The An Lushan Rebellion

Chapter 1. Before the War: How Frontier Defense Became a Center of Power

In the eleventh month of 755, An Lushan rose in rebellion at Fanyang and then led his army south. Within a few months, they crossed Hebei, captured Luoyang, and carried the war toward Tong Pass and Chang'an.

This army had been stationed on the frontier for many years. Soldiers were organized into units; horses were sustained by pasture and fodder; grain came along prefectures, granaries, and roads; and orders passed down through commanders and clerks familiar with the system. In 755, An Lushan turned the frontier armies and frontier contingents he had long commanded toward an advance into the Tang heartland.

Having decided in advance to rebel, An Lushan naturally held the advantage of striking first. But striking first alone could not enable a large army to march continuously, cross Hebei, and take Luoyang. The organization, supply, and command capacity that sustained this advance had been formed through the Tang's long management of frontier defense.

In the early Tang, the military system rested importantly on the fubing militia, rotating palace guard duty, and central dispatch. Soldiers assembled at fixed times and returned to their home districts after service. This arrangement could support many limited campaigns, but it grew increasingly unable to meet the demands of prolonged, distant frontier defense. Garrisons in the Western Regions, Hexi, Longyou, the northeast, and the southwest needed year-round familiarity with roads, water and pasture, enemy conditions, and relations among local communities. Change at the frontier would not wait for documents to travel to and from Chang'an, and a force assembled temporarily and sent home when its service ended could scarcely preserve such knowledge.

More and more soldiers therefore remained on the frontier. Long-term garrisoning increased, as did the share of recruited and professional soldiers. Their ties to their home districts, landholdings, and rotating obligations gradually weakened, while their ties to garrison settlements, rations and pay, comrades, and commanders became more concrete. Armies no longer appeared only temporarily when war began; training, supply, intelligence, appointments, and negotiation all required a local organization that operated continuously.

This change had no neat date of completion. Older military arrangements, temporary levies, and long-term recruitment coexisted, and conditions differed from one frontier region to another. Guarding the frontier gradually became a continuous occupation: people had to winter locally, horses had to be fed throughout the year, units had to retain the order forged by shared training, and new soldiers had to be led by people familiar with frontier affairs. The court thereby acquired a force available at any time, while the military towns accumulated experience that the center could not readily reproduce on short notice.

A jiedushi was, above all, a commander required by this system of frontier defense. He coordinated the armies of a region and often also influenced finance, administration, military colonies, storage, and dealings with neighboring peoples. Opportunities for action on a distant frontier vanished in an instant, so Chang'an gave frontier commanders more discretion to respond to circumstances. This system defended the empire effectively, but it also gradually allowed some frontier garrisons to acquire their own pace of action.

Its danger lay in its efficiency.

The frontier armies of the mid-eighth century had a complex composition. Han soldiers in long service, and soldiers and commanders from Turkic, Sogdian, Khitan, Xi, Tongluo, and other communities, made their livelihoods and advanced in the military towns together. Language, mounted archery, frontier connections, and a record of following a particular commander could all shape a person's position there. The military towns enrolled people of different backgrounds in one system of appointments, rewards and punishments, and supply; they retained their own languages and connections while forming new ties of comradeship and command through shared service.

The military towns gave people of these different backgrounds a shared occupational order, but did not erase the differences among them. Some excelled at mounted archery; some knew languages and roads; some entered commanders' networks through military merit; others were ordinary soldiers who received rations over long periods. Dividing them into fixed camps with later ethnic or regional labels would obscure the appointments, rewards and punishments, comradeship, and livelihoods that actually organized people.

The Tang drew in many different peoples and employed multiple military traditions. An Lushan rose gradually within this imperial order; ethnicity alone cannot explain why he rebelled. To understand the rebellion, we need to see how he concentrated several armies, multiple military regions, and personal retainers, eventually gaining the capacity to challenge the court.

Beyond the camps, many more people made frontier defense real. Raising horses required pasture, fodder, and veterinarians; military grain passed through farming households, granary clerks, cart drivers, boatmen, and the prefectures along its route; armor and weapons needed artisans' repairs; and intelligence depended on interpreters, envoys, and people familiar with frontier roads. Official accounts recorded zu-yong-diao levies—the Tang system of land, household, and labor obligations—along with grain, silk, and transport labor; on the ground, this meant when grain would be loaded, who would replace a sick horse, which households would provide laborers, and which stretch of road would remain passable after rain or snow.

This chain passed through many layers of people. Granaries had their own officials; prefectures and counties had to coordinate between local burdens and military orders; merchants weighed the risks of roads and trade; and mid-ranking officers had to turn abstract orders into the time a particular unit would depart and the place where it would collect grain. If one link was delayed, an army already “dispatched” on paper might still remain where it was. Through long operation, the frontier towns had taught many people who should take over the next link; this cooperation rarely entered grand narratives in peacetime, yet in war it determined whether an order could become action.

Commanders' power also took shape in these daily affairs. Who could issue rations and pay on time, record military merit, assign duties, and care for the sick, wounded, and their families was very concrete to people living long-term in the military towns. Loyalty rarely changed suddenly in a single day. It might begin with one allotment of grain, one reward or punishment, or one request for help, and only later settle into a more durable dependence. Soldiers thereby learned from whom orders usually came and whom to seek out when pay was overdue, illness struck, or their families faced hardship. Such dependence did not automatically make people willing to rebel against the Tang, but it gave the commander before them a quicker ability to rally them in a crisis than a distant institution possessed.

Military supply also connected the frontier towns with surrounding society. Garrisons brought defense, markets, military posts, and opportunities to migrate, but also consumed land, fodder, transport, and labor service. Some people made their living from frontier markets and military supply; some bore requisitions; some experienced both. The surviving material makes it difficult to trace these gains and losses household by household, and frontier residents did not hold one uniform position. The strength of the military towns came from networks sustained both inside and outside their gates; when they were later used in civil war, soldiers at the front and societies along the route bore the consequences together.

This inequality also existed within the camps. Ordinary soldiers, subordinate commanders, interpreters, staff personnel, and families dependent on the military towns for their lives had different ways out. People able to rise through military merit and people who long bore transport duties or requisitions might both depend on the same military town, yet their interests did not necessarily coincide. The sources have not left complete lists of them. What the narrative calls “frontier troops,” “frontier people,” or a particular ethnic group always includes people with differing situations and choices.

People in Chang'an were later shocked by the rebels' arrival because the empire's wars had long been borne by the frontier. In 755, the machinery that had directed violence beyond the borders turned inward, and the central cities began to bear costs they had previously felt less directly.

An Lushan later held the three commands of Fanyang, Pinglu, and Hedong at the same time. These three appointments corresponded to real armies and territories: the northeastern frontier, the Taiyuan direction, and the Hebei plain were thereby linked under one commander. The route from Fanyang to Luoyang by way of the Yellow River also lay nearby. Each command had soldiers experienced in long frontier warfare and controlled horses, stores, and the military resources of different communities. Holding the three commands together gradually brought their orders, appointments, and intelligence into one center; personnel or supplies lacking in one military town could more readily be moved from another. Coordination among the three commands thus accelerated, while it became correspondingly harder for the court to send another person to take over any one of them.

There were practical reasons at the time to let one commander control several directions. Emperor Xuanzong and the court needed someone able to win military successes, manage complex frontier affairs, and reduce coordination costs. The emperor's direct favor also made this vast power appear still to be granted and constrained by the center. With every new appointment An Lushan received, the frontier's capacity for coordination and the power he personally could call upon increased together; before the crisis broke out, the two were not easy to separate.

Once replacement was needed, the problem became visible. Removing an official required only an imperial edict, but taking over three commands required a new commander to receive thousands upon thousands of soldiers, subordinates, stores, grain routes, and local relationships. The army had to recognize the new official, rations, pay, and military orders had to be rearranged, and the many people who lived through the existing networks had to believe that their circumstances would not suddenly worsen. The center still had tools such as appointments and dismissals, rewards, inspection, and military designations (junhao), but these tools required trustworthy information, usable replacements, and time to carry out orders.

In ordinary times, the frontier towns still had to obtain titles, rewards, and financial support continually from Chang'an, and local commanders were far from independent monarchs. The danger arose from the widening distance between nominal power and actual takeover. When the same commander had held troops for a long time and could influence the flow of information back, the center could still dismiss him in principle, yet found it difficult to establish who would take over, how the army would respond, or how much the situation had changed before the order arrived. Central power still existed; it simply had to pass through a series of links it could no longer fully control.

An organization had already formed around An Lushan. Yan Zhuang, Gao Shang, and others took part in planning and documentation; adopted sons, confidants, staff personnel, and commanders familiar with different groups of troops occupied places within it. They interpreted orders and arranged marches, while rewards, punishments, and political claims reached the units through layer upon layer of relationships. An Lushan was able to turn a vast army because he relied on this already formed organization.

The staff office wrote documents for the commander and also preserved and directed the knowledge the military towns required. Deputy commissioners, administrative aides (panguan), secretaries, subordinate commanders, interpreters, and local contacts each had their own strengths: who knew a particular body of troops, who could deal with prefectures and counties, who knew which granary could still supply grain, and who could carry an order across a mountain pass and a ferry crossing. Through these intermediaries, a commander extended the reach of his actions. An Lushan's personal power consisted precisely in many offices and relationships becoming connected around him through long operation.

This organization could narrow the space for open opposition, but it cannot prove that all its members wanted the same thing. The Zizhi Tongjian records that when An Lushan claimed to be acting on a secret imperial order to attack Yang Guozhong, the commanders “looked at one another in astonishment and none dared voice a different opinion.” This can show only that the commanders in that account did not openly object; it cannot show how much each person knew, why they were silent, or the specific motives behind each later surrender, return to allegiance, or dispersal.

Space also constrained this power. Fanyang, Pinglu, and Hedong were connected to one another, but each camp remained subject to mountain passes, rivers, prefectural cities, winter roads, fodder reserves, and boats at ferry crossings. These conditions determined when forces could assemble and how quickly cavalry and baggage trains could travel. The frontier army's advantage came from ready-made units, established rhythms of supply, and knowledge of nearby roads, not from an unverified enormous troop total. The court could issue an edict to recruit soldiers, but people assembled temporarily had to learn their companions, passwords, and supplies anew; carts and horses put together over a few days could also easily come apart on a long march.

Rising in rebellion meant more than issuing orders. Soldiers, commanders, baggage, and horses had to be coordinated in movement by an existing organization. The available material can show that An Lushan had long controlled the military towns and rose at Fanyang, but it is insufficient to reconstruct the stores, supply chains, and every level of transmitting orders before the rising. What can be established is that an army organization originally used for frontier defense was used to advance into the Tang heartland.

Thus, what confronted the court in 755 was a set of relationships among personnel, materials, and orders capable of acting independently. Appointments and dismissals on paper still belonged to Chang'an, but actually taking over those relationships already required more than a single edict.

The Tang of the Kaiyuan and Tianbao eras remained powerful. That strength did not automatically turn resources from every region into usable frontline forces in a crisis; Chang'an and Luoyang were also among the largest and most connected cities in Eurasia. After Chang'an fell in 756 and Emperor Xuanzong fled to Shu, the Tang court and its military commanders continued to organize war against Yan; the war did not end until 763.

The problem was an uneven distribution of capacity. Elite armies were concentrated at the frontiers, while the interior had long been at peace, and prefectural and county defenses and temporary recruitment lacked experience fighting professional frontier armies. The center commanded the resources of a vast realm, but after rebel forces were already on the move it could not immediately place armies, horses, and grain in the right locations. Personnel, stores, and draft animals available in different places also had to be collected, checked, and transported. Real time and distance lay between the court's having resources and its being able to use them at once. The state could be strong overall while responding slowly in particular places; the more professional the frontier defense became, the more evident the interior's time lag was in the first shock.

This time lag also came from knowledge. Soldiers stationed in the northeast knew nearby granaries, roads, and ferry crossings, while court commanders assigned at short notice had to identify forces, weapons, and supplies anew. Fanyang looked distant from Chang'an, and courier riders still had to proceed station by station; for the local military town, the next camp and source of grain were daily business. Geography did not itself start the rebellion, but once one side had completed its organization and decided to act, it gave that side a speed the still-unorganized side could not immediately match.

Information also required time. How many troops frontier commanders actually controlled, whether local officials would still obey, which roads were safe, and whether a particular victory or defeat was trustworthy all had to pass through envoys and bureaucrats. Some people did see risks, but only part of them; reports were also filtered through factional conflict, the credibility of messengers, and the emperor's personal trust in An Lushan. The system had channels for reports, but it did not turn scattered warnings into coordinated action in time.

People inside and outside the court had long suspected An Lushan, while others continued to take his military successes, tribute goods, and obedience as credible signs. The same autonomy of a frontier commander showed his ability in ordinary times and his growing difficulty to control in a crisis. The warnings conflicted with one another; Chang'an had both to verify information and to find a replacement that would not immediately empty the frontier defenses. How personal trust further shaped judgment is the task of the next chapter; what this chapter has seen is that information and takeover had already become slower than commands on paper.

These vulnerabilities grew precisely within the empire's capacity. A vast territory required professional, permanent frontier armies; their long operation in turn formed independent systems of personnel and resources. The center depended on local commanders to handle frontier affairs quickly, and local commanders depended on titles, rewards, and finance from the center to sustain legitimacy. Neither side had separated from the other, yet they had formed a mutual dependence that was difficult to dismantle at low cost.

Later, the Tang court still had to rely on other frontier armies and local military forces to put down the rebellion. The resources supporting the rebellion were also those the court needed to keep fighting, so simply abolishing every frontier command was not feasible. The system had once been effective and still had uses in war, but the relationship among authorization, verification, rotation, and emergency takeover had broken down. This empire, still capable of so much, was ultimately injured by part of its own capacity.

The strength of the Kaiyuan and Tianbao eras had never fallen evenly on every part of the empire. The two capitals had long remained distant from war because grain, horses, labor, and risk were continually sent to the frontier. When the frontier armies turned inward, those already existing capacities turned with them.

The frontier-defense system supplied armies, roads, and resources, but did not decide in advance who would rebel. Most jiedushi continued to obey the court, and professional armies long carried out frontier defense. In 755, An Lushan and core staff members including Yan Zhuang and Gao Shang decided to rebel and mobilized the troops under their control to act. An Lushan's authority over several commands, his conflict with Yang Guozhong, the way Emperor Xuanzong trusted him, fear of purges, the planning group around him, and the concrete circumstances of the time all shaped that choice.

To write only of personal ambition would hide how the court created a commander difficult to replace; to write only of institutional failure would allow those who decided to rebel to disappear into structure. History needs to retain both conditions and choice: a war machine that could be turned in another direction already existed, and people still decided to turn it.

This distinction also explains why the war did not end with An Lushan's death. The professional lives of soldiers, the relationships between commanders and subordinates, local needs for self-protection, and trust already broken apart did not disappear with one person's life. That An Qingxu, Shi Siming, and Shi Chaoyi could continue the war shows that the rebellion quickly exceeded its original initiator; it does not cancel his responsibility for the choice made in 755.

Before the war, Chang'an depended on the frontier towns to defend a vast border; the frontier towns depended on titles and resources granted by the center; soldiers depended on the military towns for their occupational lives; and surrounding societies sustained the whole system with grain, horses, labor, and roads. These relationships had worked, but had also become difficult to dismantle quickly. Any replacement would affect titles, rations and pay, military posts, and family livelihoods at once; hesitation arose from this, and risk continued to accumulate within the hesitation.

The later outcome makes it easy to think that the danger had long been obvious. At the time, a frontier commander's independent handling of military affairs could still be merely an efficient arrangement for frontier defense, and soldiers' reliance on a familiar commander did not decide in advance that they would join a rebellion. Only when mutual trust at court broke down, replacement became dangerous, and An Lushan and his group decided to mobilize troops did long-standing conditions become compressed into a particular action. The system explains why that action was possible; those who made the decision still bear responsibility.

Responsibility also had levels. The emperor, chief ministers, and jiedushi held greater power to decide; staff personnel, mid-ranking officers, and prefectural and county officials enabled decisions to be carried out; ordinary soldiers, transport workers, and people subject to requisition often bore orders from positions with less information and narrower ways out. To write everyone as an extension of An Lushan's will would conceal coercion and limited choices; to remove the chain of execution from the narrative would make a war spanning the empire appear to have arisen from the thoughts of only a few people.

We must next enter Chang'an. The army already existed, but why it remained concentrated in An Lushan's hands for so long, and why the emperor was willing to place personal trust above institutional constraint, still need to be explained from inside the imperial center.

Chapter 2. How An Lushan Entered the Empire’s Center

An Lushan’s armies were stationed in the northeast, but part of his power was formed in Chang’an. Without the emperor’s continuing grants of offices, military commands, and political credit, he could hardly have held three commands at once; if he had been only a favored figure at court, he could not have turned favor into power sufficient to begin a war. This chapter examines how frontier military power became joined to the emperor’s trust.

An Lushan came from the multiethnic setting of the Tang northeast frontier. Accounts of his ethnicity, the origin of his name, his early experience, and court anecdotes differ in the Old Book of Tang, New Book of Tang, and later chronicles, and have been reorganized in light of the rebellion’s outcome. What can be confirmed more securely is that he knew frontier languages and relationships, rose in the military towns through military success, practical ability, and commanders’ favor, and then entered Emperor Xuanzong’s view.

This route gave him command of two political languages at once. On the frontier, achievements, interpretation, negotiation, and relations with subordinates shaped a soldier’s position; in Chang’an, titles, memorials, audiences, and the emperor’s direct recognition determined the range such abilities could attain. An Lushan repeatedly sustained imperial trust through frontier merit, appearances at court, and offerings; these actions had political effects, but cannot prove in reverse the full calculations behind each action.

Xuanzong had practical frontier-defense reasons to employ An Lushan. The northeast faced long-term military pressure from the Khitan, Xi, and other directions and needed commanders familiar with local roads, peoples, and ways of fighting. A general able to handle frontier affairs and attribute his rise to imperial favor could help a ruler bypass chief ministers and established elite families and directly influence faraway military towns. Xuanzong thereby gained a capable frontier commander, while An Lushan gained a path of advancement not wholly dependent on central bureaucratic credentials.

For a considerable period, this cooperation did work. An Lushan’s biography in the Old Book of Tang says that “each month he presented without interruption captives, camels, horses, hawks, and dogs.” This later biography shows that continual offerings were written into the narrative of his relationship with the court, while also exposing people recorded as tribute. It cannot by itself establish the frequency of offerings, all his frontier achievements, or how the court judged his loyalty because of them. The later rebellion makes every earlier appointment seem absurd too easily; the court then received more than one kind of signal.

An Lushan successively added Fanyang, Pinglu, and Hedong to the commands he held concurrently, making personal credit increasingly hard to separate from vast military office. Xuanzong saw an important frontier commander whom he had promoted and could control directly; An Lushan gained offices and a network of subordinates that increasingly depended on him personally to sustain them. The higher the office, the more armies, staff, and local relationships had to be taken over when he was replaced. Favor both enlarged his power and raised the danger when that favor was lost.

Accumulating appointments also changed how the court understood this person. A new post could be seen as a reward for success in an earlier appointment, while power already held became a reason for further commission: since An Lushan already knew several directions, allowing him to oversee more military affairs seemed to save more coordination than finding an unfamiliar commander. Such judgment need not arise from conspiracy, but it could create path dependence. Every addition of authority made the next division harder, and the harder division became, the easier it was to present continuing trust as the practical choice.

A frontier commander’s direct dependence on the emperor could also help Xuanzong reduce the intermediary power of central officials. If An Lushan proved loyalty only to the emperor, constraints by chief ministers and censors could appear as bureaucrats obstructing a meritorious servant. For the ruler, this relationship at one time enlarged personal discretion; for the system, it made it increasingly difficult for the same assessment of risk to gain authority outside the emperor. Only as the crisis approached did the court discover that the efficiency gained by bypassing intermediate layers had also weakened those layers’ shared capacity to correct error.

Audiences, rewards, grants of residences, marriage arrangements, and direct memorials could all allocate power in monarchical politics. They appeared to belong to private relationships, but actually changed who had the opportunity to explain themselves, whose reports were believed first, and who could cross usual bureaucratic levels. After entering the court, An Lushan gained not merely a feeling of closeness but political credit that could offset outside accusations.

Later generations liked to tell stories about his appearance, dancing, wit, and adoption of Yang Guifei as foster mother. These records may preserve court memories, but they also bear historians’ arrangements of character and political retribution. If they are treated directly as causes, Xuanzong seems to have handed over frontier defense merely because he was amused for a moment, while years of court appointments, assessments of military merit, and balances of power disappear from the narrative.

What makes these stories more valuable is that they show how court closeness was expressed publicly. Through exaggerated ritual, An Lushan showed that he recognized only the emperor and the noble consort, not chief ministers and regular order as being in the same position; Xuanzong’s amused acceptance let officials present see that this frontier commander possessed a special channel. Such performance may not tell us what the people involved truly believed, but it could change other people’s expectations of action.

If an official accused an ordinary frontier commander, an investigation might proceed through censors, chief ministers, and local documents. In confronting An Lushan, an accuser also had to overcome the emperor’s confidence in his own ability to judge people. The more Xuanzong believed that he knew An Lushan personally, the easier it was to interpret new warnings as court factional conflict; An Lushan could also portray doubts about him as envy of his military achievements or attempts to sow discord between ruler and minister.

Closeness required many people to sustain it. An Lushan’s staff drafted memorials; envoys carried tribute and news; relatives, guests, and officials in Chang’an helped pass information. The court, meanwhile, relied on eunuch envoys, chief ministers, censors, and local officials to observe the three commands. There was no private hotline between emperor and frontier general that escaped interference: every “firsthand observation” and “secret report” passed through particular people.

These intermediaries had their own fears and interests. An envoy’s report could be shaped by the conditions of reception and by prior judgment; whether an individual sought favor or received benefits must be shown by specific records. Opponents may indeed have seen a danger, or may have used danger to weaken political rivals. By the time information reached Chang’an through this network, facts and the credit of those who conveyed them were already entangled. The court had many channels of report, but no common procedure that reliably cross-checked conditions in the three commands.

The Zizhi Tongjian records that in the fourteenth Tianbao year the court sent Fu Qiulin to Fanyang to observe events; after receiving lavish bribes, Fu reported that An Lushan “harbored no disloyal intentions.” The Old Book of Tang likewise records that Fu took bribes and called him loyal, but dates the mission to the twelfth Tianbao year. The point is not merely that one envoy was bribed. The emperor’s judgment of whether a powerful frontier general was preparing to rebel could depend on a single personal observation, easily managed, deceived, or bought; other reports could lose credit because they came from An Lushan’s political opponents. The institutions still gathered information, but found it increasingly difficult to turn information into concerted action.

A secret envoy’s assignment had limits of its own. His stay in a frontier command was brief; the people, weapons, and ceremonies he saw might all have been arranged. After returning to Chang’an, his report was interpreted through the emperor’s existing expectations. If its result matched Xuanzong’s trust, it appeared to be verification; if not, the reporter had to show that he had not been influenced by Yang Guozhong’s side. One observation could therefore scarcely overturn a long relationship and might instead provide fresh grounds for waiting.

Other members of the court also learned from the emperor’s reactions. Those whose repeated warnings were not accepted might send fewer memorials or phrase them more sharply. Alarms did not become an effective check, but the surviving materials cannot give one uniform account of who fell silent or why. Channels of information appeared to remain open, yet officials judged from earlier handling which reports might be accepted and which might harm themselves; individual cases still require separate determination. Danger lay not only in a lack of news, but also in many people already knowing which news was unlikely to produce a result.

By the late Tianbao period, both An Lushan’s military power and his court connections had aroused concern. The court did not discover the danger only on the day he rose. Existing histories record that some people at court warned Xuanzong that An Lushan might rebel; his concurrent command of three military regions, proposals to replace commanders, and refusals to come to court were events of different times. Until collated editions and variant texts are checked, these events should not be joined into an unbroken chain of escalating crisis.

A frontier general’s adjustment of officers could be read as preparation for rebellion, or described as putting military affairs in order; presenting horses could express loyalty, or conceal the movement of people and materials; pleading illness and not coming to court might reflect real illness, or test the court. No single act readily proves a complete plan by itself. The emperor had to decide when an uncertain risk counted as evidence sufficient to remove a major general.

Removal was not merely an edict. The Zizhi Tongjian and the New Book of Tang both preserve accounts that the court considered summoning An Lushan back and dividing the commands among others, but did not carry this out. Both books were compiled late and draw on related traditions; at this stage, drafts of edicts, envoys’ reports, and the causal sequence around them cannot be written as undisputed records made on the scene. The unimplemented idea nonetheless suggests that removing a frontier general required summons, division of commands, and takeover arrangements to be handled together.

If An Lushan were summoned to the capital without a reliable successor, subordinates in the three commands might understand the order as a purge; if power were publicly divided first, An Lushan might act before the court. A replacement had to be accepted by the army, while rations and pay, documents, and local relationships also had to continue. Chapter 1 has already explained why these resources were hard to separate at low cost; Chapter 2 need only show how the cost of replacement in turn shaped political judgment in Chang’an.

The emperor also bore the consequences of wrongly judging a long-serving meritorious frontier general disloyal. If such a general were suddenly stripped of power on an opponent’s accusation, frontier defense might be damaged and other commanders might doubt whether the emperor still protected those who had received favor. But if every risk were explained away as an inability to suspect without cause, personal credit would gradually replace public verification. Before An Lushan rose, the recorded warnings, plans to summon or divide his commands, and requests to change commanders belonged to different circumstances; they show that the court was not wholly unalert, but cannot be merged into one clear measure that was simply postponed.

Time thereby changed the options. Earlier on, dividing commands, rotating deputies, cross-checking rations and pay, and limiting personal troops might have been routine administration; by 755, the same measures looked more like a public announcement that the emperor was preparing to act. The later action came, the higher the cost of takeover; the greater the fear of provoking the frontier general, the more time he had to prepare. Waiting did not preserve the status quo; it steadily narrowed the space for handling the danger safely.

This cannot prove that any earlier measure would certainly have prevented war. The materials do not allow us to calculate a path without risk for people at the time, and later outcomes cannot automatically turn every warner into a prophet who saw everything. What can be established is that warnings did not become a collective procedure of judgment able to pass beyond private favor, factional conflict, and a single envoy’s report. The court saw many fragments, but never organized them into an executable handover.

Execution meant breaking a judgment into a series of actions. Who would arrange the summons, who would first contact the deputies in the three commands, who would hold the new appointment documents, which force would ensure that the order took effect locally, and how An Lushan’s relatives and followers would be treated after he came to court—all required prior arrangement. If there was only the conclusion “he may rebel,” without an execution chain able to withstand resistance, stronger warnings were more likely to make both sides hasten their preparations.

The court might also have tried to reduce the need for a single showdown: have different channels separately check the army and fiscal revenues, gradually adjust deputies and terms of office, and prevent the three commands from continuing to accumulate around one person. The source material is insufficient to establish how much chance these measures still had in 755, and this chapter cannot design a guaranteed historical solution. They at least show that governing risk involved many finer actions than “trust” or “execute”; these actions did not form stable institutions while power could still be adjusted at low cost.

The hostility between Yang Guozhong and An Lushan compressed a long-term risk into a very short period. Both the Old Book of Tang and the Zizhi Tongjian record that Yang Guozhong repeatedly said before An Lushan rose that he would rebel. Both men tried to make the emperor see the other as the danger, and both acted in ways that forced the other to declare himself.

According to the Zizhi Tongjian, in 755 An Lushan asked to replace Han generals with thirty-two non-Han generals, and Wei Jiansu opposed it as a sign of rebellion. Details concerning the arrest of guests, pleading illness to avoid court, presenting horses, and exchanges of envoys still await checking against collated editions and variant texts in the page’s materials; until then, they are not combined into one continuous chain of “driving him into rebellion.”

These actions cannot directly reveal inner thoughts, but they show how security disappeared. The court saw a commander of three commands who refused summonses, adjusted officers, and gathered men; An Lushan faced the arrest of confidants, sons kept in the capital, discussion of dividing his military power, and ever-arriving edicts. Each side grew more likely to read the other’s next step as preparation for a purge.

The favor that had once sustained the relationship could no longer provide a safe exit. An Lushan had no credible guarantee that, after surrendering military power, he and his group would remain safe; the court had no credible guarantee that continued leniency would not make later action harder. The institutions lacked a way both to verify risk, hand over military power gradually, and restrain reprisals. Each test made the next compromise seem more dangerous.

Yang Guozhong’s warnings did bring the risk into the emperor’s view, but their public hostility also weakened the warnings. If information came mainly from one political opponent, Xuanzong could suspect it served a contest over the chief ministership; if he excluded the warnings completely for that reason, he placed personal trust above every institutional signal. Factional conflict and real risk appeared in the same memorial, making it harder for the court to establish shared credit.

Xuanzong held final power of appointment and dismissal, but could not see inside the three commands himself. He had to judge through envoys, memorials, and officials hostile to one another. A ruler’s personal trust could once speed decisions on frontier affairs, but in a crisis it made contrary evidence easier to interpret as slander. An Lushan’s political security likewise depended on Xuanzong personally; once that protection weakened, he lacked an institutional position to which he could retreat.

