In December 755 a frontier commander named An Lushan turned his army south from Fanyang, near modern Beijing, and marched on the Tang empire’s eastern capital at Luoyang. Within weeks he held the city. Early in 756 he proclaimed himself emperor of a rival Great Yan dynasty. By midsummer the western capital, Chang’an, had fallen; the ageing Emperor Xuanzong was fleeing toward Sichuan; and the most cosmopolitan court in medieval Eurasia had been broken open by one of its own generals. The fighting that followed lasted until early 763. Chinese historians still call it the An–Shi Rebellion, after An Lushan and his lieutenant Shi Siming. It did not end the Tang dynasty, which limped on until 907, but it ended the Tang that the world remembers — the open, wealthy, self-confident empire of the early eighth century — and it left later Chinese states with a standing suspicion of powerful military governors that shaped Song institutions for centuries.
This article sets out what happened, why a general of mixed Sogdian and Turkic background could concentrate so much force, how the war actually unfolded, and what the aftermath meant for China’s politics, demography and culture. It sits beside other huaxia.click history pieces on imperial turning points, from the Qin unification and Zheng He’s voyages to the examination system and the opium wars, but it does not retell those stories. The aim is concrete: dates, places, institutions and the limits of the numbers that get repeated about this war.
A Frontier Empire at Mid-Century
By the 740s the Tang court under Xuanzong (r. 712–756) had built a system of large frontier commands under military governors, or jiedushi. The old militia arrangements of the early dynasty no longer matched the scale of wars against Tibetans, Türks, Khitans and other neighbours. Professional garrison armies, often recruited from non-Han steppe and oasis peoples, held the lines. The capital enjoyed the fruits: silk, horses, music, Buddhist and Central Asian fashions, and a city — Chang’an — whose wards held something on the order of a million people when the suburbs are counted. Luoyang, the eastern capital, anchored the canal and Yellow River traffic that fed the Guanzhong heartland.
That prosperity had a price. The same policy that made the frontiers defensible also put tens of thousands of veteran troops under a handful of commanders far from the throne. After 736 the chief minister Li Linfu preferred to appoint non-Chinese generals to those posts. His reasoning, as later historians reconstructed it, was political rather than ethnic: a Chinese general who won glory might return to court as a rival; a Sogdian or Turkic commander was less likely to build a faction among the literati. Whether or not that calculation was sound, it opened a career path for men such as An Lushan.
An Lushan was born in 703 in Yingzhou, in what is now Liaoning. Britannica’s biography, written by the Tang specialist Edwin Pulleyblank, gives his ancestry as Sogdian on the father’s side — the surname An linked to Bukhara — and Turkic on his mother’s. His given name is a sinicised form of an Iranian word for “light.” After the Eastern Türk polity fractured, his family took refuge in Tang territory. He first appears in the annals in 736 as a reconnaissance officer who lost his force through rashness and was condemned to death, then pardoned. From 742 he held independent commands on the northeastern frontier. By the early 750s, with Li Linfu’s patronage and the favour of Xuanzong and the consort Yang Guifei, he held three provinces at once — Pinglu, Fanyang and Hedong — with garrisons often put near 160,000 men. He was the most powerful general in the empire.
Court culture helped him. An Lushan played the buffoon for Xuanzong’s amusement; Yang Guifei went through a mock adoption ceremony with him. Later legend embroidered those scenes into a love affair. What mattered politically was simpler. He had the emperor’s trust, the northeast’s armies, and, after Li Linfu’s death in 752, a deadly rivalry with Yang Guozhong, Yang Guifei’s cousin, who was trying to inherit Li’s dominance at court. Yang could harass An’s clients in Chang’an. He could not break An’s grip on the frontier. By late 755 An concluded — or claimed — that he had a secret imperial order to destroy Yang Guozhong, and he moved.
