That evening, however, the mobility and speed of the Turkish troops was demonstrated once more, for while a number of the Oğuz mercenaries were outside the camp doing business with local traders and merchants, Seljuk warriors appeared once again in the semi-darkness and swept in to harry those caught off guard outside the fortifications. Panic ensued as the Oğuz troops tried to hurry back inside the camp, while those inside the camp were equally confused by the fact that the Oğuz were hardly different in appearance from the Seljuk enemy, and the rumour spread that this was a full-scale assault upon the camp. But as suddenly as they had come the Turks disappeared, and although occasional raids took place during the night, no major attack was launched.
Next morning, 25 August, another Seljuk detachment attempted to seize control of the river bank opposite the Roman camp, but were driven off quickly by a concerted assault from the Byzantine infantry posted to defend the position. Shortly afterwards, however, a considerable body of the Oğuz warriors deserted to the Seljuks – their distant cousins – causing some anxiety that the remainder would follow. But they assured the emperor of their loyalty with an oath, and the worry subsided. Shortly afterwards there arrived in the imperial camp an embassy from Baghdad, from the court of the Caliph al-Muhalban, offering to negotiate. But Romanos would accept only conditions – including a Seljuk withdrawal – which it was clear the Turks could not, and so the embassy withdrew. It is also possible that he believed, or was persuaded, that the embassy was merely a ploy to delay the confrontation, allowing the Turkish Sultan to increase the forces at his disposal – the emperor having in the meanwhile sent messengers to recall the important body of troops he thought was still in the area of Khliat. In addition, and in spite of the situation of the army at the time, the Roman forces still had an impressive reputation, particularly in pitched battles where discipline and order were key elements; and it was generally felt that with the numbers of troops he had assembled and still available to him, the emperor had an excellent chance of inflicting a heavy defeat on the Turks, if he could bring them to battle.
It was, therefore, shortly after the departure of the Baghdad embassy, and on the morning of Friday 26 August, that the imperial army was drawn up outside the camp and prepared to advance on the Turkish host. The left wing was commanded by Nikephoros Bryennios and included his own western tagmata which, as doux of the west, he had commanded in Asia Minor for some time. The right wing, under Theodore Alyates, consisted of the Cappadocian and presumably most of the other Asia Minor units. The centre, under the emperor himself, included most of the guards units – the scholai included – and the Armenian heavy infantry. We may assume that the majority of the Roman heavy cavalry which remained with the main army (many had been sent off with Tarchaneiotes) were also in the centre, while the Pechenegs and Oğuz who remained with Romanos were disposed on the wings.
In spite of the arrogance and haughtiness which the emperor is reported by some sources to have displayed during the campaign, Romanos had the reputation of being a sensible and cautious general, and he knew the Turkish tactics well. The rearguard and reserve was, therefore, placed well behind the main force, able both to relieve the main line should it be forced to fall back, able also to close in and cut off any enemy units which attempted to encircle the main line and attack from behind. The rearguard included the Hetaireai and certain other elite units as well as regular tagmata, and possibly also a number of allied Turkish – Pecheneg or Oğuz – troops.
The terrain over which the battle was fought stretched over open, rocky ground from the city of Manzikert itself, with the entrenched and fortified imperial camp pitched a short distance to the south or south-east. From the city to the foothills to the south and south-east was a distance of between twelve and fourteen km of stony, rolling steppe land, the ground rising gradually before breaking up into an area of shallow gullies and stream-beds. At some distance from the Roman lines, but well in advance of this rougher land, Arslan had drawn up his own, less numerous force in a crescent formation, although he was himself not among the main body of troops, preferring to observe events from the higher ground to the rear. The Seljuk army was, in effect, divided up into a centre and two wings, but in traditional nomadic fashion these divisions in turn consisted of several smaller groupings which could, where needed, act independently.
Eastern Roman military practice demanded that, where an army of mobile horse-archers was to be attacked, the enemy should be brought to close action as swiftly as possible, to avoid undue casualties and attrition from long- and medium-range missile attack. The Roman force advanced, therefore, at a steady pace, in order with the rearguard keeping well back to protect the main line and flanks, while the Seljuks, equally disciplined, harried the Roman line with arrows while constantly but steadily moving back. While the centre withdrew, however, refusing any close engagement at all, the Turkish wings acted more aggressively, sweeping in to attack the Roman wings at close range before withdrawing again, thus forcing the Roman line to grow ever more ragged as the centre, pushing forward after the Seljuks opposite them, began to move well ahead, while the wings were held to a slower pace by the Turkish sallies.
By mid-afternoon the Byzantine centre division had reached and overrun the Seljuk camp – entirely emptied of its contents, of course – and were still pushing ahead, and when the afternoon drew to a close the emperor’s own division had reached the rougher terrain ringing the edge of the plain across which his army had steadily advanced. Losses thus far seem to have been minimal, but on the flanks things were beginning to get out of hand. The constant shower of arrows was causing considerable annoyance to the Byzantine forces and, motivated by frustration rather than by tactical common sense or good discipline, many of the units attempted to charge their foes to bring them to battle. The Turks before them hesitated and then, in classic steppe style, feigned retreat, drawing the imperial forces into the rough ground where a series of ambuscades had been prepared.
It was now dusk. The emperor had been unable to come to grips with his enemy. He was no longer in close contact with the wings, which were only tenuously connected with his own division; and his army, which had left the main camp more-or-less undefended and extremely vulnerable to a swift mounted attack behind his lines, including his rearguard, was without supplies. To continue the advance into unknown and much rougher terrain, almost certainly prepared by the enemy, would have been disastrous. His army was still well-ordered and coherent, and an orderly withdrawal – which he had himself conducted in not dissimilar situations in the previous Anatolian campaigns – was now the only reasonable course of action.