An Lushan’s sons and guests left in Chang’an meant that the relationship did not exist only at the level of office. They were part of court connections, but could also become hostages, messengers, or targets of a purge in a crisis. When news traveled from Chang’an to Fanyang, it contained not only the emperor’s attitude but the safety of relatives and allies. Conversely, whether An Lushan obeyed an order to come to the capital affected the situation of those left there. Personal ties, group security, and state orders were thus entangled on the same line of transmission.

According to the Zizhi Tongjian, Yang Guozhong ordered the metropolitan prefect to surround An Lushan’s residences in the capital, arrest retainers including Li Chao, and secretly execute them after sending them to the Censorate’s prison. This account shows that the conflict had entered coercive action; on its own, however, it cannot establish Yang Guozhong’s calculation of risk or prove that neither side could still have accepted third-party verification.

The three men did not control equal choices. Xuanzong could appoint and dismiss, divide commands, and move troops; Yang Guozhong could press for investigation and arrest; An Lushan controlled armies already organized. Ordinary envoys, staff members, middle-ranking officers, and relatives left in the two places could only transmit, conceal, or carry out orders in narrower spaces. Responsibility was therefore layered: those who controlled more resources had greater capacity to decide whether risk became a war across the realm.

An Lushan ultimately rose in the name of punishing Yang Guozhong and others. “Punishing treacherous ministers” was a familiar political rhetoric: it portrayed military action as correcting errors around the emperor, lowered the cost of directly opposing the ruler, and offered the army and officials along the route a temporarily acceptable justification. A slogan can organize obedience; it cannot absolve those who launch action of responsibility.

Once the rising began, self-protection, removing political enemies, maintaining the army, occupying cities, and contesting supreme power quickly became joined together. In the first month of 756, An Lushan declared himself emperor at Luoyang, named his state Yan, and adopted the era name Shengwu. The rebellion did not stop at a court purge; it established a different set of appointments, military orders, and demands of rule.

Xuanzong’s misjudgment also cannot be reduced to a story of an old emperor absorbed in pleasure. Xuanzong long sustained An Lushan through favor and repeatedly gave less weight to alarms of rebellion; this shows that personal judgment outweighed some institutional signals, without requiring further claims about how the emperor understood his own ability to judge people. Yang Guozhong brought real alarms and political hostility into the same channel, while An Lushan maintained the identity of a favored frontier general even as he made the military towns capable of refusing orders. The actions of all three altered one another’s costs and eventually moved the conflict from reversible tests into armed confrontation.

This relationship still did not make the decision for An Lushan. Being suspected, summoned, and seeing confidants arrested were risks he faced; mobilizing several commands and using a political pretext to send the army south were actions taken by An Lushan and his group. The court, having long concentrated military power and relied on personal favor, bears institutional and decision-making responsibility; those who controlled the armies and decided to rise cannot hide behind a story of having been “driven into rebellion.”

Not everyone in the army had the same choices. Some joined planning; some depended on office, rations, and pay; some were constrained by discipline and immediate force; others later surrendered, returned to allegiance, or dispersed. Court political slogans, when they reached the units, became questions of who called the roll, who issued grain, and who led the departure. This chapter does not unfold the specific course of those actions; it need only establish that Chang’an’s crisis of trust had found an army able to carry out orders.

Only after years in the empire’s center could An Lushan continue to receive credit when danger signals appeared; only after years commanding frontier military towns could he turn political conflict into military action when that credit broke down. The war did not come from two separate worlds. It occurred within the Tang’s own appointments, frontier defense, and court relationships, then expanded southward along the empire’s roads.

At the end of Chapter 1, the armies and resources required for war already existed; at the end of Chapter 2, the trust and avenues of retreat that sustained this power had broken down. The next question is no longer whether An Lushan would rise, but how this army crossed Hebei and the Yellow River and reached Luoyang before the court could complete its response.

This handover also defines the responsibilities of the two chapters. Chapter Two preserves the choices that still existed before the rebellion began, because people really had discussed summoning An Lushan to court, dividing his military commands, and keeping him under observation. Chapter Three begins after the choice has been made and no longer uses court motives to explain every march. In this way, the war will neither seem to have appeared out of nowhere along the roads nor be reduced to the personal grievances of a few people in Chang’an. Political relationships produced the orders; what would further spread the disaster was how those orders entered prefectures and counties, city gates, and the lives of people along the route.

Chapter 3. 755: From Fanyang to Luoyang

In the eleventh month of winter 755, An Lushan rose at Fanyang. Rebels naturally acted before the court: An Lushan knew when he would set out and where he would go, while Chang’an had to wait for military reports before confirming what had happened. What this chapter must explain is why that inherent advantage of moving first was not exhausted at the first city, river, or resistance, but rapidly became a military victory across Hebei and the capture of Luoyang.

The answer begins with an organization already in place. The Fanyang camps had formed units, familiar commanders, and a long-used system of orders; grain, horses, and baggage could continue by established means. An Lushan needed to change the purpose of these forces; the Tang, beginning when news arrived, had to appoint commanders anew, recruit soldiers, issue weapons, find supplies, and establish inland defenses. The gap was not only who departed first, but whether an army was already able to fight together.

The Zizhi Tongjian records that An Lushan assembled his commanders with a forged secret edict and said he would lead troops to court to punish Yang Guozhong; the New Book of Tang likewise says he invoked a secret edict to punish Yang. His subordinates were shocked, but none dared openly object, and strict orders suppressed dissent in the army. This compiled account is not a verbatim scene record; it nevertheless shows that a rising had to solve two problems at once: give action a claim that could circulate, and make subordinates under military discipline obey within a very short time.

“Punishing Yang Guozhong” packaged the rebellion as an action against a chief minister, so that some followers did not yet have to face the meaning of founding another regime. Such a claim need not have persuaded everyone, but it could help orders move down existing military ranks. The materials provide no common answer to how much commanders, staff, middle-ranking officers, and soldiers in each camp knew of the plan or sincerely supported it. The later surrenders, returns to allegiance, and dispersals also show that initial silence did not mean all had one will.

Histories often give enormous troop totals for this rising. Premodern military figures mix paper quotas, temporary levies, and exaggerated claims of success, and should not be read as one precise count. It can be established that An Lushan long controlled northeastern frontier commands and, when he rose, mobilized troops under his command and some frontier followers. The proportions of troop types, the scale of each contingent’s participation, and their organization cannot yet be reconstructed. His advantage was that an army already operated in an existing organization, not merely in a number on paper.

The Zizhi Tongjian records that when Taiyuan and the Eastern Surrender City successively reported An Lushan’s rebellion, Xuanzong at first suspected that enemies of An Lushan had forged the news. In that chronicle’s sequence, five days passed from his still not believing memorials reporting rebellion on the yichou day to confirming that An Lushan had rebelled on the gengwu day. This is one account of court reaction preserved in a later chronicle; it cannot by itself reconstruct real-time reporting or prove that reports from all directions were independent.

After confirming the news, the court made appointments quickly and drew on treasuries to recruit soldiers. The Tang still possessed great administrative and fiscal capacity; the difficulty was that an edict could be written in a day, while an army could not be trained in a day. Emergency recruitment required units, training, and command, but the materials do not preserve a standard account of an ordinary recruit from enlistment to first battle. The time gap gained by An Lushan thus expanded from the initiator’s head start into a difference between two degrees of organizational maturity.

The road from Fanyang to the Yellow River crossed Hebei. Yan forces advanced rapidly, but Hebei did not fall neatly to one side. Some voluntarily submitted, some abandoned cities and fled, some temporarily obeyed to preserve their prefecture or county, some secretly contacted the Tang, and some rose publicly in resistance. An army traced a southward route on the map, but every city along it faced different conditions.

For local officials, loyalty to the Tang meant immediately answering concrete questions: Could the walls hold? How long would granaries last? Would subordinates obey? Would neighboring prefectures assist? Had their families already been controlled by Yan forces? Residents generally had narrower choices. Closing gates might bring siege; opening them might temporarily avoid fighting but also bring demands for grain and horses and later risk Tang punishment; leaving could divide property from kin. Specific losses for any group still require household registers, contracts, epitaphs, or personal records.

Yan forces did not always set up wholly new offices in occupied places. The Zizhi Tongjian says that after An Lushan reached Changshan, he gave Yan Gaoqing the gold-and-purple rank, detained his sons and younger relatives, and ordered him to remain as defender; along the route he also appointed Zhang Xiancheng, Li Tingwang, and others to manage occupied places. These are several cases of occupation, insufficient to reconstruct a uniform system of supply or transmitting orders, and cannot replace local officials’ choices in different circumstances.

Yan Gaoqing’s Changshan initially also welcomed Yan forces. He later secretly joined subordinates in planning a return to Tang allegiance, seeking to answer forces still loyal to the Tang. At Pingyuan, Yan Zhenqing organized resistance through earlier preparations of repairing walls, dredging moats, and storing grain. Their choices later entered narratives of loyal martyrdom and have a clear basis in action; but courage did not create a city defense from nothing. Food, able-bodied people, documents, messengers, and responses from neighboring prefectures determined whether a wish to resist could endure as military action.

The Yan family’s fame should not conceal people whose names were not preserved. The surviving material shows prefectural cities and roads changing hands quickly, but cannot separately describe Hebei residents as carriers of grain, gate guards, refugees, or people forced to obey; these actions require proof case by case. Chronicles preserve more readily the names of officials who held out, surrendered, or were punished than the owners of requisitioned carts, people serving on walls, and families separated when roads were cut. Fewer names surviving does not mean fewer people bore the war.

Local positions also changed with the situation. A city that opened its gates to Yan today could reconsider tomorrow as Tang forces approached; an official wishing to return to Tang allegiance first had to know whether messengers could pass and whether help was nearby. To say Yan “occupied Hebei” often means only a provisional result of controlling prefectural cities, roads, and documents, not that all local society had accepted a stable new order.

Using “popular sentiment” to summarize such change requires particular caution. Someone might defend the Tang from principle, but also from family, office, local standing, or neighborhood ties; obeying Yan might involve coercion, survival, or pursuit of office. This chapter lacks cases sufficient to decide proportions, so it does not assign motives to unnamed people. Coercion, calculation, fear, and persistence could coexist. History must still judge the consequences of action, but cannot erase unequal avenues of retreat under one label.

Hebei’s differences soon mattered militarily. As Yan forces continued toward Luoyang, the prefectures, roads, and ferry crossings behind them were not all stable. If the Tang could connect scattered resistance to troops and a political claim, An Lushan had to attend both to the westward vanguard and to the rear route toward Fanyang. A second front was not yet mature, but had begun to emerge.

After confirming the rebellion, the court appointed Feng Changqing, Gao Xianzhi, and others to organize defense and recruited in the capital and elsewhere. On reaching Luoyang, Feng Changqing hastily recruited people from the streets and markets. Rolls could grow quickly, but the training, mutual trust, and battlefield experience an army required could not grow with them. The Tang had grain and funds, official titles, and the capacity to gather many people, but for the moment lacked an inland army able to face professional frontier forces directly.

Gao Xianzhi and Feng Changqing had frontier combat experience, but they did not take over troops they had trained for years. A commander’s experience could help choose positions and arrange retreats; it could not substitute for soldiers’ shared training. Temporary forces found it harder to retain organization under cavalry attack and flight; what new recruits saw or whether they understood orders cannot be inferred from the outcome. An appointment cannot reissue experience of war.

This did not make emergency mobilization meaningless. It shows that after the capital lost the initiative, the Tang could still call upon treasuries, officials, existing frontier armies, and local organizations. Guo Ziyi, Li Guangbi, and others also gathered forces from Shuofang, Hedong, and northern routes to threaten Yan’s rear. But every appointment had to travel: commanders had to reach posts, soldiers be placed in units, and supplies reach camps; winter weather, ferry crossings, and local cooperation determined when an edict actually became military force.

Emergency army-making suddenly pressed the burdens of frontier war onto the interior. Merchants, artisans, and farmers might be drawn into camps; prefectures and counties had to provide food, carts and horses, weapons, and people for city defense. For the court this was a restoration of state capacity; for the families of those drafted, it first meant labor leaving, income interrupted, and older people and children requiring other care. Documents rarely let us follow such changes household by household, but “recruitment” cannot therefore become an administrative word without costs of life.

Two times thus ran together. Yan forces moved forward along existing relations of command and supply, while the Tang, amid defeat, reconnected scattered resources to war. The temporary forces’ retreat around Luoyang was real, as was the gathering of Hebei and northern strength. To write the first defeats as if the state had nothing left would make the next eight years of war inexplicable; to write that the court still had resources as if it could strike back at any time would ignore the time organization required.

The court also had to divide attention among directions. Chang’an feared Luoyang and Tong Pass; Hebei resisters awaited support; Guo Ziyi and Li Guangbi hoped to contain Fanyang from the north. If all resources went to the front, Hebei nodes might break again; if the rear alone were cultivated, Guanzhong would feel its gateway was being lost. The later decision at Tong Pass mattered so much because the Tang did not face merely an arrow moving east to west, but was trying to join several fronts that were not yet stable.

Failure of temporary forces also altered the next mobilization. Those who returned brought real military news but also fear; new recruits heard how the preceding force had collapsed, while commanders had to decide whether to gather defeated soldiers or treat them as carriers of disorder. Defeat lost weapons, horses, and transport capacity, though this chapter has no itemized inventories proving that all these losses occurred in every retreat. “Retreat” in a war report was not a unit moving from one point on a map to another: the court had to identify anew how many people, weapons, and relationships of command remained usable.

Local offices also bore double demands. They had to give households, granaries, carts, and horses to armies while maintaining urban order, settling those who fled in, and judging which document remained valid. Officials who fled left a vacuum of authority; officials who stayed could be pursued by the other side for supplying one party. Rebuilding Tang armies was therefore also work of rebuilding credible orders: people were more likely to yield scarce grain and labor only if they believed the army would not immediately disperse.

After Yan forces crossed the Yellow River, cities along the route met different fates. Chenliu’s preparations soon collapsed and Guo Na finally opened the gates and surrendered; the Old Book of Tang explicitly says he had first offered limited resistance. Both the Zizhi Tongjian and Old Book of Tang record An Lushan’s killing of many Tang troops who had already surrendered at Chenliu; the former says “nearly ten thousand,” the latter six or seven thousand, so no number can be fixed from them. Exact totals lack reliable cross-checking and cannot simply be accepted; the political effect of the executions is clear: they punished those already surrendered while showing the next city that obedience might not ensure safety and that resistance could carry costs.

Such violence could have opposite effects. Terror might make one city open its gates quickly, or persuade another that surrender could not preserve life. Wartime orders thus continually changed local calculations. Yan forces depended on prefectures and counties to keep supplying roads, food, and administration in order to advance rapidly; massacres and plunder damaged such cooperation. The more successful the military advance, the more places an occupier had to maintain.

After resistance at Xingyang, the city fell, and Feng Changqing’s newly recruited forces retreated repeatedly. Open ground favored Yan cavalry, while repeated setbacks made temporary forces still less able to retain order. Their retreat toward Luoyang shows Tang defenses continued to withdraw; the chronicle materials cannot separately answer how soldiers, residents, and officials understood defeat.

Luoyang finally fell. The Zizhi Tongjian says that when Yan forces entered Luoyang they “let the troops kill and plunder”; the Old and New Books of Tang both record the deaths of officials including Li Cheng and Lu Yi. These materials show that occupation brought violence and official deaths, but cannot estimate the scale of harm to ordinary residents or reconstruct the condition of offices, markets, and homes across the city in a few days.

The fall of a capital-level city had its own duration. Breaking the gates was one moment; offices ceasing to execute old orders, guards changing at granaries, markets reopening or closing, and refugees seeking roads continued for many days. Yan forces had to decide which Tang officials could remain at work, which wealth belonged to the army, and whom to hunt down. A change of regime altered gates, documents, and security, yet no body of material lets us reconstruct how residents weighed leaving or staying item by item. The military result had occurred; rule was still contested in confusion.

Luoyang was not merely a large city. It connected the Yellow River, the canal, Guanzhong, and the empire’s eastern wealth, and possessed palaces, offices, stores, and ritual space able to proclaim dynastic order. Once Yan forces held Luoyang, they gained a center that could store grain, appoint officials, issue orders, and receive submissions. Chang’an now faced not simply rebels on the march but a rival regime trying to govern territory.

In early 756, An Lushan declared himself emperor at Luoyang and named his state Yan. The slogan of “punishing Yang Guozhong” could no longer contain his action. Declaring an emperor meant establishing an era name, offices, and an order of rewards and punishments; it also required followers to make a clearer choice between Tang and Yan. An armed action named as cleansing the ruler’s side openly became a contest between two regimes for all-under-heaven.

Declaring an emperor did not automatically stabilize rule. The Yan regime had to feed its army, control ferry crossings, place officials, and judge whether those who surrendered were trustworthy. With each westward advance, the route back to Fanyang lengthened; with every city occupied, people had to be left to maintain requisitions and order. Victory at Luoyang gave An Lushan the form of a regime, but also brought him to a task harder than launching rebellion.

By early 756 the war could be drawn as two lines. One ran from Fanyang through Hebei, across the Yellow River, to Luoyang, and on toward Tong Pass and Chang’an. The other lay between Hebei prefectures, Hedong, and Shuofang, where local forces including the Yan family and troops under Guo Ziyi and Li Guangbi sought to cut Yan’s rear. The two lines were unequal in strength, but they prevented An Lushan from looking west alone.

Reprisals struck Hebei resistance. The Old Book of Tang includes Yan Gaoqing in its biographies of loyal righteousness and preserves an edict of posthumous promotion in the first Qianyuan year, explaining his death with the words “his body perished but his name remains, truly displaying loyal martyrdom.” This is enough to show that Tang commendation and the official history compiled under the Later Jin dynasty together shaped a framework of loyal martyrdom, but not to substitute for later rituals and local memory. Other people connected to him often appear only as categories such as relatives, officers, or soldiers and civilians in the city; without more local or personal material, their individual experiences cannot be reconstructed.

The Tang’s northern counteroffensive offered another possibility. If Guo Ziyi and Li Guangbi could threaten Jingxing, Fanyang, and Hebei routes, Yan’s westward advance had to fear its supplies and retreat. This possibility guaranteed nothing and cannot yield a certain winning plan directly from a map; orders, provisions, coordination among commanders, and local response determined whether the two Tang forces could actually join strength.

An Lushan’s first round of victories was enormous. Tang forces retreated in succession, Luoyang fell, the Yan regime was founded, and Chang’an was directly threatened. Yet this did not decide the war’s outcome. The more territory Yan held, the more it had to keep granaries, prefectures, counties, and roads obeying; after losing the eastern capital, the Tang could still reorganize forces from other frontier armies, local official authority, and areas not yet occupied.

For the Tang court, the danger was treating the collapse of temporary forces and compliance by prefectures and counties along the route as proof that the whole state had lost capacity to resist; for Yan, it was treating rapid occupation as proof that local society had accepted a new order. Both sides could judge from incomplete news only. Whether a city still held, a road remained open, or a return to allegiance was real had often changed by the time the news reached commanders.

War also changed what “rear areas” meant. Hebei behind Yan forces could still resist, while Guanzhong, thought safe by the Tang, was about to face attack directly; a prefectural city supplying grain today might close its gates tomorrow. Control expanded and contracted along walls, stores, ferry crossings, and hostage relationships. Declaring an emperor at Luoyang enlarged Yan’s ambition, but did not remove this instability.

This is why taking the eastern capital remained far from winning all-under-heaven. Yan could issue official designations, but actual taxes and grain still had to be collected by prefectures and counties; it could order roads open, but merchants and grain carriers first judged whether routes were safe; it could punish resisters, but excessive punishment gave local people more reason to turn toward Tang forces. The Tang likewise could not recall every region with its old dynastic name alone. Both sides had to make political titles real through predictable protection, requisition, rewards, and punishments, so the war passed from rapid marching into a longer contest of governing.

On the dingyou day of the twelfth month of Tang Tianbao 14, Yan forces captured the eastern capital Luoyang; this date falls in early 756 in the Gregorian calendar, and a final version must fix the source for calendrical conversion. As Yan forces approached and captured Luoyang, the Tang court recruited troops, appointed commanders, and created new military offices in the eastern capital, Hedong, Shuofang, Henan, and the capital, while anti-Yan armed forces began to emerge locally. The crucial result was not that victory was decided, but that war could no longer end in one surprise attack. Yan had to pass Tong Pass, and Tang forces had to join northern containment into a real counteroffensive; both sides still had to solve questions of armies, roads, grain, and obedience.

The next decisive crisis came at Tong Pass. There, a pass temporarily offset Yan’s speed and concentrated Tang internal suspicion of commanders, fear for the capital’s safety, and choices among different theaters in one place.

Chapter 4. 756: Tong Pass, Chang’an, and a Court in Flight

After Luoyang fell, the critical obstacle on the road to Chang’an was Tong Pass. The pass lay between mountains and rivers, linked Guanzhong to the west and faced the Shaan–Luoyang corridor to the east. If defenders used the terrain to hold, rebels could not directly carry cavalry and rapid advance into Guanzhong. That An Lushan did not enter Chang’an immediately after taking Luoyang shows how one pass could change time.

The Tang successively had Gao Xianzhi and Feng Changqing retreat to Tong Pass. After their defeats they held the pass with their remaining troops, and Gao repaired defensive equipment. Bian Lingcheng returned to court accusing both of retreat and of withholding grain and rewards, so Xuanzong ordered them executed in the army. The court needed responsibility for repeated failures and feared that front commanders concealed news. Thus, when experience was most needed, two commanders with frontier records died within their own command system.

Their judgments were not necessarily all correct; more important, Tang fear of military power had become entangled with military crisis. The emperor needed generals to judge independently but did not trust them to report faithfully; eunuch military supervisors could oversee them, but could also turn personal conflict into life-and-death accusation. After defeat, accountability should have helped an institution correct errors, but in fear it could first remove experience still usable.

Xuanzong then appointed Geshu Han to hold Tong Pass. Geshu Han had long fought in the northwest and had great prestige, but was old and ill and had not sought the assignment. The court placed Chang’an’s safety on one pass and a physically weakened commander, while bringing to him all its earlier suspicions of frontier generals.

At the same time, the Hebei campaign briefly favored the Tang. Guo Ziyi and Li Guangbi pressed the rebels’ rear and some prefectures responded. An Lushan even feared that the connection between Luoyang and Fanyang would be cut. Holding Tong Pass and waiting for various armies to complete an encirclement could use time gradually to consume the rebels’ advantage.

The decision chain around Tong Pass must be separated. In Chang’an, Xuanzong heard differing reports from the front, chief ministers, military supervisors, and envoys; in Guanzhong, Geshu Han saw terrain, troop condition, and enemy movements; Yang Guozhong saw Geshu Han’s troops as a new political risk. Materials establish these accusations and actions, but cannot fully reconstruct inner thoughts. Officials and residents who remained knew only grain prices, rumors, palace gates, and whether defenders were still there. These did not automatically become one battlefield map. An empire could possess much information while its highest decision-maker still could not identify in time what would bear checking.

Geshu Han advocated holding the dangerous ground; Guo Ziyi and Li Guangbi also advised that Tong Pass not venture out, while threatening Fanyang from the north. The logic was that Yan forces had come far and sought a decisive battle; if Tang forces held them outside the pass and cut their rear, the outcome need not be decided at the front of Tong Pass.

But military judgment was soon surrounded by court politics. Claiming that the Tong Pass army lacked a rear support, Yang Guozhong trained sons and younger relatives of those serving in the imperial horse-pasture administration in the park and recruited ten thousand to garrison Bashang; the Zizhi Tongjian and related biographies both interpret this force as also guarding against Geshu Han. Geshu Han, noticing it, likewise took measures to control that force. Neither openly rebelled, yet each prepared for the other’s possible rebellion.

War produced a vicious circle: An Lushan’s rebellion proved that frontier generals could threaten the center, so the center trusted new commanders less; distrust then forced commanders to protect themselves, making their actions look more suspicious. Yang Guozhong viewed Geshu Han through the risk of military power, while Geshu Han took the treatment of Gao Xianzhi and Feng Changqing as a warning. “Fear” can explain relations among actions, not serve as a homogeneous psychological record of the two men. The more both tried to prevent betrayal, the more they damaged trust needed for common defense.

The court received reports that rebel troops around Shan were few and weak, and Xuanzong repeatedly sent envoys ordering battle. The Zizhi Tongjian and New Book of Tang both record Geshu Han’s judgment that Yan used weak troops to lure battle and his request to hold Tong Pass until armies gathered; after repeated imperial envoys urged battle, he finally led troops out. Chronicles dramatize his grief at leaving the pass. The surviving histories record that he favored continued defense and Guo Ziyi and Li Guangbi also asked that the Tong Pass army not venture out; Xuanzong ultimately accepted Yang Guozhong’s judgment and kept pressing Geshu Han to leave the pass.

One mistaken order cannot alone explain this defeat. The order was issued because the emperor could not judge whether front reports were faithful, the chief minister viewed the general as a political threat, the supervisory system mixed military affairs with loyalty review, and the Tang had just been betrayed by a frontier general. One battle bore the regime’s whole fear of soldiers.

Incomplete information explains only part of the defeat at Lingbao. Holding Tong Pass placed hope in time, terrain, and northern containment, at the cost of prolonged strain on Chang’an and the emperor’s tolerating a powerful general gathering troops inside the pass; seeking battle outside might quickly remove danger, at the cost of taking forces not yet integrated onto ground where rebels excelled. Standard histories preserve the outline of Geshu Han’s opposition and Xuanzong’s urging, enough to show another option existed; precise speech, orders, and what each knew of enemy conditions remain shaped by compilation and cannot be completed into a closed-room drama.

Another layer of the Tong Pass decision was who could bear responsibility for failure. Geshu Han lost at Lingbao after leaving the pass, was held by subordinates including Huoba Gui-ren after returning inside, then handed to Yan and sent to Luoyang. Yet supervisors’ reports, the chief minister’s fears, the emperor’s urging to fight, the state of temporary forces, and enemy dispositions together determined the conditions under which a commander could act. Accountability can correct error; if in crisis it first becomes a search for scapegoats, successors find it harder to report truthfully or voice unwelcome judgment.

Generals still bore responsibility for their judgments. Geshu Han, Feng Changqing, and Gao Xianzhi each had judgments, resources, and constraints, and later people may continue to debate their military gains and losses. But only by returning responsibility to the decision chain can we see that front command was not an isolated profession. A commander had to prove loyalty to the emperor, explain military power to the chief minister, account for news to a supervisor, and provide soldiers with orders they could trust; where these relations were unstable, even terrain’s advantage could be consumed by political fear.

After leaving Tong Pass, Geshu Han met Cui Qianyou in narrow terrain around Lingbao. Rebel forces used corridors between mountains and rivers, ambushes, smoke and fire, and cavalry flanking attacks; the Tang’s large formation could not deploy. Tang troops suffered a major rout on the western plain of Lingbao; after Geshu Han returned inside the pass, subordinates held him, Yan took Tong Pass the next day, and Guanzhong’s defense line immediately collapsed.

Once the front of an army on a narrow route was blocked and those at the rear did not know what was happening, greater numbers were not necessarily an advantage. Smoke, dust, and fire obscured visibility; soldiers might misjudge enemy positions, units crowded into one another, weapons could not be used, and rivers and ditches cut off the routes of retreat. The battle rapidly changed from a clash between two armies into trampling, flight, and an inability to recognize one another within a single army.

Geshu Han returned to the pass with a few men, then was immediately held and surrendered by Huoba Guiren and other subordinates. His Old Book of Tang biography says that before An Lushan he offered to write to win over generals including Li Guangbi, who replied reproaching him for failing to die for loyalty; this can only be treated as a later biographical narrative of capture and attempted persuasion. The earlier draft’s statement that rebels executed “those who betrayed their commander” lacks supporting original text verified in this round and is removed. The established irony is grave enough: a commander ordered out of the pass was defeated, handed to the enemy by his own subordinates, and then written as a tool of persuasion while captive.

Once Tong Pass broke, Guanzhong’s defense rapidly collapsed. Chang’an did not fall after a long siege, but suddenly faced rebels after its gateway was lost. The court lacked time for a second defense and many officials did not even know the emperor intended to leave. In the cyclical-day sequences of the Zizhi Tongjian and Old Book of Tang annals, Lingbao occurred on gengyin and Xuanzong left Chang’an at dawn on yiwei, five days later.

Army composition made choices harder. Geshu Han commanded frontier troops, palace guards, and hastily gathered people; their training, equipment, languages of command, and trust in the commander differed. Large forces found deployment in perilous narrow terrain difficult, and marching columns could readily become a long line blocking itself. Once misinformation, smoke and fire, or a setback in the vanguard occurred, the rear might not know whether to advance, hold, or retreat. Without accepting inflated casualty figures, one can understand why a retreat quickly became organizational collapse: what war first lost was the order through which people could see, hear, and confirm one another.

Xuanzong’s decision to enter Shu had clear geographical and political grounds. Roads to Shu were difficult, Jiannan was comparatively intact, and Yang Guozhong also held Jiannan, which could give the royal house refuge and resources. Preserving the emperor and dynastic line had practical political value. The problem was that departure was extremely secret and hurried: the court did not explain to the whole city who would defend, where officials should go, or how residents should take refuge.

