Lynn White, in his Medieval Technology and Social Change, offers the interesting hypothesis that the feudal class of the European Middle Ages derived ultimately from the stirrup. After the arrival of the stirrup in Europe by the eighth century, and the primacy this gave the horse and armor in warfare, the state made, land grants in return for the pledge to provide armored knights on horseback when called. The freeman with his battle-axe no longer was the mainstay of the military might of the state, though he was still subject to general muster. Charlemagne's attempt to raise horsemen by ordering the less wealthy to pool their resources was eventually unsuccessful because of the difficulties in administering such a program. The eventual result was the distribution of land to vassals on condition of knight's service; from this followed the creation of a fighting elite, which was to have a profound effect on Western society and history.

In Europe the introduction of the stirrup and the resulting technology, that of the horse and armor, thus made fighting a matter of class. In China, the appearance of the stirrup also led to significant changes especially in the method of warfare and in military organization. The initial impact may well have been to produce a military elite, that of the tribesmen, Xianbei" and others, which ruled northern China from the fourth to the sixth centuries, but the effect on society was perhaps less pronounced because it was possible to administer effectively a program such as Charlemagne found impossible to implement. The very success of the new military institution, however, led to the disappearance of the tribesmen as the military elite. Thus the invention of the stirrup and its ultimate effect on warfare and society in China is a topic well worth examining.

It is curious that something as useful as the stirrup should have appeared so many centuries after man began riding the horse. A. D. H. Bivar, in his admirable article on the stirrup, points to the danger involved in mounting a horse while carrying weapons, citing the example of Cambyses, king of Persia in 522 B.c., who stabbed himself fatally while leaping onto his horse.2 Yet the earliest forms of stirrup, found in the sculptures of Sanchi and elsewhere in India, are only of the late second century B.C. These took the form of a loose surcingle or saddle-strap behind which the rider's feet were tucked; later, one finds the depiction of a tiny stirrup for the big toe.3 As White observes, shed riders of colder climates could use a big-toe stirrup only with difficulty. He also cites a Kushan engraved gem of ca. A.D. 100 that appears to show some sort of hook-stirrup, but this seems to be an isolated example.

It is not known whether the Indian experiments had any influence on China. It has been claimed that the stirrup was in use in the Han,s but the earliest reliable representation comes from a Jin tomb of 302, near Changsha (Fig. 1).6 This was a mounting stirrup, that is, a stirrup placed on one side of the horse, too short to be of use once the rider had mounted. What may be the earliest representation of a proper stirrup, of full length and on both sides of the horse (Fig. 2), comes from a tomb near Nanjing that has been dated to 322 (this date is not certain, but the other objects found in the tomb make an early Eastern Jin date probable).' Both the mounting stirrup of Changsha and the full-sized pair depicted on the Nanjing horse are triangular in shape, but the modeling is too generalized to permit details of construction to be discernible. It is therefore necessary to turn to examples of actual stirrups for such information.

What is quite possibly a group of Xianbei graves dating to the early or mid-fourth century was uncovered in 1974 at Xiaomintun,b near Anyang. In one of these (no. 154) was found a set of equipage for a horse: the skull of the deceased rested on the remains of a saddle, and bells, buckles, and ornaments lay atop his body. Next to the saddle was one gilded bronze stirrup (Fig. 3). Few descriptive details other than dimensions have been published: overall length, 27 cm.; thickness, 0.4 cm. The handle is 14.5 cm. long and 3.1 cm. wide, with a horizontal slot at the upper end measuring 1.2 by 1.5 cm.; the ring has an outside diameter of 16.4 cm. and the body of the ring is 1.8 cm. thick. The stirrup was apparently attached to the left front of the saddle.8 If, in fact, there originally had been but one stirrup, then it might indeed represent a transitional stage from the mounting stirrup to the full pairs which follow, as the excavation report suggests. One might also conjecture that for purposes of pulling a strong bow while on horseback, it is precisely the left stirrup that facilitates the draw. However, a figurine of a horse with similar trappings was recovered from a Xianbei tomb of about the same period near Huhehot, and this frgure, I am told on good authority, is depicted with a pair of stirrups. Thus it may be due to some oversight that the Xiaomintun site yielded but one of the presumed pair.g A pair of stirrups found recently in a tomb at Yuantaizi,c Chaoyang,d Liaoning, is believed to be contemporaneous with that of Xiaomintun, but the construction is unlike any found elsewhere. A wooden core, perhaps of rattan, is bound with leather and then lacquered, with a painted cloud design in red. Again the handle is long, with a rounded end and a horizontal slot for attaching the stirrup to a leather strap. The bow is somewhat triangular, with a trapezoidal cross- section, broader on the inner rim than on the outer one. A triangular piece of wood was placed at the upper part of the bow, probably to help the bow keep its shape. The handle is 14 cm. long and 3.7 cm. wide, while the slot is 2.4 cm. by I cm.; the diameter of the bow is 15 cm. and it is 2.5 cm. thick.