From Fanyang to the Capitals
On 16 December 755 An Lushan’s army left Fanyang. He treated surrendered Tang officials with enough respect that more joined him. Moving along the Grand Canal corridor, he took Luoyang on 18 January 756. On 5 February, the first day of the lunar new year, he declared himself Emperor Xiongwu of Great Yan. The Tang’s best remaining hope was geography: the Tong Pass approaches that protected Chang’an from the east. Loyal forces under Gao Xianzhi and Feng Changqing held strong ground there until court intrigue and the eunuch Bian Lingcheng secured their execution. Their replacement, Geshu Han, was pushed by Yang Guozhong into fighting on open ground. On 7 July 756 the Tang army was routed. The road to Chang’an lay open.
Xuanzong fled west toward Sichuan with much of his household. At Mawei Station, not far from the capital, his hungry escort mutinied. They killed Yang Guozhong and demanded the death of Yang Guifei. The emperor consented; she was strangled. The scene became one of the most worked motifs in later Chinese and Japanese art and poetry — Bai Juyi’s “Song of Everlasting Sorrow” is only the most famous treatment — but at the time it was a military crisis, not a romance. The crown prince, Li Heng, broke away north to Lingwu (then Lingzhou) in what is now Ningxia and was proclaimed Emperor Suzong on 12 August 756, while Xuanzong, still on the road to Chengdu, abdicated in form. An Lushan’s forces entered Chang’an. Much of the population had already fled. The city that had been Tang’s showcase was looted and emptied of its court.
An himself stayed in Luoyang. By then he was seriously ill — later sources speak of blindness, skin disease and violent paranoia — and on 29 January 757 he was murdered by attendants with the connivance of his son An Qingxu. The rebellion did not end. It passed to An Qingxu, then to Shi Siming, then to Shi Chaoyi’s short and hopeless reign, each succession marked by murder inside the Yan camp. That internal bloodletting mattered as much as Tang counter-attacks.
Stalemate, Allies and the Long Suppression
Two operational facts shaped the middle years of the war. First, the rebels never securely broke into the rich Yangzi and south China basin. The Battle of Yongqiu in spring 756 and, above all, the Siege of Suiyang from January to October 757 — where Tang defenders under Zhang Xun held out through famine and, according to the sources, cannibalism — blocked the road south long enough for the dynasty to regroup. Second, Suzong’s government borrowed strength from outside the old Tang order. Contingents from the Uyghur Khaganate, then the dominant power on the eastern steppe, and several thousand Abbasid mercenaries sent in the mid-750s, fought alongside Tang generals such as Guo Ziyi and Li Guangbi. With Uyghur help the Tang retook Chang’an and Luoyang in late 757, but the rebel heartland in Hebei remained.
Shi Siming recovered Luoyang after killing An Qingxu in 759, then was himself killed by his son Shi Chaoyi in 761. By 762–63 Yan officers were defecting. Luoyang fell to Tang and Uyghur forces again in the winter of 762. Shi Chaoyi fled and died by suicide on 17 February 763. On paper the rebellion was over. In practice the Tang court was too weak to remake the northeast. Surrendered Yan generals were left as military governors of the territories they already held. Other regions fell under Tang generals who were scarcely more obedient. The second half of the dynasty became an age of fanzhen — semi-autonomous provincial commands — punctuated by later mutinies and, in the 880s, by Huang Chao’s rebellion, which finished what An Lushan had begun.
The western empire was an early casualty. Troops recalled from Gansu, the Tarim and the Tibetan frontier to save the capitals left a vacuum. Tibetans briefly occupied Chang’an in November 763. Over the following decades Tang control of the Western Regions eroded; by around 790 Chinese authority in the Tarim Basin was gone. The Guanzhong plain, political centre of China since the Han, began a long relative decline as population and economic weight shifted toward Jiangnan in the south — a migration pattern Chinese historians summarise as yiguan nan du, the southward passage of the capped and gowned elite, though ordinary farming families moved too.
How Many Died
The death toll of the An–Shi Rebellion is one of the most misused statistics in world history. A Tang census of 755 recorded about 52.9 million people in 8.9 million taxpaying households. A register from 764, the year after the war’s official end, recorded about 16.9 million people in 2.9 million households. Subtract one from the other and you get roughly 36 million “missing” people — a figure that Steven Pinker, drawing on Matthew White’s atrocity compilations, once treated as proportionally the worst catastrophe in human history. White himself later revised his working figure down to about 13 million. Neither number is a body count.