Before dawn Xuanzong left through Yanqiu Gate, chiefly with the royal house, Yang family, close officials, eunuchs, and part of the palace guard; many consorts, princesses, imperial grandchildren, officials, and their families could not go. The next day, people still arrived at palace gates as usual and only then found palace order empty. The capital’s political center did not move through public handover, but seemed suddenly drawn out of the city.

Flight therefore became more than an emperor’s journey. Nobles and residents scattered; after palace gates opened, people entered imperial precincts and powerful households to take goods, storehouses burned, officials left behind barely organized order, and others quickly sent keys and signals of submission to the rebels. For those who remained, loyalty was not an abstract declaration, but whether armies, orders, and food still supported it.

After leaving Chang’an, Xuanzong’s party quickly met shortage of supplies. Local officials fled, relay stations lacked staff, and the emperor and imperial grandchildren could receive only coarse food supplied by common people; followers dwindled, and at night distinctions of rank briefly lost force. Once cooks, ceremonial guards, stores, and relay systems that sustained a grand court broke, the emperor too faced hunger, roads, and a dispersing party.

These details reveal power’s material foundation. The emperor’s title could not itself move a party; it required horses, bridges, food, light, and soldiers willing to continue escorting. The capital’s fall exposed every dependency usually hidden behind ritual.

The rebels did not immediately pursue the fleeing court. An Lushan’s side had not expected Xuanzong to leave so quickly and, after entering Chang’an, was occupied with takeover, searches, appointments, and plunder. This left time for Xuanzong’s entry into Shu and the crown prince’s move north. Yan captured the capital but did not at once turn political victory into the complete destruction of the Tang royal house.

The ensuing withdrawal was far from a united and dignified “western progress.” The imperial party depended on relays, stores, supplies along the route, and escort troops. The Zizhi Tongjian explicitly records Xuanzong at Fufeng distributing spring cloth sent as Chengdu tribute to soldiers who followed in haste without time to bid families farewell, and worrying that prefectures and counties on the Shu route could not supply the large party. This shows supply pressure, not overall intentions of local officials, relays, soldiers, or civilians, nor how the entire road system failed. The earlier claim that civilians supplied food lacks a precise volume-and-passage reference and remains removed.

Order after Chang’an’s fall also changed in layers of time. First could come dispersing guards, stalled offices, and spreading rumor; only later came how a new regime placed troops, opened stores, requisitioned goods, and issued documents. Different neighborhoods, occupations, and statuses did not necessarily feel the same change on the same day; accounts left by officials who escaped and people who remained would differ. Later memory most easily retains court panic and famous individual fates, but a great city’s change of hands also includes whether markets open, wells are maintained, roads passable, and families can find one another. These questions do not reduce history to daily trivia; they explain how military occupation can persist and meet resistance.

Tong Pass’s failure also changed the distance between Guanzhong residents and the court. The order usually sustained by offices, wards and markets, and stores suddenly looked fragile after troop retreat and the emperor’s departure. Some may first have encountered new requisitions, some price and goods fluctuations, and some tried different roads to avoid armies; such experiences do not form a neat survivor narrative. Losing the capital damaged more than a political symbol; it interrupted urban supply and governance. This chapter lacks resident cases and cannot go on to say how they chose food, protection, and rules.

At Mawei, the escort soldiers mutinied. When Xuanzong’s party reached Mawei relay station in 756, the escorts killed Yang Guozhong; the Zizhi Tongjian gives the cyclical day bing-shen, while the Old Book of Tang annals’ web transcription gives bing-chen, a variant still requiring collation. The soldiers then did not lift their encirclement and demanded that the consort not continue with the party; under this coercion Xuanzong ordered Yang Guifei’s death. That is the stable factual frame. Details of means of death, final dialogue, and Xuanzong’s feelings have been repeatedly rewritten in poetry, tales, paintings, and drama. Later people readily compress dynastic rupture into a love tragedy because it is easier to remember than frontier commands, military rations and pay, and command systems.

Yet Mawei was first a political event in which soldiers presented the emperor with life-and-death conditions. Accounts of supply, resentment, and Yang Guozhong’s responsibility overlap with later narratives; this version preserves only locatable actions and public demands, not a shared psychology for the entire army. Entering Jiannan might also have meant entering territory controlled by Yang Guozhong. The army would no longer proceed on imperial command alone. Xuanzong remained emperor, but had to choose between his closest court ties and the only force on which he could rely nearby.

Yang Guozhong’s death cut off a political center that had lost the army’s trust; Yang Guifei’s death was demanded as further assurance. Whatever Xuanzong felt personally, the political result is clear: imperial power could not protect the most favored person, while the army could decide whether the party moved on. At this moment politics compressed private relationships into the cost of sustaining obedience.

After Mawei, elders, scholars, and common people asked the emperor or crown prince to remain and organize resistance. Xuanzong resolved to continue to Shu; the crown prince Li Heng separated from his father when people asked him to stay. The accession at Lingwu and the formation of a counteroffensive center involved armies, officials, edicts, and the military situation; whether and how ordinary people influenced the decision requires separately locatable evidence.

Xuanzong continued west while Li Heng turned north. Chang’an was occupied: one emperor went toward Sichuan to preserve the old court, while a crown prince with very few men sought a frontier command still able to fight. The Tang had not perished, but no longer had one center.

The later painting Emperor Minghuang’s Journey to Shu turned this flight into a procession in landscape. The National Palace Museum’s work takes Xuanzong’s entry into Shu as its subject; the museum considers the extant image to have features of eleventh- and twelfth-century copies, so it is material of later historical memory, not an image made on the scene in 756. Placed in this chapter, it must be described as memory: a refined picture preserves a wretched escape, while to some extent moving from view urban disorder, hunger along the road, and families outside its frame.

Chang’an’s fall ended the condition in Xuanzong’s reign in which emperor, court, capital, and imperial center coincided. Xuanzong remained, Chang’an remained, and official designations remained, but no longer in one place. Flight preserved the emperor but left a question for the northbound crown prince: if the old court could not immediately direct the war, could the Tang establish a court elsewhere?

The material hardship of flight also changed authority’s visible form. In Chang’an’s palace, the emperor’s commands were amplified by ceremonial guards, offices, storehouses, and fixed ritual; after leaving the city, an order without horses, food, guides, and guards to continue it was only words. Mawei soldiers could make demands not simply because they suddenly gained political courage, but because they held force essential to the party’s movement. Xuanzong’s choice was therefore personal tragedy and a moment when state capacity was redistributed on the road.

After father and son took separate routes, the two paths faced different obstacles. Those entering Shu had to cross mountain roads and sustain the royal house and old officials; those moving north had to skirt war and find frontier forces still operating. Neither safety nor danger alone summarizes the routes. They preserved two resources: the old emperor’s symbolic continuity and Jiannan’s geographic depth on one side, and Shuofang strength able to reorganize war on the other. The Tang later avoided immediate collapse because resources overlooked by battle maps did not all break at once.

Violence after Mawei especially should not be folded into a moral fable. Soldiers’ demands reflected anger at the Yang group and explanations of the war, but also bargaining power gained by armed people amid hunger, unpaid rewards, flight, and confused command. Xuanzong’s acceptance had personal and ritual consequences, and was a political decision under pressure over whether the party could still move. Specific language and details in histories have strong later narrative coloring, so this chapter does not stage a scene from them; but they point to a clear structural fact: when supply and command broke, decision-making once monopolized by court was redistributed to those bearing arms. Chapter 5’s reorganization at Lingwu begins in this environment of divided power.

Flight also changed the form of knowledge. In Chang’an, the court could rely on fixed offices, archives, and close attendants; after leaving the city, news ahead came more from passersby, local officials, and scattered soldiers, and true and false could not be quickly distinguished. Every stop could change advice, route, and supplies; materials do not continuously record who felt what fear. This instability directly affected choice and is an important condition for the later separation of Xuanzong and the crown prince: they did not face a clear security map, but options continually rewritten by defeat, roads, and people’s responses.

Events from Tong Pass to Mawei thus form one chain: the court first overturned a plan to hold the pass amid suspicion, Lingbao then shattered the military barrier, secret flight deprived the capital of public handover, and escort soldiers finally forced the emperor on the road to accept conditions. The Tang preserved Xuanzong and the crown prince, but lost the center where emperor, court, capital, and military command had coincided. After Li Heng moved north, the next chapter asks whether dispersed claims, troops and horses, and funds and grain could again compose a court.

Chapter 5. Lingwu: One Dynasty, Two Political Centers

While Xuanzong’s party moved toward Sichuan, Crown Prince Li Heng did not remain with it. After Mawei he separated from his father and turned northwest. Later orthodox narratives treat this route as the beginning of restoration, but on the day of separation Li Heng could not know whether he would reach Lingwu or whether other places would recognize him.

At first, the crown prince had very few people with him. His party crossed the Wei River and traveled north through the night. Along the way, a clash broke out because defeated troops were misidentified. The party continued to lose soldiers, equipment, and horses. Those without horses could not continue; some local officials fled, and the party could only recruit people and procure food anew along the route. The future emperor of a dynasty was leading something far from an orderly army at this point. He was searching for the next place that could still provide horses, grain, and a network through which to issue orders.

The journey shows that royal status could not create state capacity by itself. Li Heng could call to people loyal to the Tang, but still needed to know who held stores, which army remained intact, and which road was safe. Histories say that after he reached Pingliang, acting Shuofang commander Du Hongjian and others sent Li Han to meet him and report figures for Shuofang soldiers, horses, armor and weapons, stores, grain, silk, and military supplies. Lingwu could become a center first because it still had resources that could be counted.

The Shuofang army was an important Tang frontier force. It was relatively intact and could connect northwestern strength, horse pasture, and the Guanzhong direction of counterattack. Li Heng chose Lingwu in order to be near a military organization still functioning. This formed a central paradox of the An Lushan war: to suppress a rebellion launched by a powerful frontier general, the Tang had to rely more deeply on other frontier armies and local commanders.

To see Lingwu as merely a city where the crown prince happened to stop understates its military world. Shuofang faced northern steppe and northwest frontiers over time, while Hexi and Longyou forces faced Tibet and wider northwest communications. Like Fanyang and Pinglu, they were permanent systems formed to manage distant borders. Suzong could make Lingwu a counterattack center not because frontier forces naturally obeyed the center, but because commanders, soldiers, stores, and road networks there were willing or compelled to link their own safety to Tang continuation.

These military commands could assist and constrain one another. Hexi and Longyou were not rear areas that could be emptied at will; Tibetan pressure required defenses to remain. If Shuofang sent troops east, it also had to consider the northern steppe and its own supply. For Suzong, the real questions were which troops could move, who would hold the original frontier afterward, whether provisions could follow, and whether commanders accepted the new court’s restraint. Lingwu made clearest the paradox of Tang frontier forces suppressing a frontier-force rebellion.

After Li Heng reached Lingwu, officials repeatedly asked him to accede. He first refused, perhaps from ritual concern and political risk: Xuanzong still lived, and accession without his father’s direction could be called a seizure. But many accompanying soldiers came from Guanzhong and urgently wanted counterattack; without a highest title able to issue edicts, appoint generals, and gather obedience, the party might disperse again.

In 756 Li Heng became emperor at Lingwu, Emperor Suzong of Tang; the Old Book of Tang annals place this on the jiazi day of the seventh month of the first Zhide year. The rite occurred outside Chang’an palaces, with few civil and military officials, and court institutions could only be improvised. Chronicles describe people talking with backs turned from palace gates and a censor attempting to rebuild court ceremony through impeachment. Such details may be moralized, but they capture the new court’s problem: it needed not only an emperor, but soldiers, officials, and newcomers to believe an order existed that could restrain them.

Under very limited conditions Suzong had to do four things at once: show dynastic legitimacy for accession; appoint people able to organize war; have officials elsewhere recognize Lingwu edicts; and obtain grain, pay, troops, and allies. Failure in any one could leave Lingwu only an exiled party calling itself a court.

Honoring Xuanzong as Retired Emperor was a key arrangement for legitimacy conflict. It placed a power transfer already made within dynastic continuity: the old emperor was not publicly deposed and the new emperor undertook to suppress rebellion. Only when Xuanzong later sent credentials and recognized the fact could a possible father–son struggle for the throne be contained within ritual and power negotiation.

Tension remained. Xuanzong in Shu retained old officials and great symbolic weight, while Suzong held the rebuilding military-administrative center; appointment, representation of Tang, and final authority after Chang’an’s recovery could all cause conflict. The Tang survived partly because it did not publicly begin another father–son civil war at the most dangerous point of external war.

The coexistence of Lingwu and Sichuan also required two senses of time. Those entering Shu prioritized sustaining the royal house, settling old officials, and holding mountain roads; Lingwu needed quickly to turn military news, goods, and appointments into counteroffensive capacity. Both acted in Tang’s name but need not give the same priority to one matter. Titles and documents could temporarily sustain common legitimacy, while the distribution of actual resources continued to create tension. This difference explains why recovery of both capitals would not naturally restore the prewar court.

Writing Lingwu only as the point “Suzong acceded” misses how the state reappeared. Du Hongjian, Pei Mian, Wei Shaoyou, and others turned local military-administrative resources into usable court organization. Who managed documents, finance, counts of troops and horses, grain and silk, and repairs to the station determined whether orders traveled from the emperor to armies.

The new court constantly received memorials from all directions: war news, beacon signals, local submissions, military grain, and requests for office, arriving day and night. Someone had to prioritize, seal, forward, and check them. Suzong could not handle every matter himself or rely entirely on old court procedure. Li Bi’s participation in military-state planning and Prince of Guangping Li Chu’s entry into the marshal’s system show decision structures being remade under war pressure.

Order was also built through small acts: whether officials could impeach arrogant generals, soldiers had to observe court rites, rewards could be restrained, and documents had a fixed entrance. These may seem less dramatic than battles, but determined whether Lingwu was an emperor in a camp or a court able to restrain the camp.

Lingwu also had to turn orders into political signals places could choose to accept. Occupied regions and distant prefectures did not submit merely because a new emperor acceded; they asked whether Lingwu had troops, could endure, and issued authentic edicts. Each official who came over and army that accepted appointment added credit to the new court. An accession edict wrote legitimacy; others’ willingness to risk recognizing it tested legitimacy.

For ordinary people, reorganization first appeared in who recruited troops, collected grain taxes, decided what document was valid, and whether refugees could be settled. With two political centers, one family might have members entering Shu with Xuanzong, staying in Chang’an, or requisitioned by Shuofang. Dynastic continuity was not abstract legitimacy to them, but whether different orders would fall on them at the same time.

Histories preserve the story of Du Hongjian and others reporting resource figures, not a full surviving inventory. Figures for armor and weapons, horses, grain, and silk need not be exact stock; they first show a new political center needed measurable goods to prove it was not an empty title. Counting converted local resources into state resources that could be rewarded, allocated, transported, and accounted for. Without inventories an emperor could not judge how many people could be supplied; without documents and responsible people, military grain was grain in a store, not grain deliverable to a unit.

Lingwu could issue orders, but it still could not make them cross the battle lines immediately. Edicts had to be carried by envoys and relayed through prefectures and counties, military camps, and transport hubs. Whether an edict could take effect depended on its transmission and authentication, those receiving the command, and the military force available to carry it out. Whether an individual obeyed out of belief or fear requires specific evidence. The reasons local officials submitted must be judged case by case from their memorials, biographies, and the particular circumstances; a single set of motives cannot be inferred from the fact of submission. To hold these different motives together within a shared political alignment, the new court could not rely solely on repeatedly proclaiming its legitimacy. It also had to make appointments, rewards, pardons, and supplies of military grain reasonably predictable. What is called the restoration of legitimacy took place gradually through these small arrangements that people could test for themselves.

At Lingwu, the court also faced mutual dependence of military action and political expression. An appointment with no troops able to receive it was only a paper title; an army without recognized rewards, punishments, and sources of rations and pay could not fight long. Civil officials, generals, and local people around Suzong handled rites and battlefield affairs together because they were solving how to make orders credible. Edicts, inventories, envoys, and camp discipline looked minor but showed the new center could endure.

Shuofang became the new center’s pillar, and commanders such as Guo Ziyi and Li Guangbi gained key positions. Suzong could not rely only on personal loyalty but had to make different armies act toward a common goal. Yet those armies had their own commanders, supplies, and regional ties; the center was rebuilt in war but could not return to an old pattern in which the palace directly controlled every unit daily.

Military mobilization required finance. Grain, horses, weapons, transport, and rewards did not appear through the word “restoration.” Jiang-Huai and other regions not controlled by Yan gradually became key sources of fiscal revenue, and transport routes had to cross war-affected territory. Later proposals by Diwu Qi for new transport and salt policies show traditional fiscal routes could no longer easily meet war’s demands.

Fiscal change had already begun early in the war. Each recruitment and alliance needed immediate payment, and each lost place could cut an old tax source. While organizing counterattack, the Lingwu court learned how to extract resources from an empire of broken territory and unstable roads. Many fiscal changes of the middle and late Tang should not all be attributed to the An Lushan Rebellion, but war did accelerate the center’s search for new revenues and transport methods.

Suzong also had to balance military expertise and political reliability. An Lushan’s betrayal made the court wary of strong generals, but counterattack required actual command for Guo Ziyi, Li Guangbi, and others. Military supervisors, the marshal’s office, and civil participation in military decisions could coordinate but also cause new friction. The Tang could not escape this problem thereafter: a general most able to save the state could also be suspected for great merit and troops.

Lingwu’s fiscal problem was far more than “lack of money.” Grain, horse fodder, weapons, carts and boats, rewards, and gifts to envoys needed different resources; a reappearing court had to decide who could collect, transport, receive temporary exemptions, and which stores supplied armies first. Jiang-Huai revenues later mattered more as northern war cut old supply. Diwu Qi’s transport and salt reforms belong to later development and are not presented as mature Lingwu institutions; but from the beginning war forced a search for revenues not dependent on old capitals and roads.

For soldiers and local people, new fiscal arrangements fell directly on places: where grain was demanded, who escorted transport, which unit received reward first, whether merchants crossed checkpoints, and whether refugees gained settlement. After the two capitals fell, if the court could not make requisitions seem regular and rewards actually arrive, armies and prefectures could hardly believe Lingwu orders were worth risk. Finance was inside battle; it determined whether an army could wait, march, hold a city, and avoid disintegrating from hunger.

Once frontier forces entered the new court, military-civil relations took a new form. Generals needed the court’s titles, grain and pay, and rewards of merit to enlarge followers; the court needed generals’ troops, local knowledge, and established prestige to restore capacity. They needed and guarded against each other. An Lushan’s experience prevented Suzong from wholly trusting strong generals, while the immediate war did not allow command to be withdrawn into paper institutions. Supervisors, the marshal’s office, and civil involvement could coordinate or cause delay and suspicion. In 756 this had no answer. “Restoration” was not one-way recovery of central power, but a redefinition of what the center could do within new dependencies among armies, localities, and allies.

For refugees and local residents, a new center also brought heavy demands. Garrisons needed lodging, firewood, fodder, and labor; new officials checked households, stores, and roads. War risk affected transport, but this chapter has no merchants’ accounts or travel records and cannot write this reasonable inference as observed common choice. Some gained work in supply, transport, or office, others bore heavier requisitions and uncertainty. Materials cannot quantify these experiences as one fate for all, and Lingwu need not be depicted as universal rescue. It established a center through which war could continue and some people had to keep bearing its costs; later counterattack grew from these limited, costly capacities.

The new court began seeking aid from the Uyghurs and other powers. In 744, after defeating the Basmils, the Uyghurs established a steppe empire centered on the Mongolian Plateau and later expanded over former Turkic lands and several peoples. In seeking troops, the Tang faced another strong power with its own goals and had to negotiate.

Envoys crossed unstable areas carrying titles, marriage, and promises of return. Whether the Uyghurs would send forces depended on Tang-Yan war conditions, negotiation, and available returns; this chapter does not make these conditions a recovered single formula of decision. The Tang needed northwestern cavalry and had to accept that allies would not fight free of charge. Marriage, silk, horse trade, and allocation of benefits after recovering cities became parts of alliance.

Seeking outside aid also placed urban residents at new risk. Allied troops could help defeat Yan but might plunder after entering cities; terms Tang promised to recover capitals would be borne by cities and finance. To write “allies arrived and saved the Tang” removes alliance costs and reduces Uyghur political choice to a supporting role in a Tang story.

Negotiation with the Uyghurs had a clear earlier history. After 744 they replaced the Turks as a major power on the Mongolian Plateau, and the Tang had already maintained tribute, marriage, trade, and military relations with steppe polities. The An Lushan war made this relationship more urgent but did not turn the Uyghurs into forces Tang could command. Uyghur rulers assessed Yan and Tang strength, steppe competition, gains from trade, and risks of sending cavalry south. Tang envoys offered not simple pleas for rescue but negotiations over titles, marriage, silk, markets, and future booty.

When Li Heng acceded at Lingwu in the seventh month of 756, the Uyghur auxiliaries later involved in recovering both capitals had not yet arrived; exchanges of envoys, Tang recruitment, and their arrival all followed accession. Large-scale cooperation and recovery came in later years; this chapter concerns how terms of negotiation formed. Pouring later victory back into Lingwu makes it appear Suzong had ready external aid. More truly, the Tang tested who might become an ally and reserved resources for prices an ally might demand. Alliance was part of counterattack capacity but always had price, uncertainty, and consequence.

This international and frontier setting changes what “restoration” means. Suzong had to reorganize not simply Guanzhong but an empire moved by northeastern rebels, Tibetan pressure, steppe alliance, and Jiang-Huai fiscal revenues. Each troop moved from the northwest to civil war could expose a gap to external opponents; each promise to an ally could impose future costs on finance and city residents. From the beginning recovery could only seek survival in a new balance of power.

Uyghur negotiation further shows that Lingwu’s choices always involved forces beyond Tang. Allies’ demands for silk, marriage, trade, and postwar benefits entered city, fiscal, and ordinary life, while their cavalry brought mobility Tang troops could not supply themselves. Calling this “foreign assistance” erases both sides’ calculations. More accurately, as Tang approached collapse, it tried to connect another polity’s strength to its war on terms it could bear but which would leave consequences.

The Uyghur relationship should be judged by the same measure. Tang needed its cavalry and political support, while Uyghurs weighed benefits and risks of sending forces and whether Tang could fulfill promises. Alliance was two-way exchange and cannot insert later recovery of the capitals into negotiation in 756. Lingwu first gained the standing to negotiate again: it could send envoys, promise exchange, allocate resources, and have outside polities see a Tang future. That standing required Tang to meet costs, but was a key step from an exiled dynasty to a war center.

Lingwu’s value lay in reconnecting scattered threads: the northbound crown prince brought dynastic succession, Shuofang a relatively intact army, He-Long frontier experience and strategic depth, civil officials turned goods and edicts into order, and Uyghur negotiation placed Tang crisis in wider Eurasian politics. None guaranteed success; together they let a dynasty that lost both capitals again have conditions to fight, negotiate, and be recognized.

Lingwu gave the Tang a center able to issue orders, allocate existing military-administrative resources, and organize counterattack. It did not recover capitals or resolve new dependencies among military power, fiscal revenues, and allied armies. The next volume turns to Hebei and Henan: whether war could continue depended not only on recovering Chang’an, but on how armies, cities, waterways, and local resources in the north were contested by different regimes.

Chapter Six · Hebei and Henan: One War Has Two Maps

After Chang'an fell, what the court could most readily see was the capital: whether palace gates were shut, whether edicts could be issued, and where the emperor was. But the war was not decided in the capital alone. Guanzhong and the roads to Shu bore on whether the imperial house and court could maintain a foothold; the plains, waterways, towns, and military garrisons of Hebei and Henan formed the battlefield through which every side moved troops, grain, and people. An Lushan's rising was rooted around Fanyang. After the rebels entered the two capitals, Hebei and Henan remained regions that had to be confronted if the war was to continue and be brought to a close. Recovering a capital could restore a political center; whether stable control could be established in these regions still has to be shown through the later course of fighting and local change.

What the Tang faced here was a military network that could not be eliminated in a single decisive battle. An Lushan carried the personnel, officers, transport relationships, and local control of the former frontier commands into the war; his subordinates formed new networks of interest through campaigning and occupation. The allegiance of prefectures and counties bore on gates, stores, requisitions, and military protection, but local calculations differed, and this chapter does not compress them into one choice shared by every locality. Local officials, powerful local families, defenders, and scattered soldiers could all be forced to choose within a very short time; one submission need not have been the last choice. Later standard histories often organize such stories by judging people as loyal or treacherous, but at the scene of war a change of regime first meant whose hands took the grain, by whom labor was requisitioned, and who was entitled to carry weapons.

For the Tang court, the deeper difficulty was that it could not settle all this with central standing forces alone. Before 755, Tang military institutions had already moved from the earlier fubing arrangement toward long-serving professional armies and frontier-command forces; after war began, temporarily recruited troops, local armies, and forces held by different commanders expanded rapidly. The court needed them to fight, yet found it increasingly difficult to return them fully to the old administrative order. It is still too early to write the later pattern of military governorships back as an established fact, but the war had already made ‘who commands troops, who supplies them with grain, and who can keep soldiers’ more urgent questions than official titles.

Hebei and Henan in 756 already foreshadowed the war of the following years. They foreshadowed its most perplexing fact: rebel leaders would die and capitals would be recovered, yet the local networks, military organization, and mutual suspicion that sustained the war would not disappear with them. If the Tang wished to end the war, it had to re-enter these places; the way it did so would in turn change the dynasty itself.

After An Lushan's rising, Hebei quickly became a set of interlocking battlefields. Fanyang was the rebels' most important former base, and Luoyang the eastern capital they captured; between them lay a region linked by prefectures and counties, waterways, bridges, storehouses, and former military camps. Whoever controlled these nodes could feed a force on time, send the wounded, sick, and captured goods to the rear, and make new orders arrive before an opponent's. Tang resistance in Hebei was therefore not simply an effort to recover several commandery names on a map; it sought to cut the rebels' ability to sustain their flow of strength from the northeast into the central plains.

The restoration of Tang allegiance around Changshan and Zhao Commandery especially reveals this relationship. Local officials and soldiers and civilians sometimes rebelled against Yan forces out of nominal loyalty to the Tang, and sometimes because the rebel main force had already moved south and defenses in the rear were relatively weak. Both explanations can be true at once. For a prefecture or county to change its banner openly required people willing to bear political risk, but also usable weapons, grain, clerks, and routes for sending news within the city. To write local resistance only as the integrity of a few officials erases the residents who supported them and were first exposed when reprisals came; to write it only as a vacuum of power erases the judgment by which people still chose sides amid uncertainty.

For a time the situation in Hebei favored the Tang. Guo Ziyi and Li Guangbi led the Shuofang army from Hedong through Jingxing into Hebei and fought Shi Siming's forces around Changshan. An Lushan then sent additional northern reinforcements, including the forces of Cai Xide and Niu Tingjie, from the directions of Luoyang and Fanyang, showing that Yan had to commit more troops to the northern theater. Yet this advantage had conditions: the Tang main force had to remain in the north, and Chang'an also needed sufficient time to integrate newly recruited troops and forces defending the passes. After Tong Pass fell and Guanzhong was endangered, forces that had been able to threaten the rebels from the flank and rear were compelled to turn. This was not the simple chain of ‘one battlefield was lost, so another was lost,’ but the result of a center with limited resources having to decide which place to save first. One of war's cruellest features is that every reasonable rescue may abandon another opportunity that had been taking shape.

For Yan as well, Hebei offered no natural loyalty. An Lushan's old followers, frontier-command soldiers, local officials compelled to submit, and contingents assembled at short notice did not expect the same things from the new regime. Some saw Yan as a force that could bring military merit, rewards, and avenues of advancement; some submissions occurred under pressure from killing, plunder, and requisition; explaining an individual's motives still requires case-specific material. Others waited for Tang forces to come near again. To sustain its advance, the rebels had constantly to appoint garrison commanders, requisition or seize grain and fodder, and punish the suspect. The farther they advanced into the interior, the more they had to turn the rear into resources usable by the front; and the more they governed in that way, the more easily they produced fresh resentment and flight.

In documents from the center, a prefecture's ‘submission’ appears to be a clear point: a memorial is presented, office received, banners changed, taxes paid. In Hebei, however, submission was often first a dangerous stretch of time rather than a completed state. After a prefecture or county turned toward Tang forces, it had to decide at once who would handle defense, how former Yan troops would be treated, whether stores remained sufficient, and who could carry messages for the new regime. If the Tang main force could not come near promptly, a city that had just restored allegiance might be isolated among enemy-held forces. Its officials and residents would hope for relief, but would also be compelled to consider how to survive if relief did not arrive.

This helps explain the repeated submissions, renewed rebellions, and submissions again so common in historical writings. They certainly included the opportunism of powerful figures, grudges among former followers, and calculations over rank and office, but they cannot all be written as a locality's ‘lack of loyalty.’ Wartime allegiance had to be sustained by an army able to garrison, a storehouse able to pay, and a road that was not cut off. Without these conditions, public loyalty might merely expose an entire city to the next reprisal. Conversely, Yan did not maintain its rear only through terror; the Yan regime used posts, spoils, and relationships of trust to hold its armies together. These arrangements can explain mechanisms of obedience, but cannot prove that every follower made the same calculation about safety. Politics on both sides had to come down to city gates, granaries, and whether people stayed or left.