Another of the early examples of stirrups was found in the tomb of Feng Sufue (d. 415), a member of the ruling family of Northern Yan, which adjoined the Koguryd state. The construction of these stirrups is rather complex. A piece of mulberry wood with a triangular cross- section was bent to form the bow of the stirrup, with a ridge facing outward. The two ends meet to form the stirrup handle or "ear." At the juncture a triangular chip of wood was inserted so that the shape would not be distorted under the weight of the foot. Gilded bronze plates were nailed to the surfaces, except for the inside edge of the bow, which was lined with thin iron plates and a layer of black lacquer. The original dimensions were 23 cm. high and 16.8 cm. wide (Fig. 4).'1 It would seem that the Feng stirrups are the earliest of known date." Although the stirrups in Feng's tomb are described as being triangular and derived from those of the Eastern Jin,'3 it is more accurate to say that the bow is oval, and that this feature relates it to stirrups recovered from the nearby Koguryo sites in the area of Ji'an, in the Yalu River valley.

Stirrups found in Koguryo tombs may antedate the pair from the Feng tomb since these sites are usually ascribed to the fourth and fifth centuries, but there are no certain dates.'4 The stirrups are generally oval in shape with a long, flat, rectangular handle at the top, to which a strap would be connected through a horizontal slot. Two pairs of such stirrups were found in Tomb 78 at Wanbaoting,s which is ascribed to the first half of the fourth century on the basis of its construction (Fig. 6).'5 These stirrups are described as follows:

Other sites in this area where stirrups were found include tombs at Qixingshan,h Yushanxia,' and Maxian'gou.

Tomb 96 at Qixingshan is dated to the early fifth century. The manner of construction of the pair of stirrups found here is not described in any detail, but again we are told the stirrups had a wooden core and were covered with gilded bronze plate fastened by nails. From the drawing available to us (Fig. 7), as well as the published photograph, it would appear that this pair resembled closely that of Tomb 78 at Wanbaoting.

Examples of stirrups from a slightly later period are not as well preserved but display the same basic format. Tomb 41 at Yushanxia, dated to the mid-fifth century, yielded only one stirrup (Fig. 8). What is different here is that the metal covering is of iron.'" Yet gilded bronze continued to be used. As evidence of this, one can cite the fragment of a cover plate found in Tomb I at Maxian'gou, which was not originally identified but is now recognized as coming from a stirrup.'" Stirrups of this same shape, covered with gilded bronze or iron plate, or forged entirely in iron, turn up farther south in Korea, in tombs of the Old Silla period (fourth- seventh centuries). In a study by Akio Ito these stirrups, some seventy-six pairs in all, have been divided into five types, the first two of which resemble the kinds of stirrups just described, with wooden cores covered either with overlapping plates (Type I, four pairs), or by plates attached with rivets (Type 2, sixteen pairs). The third type, represented by one example, is pear-shaped, and the fourth and fifth types, of iron, have foot-rests that are broadened, sometimes split lengthwise, and often have slight protuberances for added traction. Types 4 and 5, probably to be dated toward the end of this period, differ from each other in the shape of the end of the handle.'" One pair, of Type 2, from the Gold Crown Tomb (no. 128), is exhibited in the museum in Kybngju (Fig. 9). One may observe the same bands around the inside and outside edges, while the surface plates have incised decoration. The usual horizontal slot on the handle to hold a strap is obscured.

The stirrups depicted in various contemporary Korean ceramic figures and murals are consistent with the shape that emerges from the Koguryo and Silla tombs. The murals in the Dancing Figures Tomb are especially important in this regard. One of the two saddles depicted on the wall of the left wing seems to have stirrups of the oval shape.22 At the least, the painting indicates very clearly how the strap was attached to the saddle. In the hunting scenes in the main chamber, the oval stirrup is very clearly shown on the right wall (Fig. 10)23 and appears somewhat less clearly in the racing figure on the left wa11.24 In these cases, however, it would seem that only the front portion of the foot entered the stirrup and it was the ball of the foot which rested on that support. This is also the case with the few examples known in Old Silla ceramic figures. The most famous, of course, is the mounted warrior from the Gold Bell Tomb of the fifth-sixth centuries (Fig. 11).25 The museum at Kybngju has another mounted clay figure displaying stirrups, as well as some depicted on an unmounted horse, but the modeling is too vague to be useful.