It was at this point that disaster struck. Having given the order to withdraw, the centre began to pull back. But on the right wing the signal to withdraw was misunderstood by some soldiers and officers, who believed that the emperor had fallen. The rearguard was essential to the success of this withdrawal, of course, for it was the rearguard which would cover the units falling back and enable the manoeuvres to be carried out without undue haste and panic. But Andronikos Doukas was a sworn enemy of Romanos IV, and it is clear from all contemporary accounts that he must have deliberately failed to follow the normal procedures. Rather than wait to cover the retreat of the main line, he reversed his own lines and simply marched back towards the camp, leaving Romanos’s division exposed, and the wings entirely isolated. Although there is one near-contemporary account which tries to explain this action away, most accept that this was a deliberate act of treachery, designed to leave the emperor in the hands of the enemy and, Andronikos may have hoped, bring about his death – an event which would leave the path to the throne of a member of the Doukas clan uncontested. By the same token, these accounts all report the fact that Andronikos deliberately spread the rumour that the emperor had fallen in order to persuade the remaining divisions to abandon the field.
Seeing what was happening to the rearguard, and recognizing the confusion on the right wing, the Seljuks now launched an all-out attack. The right wing crumbled and dissolved in rout, fleeing back across the plain; the left wing, under Bryennios, seems to have withdrawn in order, until the Turkish units which had cut the right wing off from the centre swept into its rear and forced it, too, to break up in flight. Realizing the chaos the original misunderstanding had brought about, Romanos attempted to recall the panicking units and rally them to his own standard, but his signal was ignored and the central division too began to break up and withdraw, apparently in some order still, under the incessant barrage of Turkish arrows which now fell from all sides. The emperor and a small part of the centre were now entirely surrounded and, attempting a final stand, Romanos fought on, injured, until his horse was killed under him and, after a brief struggle on foot, he was knocked down. Only the next day, as the Turks went through the bodies collecting weapons and armour, was he recognized and taken before the Sultan.
Although the imperial army had dissolved at the end of the day’s fighting, contemporary or near-contemporary claims that casualties were very high seem largely exaggerated, and the evidence suggests, on the contrary, that overall casualties were, in fact, quite light. The rearguard and reserve units under Andronikos Doukas escaped entirely unscathed and marched back to Constantinople – where Andronikos was immediately involved in the deposition of Romanos. Bryennios escaped similarly unscathed, and the left-wing units, in particular the tagmata of the west, were fighting effectively against Slavs and Pechenegs in the Balkans the following year. Of the units with the emperor in the centre and on the right wing, the Cappadocian tagmata seem to have been able to withdraw in some order, as did several of the elite units with the emperor himself, such as that of the stratelatai. Indeed, immediately after the battle, substantial numbers of units from these divisions had retreated on Dokeia (mod. Tokat, an important fortress on one of the main routes back to Constantinople, and to the south-east of Amaseia). There the emperor found them eight days later, after the Sultan had released him. Along with their commander Alyates, who had also escaped, the Cappadocian and other Asia Minor contingents vowed to support the emperor against the usurper who had been proclaimed in Constantinople upon the circulation of the false news of the emperor’s death in battle.
It would seem, in consequence, that the actual losses incurred during the battle were quite limited, and affected largely the emperor’s immediate retinue and bodyguard. A recent sensible analysis of the battle has estimated that, even if as many as 20% of the army were taken prisoner in the final hour before darkness fell, the overall losses in dead and wounded probably amounted to no more than 10% of those present at the beginning of the engagement. The most significant loss, apart from the capture of the emperor himself, seems to have been the imperial encampment and baggage train, which was abandoned as the various divisions and units withdrew towards Manzikert and beyond, and which was extremely rich – the sources emphasize the enormous quantity of booty that the Turks found there.
The reasons for these relatively light casualties are not hard to find. In the first place, the emperor’s command was not finally cut off and surrounded until it was almost dark, so that the pursuit seems not to have been pressed beyond the imperial encampment itself. In addition, even as the other divisions withdrew, most of the Turkish troops were concentrated on the defeat of this body of troops around the emperor. This is significant from another point of view also, for, taken together with the Turkish tactics throughout the battle, it implies very strongly that the Seljuk forces were themselves nowhere near as numerous as many of the (later) medieval sources claim. Furthermore, Manzikert had been garrisoned and served as a refuge for many of the troops, and since the Turks did not attempt to retake it, it remained a safe haven until after the emperor’s release.
In contrast to most popular judgements, therefore, the defeat at Manzikert was not a military disaster and did not entail the destruction of the eastern Roman army. On the contrary, the greater part of this army managed to make its way from the field with relatively few losses and in enough order to withdraw, by separate routes and at different rates, to comparative safety. The real disaster was of a political nature, for the capture of the emperor, which was soon broadcast throughout the political circles of the near and middle east, dramatically affected the general view of the Roman empire – a view certainly accepted until that point by the Turkish leader himself – that the Roman empire was a permanent, stable and unshakeable element of the political universe of the era. Romanos’s capture showed that this was not the case, and encouraged a sea-change in attitudes to the power of the eastern Roman state.
By the same token, however, the ensuing civil war, which sapped the resources in troops and money on both sides, left Asia Minor undefended against the continuing incursions of Turkmen nomads and raiders, which Alp Arslan himself was relatively powerless to prevent. It also entailed the wearing down of the divisions which survived and the destruction of many units, so that the army of the eastern Roman empire suffered far more from this internecine strife than it had hitherto suffered at the hands of external enemies.
