie 契丹界) to the east. The border to the south with the Song is treated differently: the mapmakers use a long and continuous line that runs from the west bank of the Yellow River westward, across the entire map, into the Tiandu mountains (tiandu shan 天都山), where the line is described simply with one character “border” (jie). In addition to being clearly delineated, the Song border also differs from the other three borders in that it is prominently placed inside the map, rather than on its edges. In this way, the mapmakers created a map that is centered on the Xia, but with an emphasis on its southern border with the Song.

Within the Xia border, the most prominent feature is the Yellow River that, drawn with two parallel lines, runs from the lower left corner near Lan Prefecture in a northeastward direction, taking many tributaries and bending sharply southward when it reaches the edge of the map. On both sides of the Yellow River, many roads, represented with dotted lines, connect various Xia places. Among these roads, two—“the road by which the Xia bandits invade our border” (Xiazei fanbian zhilu 夏賊犯邊之路)” and “the postal road for [delivering] state letters (guoxin yilu 國信驛路)—signify the ways the Xia domain was connected to the Song. To the west of the Yellow River, another similar set of parallel lines run from the Uyghur border at the center-left of the map northeastward into the northern mountains. They must have signified the Ruo (弱水) River that was believed to have been near Dunhuang.¹²⁷ The areas on both sides of this river, from Liang Prefecture in the east to Sha Prefecture (Dunhuang) to the west on the edge of the map, is the Hexi Corridor in modern Gansu. Compared to the southern and eastern sections of the Xia domain, which are fairly accurately represented, the Hexi Corridor shrinks in scale, no doubt so that it can be fitted into a map without unnecessarily extending it. In the Hexi Corridor, we also do not find the tightly squeezed sets of military establishments and postal stations that are present in the southern and eastern sections of the map.

Names of major administrative units in both the Song and the Xia domains are placed in squares, while places of lesser importance are simply copied onto the map without the square. Near to the Xia Prefecture, for instance, a note is added to say that “this was the Shuofang Commandery” (ji Shuofang jun 即朔方郡). Another note, near Helan Garrison (Helan jun 賀蘭軍), explains that it used to be “Feng'an Garrison of the Tang.” Evidently, the mapmakers felt the need to explain the relation between current Xia names of places and their older, possibly Han or Tang dynasty names. On the other hand, the map also does not exclusively present the Xia land as rooted in the long history within these dynasties, but reveals its non-Han nature as well. Especially among the names not in squares, we find many that might have been transcriptions of Tangut names, such as Chiluo Postal Station (chiluo yi 吃囉驛), Qiduo Postal Station (Qiduo yi 啟哆驛), and Huyi luoman 鶻移啰漫, the meaning of which I could not determine. These names never appear in national geographies surveyed above, and they must have more closely reflected the names used by non-Chinese-speaking peoples in the northwest.

The makers of this map attempted to chart the actual roads through the Xia land and took care in recording non-Chinese toponyms. Unlike the large-scale maps surveyed above, which predominantly revealed how Song literati imagined the place of the Song state in history and the contemporary known world, this Xia map was thus created for a more practical purpose. In this sense, it might have resembled the maps emperor Zhenzong viewed at the palace in 1001. The northwest region was entirely included in this Xia map. Importantly, it was separated from the Song region by a border line and a series of forts (bao 堡 and zhai 寨). In this practical map made for military and diplomatic use, we do not see the use of the Great Wall.

In the previous section, I mentioned that the Southern Song geographies essentially ignore the northwest. Southern Song maps also reflected this growing distance. Southern Song cartography also continued to develop on the basis of Northern Song models.¹²⁸ It is impossible to survey this development here due to space constraints. So I will simply mention the most renowned Southern Song map, the Dili tu 墜理圖. It was created and submitted by Huang Shang 黃裳 (1146–1195) to emperor Guangzong (1147–1200) in 1190 and carved on stone in 1247 (Figure 10). As Hilde De Weerdt suggests, this map reveals a strong irredentist sentiment regarding the Sixteen Prefectures, which is partly revealed in the colophon carved under the map.¹²⁹ As a result, the northeast in the map is depicted with much greater detail and attention to proportionality than all of its Northern Song predecessors. In contrast, the far west is, unlike the Huayi tu, entirely erased from this map, and only one place from the far west, namely Kucha (Qiuci 龜茲), made it into this map, albeit placed erroneously next to Liang Prefecture. The northwest is represented in a way similar to the Huayi tu.¹³⁰ A long structure representing the Great Wall separates the northwest from the Song domain. But there is a key difference: while Northern Song maps like the Huayi tu and the maps in the Handbook of the Geography consistently placed a further stretch of the Great Wall to the north of the northwest region, no such “northern” Great Wall is found in this Southern Song map, which drew only one singular and continuous Great Wall. If, as Tackett argued, the Great Wall was primarily a cultural construct, then the cultural ambiguity revealed in the Northern Song maps disappeared in the Southern Song. The northwest, by the Southern Song, was securely placed beyond the Great Wall.

Figure 10. Line redrawing of the Dili tu created in the Southern Song.

Conclusion

This article examines the shifting ways in which the “northwest” was discussed and visualized in the political discussions, geographical treatises, and maps from the Northern Song. The central question that all of these works had to grapple with was that of how to present the relation between the Song domain and the northwest, an area that was recently part of the Tang Empire but was then mostly governed by the Xia state.