Resistance around Changshan was therefore not an insert on loyalty unrelated to the main line. It forced Yan forces to turn back to deal with their rear, gave Tang forces openings they could use, and showed whether the Tang court could connect local initiative to a sustainable military-administrative network. After Tong Pass fell, Tang forces could no longer support this northern line with the same intensity, and local choices lost an important external guarantee. This cannot be explained only by saying that local people were insufficiently steadfast; one key constraint was the center's declining ability to fulfill promises in crisis. Choices made under such conditions were later assigned loyal-or-treacherous labels in the standard histories.

Henan was different again. It contained both a political center like Luoyang and water-transport routes such as the Bian Canal and towns linked to Jianghuai. Holding Henan meant more than holding a stretch of wall: it meant preventing the war, as far as possible, from cutting off northern military grain, fiscal revenues from Jiangnan, merchants along the canal, and the court's ability to reassemble. The rebels understood this as well. Fighting in Henan therefore often centered on unremarkable county seats, crossings, and grain routes; they may not stand out in later narratives as Chang'an does, but they determined whether large armies could keep moving in the next season.

Henan's significance was different in another way. Luoyang was the eastern capital, and its fall first shook the symbolic order of the dynasty; but it also lay in a transport world connecting Guanzhong, Hebei, and Jianghuai. The Bian Canal, Yellow River tributaries, crossings, and towns along the routes sent grain, cloth, soldiers, and documents in different directions. Many operations survive in histories only as a short sentence that ‘a certain army reached a certain prefecture,’ yet behind it lay whether enough boats were available, bridges still passable, and local stores already emptied. A large army cannot exist on one victory alone; it must eat each day, replace horses, repair weapons, and accommodate the wounded. Control of Henan was therefore part of the capacity for continuous warfare.

This also explains why cities such as Suiyang and Yongqiu repeatedly became points of contest. They were less renowned in later times than the two capitals, yet lay on routes toward Jianghuai. If the rebels expanded south, the regions from which the Tang raised revenue and moved grain would come under more direct pressure; if Tang forces held these hinges, they could still channel southern resources to the northern theater. Military grain had to pass through production, collection, storage, and transport before reaching a camp. David Graff's research supports this institutional chain; local account books are still needed to establish which farmers, merchants, boatmen, or corvée laborers bore it in particular. War linked these people into one chain, and made each person in it bear orders arriving from different directions.

The Tang position in Henan was especially difficult because it had both to rebuild local trust and to obtain supplies from localities immediately. Newly arrived officials needed to show that they were not simply another army come to requisition grain; defenders needed to convince residents that continued resistance was not a certain death; yet the court lacked time to wait for trust to recover naturally. Pardons, rewards, temporary appointments, and requisitions therefore often appeared at the same time. For the center, all were tools for restoring the dynasty; for local people, they could mean a new order that had not yet stabilized but had already begun to demand obedience.

Henan's waterways should not be drawn as a blue line that was always usable. Rivers were affected by season, water level, vessels, bridges, and military control on both banks; once a crossing fell into the hands of enemy soldiers or bandits, a shipment might not merely arrive days late but fail to arrive at all. The same was true on land. Horses, convoys, and corvée laborers needed supplies in villages and towns along the route, and when armies were numerous, roads themselves became packed with baggage, the wounded and sick, and refugees. The ‘advance,’ ‘retreat,’ ‘rescue,’ and ‘relief’ in histories always differed in speed in such spaces: orders could be written quickly, but grain and people could move only step by step.

This difference in speed was precisely what the Tang court faced in restoring administration in Henan. It hoped to move the fiscal revenues of Jianghuai north while preventing excessive requisition from immediately breaking the route; it hoped newly submitted prefectures and counties would provide defense while leaving them enough grain and manpower to maintain themselves. No single edict could resolve these aims. If local officials requisitioned too quickly, they might produce flight and concealment; if too slowly, armies at the front might run out of grain before relief arrived. Historical materials preserve these tensions as edicts, memorials, and judgments of commanders' merits and faults, and less often record how carriers, boatmen, and farmers negotiated or evaded them. We can therefore confirm that the pressure existed, but cannot invent a local response that everyone saw or accepted.

The two maps ultimately meet here: Hebei's sources of troops and military commands, and Henan's towns and water transport, both required the Tang to turn nominal restoration into continuously repeated organizational work. From 757 onward, when the two capitals were recovered, the court did not receive a blank map, but these two maps already cut apart by war and still connected to each other. Later northern campaigns still had to pass through regions where control had repeatedly changed hands, stores were unsettled, and local military orders had yet to stabilize; the difficulty lay in particular roads, supplies, and relations of control.

From this period onward, pressure on ordinary people in war should not be understood only as ‘being caught in the flames.’ Conscription, defending cities, carrying grain, building fortifications, flight, and providing temporary supplies were themselves parts of the machinery of war. Some were enrolled in armies; some were compelled to transport provisions; some temporarily preserved life behind a wall; others lost their route of escape when gates were closed. Armies did not necessarily come from afar: when a city faced siege, local people were mobilized as defenders, local militia, or drawn into different armed forces. Modern military history reminds us that the burden of Tang warfare did not fall equally on ‘all common people’; in major theaters such as Hebei and Henan, however, it was enough to remake a generation's household arrangements and local relationships.

In major war zones, ordinary people's movements were difficult to separate from armies' movements. Some followed troops in search of protection or food; some fled toward prefectures and counties where relatives lived before roads were sealed; some were conscripted for corvée labor, and military orders determined their routes. Others remained in villages, not because they chose a side, but because the elderly, children, sick people, fields, and houses made departure impossible. It is difficult to reconstruct these choices one by one from surviving standard histories; they are better at recording commanders' advances and retreats, the possession of cities, and court edicts. Precisely because the material is so uneven, the unnamed must not be written as a uniform ‘common people.’

Local society was not simply waiting for armies to decide its fate. Who guarded a gate, held a granary, admitted refugees, or organized a village's self-defense affected whether an army entering encountered emptiness, resistance, or temporary cooperation. Such action need not be romanticized as concerted solidarity. When resources were tight, local people and newcomers, wealthy households and the propertyless, those able to bear arms and those unable to flee never received equal protection. How long a city could hold may have depended on its commander, but also on who was allowed grain, who was sent into danger, and who was left outside the gates.

Prefecture names, arrows, and cities on a war map easily write flows of people as armies' appendages. Armies passing through could produce movements of goods and people, but existing material is insufficient to reconstruct the order in which ordinary households carried belongings and cared for members. Some entered cities for refuge, some migrated toward kin, and some were stopped by roads and checkpoints. Surviving material rarely provides continuous migration records, and especially does not tell us on which day a household made its decision. This text therefore does not invent a uniform route of flight for them; but the repeated appearance of empty cities, city defense, requisition, and submission in Hebei and Henan shows that movement of people had become part of controlling space.

The same map cannot show who possessed greater room to choose. Those with food, vehicles, or connections may have found it easier to leave or negotiate; conscripted laborers, the poor, the elderly and children, and the dispersed were more likely to be left out of decisions. Such inequality cannot be measured precisely from one edict, but it reminds us not to write ‘local society’ as a single actor. Both Tang and Yan needed it to provide manpower and material, yet neither could fully record what it endured.

Maps also flatten time. By the time the court received a victory report, local conditions may already have changed days earlier; a city may look positioned between Tang and Yan forces, but whether it could wait for reinforcements depended on rivers, roads, seasons, and the speed of troops. Late information changed choices themselves: a local official might defend a city or surrender on the basis of an already outdated military report, while a commander might request more grain and soldiers on incomplete information. The standard histories let later readers see a connected chronology, while people at the time often judged amid mutually contradictory news. Only by writing this gap into Hebei and Henan can we understand why the same order had different results in different places.

A further element maps often overlook is the speed at which news arrived. A report received by the court in Chang'an or Lingwu that ‘a certain commandery has surrendered’ or ‘a certain army has arrived,’ after memorials, relays, and retellings, did not necessarily equal the local situation that day. Commanders also chose how to report losses, spoils, and danger in order to seek reinforcements or shift responsibility. Wartime politics was therefore not a one-way line of orders falling from center to locality, but an information network continually delayed, cut off, exaggerated, and rewritten. Hebei and Henan were hard to bring back under Tang rule not only because they were distant, but because other military orders, other stores, and other channels of information could already determine action first in the localities.

Hebei and Henan were therefore not a single space covered by the term ‘main theater.’ Hebei lets us see how frontier-command armies possessed a rear that could sustain continued fighting; Henan lets us see how capitals, canals, and Jianghuai were drawn into a supply network. Together they show that the war of 755–756 was not complete when An Lushan captured Chang'an. The farther the rebels advanced, the harder their rear became to control; the farther Tang forces retreated, the more they still had to reassemble armies, transport, and orders from localities. One root of the repeated recoveries, renewed losses, summons to submit, and renewed rebellion of subsequent years lies here.

Hebei and Henan were also places where the power to levy was repeatedly contested. For an army, taking a city did not only mean acquiring its walls; it also meant taking over or reopening stores, markets, and labor service. In war, however, earlier accounts might be incomplete, while flight and abandoned land weakened the resources that could be levied. If officials demanded only according to prewar quotas, they could force people unable to bear them to flee further; if they relaxed completely, garrison troops would lack grain and pay. Both regimes might use temporary rewards, additional levies, confiscation, and loans to fill the gaps. These could meet an immediate need, but would also make local people more wary of the next army to arrive.

We do not possess complete account books sufficient to compare the actual tax burdens of every prefecture, and cannot write this section as a precise fiscal curve. Its point is to explain a connection often omitted: military control had to be supported by everyday extraction, and everyday extraction in turn changed people's attitudes toward military control. A prefecture or county might appear obedient, but if soldiers went unpaid, stores were empty, and residents scattered, it could hardly become a stable rear. Conversely, a place that preserved markets, boat transport, or local negotiation amid upheaval might enable one army to hold on for another season. Thereafter, what the Tang court and Yan contested was not only to whom Hebei and Henan belonged, but who could keep this fragile continuity from breaking first.

Chapter 07 — English candidate

Zhang Xun and Xu Yuan both took part in the defense of Suiyang. When the city fell in 757, both were captured by Yan forces; Zhang Xun and others were killed locally, while Xu Yuan was sent to Luoyang and later killed. From official biographies and posthumous honors to later traditions of commemoration, their defense has long been understood through loyalty and dying for principle; that frame is itself part of postwar political memory. Suiyang's memory of loyal martyrs relates jointly to accounts of the defense, later praise, and its transport position; the workings of those factors cannot be deduced from geography alone. Had rebels been able to enter Jianghuai more smoothly by this route, the fiscal revenues and canal-transport lines on which the Tang depended would have faced greater pressure. Holding a city was therefore given the meaning of protecting the state's revenues and wartime rear.

To ‘hold’ a city required a whole set of concrete actions. Under siege, people inside first had to face how food would be distributed, who would be drafted for defensive labor, how the sick and elderly would be placed, and whether relatives outside could still enter. Zhang Xun and Xu Yuan bore military and administrative responsibilities at different stages. Histories record their means of recruiting, encouraging, dealing with those who surrendered, and responding to exhausted provisions, and also preserve harrowing accounts of extreme hardship. Both the Old Book of Tang and Zizhi Tongjian recount the late siege's lack of food and cannibalism; from debates over Tang posthumous honors to modern scholarship, numbers, the scope of conduct, transmission of the material, and moral judgment all require separate discussion. These sources should neither be repeated indiscriminately because they shock, nor treated as if nothing happened because they are hard to accept. A more cautious formulation is that the siege drove urban society to a point at which ordinary ethics and the distribution of resources were hard to sustain, while records of exact numbers, acts, and the experience of each group are neither equally reliable nor equally complete.

Suiyang was closely connected with surrounding towns and roads. It linked to Yongqiu, Songzhou, and towns along the canal; whether it could hold for another day depended on whether relief existed outside, whether roads remained open, and whether nearby areas could still raise food. Zhang Xun and Xu Yuan are known for defending the city, but defense was not accomplished by shutting a gate. It required assigning those able to fight to the walls, sharing limited food among soldiers and civilians, and sending people to risk their lives carrying messages; it also had to confront people wishing to flee, people urging surrender, and people no longer able to work. Histories usually record how principal commanders gave orders, and much less often how each household bore them.

Zhang Xun first organized resistance at Yongqiu. After rebels cut off grain relief, he moved to Ningling and then entered Suiyang, defending it with Xu Yuan and others. The city was connected to surrounding villages, towns, roads, and waterways. Defenders needed news, people, and small supplies from outside, while rebels needed to block those connections rather than merely attack one wall. Surviving narratives give commanders clear names but rarely record the identities of messengers, guides, boatmen, and people temporarily drafted. We know communication and transport were conditions of defense, but not how many returned or whether they acted from the same willingness. Preserving that unknown is closer to the limits of the evidence than writing them all as anonymous but willing heroes.

Suiyang's strategic position often leads readers to think that, once the court recognized its importance, relief forces ought to have arrived. But ‘importance’ does not produce an army. Relief first had to concentrate soldiers from areas that could still mobilize them, obtain grain, boats, and carts, and cross roads contested by different armed forces. Once outside the city, it still had to decide whether to attack the besiegers, detour to deliver grain, or preserve itself while awaiting a larger opportunity. Each choice could appear too late to those inside and could also leave those making it responsible for failure and lost troops.

The longer the siege lasted, the harder the political problem of Tang relief became. Failed relief could consume forces already scarce at one point; no relief could make other cities still resisting doubt the court's promise. Suiyang thus carried both the personal choices of Zhang Xun and Xu Yuan and the result of the Tang court's ranking of several critical points with limited resources. Later praise for the endurance of its defense must not make that ranking disappear: part of the extreme cost borne by people in the city came from war dispersing the capacity for aid across vast Henan and Hebei.

War also made the line fragile between forces ‘loyal to the court’ and forces ‘protecting the locality.’ Local armies might act in the name of an edict, but might also plunder locally when supplies were difficult, discipline slackened, and command broke down. In 760, troops of Tian Shengong, nominally on the Tang side, entered Yangzhou, plundered residents' and merchants' property on a large scale, and killed many foreign merchants. This event warns us not to judge what an army meant for a locality by its banner alone. For people plundered, whether an army was rebel or official certainly affected later political affiliation, but did not necessarily at once change their experience of violence, loss of property, and dispersal.

This is also the most difficult aspect of the Suiyang material. Standard histories make the defenders exemplars of loyalty and martyrdom, while later readers are often shaken by accounts of hunger and cannibalism late in the siege. Both responses may erase the others inside: the first turns everyone into a background for loyalty, the second into a tragedy for spectacle. The available material does not permit us to decide each detail one by one, but it does require us to acknowledge that the siege brought ordinary food distribution, kinship, and urban order close to collapse. Any writing that makes this history clean, complete, or consumable is crueller than the sources themselves.

Both the Old Book of Tang and Zizhi Tongjian include accounts at Suiyang of exhausted provisions, the killing of some captives, and cannibalism. What can be confirmed here is that transmitted texts record these claims, not every number and detail in them. They come from histories for which political and moral judgment mattered, and were later repeatedly excerpted, intensified, or used to prove the defenders' virtue or cruelty. There are two especially unreliable responses: to use horrifying details to create a vivid siege scene as if we could see every household's experience; or, because the details have rhetorical and legendary elements, to remove the famine and collapse of order from the siege altogether.

A more cautious scale distinguishes levels. Suiyang underwent repeated and prolonged siege in 757, and late in it defenders and residents faced severe food shortages. Precise descriptions of numbers, sequence, and the experience of a particular group require comparison of versions item by item and cannot support dramatic conclusions. The loyal-martyr image of Zhang Xun and Xu Yuan in historical writing also affects how sources arrange causation: it tends to explain the pain of defense as a cost that loyalty can absorb. Historical narration cannot accept that absorption, because people in the city existed as more than witnesses to a political virtue.

Likewise, Suiyang and Zhang Xun's end cannot be summarized as ‘sacrifice bought Jianghuai's safety.’ Suiyang's resistance did delay the rebels' expansion southeastward and left the Tang time to reorganize waterways and fiscal revenues; but Jianghuai was not thereby spared requisition, armies passing through, and local violence. Strategic effect and social cost are not accounts that can cancel each other out. The first explains why the Tang court valued the city; the second explains why readers cannot measure it only by the court's gains and losses.

Suiyang's importance can also hide another fact: Jianghuai was not a rear area merely awaiting northern armies to decide its fate. Cities along the canal, granaries, merchants, and local officials had to keep exchange going; large numbers of refugees, soldiers, and temporary requisitions changed the conditions for doing so. For the court, Jianghuai's fiscal revenues and grain were important sources for reorganizing the war; front-line demands entered markets, canal transport, and local finance. How they changed particular household economies requires separate local evidence. A region not directly occupied by rebels was not therefore unchanged by war.

Suiyang's defense must be returned to the context of Jianghuai because it had two distinct consequences. One was strategic: rebels could not easily advance south along existing transport lines, and the Tang retained room to reorganize finance and military strength. The other was social: siege, transshipment, requisition, and refugee movement pulled connected cities and villages into war. The first readily enters court chronologies; the second lies more scattered in local memory, poetry and prose, epitaphs, and later narratives. History must not suppose that only the first is ‘real history.’

At the same time, the Tang court was not merely receiving Jianghuai's supplies. To get resources north, it had to maintain waterways, protect stores, settle or mobilize transport labor, and repeatedly coordinate local officials and military commanders. This work rarely left a distinct day like a battlefield victory or defeat, yet it determined whether later counteroffensives were possible. The so-called ‘rear’ was in fact a long chain of boats, grain, taxes, labor service, orders, and risks; a break at any point could turn a front-line plan into empty words.

The Tang court valued this area especially because it could not treat Jianghuai as an inexhaustible storehouse. Its grain, cloth, and taxes had to be collected, kept, and moved by water while avoiding war, bandits, military forced requisition, and buck-passing among local officials. When the front lacked grain, the court demanded more transshipment; the more that moved, the greater the labor service and risk borne by society along the route. ‘Keeping the rear’ did not merely mean blocking rebels outside a line; it meant preventing these daily links from all breaking. Suiyang's defense for a time mattered because it left the Tang an opportunity to reorganize this chain rather than immediately turn Jianghuai into another main theater.

Moving resources north from Jianghuai first meant that many places had to distribute limited things among harvests, ordinary tax burdens, and sudden military needs. Grain might enter the transport chain from official stores, private reserves, or market circulation; cloth, money, vessels, and labor had different sources as well. This does not establish that every prefecture was squeezed in the same way, still less permit later fiscal institutions to be projected unchanged back to 757. But it can be seen that the Tang's gradual reorganization of military strength between Lingwu and the two capitals depended not only on commanders' standing, but also on collection, transport, and exchange in the south and canal world that had not completely broken down.

That state of ‘not yet broken’ was fragile. Canal transport required cooperation among local officials, stores, waterways and boatmen, and towns along the route not being temporarily taken over by another army. When military demand increased, those responsible for transport might embezzle, delay, or be unable to finish; local residents might hide grain, flee labor service, or keep resources first for their households. Histories often summarize these phenomena as official administration, military discipline, or popular revolt, but the actual process was often a response of livelihood under pressure. Without village account books and broad personal records, we should not assign motives for them, only recognize that this supply line barely existed through continual negotiation and coercion.

Suiyang's great significance makes it all the more necessary to limit its representativeness. It lay in a key transport area, and the length and severity of the siege left dense narratives; many counties and towns not besieged for long appear only incidentally in records of troop movement and requisition. Suiyang's holding for a time does not establish that Jianghuai generally supported Tang forces, nor does Yangzhou's later suffering make every southern city the same kind of disaster zone. Waterways, markets, troop garrisons, and local administration differed from place to place.

The more reliable conclusion is that Suiyang kept a line that Yan might otherwise have threatened more quickly from immediately collapsing, while Jianghuai still gave the Tang room to continue assembling resources. That room was not a naturally secure zone; it had to be maintained through unstable arrangements among local officials, carriers, merchants, farmers, and armies. It explains both why counteroffensive became possible and why each increase in military demand pushed risk back onto these places. Read on this scale, the loyal-martyr narrative takes its historical place without swallowing other experiences of war.

This also changes how we understand a ‘rear area's contribution.’ Jianghuai did not passively send grain to the front; local society bore part of war's organizational cost through collection, storage, shipping, and market fluctuation. Some may have gained income or protection from transport and supply, while others lost labor, goods, and freedom of movement; gains and losses were far from equal. Since most material survives from the perspectives of offices, commanders, and literati, we cannot identify these positions one by one; people without names cannot all be written as supporters, nor all as victims. What can be confirmed is that the time Suiyang gained became usable Tang capacity for counteroffensive only while these daily links continued to work.

Putting the city and waterways in one chapter also avoids converting the cost of defense into a definite military result. Suiyang's resistance changed possibilities: it forced Yan to spend time and troops there, leaving the Tang room to reorganize Jianghuai resources; whether that room became counteroffensive capacity still depended on other towns, commanders, and links of transport. Its strategic value was real, but cannot be converted directly from sacrifice within the city into a string of numbers. Keeping this uncertainty neither lessens the hardship of defense nor turns unnamed people's pain into a negotiable token.

Jianghuai's continued supply of the Tang did not mean it was a naturally safe, inexhaustible rear. Grain, cloth, and taxes had to pass through collection, storage, shipping, and protection along the route; any link could break through armies taking supplies or local disorder. The particular violence at Yangzhou in 760 belongs to Chapter Ten. This chapter keeps only a judgment of scope: what Suiyang gained was time to keep organizing these routes, not an exemption of all Jianghuai from war thereafter.

This also requires caution against a comfortable account: writing Zhang Xun and others' deaths as a morally complete ending and then moving directly to Tang counteroffensive. A loyal-martyr narrative can show how people chose under extreme pressure, but cannot replace the circumstances of unnamed people in the city or the long daily life of Jianghuai as it faced requisition, broken commercial routes, and troops quartered there. One person's death defending a city does not automatically remove the fact that others lost food, kin, or livelihoods; it makes clearer the situation into which war pushed people.

Suiyang finally fell. To call it ‘failure’ is insufficient, because it delayed the rebels, and because that delay came at an extreme cost borne by those in the city. To call it ‘victory’ is equally insufficient, because one city's resistance did not end upheaval in Henan and Jianghuai. For readers, the more honest understanding is that Suiyang shows strategy and life are not two separate histories. Strategic value was borne by countless particular people amid hunger, requisition, fear, and uncertainty; historical materials are far more sparing of their voices than of commanders and edicts.

After the siege ended, the locality inherited more than ruins. Survivors faced registration again, taxation again, and searching again for kin and occupations; those who fled had to decide whether they could return; people who had shifted positions among different armies faced reprisals or pardon. The chronology of victory and defeat ended here, but society's postwar did not. Suiyang cannot be only a biography of loyal martyrs because it opens onto an entire canal world rearranged by war.

After Suiyang fell, the court and later literati repeatedly wrote of Zhang Xun and Xu Yuan; temple worship, biographies, poetry and prose, and local memory gave the narrative a long life. Such memory is not without value: it shows which choices later generations found worthy of respect and preserves material on how the war was explained. But it cannot directly reconstruct Suiyang in 757. Words commemorating a defender deliberately select passages that prove character; they preserve far less of the voices of children, the elderly, fugitives, drafted laborers, or people excluded outside the gates.

This chapter therefore uses the loyal-martyr narrative without letting it become the only lens. For many people then, Suiyang may have meant not a story that would later enter history, but whether there was food today, whether family remained, and whether there would be a road out tomorrow. We cannot supply their names and endings; for that reason we should avoid writing the unnamed as a neat, silent collection existing only to set off famous commanders. Suiyang changed the Tang court's strategic room in Henan and Jianghuai; for the specific effects on unnamed lives, the surviving material establishes only that there are questions to address, not a basis for narrating their experiences on their behalf. Both levels of change must be written together.

Another question cannot be skipped through the loyal-martyr narrative: siege does not end the relations it made as soon as a city falls or relief passes by. Those who fled had to find food and shelter elsewhere; those remaining faced new exactions and security; surviving defenders and their families might not be placed according to merit. To restore order, local officials might need to verify household registers again, distribute grain, and repair bridges and roads; this work itself required labor and documents, precisely what war had first scattered. We do not have a complete post-disaster archive for Suiyang and cannot describe this as an established administrative picture. What can be said is that strategic ‘delay’ did not mean social suspension: postwar work began before the siege had fully ended.

This gives Zhang Xun and Xu Yuan a double historical identity. On one hand, their decisions and later commemoration did become an entry point for understanding Suiyang; on the other, an entry point cannot determine the entire field of vision. Remembering named individuals helps fix a siege in history, yet may leave the unnamed majority only as background. This text retains that tension rather than resolve it with a moral conclusion: defenders' choices may be respected, while people inside who suffered hunger, separation, and coercion must not be treated as silent material for those choices. Only by recognizing that the two cannot substitute for each other can Suiyang avoid becoming a convenient prehistory to victory.

Suiyang fell in 757, but the relations left by the siege continued. After it ended there were still questions of migration, registration, exaction, and security; this chapter lacks sufficient cases to state what people who fled, remained, or had obeyed different armies respectively experienced. The sources provide no complete list of these experiences. The chapter therefore takes Zhang Xun and Xu Yuan as an entry without letting the named speak for the whole city. At the same time, the Yan regime had already changed internally early in 757: An Lushan was killed, and the situation of succession and counteroffensive opened anew. The next chapter turns to that political-military line without making it an answer to ordinary people's experience in the siege.

Chapter 08 — English candidate

Early in 757, a coup took place in the Luoyang court of the Yan regime, and An Lushan was stabbed to death in his residence. The principal surviving histories all name Yan Zhuang, An Qingxu, and Li Zhuer as participants in the assassination: Yan Zhuang advanced the plot, An Qingxu agreed to it, and Li Zhuer carried out the stabbing. They do not entirely agree, however, about the roles of those who entered the tent. Each of the three had his own circumstances, and together they turned the space closest to An Lushan into the scene of his murder.

Zizhi Tongjian also says that An Lushan's eyesight had deteriorated severely, that he was ill, and that his temper had grown ever more violent. Senior commanders could seldom see him directly, and memorials generally had to pass through Yan Zhuang; an attendant might be punished for the slightest disagreement. This is a highly dramatic account, and it also bears the Northern Song historian's way of arranging character and political retribution. We cannot treat it as an unselected record from the scene. Yet the sequence of action it preserves remains important: the commander increasingly depended on a few close attendants to carry information, the conflict over succession entered the court, and Yan Zhuang and An Qingxu obtained the positions needed to carry out the assassination.

Zizhi Tongjian explicitly says that Yan Zhuang and An Qingxu stood outside the tent at night holding weapons, while Li Zhuer entered carrying a knife to carry out the assassination. The biography of An Lushan in the Old Book of Tang says instead that Yan Zhuang, holding a knife, led Li Zhuer inside, but does not specify night in this passage. Thus, ‘at night’ belongs only to the Zizhi Tongjian version and cannot be presented as confirmed by both books; the two works also preserve different arrangements of duties inside and outside the tent. Both Zizhi Tongjian and the An Lushan biography in the Old Book of Tang record that the assassins wrapped An Lushan's body, buried it secretly beneath the bed, and temporarily concealed news of his death. What is confirmed here is that two later historical works preserve this claim. After An Lushan died, Yan Zhuang and the others first installed An Qingxu as Yan emperor, and only then announced the change of power outside as a transfer caused by illness.

This staged release of news was necessary because troops still stood inside and outside Luoyang. If senior commanders first heard that An Lushan had been assassinated but did not know who had taken control of the court, they might watch and wait with their forces, or move first against their own enemies. Announcing grave illness first could buy time for the palace; establishing the succession before announcing the death sought to present every army with an accomplished fact. Yan Zhuang then used offices and titles to reassure the commanders. The new regime first had to establish who was still willing to carry out its next order.

Securing command required further arrangements. The assassins controlled when news was issued, made the succession appear before news of the death, and then used offices and titles to reassure commanders who held troops. Once An Qingxu acquired the imperial title, Yan Zhuang and others still had to handle military orders; whether commanders in different places obeyed depended on the forces they held, their garrisons, and their future security. Shi Siming was far away in Hebei and had followers of his own. An Qingxu could not make him immediately surrender his power through the claim of succession alone.

An Lushan's death was therefore a genuine turning point. The man who had launched the rebellion and could bind different commanders to himself was gone, and the Yan regime entered a succession crisis. But Luoyang's offices remained, as did the armies in Hebei and Henan; granaries and roads did not cease operating that night. For ordinary soldiers, the first questions may simply have been whether their former officers would continue issuing orders and whether provisions and pay would still arrive. The Tang had gained an opportunity to attack, not victory itself.

The effects of the death soon crossed the palace gates. Shi Siming was then besieging Taiyuan, but An Qingxu ordered him back to Fanyang, leaving other commanders to continue the siege. Succession changed the deployment at the front, and Li Guangbi's chances of holding Taiyuan increased accordingly. The power relationship between Shi Siming and An Qingxu remained unstable, but the order sent from Luoyang at least made one important commander alter his actions. The Yan regime had been wounded; its military orders had not yet ceased to operate.