Specialists have been saying so for decades. The postwar registers reflect collapsed tax administration as much as corpses. Whole northern provinces dropped out of the imperial revenue system. Religious orders, merchants and foreigners were often excluded from the counts. Hans Bielenstein argued in 1947 that only a “comically naive” reading of the sources could treat the drop from 52 million to 17 million as slaughter. Charles Patrick Fitzgerald noted that contemporary narratives describe intermittent fighting across a limited set of provinces, not a continent-wide extermination. The honest statement is narrower and still bleak: the war killed large numbers through battle, massacre, famine and disease; it displaced millions; it wrecked fiscal capacity in the north; and no modern historian has a reliable total for the dead. When a popular book quotes “36 million,” it is quoting a census artefact, not a demographer’s estimate.
Cosmopolitan Tang and the Turn Inward
Early Tang Chang’an was famous for its openness. Sogdian merchants, Türkic soldiers, Persian and Arab traders, and Buddhist, Manichaean, Zoroastrian and Church of the East communities were part of the capital’s ordinary life. An Lushan himself was a product of that world: a frontier officer whose career made sense only in an empire that used steppe and oasis manpower. The rebellion poisoned that arrangement without abolishing it overnight.
During and after the war, foreigners could become targets. In 760 the defecting Yan general Tian Shengong massacred thousands of Arab and Persian merchants in Yangzhou — a blood sacrifice, in the political logic of the moment, to prove loyalty to a Tang court that was framing the struggle as Chinese dynasty against Hu rebels. In 761 Gao Juren slaughtered Sogdians in the Fanyang region, identifying victims by high noses, as he jockeyed for Tang recognition. Sogdian traders continued to work in China, but many hid or changed identities; the minister An Chongzhang even asked in 756 to rename his whole lineage Li Baoyu so as not to share An Lushan’s surname. Pulleyblank noted the double legacy: among non-Chinese soldiers on the northeastern frontier An became a cult figure, while among Chinese elites the “barbarian” character of the revolt fed a sharper xenophobia than the early Tang had shown.
Intellectual life registered the shock. Du Fu spent a winter as a captive in rebel-held Chang’an and wrote the poems that still define the age for Chinese readers. Li Bai tangled with princely politics and exile. Wang Wei was forced into a Yan post and later punished for it. Decades on, writers such as Han Yu attacked Buddhism as a foreign distraction — a polemical style that would have sounded odder in Xuanzong’s youth. The dynasty’s famous cosmopolitanism did not vanish in 763, but the mood that sustained it had cracked.
Institutions After the Fire
The fiscal system of the early Tang, built around the zu-yong-diao head and labour taxes tied to land allocation, could not be rebuilt on the old basis. The court leaned more heavily on the salt monopoly and on taxes assessed by wealth rather than simple headcounts. Revenue collapsed in provinces that stopped remitting. Military governors kept the taxes they raised. Eunuch power at court grew: the wartime command apparatus under men such as Li Fuguo was an ancestor of the later Bureau of Military Affairs (Shumiyuan), which by the late Tang rivalled the chief ministers.
Politically, the lesson drawn by later state-builders was not subtle. A general who combined frontier troops, tax income and personal ties to the throne could destroy the centre. The Song dynasty that reunified much of China after 960 institutionalised the opposite preference: keep a large army, but keep it under central civilian and palace control; rotate commanders; separate military from fiscal authority in the provinces; elevate the examination-trained official over the charismatic general. Song armies were often criticised by later patriots for failing against the Khitans, Tanguts and Jurchens. That criticism usually omits the problem Song Taizu thought he was solving — the century of coups and governor-emperors that followed An Lushan. The “An Lushan reflex,” as some modern writers call it, was a real institutional memory: better a constrained sword than another Yan.
None of this means the Tang were finished in 763. The dynasty recovered enough strength to destroy the Uyghur Khaganate in Mongolia in the 840s and to conduct serious campaigns into the ninth century. What it could not recover was the combination of open frontiers, obedient jiedushi and a Guanzhong core that funded a confident universal empire. When Huang Chao took both capitals in the 880s and Zhu Wen deposed the last Tang emperor in 907, they were finishing a disintegration whose turning point lay 150 years earlier.