Thus one finds from the fourth century onward, in North China, in northeast Asia, along the Yalu, farther south into Korea, and even into Japan, the appearance of a particular type of stirrup: oval, flat, and with a rather long handle, with either a wooden core covered by gilded bronze or iron plate or one forged entirely from iron.26 Unfortunately, the consideration of the early history of the stirrup does not bring us any closer to establishing where it was invented, since it seems to appear at the same time among the Chinese and their northern neighbors. As early as 1926 A. von Le Coo pointed out that, on the basis of conjecture, it is as reasonable to say that the stirrup was the invention either of a mounted people who sought with it to make riding less tiring, or of a people unused to riding who sought thereby to obtain quickly the skills necessary to meet the needs of cavalry warfare."' White brands this remark by Le Cog as a priori speculation,28 but his own conclusion that it was a Chinese invention rests on no more solid foundation than does the opinion of other scholars that it was the nomad's skill with the horse which led to the improvement of equestrian gear, including the stirrup."' At this point, we are not much further along in the solution of the problem than was Le Coq in 1926.

In China the shape of the stirrup underwent some changes during the sixth century. A pair of iron stirrups from the Northern Wei tomb at Guyuan,k Ningxia, appears to have an oval bow and a rather short handle, still with a slot at the upper end. The height of the entire stirrup is 18.7 cm., but no other details have been published and the printed photograph of the find is not clear."" It was stirrups with the oval bow and short handle which were used in the Tang, it is the type which is encountered across Asia, and it may well have been the type brought to Europe in the eighth century"' The form is seen depicted on the horses of Tang Taizong32 (Fig. 13) and stirrups of exactly the same form occur in Avar graves in Hungary, in both bronze and cast iron,33 Similar stirrups in the museums in Irkutsk and Frunze are dated sixth-eighth centuries and are identified as Turkish. There is no intent here to trace the history of the stirrup in any detail nor any farther in space and time--this history may be found in the writings of Bivar, Ambroz, and White already mentioned (see notes 1-4). For our purposes it can be demonstrated that the stirrup appears in China and the adjacent areas to the north by the fourth century and its use was well established shortly thereafter.

The earliest references to the stirrup in the literary record are not easily located. White's review of early mention of the stirrup in Western records falls back on such indirect evidence as the use of the verbs scandere and descendere among the Franks in the early eighth century to replace verbs denoting "leaping."34 In the same way, written sources in China are not particularly useful for the date or place of invention, not to speak of style or shape. The graph for stirrup, pronounced deng,' as it occurs in the Shuowen jiezi" of A.D. 100, represents a word meaning "lamp," and comes to be used for "stirrup" only later."" This use of the graph must be independent of its earlier usage, since it is rather clear that in the latter case the noun is to be related to the verb deng "to mount, to rise." It would seem that the word referred first to a mounting stirrup and that it use was then extended to the ordinary riding stirrup. The literary reference to the stirrup most often cited as the earliest occurs in reference to an event of 477 in which the dispatch of a stirrup was intended to signal an uprising.36 The recent reprinting of the collected works of P. A. Boodberg has made more generally known his citation of an even earlier reference." Wang Luano (fl. 380- 400), a giant of a man, was said to be able to mount his horse while wearing armor, without need of grasping the saddle or using a stirrup! Boodberg also suggested an emendation of a text to read that a carnelian stirrup was pilfered from a tomb dating to A.D. 346.

If the stirrup were an item of the ordinary utilitarian sort, one would not be surprised at the paucity of the written material concerning it. But the stirrup has been claimed to be of enormous consequence in the history of warfare and, indeed, in social and institutional history. Lynn White, in his very provocative treatment of the stirrup, says its introduction marked the third significant phase in the use of the horse in battle, the first two phases being the chariot and the mounted rider. The stirrup was important because it provided the rider with a secure seat and enabled the horseman to become a better archer and swordsman; more importantly, it made possible the effective use of the lance in the charge. No longer was the rider in danger of being lifted from his horse on impact. The stirrup, therefore, "made possible mounted shock combat, a revolutionary way of doing battle.

The logic of White's argument from physics seems irrefutable, and so it is with some surprise that one notes that the rise of "heavy" cavalry preceded rather than followed the appearance of the stirrup. Edward Luttwak, for example, in discussing the auxiliary cavalry of the Roman army in the late Republic and into the first century A.D., mentions that there are those who maintain that since the stirrup was not available these cavalry must have been "light," that is, that they were armed with bow or javelin and might harass the enemy at closer quarters with spear or sword but could not be "heavy" cavalry, those warriors who were armed with the long lance and could charge the enemy with shock tactics. On the contrary, he goes on to say, cavalry charges even without the stirrup were useful in breaking up formations of light cavalry and in attacking unmassed or undisciplined foot soldiers. At Carrhae, Luttwak points out, it was the combined use by the Parthians of heavy and light cavalry that annihilated the seven legions under Crassus in 53 B.c. The light cavalry sent a heavy volume of arrows onto the massed Romans while the threat of a charge by the heavy cavalry prevented the Romans from opening ranks."" One may see depictions of Parthian armored cavalry of this period in the friezes from Khalchayan (Fig. 14).41 Bivar, in his article "Cavalry Equipment and Tactics on the Euphrates Frontier," estimates on the basis of Plutarch's account that the ratio of horse-archers to lancers at Carrhae was about ten to one, and that in the following centuries the numbers of the heavy cavalry may well have increased while the role of the horse-archer seems to have declined.42 The main thrust of the Sgsanian armies, which overwhelmed the Parthians in the third century A.D., was the cavalry charge by heavily armored knights who were relatively immune to long-range archery. One sees the evidence of this "superiority of armor over missile"43 in the friezes of Flrazabad, which depict the heavily armored Sasanians unseating their Parthian foes. The culmination of this development in Iran can be seen in the relief of Khusrau II (A.D. 591-628) at Tgq-i Bustan (Fig. 15), in which one finds a fully armored knight, very close in appearance to the mediaeval knights of Europe, and yet still without the stirrup. The evidence in Iran seems to contradict White's assertion that the stirrup was an essential precursor of the heavily armored knight, even though in Europe it appears that White's observation holds true. It was only after the introduction of the stirrup into Europe in the eighth century that the armored knight began to replace the axe-wielding freeman as the mainstay of the military. What then was the case in China?