In spring 757, after suffering defeat near Qingqu, Guo Ziyi’s forces withdrew to Wugong and gathered defeated soldiers. The historical record supports this retreat and collection of the remaining troops. The present material does not allow us to explain item by item where and how Suzong’s court replenished personnel, grain, or equipment. This chapter treats the defeat at Qingqu only as a point at which the counteroffensive forces adjusted, and does not write it as a clearly bounded, nationwide supply operation.

By autumn, the military center of the counteroffensive was at Fengxiang. Prince of Guangping Li Chu directed operations as marshal, while Guo Ziyi and other commanders were responsible for the actual forces. The Shuofang army formed an important main force, and the Uyghurs and forces from other frontier regions also joined it. Chronicles preserve enormous troop figures for this army, but their different standards and modes of exaggeration cannot be fully sorted out. This chapter does not use a precise number as proof of strength. What can be confirmed is that the Tang court had concentrated previously dispersed armies and allied forces in a single direction of advance.

The Uyghur army entered the war on explicit terms. Led by the Yabghu, it arrived at Fengxiang and established a relationship with Li Chu under the rites of brotherhood. The two sides had to agree on positions on the battlefield and address food, livestock, rewards, and spoils. The Uyghur material in the New Book of Tang even records day by day the cattle, sheep, and rice supplied to the allied army. The particular amounts may not all be treated directly as precise accounts, but they show that the speed of cavalry rested on sustained and costly supply.

Several command relationships existed within this army. Tang forces followed their own commanders, the Uyghur army had its own leaders and discipline, and troops from different regions had fighting methods familiar to themselves. Envoys, interpreters, grain carts, and camp arrangements had to keep them moving along the same route. The court also had to ensure that military merits could be recorded and rewards paid; otherwise, after a joint operation, the various forces might suspect one another over their losses and gains.

In the ninth month of 757, Prince of Guangping Li Chu led Tang forces and Uyghur and other reinforcements from Fengxiang toward Chang’an. For Suzong, this departure concerned whether the exiled court could regain the capital. For those marching with it, the grand objective quickly became a daily itinerary: the advance troops scouted the route, the central force stayed in contact with the baggage train, cavalry sought fodder, and each contingent reassembled at appointed places. The closer the army came to Chang’an, the more likely rebel forces were to offer battle, and the longer its supply line became. The counteroffensive had finally begun, but no one could determine its result from the force of the departure alone.

This army also carried problems left by earlier defeat. After the setback at Qingqu, the Tang court had used offices and titles to gather scattered soldiers again. Titles were consequently issued in great numbers, but real discipline in the army still had to be built through commanders, units, and fighting together. Whether those regrouped could hold their ground under a rebel attack, and whether the allied forces would act as agreed at a critical moment, were questions that could only be answered on the battlefield. The departure from Fengxiang was the Tang's renewed wager on Chang'an with forces not yet fully integrated.

After Tang forces reached the west of Chang’an, they drew up north of Xiangji Temple and east of the Feng River. Zizhi Tongjian records a three-tier deployment: Li Siye led the vanguard, Guo Ziyi held the center, and Wang Sili was in the rear. Rebel forces also drew up to the north. This was no longer the one-way advance from frontier garrisons of several months earlier. It was a contest near the capital between two armies already shaped by a long war, for a victory or defeat capable of changing the political situation.

After the battle began, the rebel general Li Guiren came out to challenge the Tang army. Tang troops gave chase and gradually approached the rebel main formation; the rebels then advanced as a whole, the Tang forces fell back, and disorder appeared in their ranks. The rebels began to charge toward the baggage. Defeat had become concrete: if the forces continued retreating, troops before and behind could block one another, while food and equipment could fall into enemy hands.

In the confusion, Li Siye steadied the vanguard again. Zizhi Tongjian uses intensely forceful language to describe him standing before the formation with a long sword, followed by long-sword troops advancing like a wall. The specific casualty figures carry the enlargement typical of traditional accounts of military achievement and need not be accepted one by one. The outline of the action is clearer: the vanguard stopped retreating, met the rebels head-on with dense infantry, and gave the forces behind time to form up again.

On the other side of the battlefield, a rebel cavalry force was preparing to strike the Tang rear from the east. Once scouts discovered it, Pugu Huai’en led Uyghur cavalry to meet it. After removing this flank and rear threat, the Uyghur force rode around behind the rebels and joined Li Siye’s vanguard in a pincer attack. The battle lasted from noon until evening, when the rebels finally broke and retreated. The numbers of heads taken recorded in chronicles should not be treated directly as exact statistics, but the result on the battlefield is clear: the Yan army defending Chang’an lost the formation with which it could continue resisting outside the city.

Once victory and defeat had been decided, whether to pursue immediately became another concrete choice. Pugu Huai'en asked to send a small cavalry force after the rebel generals An Shouzhong, Li Guiren, and others through the night, arguing that the enemy had just been defeated and might regroup if given one night. Li Chu, concerned that the army was exhausted, insisted on resting first. Pugu Huai'en asked repeatedly through the night but was not permitted. The next day, scouts reported that the Yan generals had already withdrawn.

This disagreement has no standard answer apart from the circumstances. Pursuit might have enlarged the victory, but might also have separated exhausted cavalry from the main force; stopping could restore order to the army, but gave time to the defeated force. After the Tang and Uyghur and other allied forces defeated the Yan army in the area of Xiangji Temple, the Yan army withdrew from Chang'an, and Li Chu, Guo Ziyi, and their forces subsequently entered the western capital.

Zizhi Tongjian records that, before Tang forces entered Chang'an, the Uyghur Yabghu demanded the right to plunder under their agreement. Li Chu asked that this be postponed until after Luoyang was taken, on the grounds that if plundering began at once, Luoyang's soldiers and civilians would hold out. The Yabghu accepted, and the Uyghur army temporarily camped outside the city. This is a political dialogue preserved by Zizhi Tongjian; there is no independent material from the same setting, so its motives, wording, and process cannot be treated as a verbatim record.

Li Chu remained in Chang'an for several days, leaving people to garrison and pacify the city before the main force continued east. The next military move was already urgent, but the city also needed its gates, warehouses, offices, and roads to be taken over. Three days were enough only for the most urgent handovers. How people who had accepted posts under the Yan regime during the war would be treated, how officials who had fled would resume their posts, and how residents would prove prior ownership of houses and property were left for later handling. Only when those left behind took up these matters could the main force set out again.

After Suzong received news of the victory, he immediately sent it to Xuanzong, who was still in Shu, and also arranged for officials to enter Chang'an to report the victory at sacrifices and pacify the city. The recovery of the capital thereby became both a military fact and a fact of imperial politics: the army had opened the gates, but the new court also had to show that it could continue the ancestral temples, offices, and relationship of the two generations of emperor and son there. The front did not wait for these rites to be completed; the war had already continued eastward in the direction of Tong Pass.

Tang forces did not stay long in Chang'an. Li Chu left garrison troops and officials to manage affairs in the western capital, then the main force advanced east. Guo Ziyi passed through Tong Pass and recovered Huayin and Hongnong. For an army that had just won, the purpose of pursuit was direct: it had to cross the passes before the Yan army regrouped and carry the victory at Chang'an on to Luoyang. At the same time, every place occupied required people left to guard roads and stores, and the advancing force was dispersed accordingly.

During the pursuit, Tang forces were also escorting a group of prisoners. Zizhi Tongjian preserves two different dispositions: Suzong at one point ordered prisoners from east of the pass killed, but they were pardoned after the investigating censor Li Mian remonstrated; after the two capitals were recovered, Cui Qi and Lü Yin proposed executing without exception officials who had served the rebels, while Li Xian opposed them. The court ultimately assigned guilt in six grades and still executed many people. The original sentence conflates the persons, times, and Li Mian/Li Xian into one event. The order was later withdrawn. This episode occurred between two major battles, yet directly affected how the war could continue. Tang forces had to advance toward Luoyang and also make people along the route believe that surrender might still allow them to live.

Yan forces regrouped in the direction of Shanzhou. An Qingxu sent troops out from Luoyang, with Yan Zhuang and others resisting the Tang army. The two sides met near Xindian, where the rebels formed their lines against the mountain. The Tang army fared badly at first and was chased downhill by the enemy. The earlier victory at Xiangji Temple had not made this army victorious from then on; terrain, formations, and the extent to which each side had reorganized could still change the situation in the next battle.

The Uyghur army again changed the direction of the battle. One Uyghur contingent moved along the southern mountains and appeared behind the rebels. The Tang army, under pressure from the front, then joined it in a pincer attack, and the rebel formation collapsed. This resembled Xiangji Temple: the mobility and flank-and-rear action of allied cavalry again became part of the outcome. While Tang forces pursued toward Luoyang, they were at one point defeated near Xindian; then Uyghur cavalry attacked from behind the Yan army, and the battle turned into a Yan rout.

Yan Zhuang returned first to Luoyang to report. That night, An Qingxu led his close followers out through the garden gate and withdrew toward Hebei. Before An Qingxu left Luoyang, Yan forces killed Tang personnel they had held, including Geshu Han and Cheng Qianli; both Zizhi Tongjian and Tang biographies preserve the outline of this event. From the following day, Tang forces entered Luoyang. After Luoyang was lost, An Qingxu led the remaining Yan court across the river northward, withdrawing to the Xiangzhou area.

The recovery of Luoyang immediately exposed the cost of the alliance. Both Zizhi Tongjian and the Uyghur biography in the Old Book of Tang record that Uyghur forces obtained property in Luoyang. The former stresses that the Yabghu was not immediately satisfied and that Luoyang's elders bribed him with silk brocades; the Uyghur biography organizes its account around the allied force's military achievement and gains. It can be confirmed that demands and plundering occurred, but the two passages cannot be compressed into a single complete record. A related translation from the New Book of Tang says the plundering continued for several days and that the city's elders asked it to stop with a large quantity of silk textiles. Both materials survive within Tang writing traditions and cannot substitute for assigning responsibility item by item. Together they support a minimum conclusion: the Tang court used allied forces to regain the eastern capital, but could not make victory fall upon the city wholly according to its own order.

The promised spoils whose fulfillment the Tang court had delayed at Chang'an became a real conflict at Luoyang. For Uyghur commanders, the agreement to send troops and their followers' expectations of gain had to be fulfilled. For Li Chu, the armies had to continue cooperating, yet he also had to protect the city as its recoverer. For the city's elders, stopping the plundering could only mean handing over silk fabrics again. These three positions existed at once: Tang victory depended on the alliance, and part of the alliance's price was borne by city residents without the power to negotiate.

For Luoyang's residents, the return of the city to Tang rule and the restoration of life moved at different speeds. Offices could hang their signs again, but those who had fled, those who remained, those who held posts under the Yan regime, and those compelled into service faced accountability, pardon, occupied homes, and separated family members. The Tang court soon began dealing with people who had accepted Yan offices. The same investigation had to identify voluntary collaboration, but could also draw in people who had had no effective choice in wartime. Recovery ended Yan's open occupation of the eastern capital; it did not end the relations war had left in the city.

Chronicle material preserves a scene in which several hundred people who had accepted Yan offices collectively asked forgiveness, as well as the Tang court's subsequent edict distinguishing different circumstances: those who had actively served the Yan regime were to be tried; those compelled into contact because they were captured in defeat or lived nearby could surrender themselves and be exempted from punishment; family members humiliated by the war would not be pursued further. Institutional provisions cannot tell us how each person was finally judged, but they show that one of the first political tasks after recovery was to reclassify the obedience of the preceding year and a half. Whether the classifications were fair in turn depended on evidence, those implementing them, and local relationships.

In autumn 757, Tang and Uyghur and other allied forces successively recovered Chang'an and Luoyang, while the Yan court and remnant forces led by An Qingxu withdrew to Hebei. Yan forces in different places did not disappear at the same time as a result. Suzong could return to Chang'an, and offices and edicts again possessed the capital. Recovery of the two capitals changed the military situation facing different places and commanders; whether they changed sides, and why, still has to be judged case by case.

An Qingxu was in a very difficult position as he fled north. Zizhi Tongjian first describes him reaching Ye Commandery with very few followers, then immediately says that commanders such as Cai Xide, Tian Chengsi, and Wu Lingxun came to submit with their respective forces, while new recruitment also appeared in Hebei. Exact numbers remain limited by the chronicle's methods of counting, but the change before and after shows one thing: losing the two capitals dealt the Yan regime a heavy blow, but did not sever the links among former commanders, local strongpoints, and military populations.

Ye Commandery gave An Qingxu what Luoyang had lost: a city in which to establish himself, roads leading to Hebei's military garrisons, and commanders who still recognized Yan orders. Different contingents returned from different directions: some withdrew from Shangdang to Hebei, some brought their own forces back from the Henan battlefield, and others had already commanded armies in the north. Only after several forces arrived in succession did An Qingxu again possess the ability to defend a city and levy resources. Dispersed armies could still converge on a common strongpoint, and the Yan regime thus regained the capacity to act.

These commanders' return to An Qingxu's side can prove only that their military affiliation changed; it cannot automatically prove that they accepted his personal authority. They may have needed a common name under which to continue resisting Tang forces, needed to remain close to one another to avoid being defeated separately, and needed to find food and garrisons for their own followers. An Qingxu's continued use of the Yan name and the continued gathering of contingents in Hebei, temporarily combined, were enough to keep the war going.

Nor could the Tang court immediately disband all Yan forces after recovering the two capitals. Yan Zhuang and others surrendered, some places returned to Tang rule, and other commanders repeatedly moved between accepting surrender terms and rebelling again. Acceptance could quickly reduce resistance, but might leave former commanders in control of their followers; punishment pursued too harshly could make those still undecided fear that surrender meant self-destruction. Victory on the battlefield presented this choice to the Tang; it did not make the choice for them.

Shi Siming's position was especially consequential. In Fanyang he possessed his own army and territory; he could both be suspected by An Qingxu and negotiate with the Tang court. An Qingxu tried to regain control over him by summoning troops, while Shi Siming used the followers he commanded to alter his bargaining position. At the end of 757, he at one point submitted to the Tang with his forces. Such submission reduced the enemies immediately facing the Tang court, but did not break up the Fanyang army or transfer it to the center. Who possessed troops, who retained garrisons, and what titles the court could provide would still determine the next change.

The victories of 757 need to be understood through the way they actually unfolded. The two battles at Xiangji Temple and Xindian changed the armies' positions; Li Chu made choices between pursuit and rest, between alliance obligations and urban security; and after losing Luoyang, An Qingxu took his remaining forces back to Hebei. Every step that took place within a few weeks narrowed the Yan regime's space and also exposed the fact that the Tang still depended on allied and local armies.

In 758, the Tang court assembled the forces of nine military governors, including Guo Ziyi and Li Guangbi, to attack Xiangzhou, held by An Qingxu. The armies had no unified marshal; the eunuch Yu Chao'en served as commissioner for observing military affairs. This northern expedition appeared capable of ending the war precisely because the victories of 757 had recovered the centers and forced Yan forces to concentrate. Why it would nevertheless fail must be told again through how these armies came together, who commanded them, and how they were supplied.

Chapter 9

After the two capitals were recovered, the Tang court needed an ending that could be seen. An Qingxu withdrew to the Xiangzhou area; if Tang forces could surround and destroy him, they could declare that the rebellion's core had been removed. In the autumn of 758, the court assembled the forces of nine jiedushi, including Guo Ziyi and Li Guangbi, to attack An Qingxu, and then besieged Yecheng, the seat of Xiangzhou. The ‘Battle of Xiangzhou’ and the ‘Siege of Yecheng’ in historical sources usually refer to different geographical scales of this campaign. On paper, this was almost exactly the formation the Tang needed: forces arriving from different directions, enough momentum to overwhelm a trapped enemy, and a chance for the center to draw the various armies back into a narrative of victory.

The danger of the northern campaign at Xiangzhou lay precisely in how much it resembled an operation that ought to succeed. The two capitals had been recovered, An Qingxu was trapped, and the Tang court could assemble forces from several routes and compress the war, in the court's view, into one central objective: encircle Yecheng, destroy the remaining forces, and restore the north. This view had a real basis, but it underestimated everything that the previous two years had changed. The armies did not set out from the same camp; they arrived with different commanders, modes of requisition, relationships with their soldiers, and supply difficulties. The center could put their military designations in the same edict without making grain and fodder, intelligence, and risk the responsibility of a single system.

The northern campaign at Xiangzhou also carried political pressure. After the two capitals were recovered, the Tang court needed to prove that restoration was not a brief episode; the encirclement of An Qingxu seemed to provide a stage sufficient to end the war. Under that expectation, a siege that dragged on was not only a military problem but also a drain on the court's newly recovered prestige. The court pressed for results, while the armies retained their own forces and supply relationships; how each side understood the other's intentions cannot be written as a fully symmetrical record of inner thoughts. Every decision on the battlefield could therefore be judged by whether it would end the rebellion quickly.

This does not mean that commanders erred only because they sought fame. More importantly, political expectations changed the time available for reasonable action: choices that might otherwise have allowed time to wait for supplies, strengthen the blockade, or scout again could become difficult to bear for fear of losing an opportunity; costs of coordination that the center ought to have borne were shifted to commanders already under pressure at the front. The wind and rain, retreat, and mutual accusations recorded in the histories become intelligible within this structure. They were neither accidents appearing from nowhere nor sufficient on their own to explain the collapse of a siege that lasted across years.

The first question was not whether there were enough men, but whether they could become an army. The forces from different routes each had their own commanders, sources of troops, supply lines, and calculations of military merit, and they might not have been willing to place all their soldiers and grain under another's command. The commander appointed by the center had to coordinate them but might not have had practical control sufficient to override them. Each force faced pressure to preserve its own strength, and the division of military merit was also at issue; how particular commanders weighed these matters must be judged from their actions and memorials separately. Siege warfare most requires stable logistics, unified timing, and assessment of outside relief—precisely what a coalition army finds hardest to sustain over time.

A siege also magnified questions of timing in command. Should the first troops to arrive attack at once or wait for the others to join them? Should provisions go first to the vanguard, the troops on guard, or those transporting them? Was one force's withdrawal a tactical adjustment, or would neighboring forces read it as flight? Numbers alone could answer none of these questions. The disastrous defeat at Xiangzhou has institutional meaning not because premodern armies did not understand coordination, but because coordination required a command relationship accepted by all parties and supplied continuously. In its haste to complete the counteroffensive, the Tang court exposed its limits at the very moment patience and unity were most needed.

As the siege lengthened, the most practical questions among the armies came to the surface. Who would attack the city first, who would guard the outer perimeter, who would undertake the most dangerous night assault, who could keep captured goods, and whose grain route would receive priority all bore on whether soldiers would continue to obey. Front-line commanders did not necessarily hesitate from self-interest; they also knew that a failed assault could rapidly deplete the troops they knew best, and replacements were not easy to obtain. The problem was that, if each army first calculated losses it could bear itself, the coalition could hardly sustain consistent action. Xiangzhou's failure cannot be reduced to one commander's cowardice, but neither can it be waved away with the phrase ‘difficulty of coordination’: that difficulty was itself a fact of the Tang's military and political structure after the war.

A multi-pronged siege first offers the reassurance of numbers: armies from many places converge, and it seems that no Yan force can escape. Yet troops become a common force only when they act at the same time, in the same place, and in ways that allow mutual trust. The Tang armies around Xiangzhou came from different military commands and routes; their times of arrival, provisions carried with them, and familiar modes of command all differed. Those who arrived first had to camp in unfamiliar territory, while later arrivals had to pass through roads and markets already used by the advance forces; if any army suffered deaths from illness, desertion, or lack of grain, the burden was not necessarily shared equally.

Siege warfare especially tested this inequality. In field combat, an elite unit might change the situation through a short burst of force; a siege required maintaining a blockade, vigilance, siege equipment, reconnaissance, and supplies for weeks or longer. Besiegers had to defend their own camps and transport routes while guarding against sorties from the city and relief forces from outside. Who bore the outer perimeter, who attacked the city, and whose grain route was protected first all became questions of how military merit and losses would be distributed. The center could give a commander a title, but a title could not immediately remove each contingent's responsibility for its own sources of troops.

Here too, it would be wrong to explain every hesitation as self-interest. Commanders knew that soldiers were not numbers that could be replenished at will: a force that had long followed them might consist of people from the same locality, former subordinates, hired men, or dependents; a failed assault would cause casualties and weaken later operations; whether it also harmed soldiers' trust in their commander requires military evidence. But when every commander first had to ensure that his own force could return alive, the coalition lacked a shared mechanism willing to bear decisive risk. What Xiangzhou exposed was not that Chinese armies were ‘naturally unable to coordinate,’ but that war had dispersed military authority, supply, rewards, and punishments beyond the center's ability to reintegrate easily.

The experience at Xiangzhou also compels a reconsideration of what a ‘central order’ meant. An edict could designate commanders, military titles, and rewards, but it could not replace an army's daily food, fodder, medicine, or a believable route of retreat. The more orders were issued, the more likely separate forces, if they had no way to share losses, were to understand them as demands that someone else take the first risk. And if local people saw only armies coming and going and heavier requisitions, they might not see their own safety in the emperor's aims. In 758 the Tang court could still assemble nine armies to besiege Xiangzhou, but that assembly did not form a stable unified command. Xiangzhou's outcome shows that the capacity to act and the capacity to keep different forces directed at one goal are not the same thing.

For residents of Hebei and Henan, a siege did not come closer to peace merely because it involved ‘nine official armies.’ Large forces brought enormous demands for grain and fodder, roads, camps, and labor service; rebel and government forces competed for supplies in the same region, and villages outside the city often bore the first burden. We lack continuous records sufficient to reconstruct conditions village by village and cannot write every locality's experience as the same disaster. But one thing can be established: the longer the war lasted, the less ‘a siege’ was simply a confrontation between inside and outside a city, and the more production, transport, and flight across a wider region were compelled to revolve around armies.

The stationing of large armies turned the surrounding area into an expanded military space. Camps required timber, fodder, water, and fuel; siege operations required earthworks, equipment, and transport; the sick and wounded required accommodation. All of this was drawn from nearby villages and towns. Residents might at times sell supplies to troops or obtain protection in exchange for labor service; at other times they directly lost grain, livestock, and labor. The sources make it hard to tell what kind of requisition each village faced, so the area around Xiangzhou cannot be summarized as a uniform wasteland. But the evidence is enough to reject another image: that the cost of the siege was borne only by An Qingxu inside the city and the nine armies outside it.

The longer war dragged on, the more easily its timetable disrupted production. Those conscripted for labor during the busy farming season could not cultivate at the same time, and displaced households found it difficult to sow as usual; reduced food in the next season then made armies and local governments demand from one another amid tighter resources. This cycle did not occur automatically everywhere, nor can the An Lushan Rebellion alone explain all later Tang changes in population and finance. It indicates only a more immediate causal connection: when a siege long occupied people, animals, roads, and storage for military use, opportunities for local production to recover were compressed, and the next mobilization became harder as a result.

For people inside and outside the city, the identity of friend and enemy did not always immediately determine safety. If Yan forces searched for grain and conscripted troops inside the city, and Tang forces took supplies and built defenses outside it, households could be forced to navigate between different military orders. Some might try to hide grain, some seek the protection of armed powerful families or monasteries, and some move with the armies; these are possible ways of survival, not a uniform social response that existing evidence can establish generally. What must be retained is the asymmetry: commanders' memorials tell who advanced and withdrew, while ordinary people's strategies leave only scant traces.

For local society, an army's withdrawal did not mean that pressure was lifted. Departing troops might still take the last supplies, leaving ruined roads, uncultivated land, and sick or wounded people with nowhere to go; when Yan forces returned, they could investigate people and goods according to earlier allegiance. Residents around Xiangzhou were not a homogeneous group; without continuous local materials, this chapter does not generalize for them about the degree of harm or their capacity to act. After Xiangzhou, war moved from the idea that it could end at once to a reality in which no one could know when it would end.

The other side of Xiangzhou was that Yan forces were themselves unstable. After An Lushan was assassinated in 757, Yan Zhuang, Gao Shang, and others installed An Qingxu to succeed to the Yan imperial title; this was a transfer of power after a palace coup, and its process should not be concealed by calling it ordinary ‘succession’ without qualification. Shi Siming, meanwhile, had his own troops and base in Hebei. The Tang siege of Xiangzhou offered an opportunity to exploit this rift. But a rift did not automatically serve the Tang. A besieged regime could seek outside help and could also temporarily turn internal competition into a reason for common resistance. Early in 759, Shi Siming advanced from Weizhou toward Xiangzhou. While using camp drums, elite cavalry, and disguised men to harass Tang forces and disrupt transport, he brought his main force closer. Tang troops shifted from besieging the city to having to face both relief forces and a supply crisis, and the situation changed accordingly. After the Tang army collapsed and retreated at Xiangzhou, An Qingxu first sent men to submit to Shi Siming and then went in person to Shi's camp. Shi Siming killed An Qingxu and his close associates, took over Yecheng's troops, prefectures and counties, and treasuries, and then declared himself Yan emperor. Rather than ending the war by exploiting its opponent's division, the Tang helped the battlefield select a new leader better able to command troops.

Traditional narratives like to find a clear moment for the defeat at Xiangzhou: wind and rain, a formation, one commander's hesitation, or a misjudged order. Such a moment may have existed and may explain how the battle collapsed; but reducing the whole defeat to one accident misses a more troubling fact. After the campaign, the Tang court still needed to rely on the various military commands to sustain its strength, and could not easily punish all who failed or bring all their troops back under the center. Xiangzhou involved not only tactical errors but also problems in relations between the center and commanders, multi-army command, and supply coordination. ‘Institutional failure’ here is limited to these observable links.

After the Xiangzhou situation collapsed in 759, Tang forces withdrew and An Qingxu was killed by Shi Siming. Many readers see this as another bout of infighting among the rebels, as if infighting itself were already good news for the Tang. In fact, Shi Siming took over an army reconcentrated through crisis and a Hebei base that could still be mobilized. He did not need to reproduce An Lushan's personal image or win everyone's loyalty; in continuing war, control of core troops, supply routes, and the ability to reward and punish followers was enough to sustain a new regime. The Yan regime's shift from the An father-and-son line to the Shi father-and-son line shows that it no longer depended entirely on the original leader of the uprising.

Yan's predicament differed from the Tang's, but it was not lighter. An Qingxu lacked his father's prestige, while Shi Siming had his own followers and judgment. The Tang siege should have magnified this instability; but under outside pressure, internal competition could temporarily yield to common survival. Shi Siming's relief of Xiangzhou did not mean that Yan suddenly had no rifts; it showed that the direction of those rifts was not determined by the Tang court. If Tang forces could not end the siege quickly, they gave Shi Siming time to regroup, renegotiate, and reestablish his position.

Shi Siming's arrival to give relief was more than the simple news that rebel reinforcements had appeared. After An Qingxu asked for aid, Shi Siming first sent Li Guiren's force to Fuyang, then assembled troops from the direction of Fanyang, took Weizhou, and moved south to Xiangzhou, showing that he could still mobilize Hebei troops and strongpoints along the route. For the besieged An Qingxu, relief offered a chance to escape isolation; for Shi Siming, it was also a chance to gain greater bargaining power and prestige in the army. The two aims could overlap for a time without requiring mutual trust.

The killing of An Qingxu after the Tang army's collapse and retreat shows that outside relief had not repaired all the Yan regime's internal contradictions. Yet the Tang court could not treat the result simply as the enemy's self-destruction. Through relief and takeover, Shi Siming acquired the most crucial wartime assets: troops still able to fight, territories from which resources could be collected, and a position from which to promise rewards and punishments to followers. Xiangzhou's failure thus inflicted a brutal reorganization on Yan. The opponent the Tang had sought to destroy instead continued in a more concentrated form, and one more dependent on Hebei.

Weather, terrain, and sudden changes in battle during a siege could certainly change an outcome. The wind, rain, and confusion in military formations recorded in the histories should not be erased because we stress structural questions; for soldiers on the spot, a moment of disorder could mark the boundary between life and death. At Xiangzhou, wind, rain, and sudden disorder had greater effects because command among several armies, supply, and routes of retreat had not formed stable coordination. This is an explanation for this campaign, not a rule for all routs. An army able to pass orders steadily, support one another, and retain a route of retreat might still be harmed by the same weather, but would not necessarily disintegrate in the same way. Xiangzhou layered contingency over long-term conditions, which is why it deserves repeated explanation rather than treatment as only a dramatic climax.

After Xiangzhou, the Tang court faced harder choices. To continue an all-out northern campaign meant investing more finance and troops in a battlefield that had already revealed coordination difficulties; to ease pressure would give the new Yan regime room to recover and consolidate. It could not simply choose one or the other, because although the two capitals had been recovered, the war in the north had not gone away, and border defense in the northwest was also under new pressure. After recovering the capitals, the Tang still had to deal simultaneously with frontiers, armies, and finance. ‘Recovering the capital’ was therefore not evidence that restoration was complete.