Northern Song representations of the northwest offer a spectrum of answers to this question. At one extreme end, the early Song geographical compendium Record of the World treats areas in the northwest as if they were still part of the Song, recording the same kind of information and organizing them into the same categories as other parts of the Song domain. Only after digging into the details of each record, such as their distances to the Song capitals, can we observe the actual status of these places outside of the Song realm. On the other end, Sima Guang considered the Song realm at the end of Emperor Taizong's reign as the reunification of “Traces of the Great Yu,” which excludes the northwest entirely. Along the same line, geographical texts such as the Treatise of the Nine Regions treat the northwest as “beyond the sphere of civilization,” and naturally beyond the Song realms too. This view was visualized in cartographic works such as the Map of the Governance of the Nine Regions, where the northwest was pictured as wilderness without any place worth recording.

Most Northern Song works, however, offered answers to this question that fell somewhere between these two extremes. Almost all recognized that the northwest was not part of the Song state. The more expansionist views, such as the one expressed by Wang Shao, claimed that the northwest ought to be incorporated into the Song on the basis of their status as part of the Han and the Tang dynasties. This view imagined an idealized world that ought to be governed by the Song that resembled the depictions of the Yuji tu, which also saw the northwest as an integral part of a trans-dynastic entity that we might call “China.” Such mismatch between ideal and reality is discernable in geographical treatises such as the Extensive Record of Geography, which lists the northwest as within “the traces of Great Yu” but also calls it a barbarian land. It is also visible in the depiction of the northwest within two stretches of the Great Walls in the Huayi tu and the maps in the atlas the Handbook of the Geography. This status of “between the Great Walls” reflects the ambiguous state of the northwest's relation with the Song in the Song imagination.

When we consider these three modes of representation together, it becomes clear that they did not follow the vicissitudes of politics in any identifiable manner. Emperor Shenzong's reign is known to be a broadly expansionist time, but the geographical treatises that bore his reign name reflected the politics of Sima Guang better than that of emperor Shenzong and Wang Anshi. Such complexity of vision is understandable when we consider the authors of these works. While political treatises were produced by some of the most renowned and active officials at the Song court, often explicitly targeting the emperor's attention, the authors of geographies were generally much lower in status. Some of these authors, such as Ouyang Min, were obscure figures. The diversity of the Song intellectual world meant that at any given time, regardless of the official policy at the court in Kaifeng, there existed a variety of views among literati-officials across the Song state. Nonetheless, it seems possible to discern a very rough trend, especially in geographical treatises: as the Xia consolidated its hold on the northwest, it became harder and harder to maintain any pretense of the Song irredentist claim to the northwest. By the Southern Song, in almost all three kinds of discourse, the northwest was securely excluded from the normative Song domain.

My investigation of the northwest offers a good chance to compare it to how other parts of the Song borderlands were imagined in the Northern Song. The “Sixteen Prefectures” region in the northeast, for example, was seen in remarkably similar terms in these Northern Song works. Like the northwest, the “Sixteen Prefectures” were seen as “beyond the sphere of civilization” in the Treatise of the Nine Regions and Extensive Record of Geography; they were excluded entirely in the “Map of the Governance of the Nine Regions,” but included in the Yuji tu, and placed south of an imagined Great Wall in the Huayi tu and the Handbook of the Geography. Yet, unlike the northwest, which featured less and less prominently in the political discourse of the Southern Song, the “Sixteen Prefectures” became a rallying point for irredentist fervor. Wang Jinping showed that the “Sixteen Prefectures” were so important in the Song imagination both because of its strategic location between North China plains and the steppe, and the sense that Han people (Han'er 漢兒) lived there.¹³¹ The northwest, on the other hand, was occupied by a mixture of different ethnic groups, and was quite distant from the heartland of the Song. Such distinctions help us understand their different treatments in Song border policies and political imagination.

On the other hand, the far west—modern Xinjiang and beyond—was treated very differently from the northwest. Very few Northern Song works had any serious interest in this vast region. The Record of the World during the Taiping [xingguo] Reign is the only work that included detailed records about the far west. Other geographical treatises, such as the Treatise of the Nine Regions and the Extensive Record of Geography, only include very perfunctory information on this area. In most Northern Song maps, the far west was either entirely removed or included in garbled ways that defy proportionality. The clear distinction between the northwest and the far west is visualized in many maps by a northern stretch of the Great Wall.

That this northern Great Wall might have been Song imaginations of a Han dynasty wall speaks to a final issue I would like to comment on. Among the works surveyed in this article, the most popular view would concede that the current domain of the Song state was smaller than what it ought to be. But what did this greater, imagined domain look like? My close look at the northwest shows that the idealized domain for the Song literati-officials looked closer to the political territory of the Western Han, and much smaller than the Tang, which extended regular governance into eastern Xinjiang. This scope of the Song imagination might have something to do with exegetical traditions, developed during and after the Han dynasty, that associated the borders described in the Tribute of Yu with Western Han borders. When names like liusha and sanwei became concrete places locatable to Sha Prefecture/Dunhuang, it was easy to see them as where the imagined Chinese world should end.