What the Sources Allow Us to Say
The An–Shi Rebellion is well documented by the standards of medieval Eurasia — dynastic histories, the Zizhi Tongjian, poetry, stele inscriptions, and a growing modern secondary literature from Pulleyblank’s 1955 monograph onward — but it is still a story full of court gossip, inflated headcounts and moralising. Traditional historiography often blamed Yang Guifei’s beauty and Xuanzong’s dotage. Modern work, including Pulleyblank’s, puts the weight on frontier militarisation, ministerial rivalry and the structural bet Li Linfu made on non-Chinese commanders. Both frames contain pieces of the truth. A fat general clowning in the inner palace is vivid; three provinces under one jiedushi is the mechanism.
For readers trying to place the rebellion among other Chinese upheavals on this site — the Taiping, the war with Japan, the civil war that brought the Communists to power — the useful comparison is institutional rather than body-count theatre. An Lushan showed that a Chinese empire could be gutted from its own military periphery. Every serious regime after him tried, with varying success, to make sure no single general could do it again.
Poetry and painting kept returning to Mawei and to the hard road into Shu because those scenes compressed the whole disaster into a few images: an emperor who could not protect his favourite, a capital emptied overnight, a dynasty that survived as a name while its substance drained into the provinces. Bai Juyi’s narrative poem fixed Yang Guifei’s death for later readers; landscape painters restaged the flight through the Qinling passes for centuries. Those artworks are not reliable reportage. They are how a literate culture metabolised a political failure. The institutional record — jiedushi who kept their taxes, salt monopolies that replaced head taxes, Song rules that clipped generals — is less picturesque and more decisive. Empires remember catastrophes in verse; they prevent the next one with staffing charts.
Conclusion
Between 755 and 763 the Tang empire fought a civil war that began as one commander’s bid for the throne and ended as a permanent redistribution of power away from Chang’an. Luoyang and Chang’an fell and were retaken; emperors fled and abdicated; Uyghur and Abbasid auxiliaries helped stitch the dynasty back together; the west slipped toward Tibet; the south grew relatively stronger; and the census numbers that look like an apocalypse mostly record a state that could no longer count its people. An Lushan was dead by early 757. The world he broke — early Tang cosmopolitan confidence, Guanzhong primacy, obedient frontiers — did not come back. Later Chinese governments, especially the Song, built their military institutions around that fact. That is the rebellion’s lasting meaning: not a single shocking death toll, but a hinge between two different ways of running an empire.
Sources
- Edwin G. Pulleyblank, “An Lushan,” Encyclopaedia Britannica; and The Background of the Rebellion of An Lu-shan (Oxford, 1955).
- E. G. Pulleyblank, “The An Lu-Shan Rebellion and the Origins of Chronic Militarism in Late T’ang China,” in Perry and Smith, eds., Essays on T’ang Society (Brill, 1976).
- “An Lushan rebellion,” Grokipedia (overview of dates, campaign narrative and census debate; cross-checked against Pulleyblank and standard Tang chronologies).
- John D. Durand, “The Population Statistics of China, A.D. 2–1953,” Population Studies 13, no. 3 (1960).
- Hans Bielenstein, “The Census of China During the Period 2–742 A.D.,” Bulletin of the Museum of Far Eastern Antiquities 19 (1947).
- Denis Twitchett, ed., The Cambridge History of China, vol. 3, Sui and T’ang China (Cambridge University Press, 1979).
- Christopher I. Beckwith, Empires of the Silk Road (Princeton University Press, 2009), on mid-eighth-century Eurasian turmoil and Tang–Uyghur relations.
- David A. Graff and related essays in Journal of Chinese History special issues on Tang–Song military institutions (Cambridge University Press).
- The China Project / SupChina, “The end of China’s golden age” (2021), for a clear popular summary of the census controversy.
- Matthew White, The Great Big Book of Horrible Things (W. W. Norton, 2011), on the revised ~13 million working figure derived from the same census gap.