As Tang forces withdrew, the center had to address more than the loss of a siege position. How the separate forces would return to their own defense zones, how casualties and dispersal would be replenished, how promised rewards would be honored, and whether nearby prefectures and counties would again wait and see after the Tang departure were all continuing military and political questions. If the court heavily punished the defeated commanders, it might damage the relationship of dependence it had with those forces; if it did not pursue responsibility, it would weaken the authority of its orders. No response could restore the coalition army before Xiangzhou in its original form.

Xiangzhou also shows that ending a war did not have to mean only ‘annihilating the opposing forces.’ The Tang court later had to adjust repeatedly among forceful assault, inducing surrender, recognizing existing commanders, reappointing local officials, and maintaining frontier defense. This was not simply weakness, but a practical choice when troops, finance, and credible commitments were all limited; yet each provisional arrangement also carried old troops, old relationships, and new local powers into the next stage. Xiangzhou did not create every later problem, but it made their entanglement harder to avoid. Shi Siming's ability to build a regime again in Chapter Ten occurred precisely in a space not cleared by a single northern campaign.

In the sources, Xiangzhou should not be reduced to a single battlefield map. Existing chronological materials more readily present a continuous account of court orders, commanders' advances and retreats, and the final collapse; records of camp supply, labor service, and local life are more scattered. How grain was distributed in camps, how transport routes were maintained, and how local people avoided or bore labor service often do not have equally continuous textual records. This chapter discusses the siege's supply and social pressures on that basis, not to claim that it has reconstructed all life around Xiangzhou, but to remind readers that commanders' decisions, which can be narrated precisely, rested on material and labor recorded less precisely. Their invisibility should not be mistaken for nonexistence.

After Xiangzhou, the Tang court still had to place retreating troops in new defense zones, enter unfinished military merit and losses into systems of reward and punishment, and prove to prefectures and counties still watching that it had not left again. These were not incidental aftermaths of a campaign, but conditions for whether the next stage could continue to mobilize. If troops felt their merit would not be recognized, or local people felt that defending the Tang would only bring reprisals without relief, the center would find it harder to obtain soldiers and grain. Shi Siming's ability to gain a new central position through these gaps shows that Xiangzhou's failure occurred both on the battlefield and in the period afterward, as the question became who could still trust which side's promises.

This distinction did not apply only to commanders. Soldiers would observe whether grain rations and pay were delivered and whether companions were properly settled; grain carriers would observe whether roads remained open. The relationship between armies taking supplies and local protection requires specific records of discipline, requisition, and local conditions; a general mechanism cannot be turned into a shared observation by residents. We cannot write these observations as a documented collective opinion, still less assume that everyone made the same choice from them. Yet wartime order was made up of many such small judgments. The failure of one siege could spread outward along them: making it harder to replenish troops, making local people more inclined to wait and see, and making it easier for an opponent to present itself as the only side able to provide order.

Xiangzhou is therefore more than a turning-point sentence—‘Tang forces were defeated, and Shi Siming rose.’ It made the conditions for ending the war more complex: to end the Yan regime, the Tang court needed not only battlefield victory but also surrender, garrisoning, collection, and local administration that could continue after victory. If these links could not connect, recovering a city might only postpone the next contest for it. The repeated changes of hands at Luoyang and succession within the Shi line in Chapter Ten were the continued operation of these complex conditions. They were not an unrelated second act, but an extension across a wider space of the fractures Xiangzhou exposed.

Xiangzhou left another question: tempo. The court hoped to give a visibly final ending to a war that had already dragged on for years, while localities and armies could act only to the rhythm of harvests, roads, casualties, and supply. When the two tempos did not match, the most likely outcome was not that either side did nothing at all, but that decisions reasonable in each local setting could not be assembled into overall victory: waiting for relief consumed grain, a hasty attack lost troops, and a retreat to preserve one's own force damaged neighboring armies' confidence. To blame all these choices on individual character makes the narrative simple but obscures the conditions that prolonged the war. After Xiangzhou, the Tang court still had the capacity to fight, but it could no longer assume that scale, legitimacy, and one gathering of armies would by themselves end the northern war.

Chapter Ten: 760–761: Shi Siming and the Second Yan Regime

After Shi Siming killed An Qingxu, the rebel forces did not become a single, wholly level territory. He had only demonstrated that in this war, whoever could command the most capable troops, the most stable rear area, and the most credible system of rewards and punishments might become the new centre. Shi Siming had been one of An Lushan’s former subordinates and had deep ties to Hebei’s military networks. When he took over the Yan regime, the Tang had recovered the two capitals but still found it difficult to penetrate the north; Shi Siming’s followers needed supplies and rewards, and some troops did change sides during the succession crisis. Neither “security” nor the specific motives for changing sides can be uniformly inferred from the outcome. Thus the second Yan regime emerged. It grew from the war’s continual elimination, absorption, and reorganization of old followers, surrendered commanders, and local forces; it was not a simple continuation of An Lushan’s original plan.

Shi Siming gained the upper hand from 759 onward not merely because Tang forces had failed at Xiangzhou. After Xiangzhou, Tang forces had to redeploy, while Yan forces had an opportunity to reconnect Hebei’s troops, local strongpoints, and new relations of reward and punishment. Luoyang lay where these two sets of actions met: it led toward Guanzhong and faced Hebei and Henan. Holding it did not necessarily mean that one side could immediately control every route; but the side that lost it had to make other arrangements for the grain routes, military communications, and garrisons leading to the front. Its loss again around 760 was the result of that line being severed once more.

In the autumn of 759, Li Guangbi voluntarily withdrew from Luoyang, which was difficult to hold, and moved his main force to Heyang. Shi Siming then entered the empty city and again controlled Luoyang. The military meaning of the eastern capital’s fall was direct: Tang forces had to redeploy, and transport and grain routes in Henan again came under strain. Its political meaning was equally sharp: a capital that had just been declared recovered could still be taken back. For the court, this was not a matter of face, but a repudiation of the narrative that restoration had already been completed. Cities changing hands repeatedly may have involved migration, searches and arrests, garrisons, and requisitions, but whether each occurred and whom it affected must be checked city by city and occasion by occasion. A city is not an object that can be restored once and for all; changing hands alters occupation, requisition, and official order. Empty houses, debts, and residents’ state of mind each require separate evidence and cannot be treated as fixed consequences every time a city changes hands.

Shi Siming’s recapture of Luoyang forced the Tang court to confront again a question that had appeared in 756: whether a capital and its centre could be defended depended on its relations with surrounding troops, roads, and food, not on whether ritual had been restored within the city. The recovery of the two capitals in 757 created conditions for the Tang court to rebuild a political centre; the loss of Luoyang again around 760 reminded people that the eastern capital remained on the edge of the northern battlefield, not in a secure hinterland beyond the war. For military and administrative decision-makers, Luoyang was both a symbol and a node of storage, transport, and garrisoning. For its residents, it was first of all a place of homes, markets, kin, and whether one could sleep safely at night. Neither scale was more real than the other, but they made the same change of hands mean very different things.

Repeated changes of hands also altered people’s relationships with regimes. Some accepted office under Yan control and, when Tang forces returned, had to explain why they had stayed. Some had fled and returned to find their former homes occupied. Some made a living from the army’s needs and lost their income when it departed. Standard histories most readily preserve the names of high officials, defending commanders, and those who rebelled or surrendered; ordinary residents often disappear into a few words such as “the city fell” or “the city was recovered.” We cannot invent their fates, but we can refuse to treat a city as a square on a chessboard that can be turned over repeatedly. Each turn altered who was entitled to live, trade, pay taxes, and receive protection.

From the Tang court’s perspective, holding Luoyang could not depend on the city garrison alone. The city needed grain, firewood, horses, and recruits from surrounding areas, as well as stations and ferries along the roads and prefectures and counties willing to remain obedient. If the surrounding areas were already under Yan threat, the more administration was restored inside the city, the more external supply was needed to sustain it. Conversely, if Shi Siming took Luoyang, he too could not treat it as an object already possessed: he had to station troops, obtain military grain, deal with its former officials and population, and guard against Tang counterattacks from other directions. A city changing hands was not a full stop, but the beginning of costs that both sides had to keep paying.

That beginning was especially unstable for residents. An army entering a city might bring new opportunities for employment, trade, and protection, or it might bring requisition, investigation, and forced migration; which result a person encountered depended on status, property, connections, and military discipline at the time, not simply on the flag above the gate. In the Luoyang passages of the Zizhi Tongjian and Guo Ziyi’s biography, the narrative chiefly concerns who entered or withdrew, troop deployments, and regime milestones; continuous records of residents’ housing, trade, taxes, and family life between the two changes of hands are noticeably sparse. Here we need to slow down: each time a city was declared returned to one side, residents had to learn again who collected taxes, who heard cases, who could protect shops and homes, and whether contracts and kinship relations from the past few years would still be recognized. We cannot quantify that uncertainty, but neither can we remove it from behind the word “recovered.”

Shi Siming himself could not escape the logic of succession that the war had produced. In 761, Luo Yue, Cai Wenjing, and other commanders under Shi Chaoyi staged a coup against Shi Siming at close quarters, capturing and killing him; Shi Chaoyi accepted the result and then ascended the throne. To abbreviate this as “Shi Chaoyi personally killed his father” would conceal the actual perpetrators. An Lushan died at the hands of his son, An Qingxu at the hands of his commanders, and Shi Siming again at the hands of his son; this sequence of killings cannot simply be explained as the rebels’ “innate brutality.” More precisely, command in wartime lacked stable mechanisms of succession and restraint that were widely recognized. Rewards, kinship, old ties, and the military situation could all affect where troops gave their allegiance; different forces did not carry equal weight. When a commander fell ill, became suspicious, or lost prestige, those closest to power could turn a military crisis into a succession crisis.

Yet the war did not end immediately after Shi Chaoyi took over. Armies did not obey a single family alone, and the Tang had not suddenly gained enough strength to overwhelm Hebei completely. This matters because it shows that the so-called “Second Yan” was not a by-product of a court tragedy, but a wartime regime sustained by military commands, local resources, and continued mobilization. With a new leader, the machinery of war could keep running; it could change speed and direction, and it could continue to consume people.

Because An Lushan, An Qingxu, and Shi Siming each died in struggles for power at close quarters, it is easy to imagine the Yan regime as a camp sustained only by fear and destined to destroy itself. That story is too convenient. The harder question is why the troops could still fight after three succession crises. The answer is that military ties were never bound to one commander alone. Commanders, kin, old followers, local suppliers, and those who controlled roads all had interests and dependencies of their own. After Shi Siming’s death, Shi Chaoyi assumed the imperial title of Yan and controlled the main force at Luoyang, but fighting broke out within Fanyang and many of An Lushan’s old commanders did not obey his orders. He acquired some troops and nodes, not a complete inheritance of all his father’s power.

The same fact explains the Tang’s difficulties. The court hoped for conflict within Yan, but could not ensure that conflict would become surrender. It hoped to induce local commanders to submit, but had to make amnesty, office, grain and pay, and safety appear credible. At this stage of the war, military victory and political settlement could no longer be separated. After defeating an army, it was still necessary to decide how to settle its soldiers, deal with the places it had occupied, and allow newly appointed officials to survive locally. If these questions had no answer, the defeated might change flags, change leaders, and again become manpower for the next war.

The succession crisis of Shi Siming and Shi Chaoyi shows that father-son status did not automatically produce obedience in the army. Command depended on personal guards, old followers, other commanders, and supplies and rewards that could actually be delivered; whether a leader could protect his followers in defeat was only a possible condition, not a belief shared by everyone. This judgment was also affected by the wider military situation: when Tang forces approached, attaching oneself to an apparently stronger centre might be advantageous; when grain and pay ran out or command became suspicious, commanders might wait, defect elsewhere, or seek their own survival. Shi Chaoyi’s acquisition of part of the army after his father’s death did not mean he inherited a complete, unconditional Yan regime.

The relations within Yan forces should not be described as simple personal loyalty. Armies formed by frontier commands and war contained people of different origins: long-standing followers of old commanders, people recruited during the chaos, people incorporated because of local affiliation, and those who lived by military achievement and provisions. Surviving materials record commanders’ groups far more often than ordinary soldiers, so we cannot precisely list the proportions or intentions of each kind of person. But because the forces were not homogeneous, any succession required their obedience to be won again. That An Lushan, An Qingxu, and Shi Siming died in succession while the army could still fight shows both that its core armed force did not disappear with individuals and that “Yan” had become a set of organizational relations many people had to confront in wartime.

The Tang faced a corresponding difficulty with those who surrendered. If it merely executed or disbanded enemy troops, it might force more people to hold out to the death; if it retained their leaders and followers, it had to provide office, grain and pay, and predictable safety. Inducing surrender was therefore not a political slogan, but an institutional experiment in how armies could be resettled. It could reduce fighting in the short term, but might also preserve older retainers and local ties. This does not equate that consequence in advance with any later fixed pattern; it identifies the dilemma already present in 760–761: defeating a commander did not mean acquiring the people and resources on whom he had relied.

To write 760–761 as two episodes of fathers and sons killing one another easily creates a closed tragic structure: An Lushan dies, An Qingxu loses power, Shi Siming is then killed, and the rebels ought to destroy themselves. Yet Shi Chaoyi could still sustain the war, which breaks that structure. Wartime regimes depended not only on leaders’ families, but also on urban resources, military routes of retreat, and the actual choices of local officials and local forces; the grounds for those choices must be demonstrated case by case. Conflict between the Shi father and son could shift the centre of these relations, but could not decide for them.

This phase of the war also reminds us that the boundaries of the main battlefield were not the boundaries of its effects. In 760, Tang forces entering Yangzhou after putting down the Liu Zhan rebellion carried out plunder and killings. Relevant biographies in both the Old Book of Tang and the New Book of Tang mention the deaths of foreign merchants, but differ over numbers and labels such as “Persians” and “Dashi”; this edition does not treat either number as conclusive. The event’s precise scale, the identities of different groups, and the attribution of responsibility need separate checking through more rigorous historical work. But it at least shows that prosperous Jianghuai cities did not stand outside the war merely because they were far from Hebei; it also shows that the name “government troops” could not automatically guarantee residents’ safety. Armies mobilized by the state to restore order could, when supply, discipline, and chains of command failed, also become sources of disaster for local society.

Yangzhou’s experience especially reminds us that the An Lushan war did not unfold only on a political map of northern Han territory. Tang cities were already connected through river transport, overland routes, and maritime trade to foreign merchants and goods from borderlands; when armies entered these nodes, violence could spread along markets and routes of transport. “Foreign merchants” should not be written as a backdrop meant to suggest internationalism: like local residents, they were particular taxpayers, traders, employees, family members, and migrants, though the surviving record is especially uneven about their names, languages, and losses. What war destroyed was not an abstract prosperity, but many paths of life that had allowed people to reach one another.

The plundering of Yangzhou in 760 should stand as a warning in this volume, not as an episode inserted to intensify horror. Yangzhou lay within canal and trade networks. It was far from the Hebei theatre, but not far from the troop movements, supply difficulties, and local disorder produced by the war. For the Tang court, it belonged to a world of revenue and transport; for armies, it could be a place to obtain grain, goods, and rest; for residents and foreign merchants, it was a city whose life should have depended on markets and local order. When an army nominally belonging to the Tang side caused violence there, it made especially clear that military action to “restore order” could also damage order itself.

This does not erase distinctions in responsibility between the Tang and Yan sides. The Tang’s title, administrative structure, and postwar policies differed from those of the Yan regime, and the command chains and accountability of particular forces should be identified case by case. What must be refused here is a different kind of shortcut: removing victims from the narrative whenever an army is called government troops. Wartime supply is not a string of impersonal numbers. When grain, boats, cloth, livestock, and labour were taken from localities, individual households, firms, and villages bore the burden; even when we lack every person’s name, the narrative should retain that cost.

Existing materials are sufficient to establish the perpetrators and victim groups in the plundering of Yangzhou, but not to verify the most widely circulated death figure. This text does not adopt the very widely repeated but difficult-to-check death total, nor does it compress differences among sources about participating forces and victims’ identities into a certain story. In late 760 the Liu Zhan rebellion spread into Jianghuai, and the Tang court sent forces under Tian Shengong and others to suppress it; in early 761, the Pinglu Army, nominally on the Tang side, carried out extensive plunder in Yangzhou. Yangzhou’s place in canal transport and commerce explains why the city concentrated people and goods, but cannot substitute for an account of the responsibility of the particular forces involved. This qualification is enough to show that the war’s supply demands, risks to military discipline, and disorder in command could cross the main front in Hebei and Henan and reach the places from which the Tang court drew revenue.

Yangzhou’s position was not an abstract matter of “wealth.” A city with developed transport and markets concentrated warehouses, boats, goods, residents, and travelers. For armies urgently needing supplies, these were resources that could be requisitioned or plundered; for households, firms, artisans, and sojourners in the city, they were conditions of livelihood. If armies could not secure grain and pay reliably, or commanders could not restrain their followers, losses would first fall on these visible and concentrated resources. To call this the cost of supply is not to treat violence as an unavoidable natural disaster, but to indicate how failures of political and military organization injured people in concrete ways.

Nor should Yangzhou’s experience be turned into a total moral judgment on “Tang troops.” Responsibility requires distinctions among actual units, chains of command, local handling, and later narratives, and the surviving sources do not make those distinctions easy to complete. Its role in this volume is to correct an overly comfortable spatial division: the north as war, Jianghuai as recovery. The reality was closer to this: the more the Tang court relied on Jianghuai’s wealth and waterways to continue its northern campaign, the more likely these places were to be drawn into troop movements and contestation. Chapter Seven has shown how the line was held at Suiyang; Yangzhou shows that an unbroken line did not mean the people along it were safe.

In 760–761, the Tang court could only try to win over commanders and resources in different places while preventing areas already recovered from slipping beyond control again. It needed cooperation from external powers such as the Uyghurs, and had to face the price of that cooperation; it needed to rebuild the credibility of central commands, while recognizing the military authority actually held by commanders at the front. Mid- and late-Tang politics did not suddenly come into being at this moment, but the war had pushed several long-term trends to a more dangerous degree: professional armies were harder to disband than before the war, local military expenditures were harder to restore to older routines, and negotiation, mistrust, and mutual constraint between centre and locality all became everyday matters.

By the winter of 761, Shi Chaoyi’s regime still controlled Luoyang and was fighting in many places across Henan and Hebei. Tang forces had taken some county seats and defeated parts of the Yan army, but had not ended the war. The Tang again possessed some cities, allies, and sources of revenue, but still could not turn the north back into the locality it had been in 755. The war did not persist because An Lushan was still alive, nor would it naturally end because Shi Siming had died. It had become a war that could end only through a new arrangement of military authority, supply, towns and cities, and personal security; those arrangements would reach beyond the northern battlefield into alliances, frontiers, ports, household registration, and ordinary families.

For the Tang court, the existence of the Second Yan also forced it to abandon an illusion: recovering the two capitals would not make every locality naturally return to the old order. It had to shift repeatedly between military action and political settlement, guarding against Luoyang’s loss again while also preventing Henan and Hebei from becoming alienated anew through excessive requisition or punishment. For Yan, continued fighting was not stable victory either, but a continual exchange of grain, garrisons, and fresh obedience for survival through the next season. Both sides were being worn down, while the societies affected had no right to suspend their lives.

By 761, the war had not reached a natural boundary, but why it had not ended in 757 was already clear: a leader’s death could not disband an army, recovering a capital could not automatically take over the roads, and victory by an allied army could not exempt localities from the cost. The military conclusion of 762–763 had not yet arrived; nor would the military power, finance, migration, and urban memory formed before then disappear with the collapse of the last Yan force.

This explanation also requires restraint about foreknowledge. People living in 761 did not know how 763 would end, or how long any temporary arrangement would last. Looking back from the consequences, we still cannot treat later changes in military governorships, finance, and frontier defense as an endpoint already visible to actors at the time. Shi Siming and Shi Chaoyi faced immediate crises of troops, cities, grain routes, and trust; the Tang court faced the question of how to continue existing when it could not win at once. Only by leaving the future unknown does the continuation of the Second Yan become more than a footnote to a later narrative of decline.

The year 761 is a line in time, not a wall. Hebei’s military commands, Henan’s waterways, Jianghuai’s supplies, the restoration of the two capitals, and the defeat at Xiangzhou remained connected to one another. Those connections in turn extended beyond the main battlefield to alliances, frontiers, ports, and household registration. A civil war can continue not only because some people keep fighting, but because many places still have to provide roads, food, labour, and an order that can be trusted for fighting. After Shi Siming’s death, those conditions did not disappear with him.

By 761, neither the Tang nor Yan had acquired a natural line at which the war could stop. The Tang court had to connect military advance with inducing surrender, settlement, and local administration; Shi Chaoyi needed to keep organizing followers of different origins and maintaining supplies; whether those followers believed he could offer protection must be judged from their comings and goings and from particular evidence. Volume Two ends here: why the war had not ended in 757 can now be explained jointly through armies, roads, local choices, and political commitments, rather than by the continued life of any one leader.

Chapter Eleven: A Cross-Section of 756–760: Reinforcements, Frontier Pressures, and Ports

The recovery of the two capitals in 757 and the disaster at Yangzhou in 760 occurred in different spaces within the same war. Setting conditions in different places alongside one another reveals connections beyond the timeline: whom the Tang relied upon, what was lost elsewhere, and how a war whose principal battlefields lay in Hebei and Henan reached frontiers and ports.

The An Lushan Rebellion is most easily told as a civil war between two armies: the Tang on one side, An Lushan and his successors on the other. That account captures the initial outbreak but not why the war later lasted so long. After 755, the Tang, having lost the two capitals, had to reorganize its forces; it could not merely mobilize its existing palace guards and regional armies. It had to seek help from the Uyghurs to the north and confront Tibet to the west; it had to hold grain routes, courier roads, and cities while sustaining a network of exchange already torn by war.

The Uyghurs had long been an important power in the Tang's northern political world. This steppe polity had long been an important force in the Tang's northern world. The two sides had relations involving envoys, horses, silk, and investiture, as well as mutual probing along their frontier. The Tang army needed cavalry and rapid mobility, and the Uyghurs knew it. After Suzong took the throne at Lingwu in 756, the Tang court quickly dealt with the Uyghurs, Tibetans, and others, requesting reinforcements and arranging marriage alliances; Uyghur reinforcements took part in fighting that same year. ‘External polities’ should be rendered as named relationships, not compress different partners into one kind of alliance.

This also changed the Tang's posture at the negotiating table. A powerful empire that still possessed Chang'an, household registers, and ritual institutions could describe neighboring leaders within a tributary order; a regime that had just fled the loss of the two capitals had to offer marriage alliances, rewards, silk, grain, and promises in exchange for troops and time. Historical works often place such items in passages of institutional history, as though they were merely formal categories. For people at the time, they concerned whether they could eat, set out, survive the winter, or return home.

One difficulty of the Uyghur alliance was that the Tang could not treat it simply as a one-time hire. Before troops were sent, there had to be envoys, marriage alliances, and ritual arrangements; on campaign there had to be fodder, food, and roads; afterward there had to be rewards, markets, and continuing diplomatic recognition. In the Tang court's accounts, silk was a calculable expenditure; for those drawn in through production areas, warehouses, transshipment, and weaving, it was also a way in which war redistributed burdens toward the interior. Without these material chains, cavalry would not appear in Guanzhong merely because an alliance had been declared; with them, military aid was never simply the story of ‘outside forces helping China.’

The relationship between the Tang court and the Uyghurs also reveals a fact ill-suited to heroic narrative: political alliances need not be formed by equals, nor permanently directed by the victorious. The Tang still had immense cultural and ritual resources and could organize relationships through titles, marriage alliances, and rewards; the Uyghurs possessed cavalry urgently needed by Tang forces and the option to turn elsewhere. Each side needed the other, but not in the same way. That asymmetry is why disputes over ritual, remuneration, and military discipline kept appearing within the same alliance.

It is also insufficient to write this relationship as ‘the Tang was forced into humiliation.’ The Tang court was not without agency: it could send envoys, alter routes, choose commanders, use competition among different powers, and at times refuse the other side's demands. More accurately, war forced it to act within choices far narrower than before the war. Imperial strength was not merely a matter of how many troops it possessed; it also included whether it could set terms under pressure on multiple fronts, fulfill promises, and persuade partners that it still had a future.

The institutional constraints of alliance also appeared in time. The travel of envoys, marriage arrangements, troop assembly, and preparation of supplies could not be completed immediately on the strength of one edict from Chang'an; yet the course of war did not wait for diplomacy to be concluded. The Tang court had to choose between ‘needing reinforcements at once’ and ‘not losing its negotiating position in the long term.’ Giving too readily could increase later burdens, while hesitation could lose an opportunity on the battlefield. This was not a matter of any emperor's or general's character, but a structural constraint faced by a state that had lost the ability to coordinate calmly in crisis.

The Uyghur side likewise faced internal and external constraints. Cavalry was not an abstract military resource, but an organization with leaders, families, herds of horses, supplies, and expectations of spoils. It could change the battlefield for the Tang, while also having to calculate the risks of long-distance marching, winter supply, competition with Tibet, and the stability of power at home. The Tang–Uyghur relationship therefore should not be fitted into the one-way moral language of ‘helper’ and ‘recipient’; it was a relationship that continually recalculated its terms amid mutual need.

This recalculation changed the Tang court's own institutional language. Investiture, rewards, and marriage alliances could once display the court's central position; in crisis they also became resources that had to be exchanged. Ritual remained important, but it could no longer conceal differences in power. What envoys disputed in tents were seats, forms of address, and promises; what lay behind those disputes was who had the capacity to keep armies moving forward. Treating ritual as a mere formality misses how it participated in the practical organization of war.

In the autumn of 757, Tang forces joined with Uyghur forces and others to recover Chang'an and Luoyang in succession. After Suzong returned to Chang'an, the Tang court restored the palace, ancestral temples, and central administration of the western capital. Although Luoyang had been recovered, it remained a frontline city afterward; this does not warrant saying that the two capitals were equally stable in administrative terms.

But recovery did not mean restoration. Uyghur material from the New Book of Tang, presented by the University of Washington, explicitly records Uyghur troops plundering Luoyang and the city's elders asking them to stop with a large quantity of silk textiles. It also records the Tang court's later, continuing use of silk, official titles, and ceremonial honors to maintain the relationship. This material comes from Tang historical writing and naturally bears the Tang court's narrative position; precisely for that reason, even it cannot depict the allied army's actions as cost-free rescue. It instead reminds readers that the Tang did not recover an empty city without people.

People in the city would not understand ‘the Tang army has returned’ as an abstract political verb. The plundering of Yangzhou is enough to show that residents and foreign merchants faced violence; the four biographical accounts do not fully preserve particular losses of warehouses, return to the city, kin, and work. Historical materials cannot leave a roster for every household. We cannot write people whose names were not preserved as one kind of victim, nor, because their names seldom appear in biographies, describe a city's recovery as a triumphant return without shadows.

The Tang court's dependence on the Uyghurs also implied a longer change: it had to recognize that the other side had calculations of its own. The Uyghurs could fight alongside the Tang and could also use Tang vulnerability to seek better terms of exchange; this mutual testing would continue later. The Tang–Uyghur alliance was neither permanent loyalty nor one-way benefit, but was repeatedly sustained by negotiation, promises, and the military situation; particular choices still had to be made in each negotiation. Alliances in war are usually much more complex than the phrase ‘reinforcements arrived’ on a commemorative stele.

Ordinary people would not read these inequalities directly in diplomatic documents, but would feel their consequences in daily life. Armies needed daily supplies, cities needed to house arrivals, and local officials had to press for cloth and grain. The cost of an alliance was often dispersed through many unseen requisitions, so that later generations remembered only which city it helped recover. Writing this dispersal back in is not an attempt to set livelihood against military history; it is precisely to show why military history cannot be separated from livelihood.

Frontier pressure also means that ‘reinforcements’ cannot be understood as an affair that can disperse after one success. If the Tang court allowed Uyghur forces to advance deeply, it had to worry about supplies, plunder, and urban order; if it did not, it might lose the mobility it needed most. Every arrangement involved local stores, roads, and requisitions. We cannot list complete amounts for every supply, but we can see that the more effective an alliance proved, the more the material and political costs of maintaining it became unavoidable postwar questions.

Alliances also had to be carried out locally. Rewards promised by the court had to leave warehouses, agreed routes had to be arranged by officials along the way, horse herds needed fodder, and armies needed camps. Central documents could combine these matters into one dispatch of troops, but localities had to break them into many particular tasks. If any link failed, reinforcements could turn from a strategic asset into pressure that societies along the route could scarcely bear.

The Uyghur material in the New Book of Tang is detailed about envoys, leaders, campaigns, and rewards, but says much less about those who carried transport, translation, and supply. We can therefore establish the political milestones of the alliance, but cannot fully calculate the burden on every prefecture and county. Retaining this gap prevents us from passing off the court's institutional list as smoothly completed local experience.

When the Tang borrowed troops from the north, Tibet to the west did not stand still. The Tibetan Empire was expanding in the eighth century, and the military pressure on the Tang in Hexi, Longyou, and other places was already severe. The An Lushan war forced the centre to turn troops, provisions, and attention inward; once frontier defense was thinned, threats that had formerly been held at the margins moved nearer to the capital.