The evidence we have indicates that cavalry in China was of the "light" variety until at least the fourth century A.D. The traditional date for the adoption of mounted archery in China is 307 B.C.,44 but the first real evidence with respect to equipment and appearance is probably the extraordinary pottery figures from the mausoleum complex of Qin Shihuangdi. The second of the three pits thus far reported is estimated to contain 116 figures of cavalrymen and mounts."' The armor worn by the mounted soldiers is shorter than that of the armored infantrymen and has no shoulder guards. The robes worn under the armor also differ from those of the foot soldiers in that the overlap is to the front, probably to facilitate riding. The cavalrymen wear small caps with chin straps and are shed in boots rather than in sandals. They are believed to have been archers, since their left hands seem to have held bows, and quivers holding eighty to a hundred arrows each were found in their vicinity.46 The military array of the pottery figures calls to mind the description in the Zhan'guoce of the Qin army of one million warriors wearing armor, one thousand chariots, and ten thousand cavalry.47 The ratio supports what the pits seem to indicate, that the cavalry at that time was a relatively minor part of the military force.

Cavalry apparently gained in importance during the Han, but the uses to which it was put perhaps did not change very much. In the army of figurines found in YangjiawanP the infantry still greatly outnumbers the cavalry--by approximately 2,000 to 600--but whether this ratio of three to one was typical of Han armies is not clear.48 Some of the cavalrymen are said to be depicted as wearing armor, but there is no information as to their numbers."" Some of the cavalrymen carry quivers on their backs.

There are other indications, however, that the cavalry also carried spears and dagger-axes for close-range fighting. The remarkable procession of cavalrymen and chariots from Wuwei,s~ Gansu, shows a troop with such weapons."" How these might have been used is shown in the battle scene on the relief at Xiaotangshan,r where one sees a man being pulled off his horse by a cavalryman wielding a dagger-axe while other horsemen gallop toward each other shooting arrows. The Chinese and their nomad opponents are distinguished only by their headgear (Fig. 16).51 This cavalry can still be classified as "light," because we see no evidence of heavy armor for either rider or mount or any indication of the use of massed charges. The successful use of the dagger-axe depended more on the strength of the arms and the dexterity of its user than on the momentum of the attack. The increasing importance of cavalry as opposed to chariots and infantry came about in an attempt to deal with the Xiongnus horsemen, and in such battles heavy cavalry would not have been very useful.

In the lintel relief from the facade of the tomb at Yi'nan,' possibly late Han or slightly later (Fig. 17), there seems to be very little change in the attack.53 Death is still indicated by the head being separated from the body. Foot and mounted archers move across the scene, but here we see two riders holding short swords and shields. Large side-boards, which become characteristic of the late Ilan and Jin, are in evidence, but there is still no stirrup. The Jiayuguanu painted bricks of the Wei-Jin period are perhaps a better source, since the details are clearer, but here, unfortunately, we may observe only processions, not battles. The cavalrymen who form part of what seems to be an official entourage clearly wear armor, perhaps scale habergeon and casque helmets with plumes, carry spears, and nestle their legs in shaped side-boards for support (Fig. 18). There have been changes in the artistic style from the Han, but none in terms of equipment. In the next few decades, however, we may observe enormous differences from what went before.

One of the new developments of this time was the widespread use of horse armor, or bardings. There is a memorial written by Cao Zhi (192- 232), probably in 226, which is important for the study of the history of armor in China because it gives the names of a number of types of armor of his time. In one version of that memorial a set of horse armor (makai yiling") is mentioned. A few years earlier Cao Cao (155-220), in speaking of the odds he had overcome, says he had twenty suits of armor against ten thousand for his opponent, and ten sets of horse armor against three hundred."" It is not clear what sort of horse armor is meant here--it may simply be the sort of chest protector one finds depicted in the Changsha tomb of 302, where the mounting stirrup is also shown. A hundred years later the capture of five thousand and even ten thousand armored horses is recorded, and other terms such as "iron horse" (tiema)Y and "iron cavalry" (tieji) have also become common."" Iron plates that were probably used for horse armor were found in Feng Sufu's tomb of 415, where the earliest datable stirrups were found (see above). The conjunction of stirrups and bardings is probably not fortuitous, and it may well have been the increasing use of armored cavalry that provided the incentive and favorable environment for the development and widespread use of the stirrup.