The Uyghurs and Tibetans therefore cannot be treated merely as two pieces of ‘international background’ to the An Lushan narrative. They respectively affected the Tang's sources of troops, borders, trade routes, and diplomatic choices. Uyghur reinforcements supplied crucial strength for recovering the two capitals, while Tibetan pressure continually reminded the Tang court that it could not stake all its resources on Hebei, Henan, and Guanzhong. A regime moving troops among multiple fronts often ends not by holding every line, but by giving up room it had once taken for granted in each choice.

Pressure on the western frontier did not cease as the northern war changed. It would eventually reach the gates of Chang'an, but that belongs to another causal chain with its own frontier history and military conditions. Here it is enough to see that the civil war repeatedly drew away troops and fiscal revenues, making it harder for the empire to hold several directions at once.

Likewise, pressure from Tibet and Nanzhao should not be written as a shadow in the southwest corner of a Tang map. The Tang already required long-term garrisons, roads, and local intermediaries in the northwest and southwest; when forces and fiscal resources in Guanzhong, Hedong, and Hebei were consumed by civil war, frontier regions felt not simply that ‘the centre was busier,’ but changes in defense, supply, and the priority of orders. Nanzhao's position in this period was intertwined with Tang–Tibetan competition. It cannot be placed directly among An Lushan's rebels, still less can every southwestern frontier event be explained as a result of the war in Hebei; it reminds us that Tang choices were never made on an empty board.

For the court, the most painful decisions were often not whether to send troops, but where to draw them from, which grain route could for the moment be guarded less, and which regional commander could first be placated. Moving a force from the border to the civil war might ease a crisis in Guanzhong or Henan, but might also leave an opportunity for a neighboring polity; continuing to guard the frontier meant a slower counteroffensive at the two capitals and on the northern battlefield. Historical works often preserve appointments and battle reports, and much less often preserve who starved or fled after a particular transport line was cut. The absence of the latter does not mean it was not a real cost of political choice.

The Nanzhao direction especially requires restraint. Conditions in the southwest were connected to Tang–Tibetan competition, local power, and longer frontier relationships; they cannot all be folded into An Lushan causation merely because their dates overlap. It is included in this history to show the opportunity costs of the Tang court's coordination across multiple fronts, not to offer a universal cause for every southwestern conflict.

This pressure on multiple fronts ultimately returned to priorities within the court. Commanders, horses, grain stores, and envoys committed in one direction could not be used in another at the same time. Looking back from the outcome, later generations easily write every choice as a decision made with knowledge of its consequences; actors at the time faced delays, false reports, and conflicting requests. Retaining the incompleteness of information prevents history from becoming a chess game whose ending was known in advance.

If the An Lushan Rebellion is read only along Fanyang, Luoyang, Chang'an, and Hebei, the south easily becomes a blank map: without a large pitched battle, it seems to have no history of comparable gravity. But war does not advance only through armies. It also crosses the principal battlefield through requisitions, flight, rumors, garrisons, commercial routes, and the rearrangement of local power.

The Guangzhou event of 758 and the Yangzhou event of 760 must be discussed separately. The two places were far apart, occurred at different times, and are recorded inconsistently as to participants, victims, and numbers. Tang historical accounts of the 758 Guangzhou event use terms such as ‘besieged the prefectural city,’ ‘raided,’ and ‘seized,’ while preserving an outline of plunder, burning, and departure by ship. Modern research commonly summarizes it as an attack and brief occupation, but interpretation remains divided over whether the actors were an outside army or local and itinerant merchant communities. Both modern research and Tang biographies attribute the Yangzhou plundering to Tian Shengong and the Pinglu Army, nominally on the Tang side; it was connected to the rebellion of Liu Zhan at the end of 760, and the chronicle's major sack took place in early 761. Victims included local residents, merchants, and foreign merchants. The materials do not enable us to identify a directly demonstrable causal line from Guangzhou to Yangzhou. Combining them into the phrase ‘foreign merchants were massacred in southern China during the An Lushan Rebellion’ may be easy to circulate, but flattens two already contested local histories into a simple parable. What can be said more securely is that ports and commercial cities in wartime were not naturally immune from local military violence; people from different places who depended on trade for their lives may have found it especially hard to secure protection when state armies and local powers fell into disorder.

The Guangzhou attack occurred in 758; the Yangzhou event arose during the suppression of Liu Zhan's rebellion at the end of 760, with the major sack recorded in early 761. They belonged to different phases of the An Lushan war, and existing materials do not prove that they were part of the same action or linked by a direct causal chain. Both suggest that war's effect on a city did not end on the day its walls were breached. For coastal and canal cities, merchants, boatmen, interpreters, temples, warehouses, and local officials already formed a transregional network of life; when armies, tax obligations, and local armed forces entered, the network was compelled to choose new patrons, and people whose identities were hardest for local power to recognize could lose protection first.

The connection between port cities and the centre of civil war did not necessarily run through a rebel army advancing directly south. News, prices, military needs, refugees, and local officials' fears could all travel faster than the main force. Yangzhou stood at a node of canal transport and commercial networks; those coming and going included local residents as well as merchants of different linguistic, religious, and trading backgrounds. The city's very prosperity meant that its stores, vessels, markets, and labor could all become targets in military tension.

The violence at Yangzhou in 760 therefore cannot be treated only as a curious tale of ‘foreign merchants victimized.’ Existing accounts often focus on outside merchants; this tells us that a transregional trading community did exist in the city, but may also let unnamed local residents, hired workers, boatmen, and refugees recede from the record again. Setting outsiders against local people likewise misunderstands urban life: trade was not a group of isolated foreigners moored at the shore, but a web of relationships sustained together by intermediaries, interpreters, suppliers, artisans, landlords, tax clerks, and families.

Causation calls for still greater restraint. The attack and brief occupation at Guangzhou in 758 and the killing by government forces at Yangzhou in 760 both occurred amid the disorder of the An Lushan war, but no reliably identifiable direct causal chain joins them into one action. Their local politics, military composition, traditions of record, and consequences differed. Setting the two events side by side is only to show that ports and commercial cities in wartime were not outside the main battlefield; it is not to create a universal explanation running from Fanyang straight to Lingnan and Jianghuai.

Such distinctions also give ‘world history’ a more accurate scale. The Tang was connected with the Uyghurs, Tibet, Nanzhao, and maritime trade routes, but that did not mean every connection was clearly controlled by the centre. War made these connections seem more fragile and made different communities more dependent on finding safety among local powers. Some could use wealth, kin, or linguistic ties to leave a city; others could only wait for the next order from armies and officials. We do not know how every household chose, but because official histories say little about them, ports should not be turned into trade symbols without people.

For local officials in Jianghuai and Lingnan, a distant war also arrived through prices and administrative orders. Demand for military expenditure and silk, movement of people on the roads, and anxiety over the safety of trade routes could all change how a city levied taxes, housed arrivals, and regarded unfamiliar groups. We cannot invent a precise curve for each price change, nor attribute every local order to the centre; but ports should be understood as part of political and military supply networks, rather than merely prosperous points on a map.

A port's vulnerability also lies in its dependence on trust. Shipping, storage, and exchange need security and enforceable rules; material on Yangzhou in the An Lushan period can demonstrate violent destruction, but is insufficient to reconstruct the full changes in credit and contract markets. Once war breaks such expectations, loss appears not only after a massacre or a sack, but in the continuing consequences of merchants no longer daring to come, local people unwilling to invest, and workers losing their livelihoods. We cannot quantify these consequences precisely, but should not exclude them from the history of war.

Ports were also places where information converged. Refugees from the north, collection orders from the interior, goods from overseas, and demands from local armies could meet in the same market. News could enter a city before the main force, but stockpiling grain, fleeing, stopping credit, and profiting from disorder would each require price, contract, litigation, or personal records; this chapter does not write possible reactions as a list of events that happened. We have no reliable material to assign a proportion to each reaction, but that difference explains why the effects of war could enter urban life before the main army arrived.

By 760, the Tang did not yet know how matters would end in 763. What it could see was a world in which it had to continue fighting while finding it difficult to attend to every frontier at once. Uyghur reinforcements, Tibetan pressure, the Nanzhao direction, and Jianghuai ports were not four side lines that could be closed separately, but the reality of one empire continually rearranging resources in crisis. War crossed the principal battlefield because continued fighting required people, horses, grain, and money to be gathered from afar, and alliances and relations of command that could actually be honored to be maintained.

Chapter Eleven's cross-section therefore does not say that all regions suffered in the same way. Guanzhong, Hebei, Jianghuai, Lingnan, and the frontiers faced different military and administrative conditions. Retaining those differences prevents ‘the whole realm in chaos’ from becoming a stock phrase that covers local experience; placing them in the same chapter makes it possible to see how those differences were entangled by the same crisis of fiscal resources, transport, and alliance.

Placing the regions alongside one another is meant to show how limited resources were forced into an order of priority, not to say that they endured the same disaster: when an army, a marriage alliance, a batch of silk, or a grain route was committed to one place, another might be left thinner. The Tang court still had choices and a capacity to coordinate, but each choice increasingly had to be fulfilled jointly by different regions, allies, and ordinary suppliers.

Chapter Twelve · Who Disappeared from the Registers

There is an enormous gap between the Tang dynasty's registered household and population figures before and after the An Lushan Rebellion. This gap demonstrates that the war severely disrupted the population and social order, but it cannot be treated directly as a death toll. Household registers record two layers of fact at once: whether a person is alive, and whether the authorities can still find that person, enter them in the register of a particular prefecture or county, and use that registration to levy taxes, require service, and assign obligations.

War changed all these things at once. Some died in fighting, hunger, and disease; some left their registered home for kin, monasteries, armies, or new residences; some ceased registering to avoid taxes and service; others remained while local offices fled, archives were destroyed, or administration no longer reached the land. The count fell in each case, although different changes lay behind it.

“How many people are missing from the registers” and “how many people died as a result of the war” are related questions, but they measure two different things. The first figure is affected by where people went, administrative boundaries, registration criteria, and the authorities' capacity; the second requires separate evidence of deaths in battle, from famine and disease, and from other causes. To present the enormous gap directly as a total death toll, although striking, would obscure the processes through which population change actually occurred.

The totals in household registers first had to be assembled locally. Prefectures and counties needed to know which households lived within their boundaries, which people belonged to the local register, and who should be counted under adult males, other registered persons, or other categories before the figures could be aggregated at successive levels. Village heads, local communities, and prefectural and county officials had to identify households, check identities, preserve register pages, and then send local figures up to the next level for aggregation. When control of a locality changed hands during the war, the old documents might become disconnected from the new authorities. Some households might still be living on the land yet temporarily cease to appear in the totals visible to the court. A falling number could therefore mean that a person was no longer there, or that the process by which that person was recorded no longer existed.

After registration standards changed, figures from two years may no longer measure exactly the same object. A tax-and-service order closely tied to land, adult men, and fixed residence was already under pressure from long-term social change, and war more quickly loosened the relation among land, people, and government. Between the beginning and ending totals lies the method by which the state counted people.

The An Lushan war caused widespread and devastating death, hunger, disease, migration, and family separation. That judgment does not depend on a falsely precise total death toll. The scale of the disaster was undoubtedly immense, but the surviving gap in household registration cannot tell us how many people underwent each kind of experience. Caution in interpreting the figures is intended to clarify how the disaster unfolded, not to diminish it.

Surviving registers can show that some people left a registered home or registered elsewhere, but usually cannot reconstruct the stages of a migration. Whether a family travelled together, sought kin first, or stopped on the way needs local documents, epitaphs, or personal records case by case. Without them, this text says only that migration occurred and assigns no route to unnamed people.

People differed in the distance and speed at which they could move. Those with travel money, animals, kin, and information could more readily take family and some property; the sick, elderly, young children, pregnant people, and those without transport had fewer choices. If adult men were conscripted first, a household might lose the labor to farm or take elders and children away. One registration category of ‘flight’ therefore contains very different bodily conditions and family responsibilities.

Reregistration elsewhere can change counts at both old and current residences, but numbers cannot say how newcomers and hosts divided food, debt, and official obligations. Tang households did not always coincide exactly with families, co-residence, and common property. This account retains the institutional change, while concrete relations of life await case material.

The old household registers gradually became less accurate, while the state's demand for labor and supplies continued to grow. Armies needed grain, horses, cloth and silk, boats, road maintenance, and transport labor. Even prefectures and counties away from the battlefront could see their harvests, markets, and household divisions of labor change as they supplied the front. The authorities found it increasingly difficult to rely on the old registers and had to carry out requisitions through successive layers of prefectures and counties, temporary commissions, military commands, garrisons, transport contractors, and powerful local figures.

Each layer could change how the burden fell. Some people were ordered to procure grain; some were punished for failing to meet their quotas; some took the opportunity to impose additional levies; and some obtained exemptions through office, religious status, property, or connections. For a household, it was difficult to distinguish precisely how much of every portion of grain and cloth taken away was due to central rules, local decisions, or additional levies imposed by intermediaries. What they knew first was how much grain remained at home, and which person, which animal, or how many days of labor they would have to give up next.

Graff’s study of Tang military service and extraction shows that armies’ need for cash, cloth, and grain extended burdens across a broad taxpaying population. General evidence shows repeated requisition depleted household resources, but not that one household used stored grain, seed, tools, and credit in a particular sequence. This text therefore stops at ‘burdens accumulate’ and does not make a possible sequence a shared experience.

The distribution of burdens became still more unequal. Those with connections, property, or replaceable labor might defer, shift, or lighten obligations; clearly registered, unprotected households unable to move could be found again and again. As registration weakened, requisition relied more on local relations, and those unable to bargain could bear more.

Grain and cloth had to be stored and transported between collection and delivery to camps, and a longer war prolonged the demand. Without continuous accounts joining a particular lost store, shipment, and supplementary levy, however, a general mechanism cannot be written as one causal chain that certainly occurred.

Administrative categories such as “fugitive households” primarily indicate that the authorities could no longer continue to register people and requisition their resources and service through their place of origin; they do not automatically preserve the motives of those who left. Tax avoidance, flight from military danger, hunger, and family separation must each be established from other evidence. Where no individual case is available, this account does not infer a single shared cause from an administrative category.

Registers show how the state observed society and how state and people sought and avoided one another. Officials wanted to know where people were, what they possessed, and what they owed; people decided whether to be fixed in a locality in light of requisition, security, and available ties. When old registration failed, war burdens fell more temporarily, through intermediaries, and unevenly.

Rules of classification could also change people's strategies. Being classified as a local household, a guest household (kehu), an adult man, or a particular kind of dependent entailed different obligations and protections. People who understood the rules and had connections might seek a status carrying a lighter burden. Those who did not know how to plead their case and had no one to vouch for them were more likely to be fixed in a position from which levies were easiest to extract. These differences, too, were part of the social relationships behind household and population figures.

Tang registers organized people by household, person, adult male, and status, but a registered household cannot simply equal a private family and all common property. Dunhuang and Turfan registers can analyze household structure, not how a wartime household redistributed cultivation and care. Population loss changed household capacity, but this text will not supply unevidenced choices.

Tang registration materials show some female household heads, marriages, and family structures, but their purpose and survival limit them. They cannot make managing household finances, migrating with children, or sustaining exchange a common wartime experience of women. These questions need future work on epitaphs, contracts, and local documents, not narration in their place.

Registration and actual co-residence or common property did not always correspond one for one, but household studies leave no general script for wartime shelter, care, and child migration. Claims about kin guarantees, children's food, or separation rates await epitaphs, contracts, monastery records, or local cases. Plausible imagination is not substituted for evidence here.

Resource differences accumulated within and among households. Those retaining land, animals, commercial capital, or reliable ties might live through kin, hired labor, or exchange; people who lost their main laborer and stable shelter could lose again in the next collection. Wealth did not guarantee safety and poverty did not fix every fate, but identical levies had different effects in different conditions.

Families reappearing in registration after war often had changed form. Fugitives might return, newcomers be included, boundaries become effective again, or rich households incorporate more dependents. Rising household totals show more registered objects, not by themselves the end of trauma, inequality, or status change.

Temporary wartime identities could become permanent. People attached to armies might remain military households or retainers; sojourners might register in a new place; landless people might tenant for a long time. Their household, property, and obligations differed when they returned to official view. Two annual totals count not merely the same people changing, but changed forms of social organization.

For a household, ‘recovery’ had separate tempos for dwelling, land, registration, old debt, labor, care, and recognized kinship. Records more readily see a return when registers are remade, and seldom date every layer. Rising household figures record only part of the process.

Official histories preserve more information about reign dates, appointments, campaigns, and policies, while materials on taxes, service obligations, and household registration bring us closer to how the state located people, land, and sources of tax revenue. They can show which categories institutions recognized, but they can also reduce ordinary people to categories of famine, fugitive households, registered adult males and other persons, or burdens. Classifying a person as locally registered, a member of a guest household (kehu), a monk, a student, or a member of an official or military household meant that the state understood that person in different ways. Changes in these labels, however, did not necessarily preserve the fear, bargaining, and choices surrounding the changes themselves. Every kind of material was produced within particular institutions and for particular purposes of expression. We must first know why it was left behind and who preserved it before we can judge whom it made visible and whom it left out.

The poetry of Du Fu and others preserves another kind of visibility. Requisitioning for service, poverty, migration, family separation, and anxiety about the order of the state can enter the written record through a particular author's experience, rhythm, allusions, and ethical judgments. Poetry can preserve how an author wrote about compulsory service, bodies, and roads. Yet one poet's experience cannot encompass every region, social status, and gender, still less speak for those who left no writing.

Official histories, tax documents, and poetry complement one another but cannot cover the empire. We lack records generally joining gender, age, household resources, migration routes, and eventual destination, and village archives for every army's passage. Some regions have more documents; others survive only in epitaphs, monastery material, poetry, or later compilations. Uneven preservation makes people already lacking power to write and archive harder to see.

These materials also operate on different timescales. Battle reports and chronicles can record a city changing hands on a particular day; materials on reforms to taxes and service obligations are more likely to show why institutions changed years later; poetry may preserve a moment's experience more closely. Compressing them all into a single table of evidence with no differences in timing would obscure the process by which a household may have fled in the year of the war and still been living, many years later, under a new status and new taxes and service obligations. Connections need to be established between these materials, while their chronological sequence must also be preserved.

Materials leave gaps in different ways. Archives may have been destroyed during the war, while an administrative category may never have been intended to preserve care, fear, or negotiations within a family. People with literacy and official positions were more likely to enter historical accounts by name; others often appeared under collective labels associated with taxation, flight, or famine. “The absence of a record” may result from lost documents, or from what the record-keeping system was designed to record. Without research into the history of preservation, we cannot assign a cause to every blank.

Historical judgments can still hold even where the materials contain gaps. The war caused large-scale death and displacement; requisitions, migration, and household resources affected one another; and the state's capacity to register people and exercise actual control was also severely damaged during the war. These conclusions all have evidentiary support. But existing materials can only remain silent about the exact number of people in each category within the gap, and about the unrecorded fates of those whose names are unknown. This boundary narrows what we can conclude, while allowing the narrative to convey human experience without resorting to invention.

Record bias also means many actions are absent: people may have moved with family, bargained over levies, shared grain with newcomers, or refused demands they could not bear. Possibilities cannot all become known stories, and silence cannot prove no choice, labor, or loss. Limits should restrain narration while leaving room for unrecorded action.

The Two-Tax Law introduced in 780 registered households and people where they actually lived, no longer distinguished between resident households and guest households (kehu), graded them by wealth, and collected taxes twice a year, in summer and autumn. Commercial taxes, taxes on acreage, and the old benchmarks for tax assessments remained within the system. It reorganized postwar taxation, but did not tax property alone, nor was it an immediate measure taken when the war ended in 763. Seventeen years separated 763 and 780: the Two-Tax Law cannot be presented as an answer that appeared as soon as the war ended. By then, the questions of where people lived, what they possessed, and on what grounds levies could be imposed on them could no longer be answered solely by the old household registers. The Two-Tax Law was part of a long process of change, rather than an answer that appeared immediately at the war's end.

The gap in household registration cannot by itself tell us where each person went. Death, migration, changes of status, and omissions from the registers all require further evidence; the available figures cannot provide a single explanation for everyone. The registration gap is evidence of disruption affecting both state and society, but it does not record a cause of death for every person within it.

This silence is where a conclusion must stop, not a blank in it. It confirms war's destruction while refusing to invent one fate for people who left no record.

It also changes the scale of the war’s ‘end.’ In 763 Shi Chaoyi died with no way out; Yan commanders including Tian Chengsi, Li Huaixian, and Xue Song successively submitted with the prefectures they held, and Yan lost its shared imperial title and operational center. The war conventionally ends there, but armies and local powers were not dissolved together and prewar order did not immediately return. Chapter Thirteen returns to that military end; household life continued on another clock.

Chapter 13. The End Was Not Restoration: 762–763, the Recovery of Luoyang and the Tibetan Entry into Chang’an

Return now to the war’s end. In 762, as the war entered its eighth year, Shi Chaoyi still controlled Luoyang under the imperial title of Yan and continued fighting in many parts of Henan; yet many of An Lushan’s former generals no longer obeyed him, and the Yan regime could no longer reliably mobilize all the territory under its control. That year, Tang Suzong, Li Heng, died of illness, and Crown Prince Li Yu succeeded him as Emperor Daizong; Tang Xuanzong, Li Longji, had also died more than ten days earlier. The successive mourning rites for the two courts formed the political setting of Daizong’s accession. A change of emperor required renewed confirmation of edicts, military orders, alliances, and rewards; how it affected particular people beyond the battlefield requires local sources and cannot be filled in directly from court procedure.

When Daizong took the throne, there was no state that he could calmly take over. Shi Chaoyi had not yet been eliminated, military forces in Hebei remained, and the Uyghurs were also watching the Tang succession crisis. Both the Uyghur biography in the New Book of Tang and the Zizhi Tongjian record that, as the Tang court held successive mourning rites, Shi Chaoyi told the Uyghurs that Tang was “without a ruler” and urged them to move south. What is preserved here is Yan rhetoric aimed at winning an ally and its retelling in Tang histories, not an editorial assertion about the condition of Tang. The first task of Tang envoys was not to expound an eternal order, but to show that the new emperor truly existed and that Tang could still fulfill exchanges and alliances.

This effort at persuasion itself exposed the empire’s position. In the past, Tang decided who would receive investiture and who could enter ritual space; now it had to explain in the other side’s camp why its own succession was valid. Envoys, princes, and generals still negotiated through recognized titles and status, but behind them lay the size of armies, provisions, and urban wealth. Political ritual had not disappeared; it now had to bear a more naked relationship of force.

The successive deaths of the two former emperors in 762 left Daizong with more than a question of ritual succession. For commanders at the front, local officials, and allies, whether the new emperor could issue orders that would be carried out depended on whether he could confirm military merit, supply pay and provisions, and honor existing commitments. To the military commands in Hebei that were still watching, the mourning at the center might also be read as a new opening. Shi Chaoyi’s attempt to tell the Uyghurs that Tang had lost its ruler took advantage of precisely this uncertainty.

Tang envoys did not face a blank diplomatic table. The Uyghurs already knew of Tang’s civil war and its resources, and knew that they could receive different promises from both sides. The envoys had to reaffirm not only an old friendship but also Tang’s continuing ability to provide silk, official titles, and relationships it could honor. The process of confirming the alliance again shows that Tang still possessed organizational resources; it also shows that these resources could no longer be used solely on rules set by Tang itself.

This helps readers understand why the recovery of 762 was not a final push naturally completed by a central army. The Tang court had to connect succession, negotiation, military orders, and supply at the same time before it could redirect Uyghur forces toward Luoyang. Reading only the battlefield outcome makes it easy to think that the new emperor won as soon as he ascended the throne; seeing the negotiations in between makes clear that victory was also a precarious preservation of institutions.

In the winter of 762, the Tang vanguard led by Pugu Huai’en fought alongside Uyghur forces, defeated Yan forces at Hengshui and elsewhere, and recovered Luoyang. Uyghur troops and some Tang troops then looted inside and outside the city; military recovery did not mean that residents’ lives had recovered. In timelines of the war, this episode is often written cleanly: the allied army defeated the enemy, the eastern capital was recovered, and the rebels collapsed. It did mark the loss by the Yan regime of its most important city and organizational center, after which Shi Chaoyi’s power was difficult to sustain.

Yet the way the allied army entered the city again warns us not to understand “recovery” only as an endpoint with a clear division between good and evil. Uyghur material in the New Book of Tang records plunder and fires by Uyghur troops, and related accounts record disorder and plunder by later troops. Historical descriptions may involve exaggeration, standpoint, and selective materials, and their specific casualty figures cannot be turned directly into firm conclusions; but they at least prevent us from imagining a city that naturally recovered once rebel forces withdrew.

A city repeatedly entered and left by forces from many sides first needs to regain ordinary life: whether markets can open, whether well water is usable, who clears the streets, who claims houses, who has the right to collect taxes, and who dares light a lamp at night. These questions scarcely appear at the center of edicts proclaiming victory, yet they determine what “the return of Tang” actually meant to residents. Whether people returned to the city after the war and could recover relatives and property requires household registers, contracts, epitaphs, or local records; this chapter will not substitute imagined scenes.

Luoyang’s place in the An Lushan Rebellion was far more than that of “another capital.” Linking Guanzhong, Hebei, Henan, and the directions of the Grand Canal, it was a hub where armies, food, officials, and information repeatedly converged. Yan forces first occupied Luoyang in December 755, and An Lushan subsequently made it the center of the Yan regime. In autumn 757, Tang and Uyghur forces first recovered Luoyang after fighting at Xindian and elsewhere; in September 759, after Li Guangbi withdrew, Shi Siming’s forces entered again and controlled Luoyang; in 762, Tang and Uyghur forces recovered Luoyang once more. This was the city’s fourth major change of regime during the war, and the victory was followed by severe plunder. Each transfer changed who could control transport and storage, and changed how surrounding prefectures and counties judged to whom they should deliver grain and from whom they should seek help.

The military and symbolic value of recovering Luoyang in 762 thus existed at the same time. It deprived Shi Chaoyi of his most important political and supply center and enabled the Tang court again to declare the eastern capital restored; but this did not mean that the surrounding area was immediately settled. Retreating armies, allied troops in garrison, Tang troops awaiting rewards, and local people hoping to avoid reprisals all acted in the same space. Order does not arise automatically when a city gate changes flags.

Records of plunder and fires in the Uyghur material of the New Book of Tang must be read with their identity as sources in mind. Coming from the Tang historical record, they may selectively stress an ally’s transgressions to explain Tang’s predicament, or they may preserve urban wounds not erased by a narrative of victory. We cannot use them alone to establish every detail or number; but neither can we remove the costs borne by city residents from the narrative because the sources have a standpoint. The most cautious approach is to let “recovery” and “renewed harm” coexist, without forcing either to cancel the other.

People in the city also did not divide simply into Tang troops, Yan troops, and Uyghur troops. Some were officials who had remained behind; some had accepted assignments from Yan for their livelihood; some returned from elsewhere to claim houses; some took advantage of the disorder to occupy markets and warehouses. Post-victory investigations, amnesties, and appointments would press these complex experiences back into categories such as loyal, rebel, good commoner, and bandit. History must understand why classification was necessary to rebuild a regime, while remembering that it cannot fully express every forced choice people made in a city that repeatedly changed hands.

After losing Luoyang in 762, Shi Chaoyi first led light cavalry eastward, crossed the Yellow River north through Bianzhou and Puzhou, and then suffered repeated defeats in Hebei; this initial march cannot simply be written as a flight north. In 763, after being refused refuge at Fanyang and elsewhere and with pursuers approaching, Shi Chaoyi hanged himself. His death is commonly the chronological endpoint of the An Lushan Rebellion, but Yan generals who surrendered, its troops, and local powers did not disappear with it. This ending was not romantic: there was no day on which everyone laid down arms at once, nor a moment when the old order was immediately restored to its former place. It was more the point at which a regime lost the conditions for continued existence after its chain of command broke, resources were exhausted, and subordinates sought other paths.

To end the war, Tang also had to accept real incompleteness. It needed to induce surrender, settle, and appoint generals who had formerly been on the rebel side or had acted independently in the localities, and it needed to allow some people with troops and territory to continue to exist in Tang’s name. To describe all of this as weakness by the court is too simple; to describe it as enlightened magnanimity is equally too simple. Once Hebei could no longer be brought under control through purely central commands, this was a way to end the war, and it also left conditions for the later, long tension between the center and local military commands.

After Luoyang fell, Shi Chaoyi continued fleeing north, while his followers and local generals gradually dispersed or transferred to Tang. When he died in 763, the Yan regime thereby lost its final center able to claim supreme command. Traditional narratives easily write such an ending as the moral retribution of a traitor abandoned by all. Moral judgment is part of later readings, but cannot replace explanation: when an army suffers prolonged defeat, inadequate supply, and an uncertain future, subordinate commanders seeking another way out is a common form of the disintegration of a war organization.

The Tang court’s inducements to surrender were not simply acts of mercy. It needed to exchange offices, amnesty edicts, and permission to retain followers for an end to fighting in more regions; if everyone who had belonged to the Yan regime were driven into a desperate corner, the war might instead be prolonged. Such a policy left room for many people to survive and also enabled old military networks to continue under new names. It was a strategy for ending civil war, and at the same time one reason why the postwar center found it difficult to recover local military power completely.

The generals and local forces in Hebei that chose to surrender to Tang also turned the ending into new administrative problems. After the war, the center had to restore systems of appointment, taxation, and documents; how local society judged new officials’ protection, military discipline, and handling of debts must be established separately through prefectural and county cases. If central promises could not reach roads, warehouses, and adjudication, nominal submission might be very fragile. The “unification” after An Lushan was therefore not merely a matter of the map, but the slow process of whether these daily institutions could be rebuilt.

The disintegration on Shi Chaoyi’s side also reminds us that loyalty in a long war has levels. Subordinate commanders might be loyal to a particular leader, a network of their troops, a family interest in one place, or the immediate survival of their followers, without being loyal to an abstract regime. Tang’s inducements to surrender were effective in part because they left exits for interests at different levels. To call all such choices betrayal moralizes calculations of survival in war; to call them all rational choices ignores fear, old resentments, and threats of violence. A more honest account retains their mixed character.

For ordinary soldiers, the ending did not necessarily mean going home. Some would be incorporated into new armies, some would follow former commanders and change flags, some would return to their home communities after losing provisions, and some would leave no record again while on the move. After 763, the armies were not disbanded as a whole; some units and commanders entered a new order of military commands. Existing materials are insufficient to calculate whether particular soldiers could return home or how they were reorganized or settled, so the main text retains only continuity at the institutional level.

This also explains why Hebei was not a blank space after the ending. Armies did not disappear into thin air, nor could supply routes and local relationships be rewoven by a single edict. Inducements to surrender, appointments, amnesties, and precautions would continue to alternate thereafter. The year 763 ended a phase of war centered on the Yan regime, but left the question of how to rule regions long caught in war to the longer course of middle and late Tang.

After Hebei submitted, the center still had to join two systems of power: that of generals and that of prefectures and counties. Surrendered generals with armies could accept Tang titles without necessarily immediately handing taxation, official appointment, and public order entirely to civil officials; prefectural and county officials might likewise have formal status but no troops. If the center bypassed military generals, its orders might not be carried out; if it relied entirely on them, local military power would become further entrenched.

The choices of local people in turn determined whether this structure worked. Roads, supplies, and appointments had to function again after the war; how merchants, local communities, and former officials each made their choices requires local cases and cannot be written in their place by macro-political history. These were not formal political declarations, yet they determined whether a regime had governing capacity in the localities. Shi Chaoyi’s death ended competition at the highest level, but judgments at the grassroots about security and predictability still had to be formed anew.

Chronicles and later narratives make selections and tend toward moralization in their accounts of Shi Chaoyi’s final route of defeat, the dispersal of his followers, and details of his death. This chapter retains only the stable main sequence—“the fall of Luoyang—flight north—the collapse of support—death in 763”—and does not use uncollated troop numbers, schedules, or legendary last words to heighten drama. The meaning of the ending comes from an organization losing the conditions that sustained it, not from a literary final scene.

The asymmetry of materials at the ending also deserves to be remembered. Court edicts, chronologies written by historians, and generals’ biographies are more readily preserved, whereas why local people who submitted changed their choices and how residents judged whether a new order was trustworthy often survive only in fragments. We can write clearly the main sequence of a military regime’s collapse, but we cannot automatically fill missing local voices with unanimous welcome for Tang’s victory.

Local choices in Hebei are another key to understanding Shi Chaoyi’s end. The Yan regime was initially able to establish a base in Hebei not through one Luoyang or one leader alone; it also depended on armies already formed, local resources, and relationships among generals. By 762–763, Tang counteroffensives, Uyghur intervention, and instability within Yan forces had deprived these relationships of their former balance. The shift of certain generals to Tang can prove only a change in political and military affiliation; whether troops and grain, followers, families, and personal prospects were motives must be determined case by case.

The Tang court could not respond to these choices with punishment alone. If it demanded that all localities immediately disarm, surrender their former followers, and accept a thorough reckoning, many people would continue to resist; if it offered titles, amnesty edicts, and room to retain followers, it could bring the war to a close more quickly but would preserve local military resources along with it. This is the two-sided character of the policy of inducing surrender: it was both a practical way to end killing and a political legacy the center would have to keep dealing with thereafter.

In October 763, the main Tibetan force advanced under Gao Hui’s guidance, Tang Daizong hurriedly left Chang’an, and Tibetan troops subsequently entered the capital. This event has its own causal chain within the Tang–Tibetan war; it was not an extension of the Yan armies’ war. It often presents a stark juxtaposition: on one side the An–Shi rebels finally disintegrated, and on the other the capital once again fell into the hands of an outside army. It should not be written as a “sequel” to the An Lushan Rebellion, nor treated as an inconsequential episode. Tibetan expansion had its own frontier history, military strength, and history of Tang–Tibetan relations; but the depletion of Guanzhong and the frontier defenses by the An–Shi war made it harder for Tang to resist attack from the west at this moment.

For residents of Chang’an, the scholarly necessity of distinguishing the two wars did not lessen the fear before them. After Tong Pass fell in 756, Xuanzong left Chang’an with the court, and Yan forces then entered and occupied the capital. In 763, Chang’an again experienced the entry of an outside army; this time, as Tibetan troops approached, Daizong and his entourage fled hurriedly east to Shanzhou. This was a different war from the Yan occupation of 756. The National Palace Museum holds The Minghuang Emperor’s Journey to Shu, which depicts Xuanzong’s journey to Shu during the crisis of 756; the museum judges the surviving image to have characteristics of an eleventh- or twelfth-century copy. It is not an on-the-scene image of the Tibetan entry into Chang’an in 763. It can only help us understand how later people repeatedly remembered Tang disorder through “the emperor too was on the road”; it cannot replace evidence for the second crisis.

The Tibetan army installed Li Chenghong, Prince of Guangwu, in Chang’an and set up officials; about two weeks later it withdrew because Tang forces regrouped, decoy forces were used, and there was a response within the city. “Brief” does not mean that plunder, burning, and the flight of residents did not occur. Yet the interval between the capital’s two exposures was short enough to change how people imagined imperial security. A state can retain its institutions, era name, and capital, and still, while these symbols remain, lose the capacity once most trusted: to ensure that the center would not easily be forced to move, and that frontier defense would not suddenly arrive at the city gates.

The entry of Tibetan troops into Chang’an in 763 most clearly breaks the illusion that “the rebellion ended, so the crisis ended.” The Tang–Tibetan conflict had a longer frontier background, and the Tibetan attack had particular timing and military conditions; it should not be narrated as Shi Chaoyi’s final force, nor told as a result produced by the An–Shi rebellion alone. But the transfer of forces, strain on supplies in Guanzhong, and contraction of frontier defenses caused by the civil war were indeed background that cannot be ignored in understanding this moment.

When Chang’an faced troops entering the city for a second time, the shock to political symbolism was great. In 756, Xuanzong’s flight had already shown that the capital was not impossible to lose; in 763, Tang had just reached a military ending and yet had to confront an enemy army arriving from the west. The capital still existed and the court had not wholly collapsed, but the sense of security it represented had been pierced twice. For many residents, distinguishing which war the two crises belonged to did not alter the fact that their homes, food, relatives, and roads had again become unreliable.

Here it is especially necessary to avoid using The Minghuang Emperor’s Journey to Shu to create a convenient visual continuity. Its collection title continues to call it “Tang work,” but the National Palace Museum explains that the surviving image has a Song-style landscape structure, and some scholars regard it as an eleventh- or twelfth-century copy; its author and exact date remain under discussion. It cannot serve as illustrative evidence for the Tibetan entry into Chang’an in 763. Its value in this project is to help readers understand how later people treated “the emperor on a mountain road” as a symbol of national rupture, not to combine two different events into one image.

After Tibetan troops withdrew briefly, Tang was able to continue. This result should not make us play down the fear during that time, nor should it make us explain the renewed loss of Chang’an as proof that the dynasty was already destined to perish. Regimes in history often continue after major defeats; the question is how the resources Tang could use, the alliances it could sustain, and the boundaries it could guarantee had changed as it continued, while whether people believed a particular promise requires other materials. The year 763 gives precisely this kind of answer.

The Tibetan entry into Chang’an also cannot be inferred backward from the An–Shi war alone. Tibetan expansion itself, long competition along the Tang–Tibetan border, and the concrete conditions of the campaign all belong in specialized research. The An–Shi war can explain why Tang found it harder to defend the interior and the frontier at the same time, but it cannot explain all the causes of Tibetan action. Keeping background, catalysts, and direct causes separate is a causal boundary that must not be lost in understanding this frontier crisis.

The Tibetan entry into Chang’an made this sense of incompletion clearer. The center had just eliminated the largest enemy in the civil war, only to find that frontier pressure had not waited for it to put its internal affairs in order. For Daizong’s regime, the ending was not a chapter that could be closed, but the need to begin another round of troop movements, negotiation, and raising provisions before the wounds had been stitched. What is meant by Tang’s survival is that this capacity for continuous response had not entirely disappeared; what is meant by the end of the old Tang is that it could no longer meet every front with the margin it had possessed before the war.

Thus, 763 was an ending in military terms, but not a reset in political terms. Tang restored nominal unification, while its actual military and fiscal control was more uneven than before the war began. The later problem of the regional military governorships cannot be compressed into “as soon as the An Lushan Rebellion occurred, the regional military governorships appeared”; the jiedushi system and local militarization already had foundations before the war, and the war pushed them into relationships that were harder to reverse.

Declarations that a war has ended are usually quieter than the trumpets that begin it. Some troops were reorganized, and some surrendered generals continued to command troops; there are no overall statistics on whether ordinary soldiers returned home, so it cannot be written as a firm conclusion that “many could not or did not wish to return.” Documents show the recalculation of taxes, offices, and military merit; how families faced relatives, land, and care still requires case-by-case explanation from household registers, contracts, epitaphs, or local records. History often places 763 as an endpoint on a timeline, but for many people the endpoint was only a change from active flight to long reconstruction.

Shi Chaoyi’s death can serve as the chronological endpoint of the Yan regime; it did not reset to zero the armies in Hebei, local official titles, relations of exaction, and population movements left by the war. In 762–763, the Tang court accepted the surrender of Yan generals including Xue Song, Tian Chengsi, Li Huaixian, and Li Baochen, then allowed them to hold office in their former jurisdictions or neighboring regions and to retain followers. This reduced further fighting, but brought existing military power into the postwar order.

The Tibetan entry into Chang’an placed another frontier beside the ending. When Shi Chaoyi died in 763, the Yan regime no longer had a shared imperial title and supreme commander; but generals who had previously surrendered still held armies and territory, so a regime’s end did not mean the dispersal of military networks. After Tibetan troops briefly entered Chang’an that same year, Daizong remained on the throne, and the Tang court restored the Chang’an court and continued appointing officials, levying taxes, and moving troops. Dynastic continuity is observable institutional continuity; it does not mean that governing capacity had recovered to its prewar level. What Tang gained was not a return to 755, but the capacity to continue existing under new military, fiscal, and frontier conditions.

Chapter Fourteen · The Empire That Survived: Military Provinces, Finance, North and South, and Later Memory

After the main line of the Yan regime ended in 763, the Tang continued under Li-Tang emperors and central offices. Conventionally, the dynasty ends only in 907, when Zhu Quanzhong accepted Emperor Ai’s abdication and founded the Later Liang. This continuity did not mean that the prewar territory or central control had been restored. Chapter Fourteen does not extend continuity into a claim that every sphere of society was uninterrupted. It considers only the frontier commands, finance, regional differences, and later memory that this book can discuss one by one. The question is through what limited, observable institutional arrangements the empire continued, and what changes that single line cannot encompass.

The answer involves several connected lines. The center retained capacities for recovery and coordination, along with new fiscal and military arrangements; local military commands possessed forces, territories, and personal networks that were harder to reclaim at once. The prewar jiedushi system was not invented from nothing by the rebellion, nor did postwar military provinces become fully independent states overnight. Between center and locality there was obedience, bargaining, repeated appointment, and at times open conflict. Calling all these complex relations “loss of central control” once again writes history as a lamp that suddenly went out.

War changed the starting point for negotiation. To end the fighting, the court had to rely on people who actually controlled troops, and those people knew that the center’s dependence was more than ceremonial courtesy. For a long time afterward, Tang unity was not unity without cracks, but unity maintained within them. It required more compromise and more fiscal resources, and it required emperors and officials continually to judge which places had to be recovered and which could only temporarily be acknowledged.

The An Lushan Rebellion began when a frontier general rose in arms, passed through the loss of the two capitals, the reorganization of the throne and of alliances, and ended with the Yan regime’s loss in 763 of its shared imperial title and supreme commander. It did not end the Tang. What follows does not reduce the postwar empire to one uniform change. It asks separately how military commands continued, how finance was adjusted, how regional experiences differed, and how later people remembered the war.

This is why the An Lushan Rebellion deserves to be questioned repeatedly, rather than left as a summary of a “nine-year event.” Its consequences lay both in court orders and in a household leaving the registers, a port meeting violence, a poet writing of conscription, and a painter later repainting the road of flight. The empire survived; history leaves us the task of asking at whose cost it survived.

After the rebellion, relations between military provinces and the center were not simply a matter of one side growing while the other declined. That identifies an important direction but scarcely explains the actual political structure. Jiedushi were not always wholly hostile to the center: some were appointed by it, some needed its legitimacy, titles, and fiscal connections, while some could in practice maintain considerable autonomy through local troops, land, kin, and former followers. Nor could the center only lose. It could replace commanders, divide jurisdictions, use conflicts among commanders, control certain fiscal nodes, and at different times tighten or relax its restraints on localities.

The rebellion pushed a structure already developing into a position harder to reverse. To end the war, the court had to retain or incorporate many people who could truly command troops; to prevent another rebellion, it also had to guard against them. The manpower, supply channels, and personal networks local commanders gained in wartime did not dissolve at once because of a written appointment. Postwar politics was therefore not the disappearance of the center, but the center operating amid continuing negotiation and vigilance.

This setting also varied by region. The relations of military commands in long-contested areas such as Hebei were not identical to local rule in Jianghuai, Guanzhong, or the southwest. Painting every region with the same label of “military-provincial separatism” erases places where the center could still collect effectively, appoint, and intervene, as well as the competition that already existed among different commands. A history should not turn complex politics into the triumph of one concept.

The manpower and provisioning of local military commands brought central-local relations into requisition and security, but practices differed by region and command. Without local cases, this chapter does not generalize a shared experience for farmers, merchants, or carriers. It only affirms that institutional change entered daily life through particular arrangements for collection and protection.

After 763, one of the Tang court’s hardest tasks was to regain stable revenue. War damaged the old registers, disrupted relations between land and population, and made long-term military spending routine. The earlier logic of the zu-yong-diao system depended on relatively stable registration and obligations linked to persons and land. As population movement, land concentration, evasion, and local division expanded, its methods of collection covered reality less and less. Existing revenue could no longer pay for frontiers, armies, bureaucracy, and relief, so the court had to reorganize collection.

The introduction of the Two-Tax Law in 780 was an institutional point within long Tang adjustments to registration, land, commercial taxation, and collection times. It was neither an immediate measure after the war ended in 763 nor simply a tax only on property. Yang Yan’s plan required registration where people lived, no distinction between native households and guest households (kehu), grading by wealth, and two collections in summer and autumn. Lu Zhi later criticized conversion of property into money for producing inequality and evasion. The two policy texts establish that a design and debate existed; they cannot replace the account books of individual prefectures and counties.

Writing the Two-Tax Law as a direct product of the An Lushan Rebellion compresses the seventeen years from 763 to 780 and overlooks changes in land, registration, and collection already under way before the war. Existing material can list postwar fiscal difficulty as one background to reform. How much the war “accelerated” reform still needs finer institutional comparison, and this text does not write it as a measurable, necessary cause.

Changes in taxation also remind us that the state’s renewed ability to “see” population was not always pure benevolence. Registration could give someone a legal position, market order, and public services; it could also make that person again an object of taxation and labor service. For people who had just fled, the two meanings were often difficult to separate. We need not write all registration as oppression, but neither can fiscal reconstruction be written as a technical repair borne by no one.

Changes in finance and military provinces also had phases. After 763, the first tasks were to end large-scale civil war, reconnect tax grain, and recognize local forces able to carry out orders; the Two-Tax Law in 780 marked another adjustment in the logic of collection; afterward the center and military commands continued to pull against each other. Compressing these phases into “military provincial separatism and the birth of the Two-Tax Law after An Lushan” erases more than a decade or longer of negotiation, failure, regional difference, and policy debate.

Yang Yan’s proposal provided for registration at one’s current residence, grading by wealth, and two annual collections in summer and autumn. Lu Zhi’s objections specifically said that converting property into money could produce inequality and different outcomes for mobile merchants and settled farmers. These debates suffice to show that institutional design does not equal fairness in execution. But this chapter has no accounts for each prefecture and cannot go on to state which officials or powerful families changed tax burdens in each place and how.

The same was true of military commands. At times the court could weaken some local powers through appointment, division, fiscal control, and competition among commanders; at other times it had to tolerate their larger room for action in exchange for frontier or regional stability. Local commanders were not all equally strong, and the center was not equally powerless everywhere. A history that speaks of changed central-local relations should retain this unevenness rather than paint the whole mid- and late Tang as a picture in which the center had disappeared.

The fiscal changes of several phases also included the center relearning how to rely on markets and regional income. Grain, salt, silk, land, and commercial activity entered different arrangements for collection and transport; the state could no longer support military spending solely through the relatively orderly prewar logic of registration and labor service. There was no single road to modernization opened by An Lushan alone. There was a process in which war made old methods harder to sustain and new ones more urgent, yet more uneven.

The relation between military commands and finance especially shows why power cannot be divided only into “center” and “locality.” Local commanders without stable provisioning could hardly maintain their followers; the center, unable to obtain tax grain from localities, could hardly restrain commanders. The two sides competed and depended on one another. Some arrangements gave localities greater practical room, while others let the center reach back through finance and appointment. Treating this relation as static separatism misses the continually adjusted politics of the mid- and late Tang.

Mid- and late-Tang finance also depended on how revenue passed through localities to reach the center. Assessed tax did not equal resources actually received by the center: local retention, transport loss, priority for military spending, and temporary surcharges could all change the result. The Two-Tax Law reorganized the logic of collection but did not end central-local competition over the use of revenue. An institutional history that only lists tax names misses the question of power: who could actually control money already collected?

The continuation of military provinces was not simply the personal ambition of commanders. Armies required long-term supply, followers and their families formed local connections, and frontier defense and security made easy abolition impossible for the court. Replacing a commander did not mean that the personnel, finance, and local needs sustaining the command disappeared. For this reason, the court’s control of military provinces consisted of successive appointments, divisions, appeasements, and campaigns, not one postwar rectification.

To understand post-An Lushan changes between north and south, war damage, population movement, canal transport, and the Jiangnan economy must be placed on different time scales. This perspective is necessary because it lets us see consequences beyond the military map. Fighting damaged production and transport in many northern regions, while the south’s role in grain, commerce, and finance became increasingly prominent. People in different regions entered new economic connections through flight, trade, officeholding, and relations to land.

But “the southward shift of the economic center” is not a sentence that can replace every local story. The south was not empty ground naturally rewarded by northern destruction, and the north was not a ruin thereafter incapable of recovery. Many trends had formed before 755, and many changes took generations; opportunities gained and costs borne also differed by region and social position. The An Lushan Rebellion was a powerful catalyst: it interrupted, accelerated, and redirected some paths, but cannot explain everything from Tang to Song.

Ports and roads should be remembered especially here. Violent events in places such as Guangzhou and Yangzhou remind us that so-called southern development did not mean everyone was safe, or that commercial networks were naturally exempt from war. Goods continuing to move does not mean boat operators, merchants, interpreters, and hired workers paid no price. An economic history that draws only arrows for tax grain and transport can easily make these people disappear from the picture again.

“Northern decline and southern rise” can only be a starting point for long-term change, not a replacement for the different processes of recovery and development in each place. The generalization is explanatory, but it must be unpacked in an An Lushan study. The north suffered serious wartime damage; many people migrated southward or to safer areas; and Jianghuai and Jiangnan became increasingly important in grain, trade, taxes, and transport. These are necessary clues to long-term change.

Before 755, the south already had its own land, towns, and commercial networks, and the north did not cease recovering after the war. Macro-level research can establish long-term changes in southward migration and in Jiangnan’s economy, transport, and finance, but it cannot derive from that trend each migrant’s route, dependent relationship, or opportunity. Those concrete experiences require regional histories and case materials. This chapter retains only the long-term direction and the boundary against a single cause.

The importance of ports, canals, and local markets in this process echoes Chapter Eleven. Trade networks could be damaged in war and could also become part of reconstruction; they brought mobile opportunities to some people and made others more vulnerable to requisition, plunder, or exclusion. An Lushan is not a sufficient explanation for southern prosperity. It was an important catalyst that reordered existing differences and possibilities through violence.

Changes in the north-south configuration also moved at multiple tempos. Northern warfare pushed some people, capital, and skills elsewhere; grain, canals, salt revenues, and urban networks in Jianghuai and Jiangnan became increasingly important to imperial finance. Yet these changes had prewar roots and were shaped over a long period after the war. Southern growth cannot be directly derived from northern suffering, and northern destruction does not mean that northern history ended. Only by writing regional change as a long-term, interconnected process can we avoid beautifying compelled migration as the natural progress of a national economy.

The An Lushan Rebellion did not leave only one memory. Official histories placed it in narratives of emperors, commanders, officials, and institutions; Du Fu’s “Spring View” writes of a ruined capital, beacon fires, and letters from home that are hard to receive, while “The Officer at Shihao” writes of nighttime conscription, family casualties, and the Heyang service. These poems offer a named author’s war experience organized in poetry: they can be used to discuss the disorder he wrote, but cannot be converted directly into the scale of conscription or replace records of people in different regions and positions. Later painters sought images of national disorder in the mountain road of Xuanzong’s flight to Shu. Every kind of material has parts that let us see and parts it cannot see or chooses not to write.

Du Fu’s poetry deserves close reading, not because it gives us a scene more “real” than histories, but because it brings the felt texture of war into language: how orders enter a household, how separation is spoken, and how people still try to judge right and wrong amid the collapse of large-scale order. But Du Fu was an author with position, learning, and poetic skill. His poems are individual writings with a position and choices of genre; they can neither represent everyone nor serve as an unmediated record of the scene. Treating him as the sole ordinary voice would create another silence.

The “Minghuang’s Flight to Shu” painting is the same. It lets later viewers see the fleeing emperor amid steep mountain roads and a procession, but it was painted long after the event and may itself be a later copy. What it first records is how memory formed: why people were willing again and again to imagine the An Lushan Rebellion as the moment the emperor left the capital. It cannot serve as photographic evidence for 756, and need not lose value for that reason. The value of historical material lies precisely in being able to say clearly what kind of material it is.

When writing the postwar period, the easiest mistake is to return to large words: military provinces, finance, north and south, transformation. They are necessary words, but cannot make us forget on whom these changes fell. A local military command gaining greater autonomy meant some people could find a place in new networks of power, and others faced more layers of requisition and military service; a fiscal system changing meant some people were registered again, assessed for tax again, and seen again; a city resuming trade meant some people could reopen shops, while others were still searching for family.

We cannot supply all their names. What we can do is refuse to let missing names license careless generalization. A focused history should reconnect large changes with particular people: not force household after household into blanks in institutional history, nor enlarge a poem into a complete social picture, but always acknowledge that institutional power ultimately falls on bodies, homes, roads, and everyday choices.

Later people repeatedly returned to the An Lushan Rebellion because it often supplied a powerful political allegory: how a flourishing age lost its internal balance, how the capital fell, how the emperor left, and how the dynasty survived. Allegory has force, but it can also compress complex experience into a few memorable scenes. Yang Guifei, Mawei, Xuanzong’s flight to Shu, the image of An Lushan, and lines from Du Fu were especially likely to become symbols later people repeatedly invoked.

These symbols preserve real historical memories but cannot encompass all experiences. Du Fu’s poems are a poet’s profound writing about an age he lived through in war, but his tone cannot be made everyone’s voice; “Minghuang’s Flight to Shu” lets later people see a fleeing ruler, but is later visual memory, not news photography; official histories put Yan rebel actions into the language of dynastic orthodoxy while also preserving details later people can ask about anew. Different materials should be judged separately by source position, genre, date of composition, and relation of transmission. Differences among materials cannot be solved by choosing one “truest” book.

This identification also protects history’s openness to discussion. If a later image is treated as an on-the-scene record, a poem as statistics, registration figures as a death list, or a fiscal institution as the necessary consequence of one event, the narrative may look very orderly but will lose room for correction. A history’s responsibility is not to close every question at once, but to let readers know where conclusions come from and where they can continue asking.

This book has identified only several situations that can be discussed separately: frontier troops turning inward, the loss of the two capitals, allied participation and its costs, and the limits on explaining gaps in registration. They cannot be combined into one “deepest legacy,” nor can they directly represent the understanding of everyone in later ages. How later people understood An Lushan must be examined separately through the histories, poetry and prose, and images of their respective periods. This chapter only shows the different perspectives those materials leave.

Later memory’s repeated return to An Lushan is also connected to this incompleteness. It can be written both as a warning about the decline of the flourishing Tang and as proof of dynastic resilience; it contains stories of rulers and powerful ministers, and fragments about commanders, frontier people, merchants, and displaced people. No single narrative can possess it. A fuller historical account can put materials and questions at different levels along one route that remains open to inquiry: first letting people know what happened, then which conclusions still have boundaries, and why those boundaries are themselves part of history.

Cultural memory also takes part in this adjustment. Later readers of An Lushan can use Xuanzong, Yang Guifei, An Lushan, Guo Ziyi, and Du Fu to discuss ruler and minister, frontier defense, loyalty and righteousness, war, and people’s livelihoods; different ages choose different people and scenes as lessons. That selection is not itself a betrayal of fact, but the way memory serves present questions. What history should do is not forbid later interpretation, but return interpretation to the identity of its materials and to temporal distance, so that later moral allegory does not swallow the complex circumstances of the time.

The flourishing Tang and An Lushan in historical memory should not serve only the formula that what prospers to an extreme must decline. The formula is easy to remember, but it compresses frontier institutions, local interests, the costs of allied forces, household movement, and long-term fiscal questions into a moralized cycle of rise and fall. A better reading does not reject summary; after summarizing, it still retains the questions of which changes had already happened before the war, which were accelerated by it, and which were further shaped by later people.

Cultural memory selectively compresses these slow processes. Xuanzong’s flight, Mawei, and Du Fu’s poetry readily become clear scenes; fiscal transport, distorted registration, and the everyday maintenance of military commands enter legend less easily. A full history must bring the latter back into view, because a dynasty’s transformation more often occurs in this institutional labor that is difficult to turn into scenes.

This also marks a boundary for “using history.” An Lushan is not a formula of rise and fall that can be directly applied to the present, but it trains a kind of judgment: ask about costs when seeing victory, about statistics when seeing startling numbers, about implementation when seeing reform, and about prior history when seeing a single cause. History enters life not by turning people of the past into maxims, but by carrying this caution into choices today.

The postwar period forms an inseparable half of the history of An Lushan. If we write only the uprising in 755 and defeat in 763, we learn how the rebellion ended but not why the Tang continued in another way. Only by reconnecting finance, military commands, regions, households, and memory to the main line can “the empire survived” become a conclusion that needs explanation rather than easy praise.

This explanation also requires us to distinguish “the state remains” from “life is stable.” The Tang retained emperors, officials, and large administrative frameworks, but that did not mean every region immediately received equal protection, tax burdens, and market order. The speed of the center’s recovery varied with distance, military commands, fiscal resources, and local society. For some people, the postwar period began a recovery of roads and trade; for others, it meant only moving from one army’s requisitions to the demands of another set of accounts.

This caution also applies to the word “decline.” The Tang afterward still had periods of recovery, creativity, and expanded political capacity; the mid- and late Tang were not the afterglow of more than a century of continual descent. An Lushan changed the conditions but did not cancel later people’s actions. If the flourishing Tang alone measures everything, postwar institutional innovation, local vitality, and cultural creation will be judged failures in advance. A real history must allow survivors to continue to have a future.

The mid- and late Tang still saw institutional adjustment, local economies, and cultural activity, so the more than one hundred years after the war cannot be written as a motionless afterglow. How displaced people, merchants, artisans, and particular localities rebuilt must be established by corresponding cases. This chapter does not use a string of things that might have happened as the shared ending of ordinary people.

The An Lushan Rebellion had ended, but Tang military, political, and fiscal relations entered a new phase. Many unnamed people lost their former lives, while some rebuilt households, markets, and local order under new conditions. A certain main line and differences still awaiting specialized histories exist together: after 763, the Tang would continue for more than a century; what continued was already an empire changed by war.

Evidentiary limits and review gaps

• The English edition is derived from the accepted Chinese work; no independent source rereading is claimed.
• Formal paragraph-to-source relations have not been inferred for any of the 519 translated paragraphs. Sixty inherited chapter material notes and their limits remain visible.
• The original Chinese evidence review remains available in the source edition; chapter material references are not substituted for that review.
• Native English editor, second translator, named historian and independent reader reviews remain pending.