One need only glance at the mural from the tomb of Tong Shou, dated 357, to see how different the appearance of the cavalry has become (Fig. 19).58 The body of the rider is almost completely covered by armor. He wears a plumed helmet that protects the sides and back of the head, a habergeon with high neck and shoulder guards, and chaps. The armor was made of lamellar plate, but one cannot say whether of iron or of lacquered leather. The bardings almost completely cover the horse and include a chanfron of distinctive shape. One of these head protectors, of iron, was found in Japan."" As with the stirrup, our best evidence of these developments is to be found in the murals of the Koguryb tombs of the fourth-fifth centuries. A variety of armor is depicted there, including a stirring scene of running combat (Fig. 20). The resemblance of these figures to the cataphracti in the famous graffto of Dura-Europos (Fig. 21), to be dated to the third century, perhaps gives evidence of widespread diffusion of this sort of armor, even if the stirrup was not yet available. The armor of the riders in China underwent a number of changes in the following centuries: "double-faced," or liangdang,ab and "cord-and- plaque" armor were developed, culminating in the marvelous cuirassiers of the Sui and early Tang.60 As we have seen, heavy armor and horse bardings may develop without the stirrup, as they seem to have done in Southwest Asia and some parts of Central Asia. But in China the appearance of the two seem to coincide. It is just at the time that widespread use of horse bardings and heavy armor make their appearance, both in the archaeological record and in the literature, that we begin to find reference to the stirrup. The weight of armor must have made retention of one's seat more difficult and recovery from a fall not easily accomplished. Strabo mentions that the fully armored Sarmatians were helpless when they fell from their horses.61 The Jin figures of horses with high, straight bows on the saddles and what may be molded side-boards into which the upper leg could fit may have been the response to this problem before the stirrup became available

The emergence of the armored cavalryman had an important impact on the nature of warfare in China. The first effect was to give the nomad horseman a new importance, and the nomad dominance of North China in these years is perhaps to be attributed to the nomad's skill in using this new military weapon. The size of military forces during these years is infrequently mentioned, and when given the figures usually lump together infantry and cavalry, obscuring the ratio between these two branches of the armed forces. One may note, however, the association of cavalry with the nomads, especially those of the northeast. In 312, when Shi Leac managed to break a siege of his capital by the Xianbei, he reputedly captured five thousand kaima,ad or barded horses.62 Four years later a Xianbei force joined the hard-pressed house of Jin, but fell into a trap prepared by the same Shi Le, and ten thousand kaima were captured.63 The Yuwenae people of Manchuria (whose rule of Northern Zhou was still two centuries away) gave Shi Hu,af Shi Le's successor, ten thousand horses in 342, while he managed to collect another fourteen thousand belonging to the commandery functionaries under his rule the ease of the gift and the necessity of confiscation contrast sharply.64 In 352 the army of Yan was said to be all cavalry, while that of Ran Min,ag successor to the Shi family, was said to be largely infantry.65 At the end of the century, when the Murongah were being pressed hard by the Tuoba,a' it was argued in planning strategy that the Tuoba armies were mainly cavalry who carried but a week's rations.66 These citations are not comprehensive, but perhaps give a sense of the kind of material one finds in the historical sources. As these nomads moved into China it became increasingly difficult to maintain their numbers of cavalry, yet it would seem that effective cavalry was the key to success in battle.

It is obvious that the cavalry was becoming more important in the wars of this period, but the ways in which the cavalry was employed are not clear, for there is unfortunately little information available to us. All too often the warrior is depicted as a lone champion. Chen An,aj who fell before Liu Yao'sak forces in 323 and became a local hero, is described as brandishing a seven-foot sword in his left hand and a long, snake-bladed lance in his right, but also as being able to fire off arrows at distant foes."' Ran Min wielded a two-bladed lance to dispatch three thousand foes--obviously, a gross exaggeration or a synechdoche in which he is taken to represent his army.68 In another battle the Murong fastened together five thousand of their best mounted archers in a square formation by means of iron chains! In this case it is said that Ran Min led the attack with a double-bladed lance in his left hand and hooked halberd in his right, but despite his beheading three hundred of the enemy, their line held and Ran Min was captured."" The utility of the iron chain for a force of archers is not clear, but it would seem that in this battle we have an example of cavalry being used both as chargers and as bowmen. The elegant cavalier of the southern courts depicted on a brick of the second half of the fifth century (Fig. 23) and cavalrymen in the battle scene from Cave 285 at Dunhuang (Fig. 24) carry bows or quivers or both. This may be only additional protection for a cavalry whose primary purpose was the charge, as Bivar claims was true of the SasSLnian warriors at Naqsh-i Rustam.70 In the Dunhuang mural we see an enemy force of infantry enduring a barrage of arrows from archers on one flank and on the other cavalry lancers picking off individual foe. Just such a scene is described in an incident included in the Zhou shu. Cai you,a' one of the foremost generals of the Western Wei, took part in a disastrous campaign in 538 against the Eastern Wei. As described in his biography, he distinguished himself in battle:

Obviously, literary convention may have recast the conflict into terms of individual combat, for it would make no sense to send a single cavalryman on such a mission. The information on the movement of cavalry erz masse, however, is sketchy. For example, we are told that at the battle of Shayuan"" in 537, to be discussed below, a cavalry charge cut through the enemy flank.72 One can only suppose some sort of massed charge, but no details are given.

At this point, let us broaden the focus from the cavalryman and his function on the battlefield to the military forces in general. As we do so, it will become apparent that during the post-Han period the cavalry had become a crucial factor in the military forces, and that at the same time large-scale mobilizations of the general population came to be less useful. In the following, we will go into some detail about two periods in particular because they provide comparable data on the use of the cavalry. The first period is that of the battles between the two states led by Shi Le and Liu Yao in North China in the early fourth century, and the second is that of the similar struggle between the states founded by Gao Huan"" and Yuwen Tai in the sixth century.

In the first case, North China was divided into an eastern state under Shi Le and a western one under Liu Yao, whose capital was Chang'an. Luoyang was seen as a key point in the defense of the northeast since it controlled passage into the Central Plains.73 At the same time any army wanting to move through the Tong Passap into Shaanxi needed first to control Luoyang. The battles between these two states in the 320s therefore revolved around control of this city. It was a battle at Luoyang that in fact led to the capture of Chang'an by Shi Hu, the successor of Shi Le, in 328, a significant event during this period and one which had important lessons to offer in later years. Liu Yao, the ruler of the western state, was a Xiongnu, and most of his troops seem to have been Dia4 and Qianga' tribesmen. Even when he marched at the head of an army said to number 280,000 men, in 323, it was remarked that he had few well trained troops O'irzg zUaS) and that the Di and Qiang scattered easily.74 An army of Liu's, consisting of eighty officers and several thousand Di and Qiang, was taken prisoner in 325 when Luoyang was taken by Shi HU.75 I, 328 Shi Hu made his next move by leading 40,000 men into the Tong Pass area. Liu Yao sent Di and Qiang soldiers to guard his rear while he led his best troops to counterattack. He was successful and pushed on to take Luoyang.76 There is a statement, no doubt only rhetoric, that he commanded 100,000 armored men (dal jia"t).77 Shi Le responded to this serious threat by leading an army of 60,000 foot and 27,000 horse to relieve Luoyang.78 Liu Yao made the mistake of remaining in Luoyang, where he was trapped and captured together with his army after a fierce battle at the gates of the city. The road to Shaanxi and Chang'an now lay open, and the armies under Shi Hu soon overthrew the Liu family and established control of that area as we11.

It is clear that Liu Yao should not have allowed himself to be stranded at Luoyang, and certainly later tacticians took serious note of his blunder. One may see in later conflicts a general skittishness in this regard--it seems almost as if every clash of arms is followed by one army or the other scurrying back to safety. Luoyang was a critical but also a dangerous vantage point. During this period there is mention of a number of organizational procedures that provide interesting parallels with later developments. Liu Yao had been aware of the vulnerability of his state well before his defeat and had taken steps to strengthen his defenses. In a move perhaps directly related to the threat from the east, a command structure was established to defend the capital. Liu Yao appointed his son Liu Yin"" to be Grand Constable (da sima"") and Grand shanyua" with his own headquarters; below him were the traditional Xiongnu posts, from xiartwalzgax on down, and the leaders of the various non-Chinese groups, such as Hu,"v Jie,az Xianbei, Di, and Qiang, were appointed to fill these positions. Obviously Liu Yao looked to these various tribesmen for his support. What is to be noted here is that the form in which the organization of the defense took shape was the traditional tribal one. In many ways, the case of Liu Yao provides an illuminating parallel with the defense of the same area two hundred years later by the Xianbei under Yuwen Tai.

Turning to the sixth century, we find the situation in North China very similar to that of the early fourth century. Once more the north was divided into two contending states, again separated by that portion of the Yellow River which flows from north to south. The Tuoba state of Northern Wei had broken apart, and the intensity of the conflict may have been fueled by the fact that each side put forth its own Tuoba prince as the legitimate claimant of the Wei throne. Just as in the earlier struggle between Shi Le and Liu Yao, control of Luoyang was crucial, and the area along the Yellow River near that city became the scene of many important battles. A review of this struggle will provide the setting for our discussion of the restructuring of the military forces in the Western Wei state.

The Eastern Wei armies were the more aggressive at first, and the Western Wei seems to have been rather more on the defensive. In 537 a three-pronged attack was mounted on the Western Wei through the Tong Pass. Yuwen Tai crushed one of these armies and the others retreated. Yuwen Tai then followed the withdrawing forces as far as Hongnong where he in turn had to fall back before the main Eastern Wei army, reputed to be two hundred thousand strong, and led by Gao Huan himself. The defeat of that large Eastern Wei army at Shayuan, just west of the Tong Pass and thus inside the threshold of the Western Wei stronghold, by a Western Wei force of a mere ten thousand men was a significant event in Western Wei history, and is mentioned frequently in the history of that state. The claim was that some six thousand of the enemy were killed and seventy thousand captured. Dugu Xin,bb a veteran Xianbei general under Yuwen Tai, was then ordered to take advantage of the victory by occupying Luoyang.

The triumph of the Western Wei was short-lived, however. Dugu Xin was besieged by the Eastern Wei at Luoyang, and Yuwen Tai led a force to relieve him. The besiegers withdrew, with Yuwen Tai leading his light cavalry in pursuit. The battle took place at Heqiao,bc and the Western Wei lost. The situation was a perilous one, since Hongnong, on the line of retreat, had gone over to the Eastern Wei, and, in addition, the many prisoners taken the year before were rebelling in Chang'an. Yuwen Tai managed to overcome these difficulties and may have considered himself fortunate to have lost only the several tens of thousands that Eastern Wei claimed he had lost at Heqiao.

In 542 the Western Wei was successful in keeping possession of yUbi,bd a Stronghold on the east side of the Yellow River,83 but in the next year there occurred another setback for that state. Yuwen Tai saw an opportunity to retake Luoyang when a defector from Eastern Wei offered him entry, but he was unsuccessful in preventing Gao Huan from leading his army across the Yellow River. At a battle in the Mang Mountains the Eastern Wei line held, and Yuwen Tai was forced to lead the remains of his army back through Tong Pass. The Eastern Wei claimed that he lost more than four hundred officers and sixty thousand soldiers.84 Whatever the actual count, it must have been a serious loss.

There was little action of this sort during the next decade, but starting in 553 there appears to have been an increase of military activity and of numbers of combatants on the part of Western Wei. In that year Yuwen Tai led thirty thousand cavalry against the Tuyuhunb~' in the northwest.85 Having thus secured his rear, he dispatched fifty thousand foot soldiers and cavalry the next year to sweep over the defenses of the state of Liang to the soUth.86 There followed another decade- long pause in hostilities, until the northwestern state, now termed [Northern] Zhou, began joint operations with the Turks against their neighbor, the [Northern] Qi in 562-63. Two columns, one of ten thousand cavalry from the north and the other of thirty thousand cavalry from the south, joined the Turks in an attack, but the action was ultimately unsuccessful.87 The next effort, in 564, was a general mobilization of the state. The description of this immense effort is as ponderous as the actual movement must have been. "Thereupon they summoned to the court the twenty-four armies, the unattached and the attached troops of the left and right apartments, the soldiers of Qin,bf Long,"g Ba,bh and Shu,b' and the hosts of the various border states, amounting to two hundred thousand men."ss Unfortunately, there is no breakdown by type of the troops involved. This army returned after only one battle.

Another incident in 570-71 led to some minor activity, but in 575 a series of campaigns was mounted against Northern Qi that eventually led to the conquest of that state. The first attack was made in a number of columns (amounting to 170,000 men), but again the forward movement seems to have faltered.g" The actual conquest in 576-77 was accomplished by a force of 85,000 men, of whom 20,000 were said to be cavalry and the remaining 65,000 a mixed force of foot and cavalry.

In the foregoing account of the events during this struggle in the sixth century figures concerning the number of troops involved occur, but they are few, and any analysis based on such numbers must be tentative at best. Nevertheless, there are parallels that may be drawn between the conflicts of east and west in the fourth and in the sixth centuries. We have seen that the large-scale mobilization by Liu Yao in 323 was not considered a serious threat; similarly, the campaigns of 564 and 575 that involved large-scale mobilizations were not particularly successful. The same may be said of Shi Hu's attack on Yan by five hundred thousand men in 34292 ,,d the enormous concentration of manpower assembled by Fu Jianbi in 383.93 I, all of these cases the administrative techniques employed to mobilize such numbers far outstripped the capacity to use them effectively in the field: the significant victories were won by smaller numbers of more highly trained troops.94 The information we have suggests that the cavalry, both light and heavy, came to be an essential part of the armed forces, that greater reliance was placed on these mounted warriors and less on the poorly armed and inadequately trained infantry raised by large-scale levies for specific campaigns.

During the years of division, bearing arms in North China was mainly the responsibility of the non-Chinese invaders who ruled that area. It is said that Gao Huan told his soldiers in the Xianbei language, "The Chinese are your slaves, they farm for you, the women make your clothing, they bring you your supplies and clothing, enable you to be warm and well fed. Why do you want to oppress them?" To the Chinese he would say, "The Xianbei are your clients, you give them a measure of grain and a bolt of cloth, and they fight for you so that you may have peace and order. Why do you hate them?""" Given such a division of functions, the position of the Xianbei as a ruling elite was stable. The Chinese could be mobilized as foot soldiers for local defense or for an occasional large-scale campaign without disturbing the arrangement. But armor for man and horse had become quite heavy after the appearance of the stirrup in the fourth century, and such equipment was also expensive. At the same time, the level of warfare had intensified among the rival states in the north, and between them and the southern Chinese states as well, while the reservoir of Xianbei tribesmen dropped. It thus became necessary to bring into the army well equipped and well trained Chinese. Only those of substance and strength were admitted in this recruitment since it was expected that they would contribute toward supplying their own equipment.

In 543, after the disastrous battle of the Mang Mountains in which the Western Wei forces sustained enormous losses, the statement was made: "Thereupon they recruited widely among the local magnates of Guanbk and Long (i.e., the core area of Western Wei) in order to increase the military forces. When we look to the sources to see whether there is actual evidence of this recruitment, it is most significant that we find five references to "local troops" being raised within the state and no mention of any other sort of recruitment. As an example, we have Guo Yan,b' who in 546, "having been selected as being most eminent in his prefecture, controlled and led the local troops. The other four cases are similar. One can find an echo of this system as late as 574, when it was announced that in certain areas rank as an officer was to be given to any who would recruit and lead a group of men to serve in the army. The units thus raised at the local level were fed into a centralized command structure that had the responsibility of fielding the main military force of the state. But as the Xianbei rulers had begun to accept Chinese and other ethnic groups into their forces, it must soon have become apparent to them that they were in danger of being swamped by this new influx of recruits. At the same time, there may have been a need to introduce a more systematic organization into the army. We have noted that two centuries earlier Liu Yao made use of the traditional Xiongnu apparatus with a shanyu at its head. In this case the Xianbei revived a pattern of dan confederation that frequently served as the organizing principle of nomad empires on the steppes.

A decree was issued in 554 that the thirty- six tribes and ninety-nine clans of the traditional Tuoba state were to be reinstated, that the officers of highest merit were to be appointed the heads of these tribes and clans, and that the soldiers they commanded were to adopt the surnames of their officers.'0' This would have the effect of making the army into a tribal confederation, as the Tuoba state had been at its beginning. The bestowal of Xianbei surnames upon eminent Chinese, so common at this time, was intended to produce the same result, that of tribalization or "Xianbei-ization." The effect of these policies was to make the enlistment of Chinese into the military forces acceptable to and controllable by the Xianbei rulers. The next step, taken in 574, no doubt in preparation for the conquest of Northern Qi, was active recruitment of Chinese; those who enlisted had their names removed from the district rolls so as to exempt them from taxes and corv~e. Then, it is said, half the Chinese became soldiers, obviously an exaggeration but a significant statement nevertheless.

To return to the question posed at the start of this paper, why, if the stirrup brought feudalism to Europe, did the same thing not happen in China? Joseph Needham, in considering this problem, has recourse to the "astonishing stability" of Chinese civilization, which is "so deeply civilian in its ethics that the very conception of aristocratic chivalry was perhaps impossible."'"" The answer, rather, is to be found in the developments described above. In China the bureaucratic apparatus existed to administer the resources of the state down to a relatively low level.'"" There was little need to "broker" central power on a regional basis in order to acquire the needed resources. "Local troops" were a kind of brokered product--rank was given in exchange for the service of these personal followers, but only at the local level. The recruitment of Chinese into the centralized army gave the Northern Zhou and its successors, the Sui and the Tang, the military advantage over their rivals. The advanced bureaucratic techniques of the Chinese state enabled this recruitment to remain under centralized control--there was no need for power to be brokered by middlemen as was the case in Europe. The problem for the Northern Zhou rulers, however, was that as they permitted the Chinese greater participation in the military, they were allowing their hold on the state to slip away. The pretense of Xianbei names could not conceal the fact that the Xianbei forces were becoming a less important factor in the power structure of the state. Just as the military participation ratio (to use Andreski's term) of the Chinese increased, the Xianbei elite had to make way. Eventually rule slipped from Xianbei hands and after a transitional period of rule by men of half-Chinese, half-nomadic background, mirroring the army itself, China again came to have Chinese rulers--all this without undergoing a racial confrontation. The stirrup and the heavy armor that it made possible, therefore, seem to have had very different consequences from those which emerged in Europe.