"Man does not enter battle to fight, but for victory. He does everything that he can to avoid the first and obtain the second"
--Ardant du Picq, Battle Studies: Ancient and Modern Battle, trans. John Greely and Robert Cotton (or. pub. Paris, 1870; trans. edition, New York, 1921), pg. 1.
Of the many books and articles published explaining the tactical mechanics of ancient and medieval warfare, none have influenced my views on the topic more than a short article by Philip Sabin titled "The Face of Roman Battle." In this article Sabin attempts to draw an accurate description of the way a Roman legion and its maniples actually worked on the battlefield. He is not the only one to attempt this feat. The clearest description of the pre-Marian armies is the account found in the eighteenth book of Polybius's Histories, and historians have been squabbling over just what Polybius's rather ambiguous report means for the better part of the last two centuries. I believe that Sabin's is the best of their efforts. What makes his description so convincing is the building blocks he uses to construct it. Sabin starts his reconstruction with a few general insights about the nature of ancient combat, especially the hand-to-hand sort. His most important insight is this: close combat is absolutely terrifying. When you realize just how terrifying it is much of what we find in the ancient and medieval source starts to make a lot more sense.
Sabin's case study is the Roman legion. In his essay's first section Sabin surveys common features of battle narratives preserved in the extant histories and concludes that any description of Roman battle mechanics must be able to explain a few odd features of these accounts to be considered legitimate:
Roman heavy infantry engagements possessed several clear characteristics which must be accounted for by any model of the combat mechanics involved. If not decided at the first clash, the contests often dragged on for an hour or more before one side finally broke and fled. The losers could suffer appalling casualties in the battle itself or in the ensuing pursuit, but the victors rarely suffered more than 5 per cent fatalities even in drawn-out engagements. The fighting lines could shift back and forth over hundreds of yards as one side withdrew or was pushed back by its opponents. Finally, the Romans had a practical system for the passage of lines, and preferred to reinforce or replace tired units with fresh ones rather than maximizing the depth of the initial fighting line. As Sabin read the ancient accounts he realized that parallels for many features of Roman combat could be found in descriptions of early modern Europe's bayonet charge:
We know from eighteenth- and nineteenth-century engagements that bayonets caused only a tiny proportion of battle casualties, but bayonet charges do seem to have been decisive in triggering routs. The explanation for this apparent paradox seems to be that cold steel held a unique terror for troops, over and above that caused by the more random and impersonal perils of shot and shell. The morale of opposed infantry formations appears to have been closely interlinked, such that if one side could nerve itself to launch a bayonet charge in the conviction that the enemy would not stand, the enemy did indeed break before contact. Conversely, if mutual deterrence was maintained, then the combat could bog down into a bloody close-range firefight between the opposing lines, often lasting for hours....Why was cold steel a "unique terror" for troops in combat? On the face of it a sword does not seem any more frightening than the cannon-ball. Pop culture portrayals of small imperialist forces putting hordes of backward natives to flight with nothing but gun and powder suggest the opposite conclusion. Images of countless thousands led to the slaughter on the banks of the Somme or hills of Verdun only strengthen the impression. But those men who actually withstood both the bullet and the bayonet overwhelmingly preferred to face the former. A similar preference for arrows and cross-bows shot from afar over spear thrusts and sword strokes closer to home pervades the ancient and medieval sources.
There are striking parallels between the psychological role of bayonet charges in modern warfare and the way in which many ancient combats were decided at or before the first shock, with a charge by one side prompting its enemies to take flight at once. Hoplite engagements seem to have been particularly susceptible to such an early resolution, sometimes even producing 'tearless battles' when one side fled so soon that it outdistanced any pursuit. Goldsworthy claims that late Republican and early Imperial legionaries exploited their professionalism and esprit de corps by winning similar swift victories against less resolute opponents through a coordinated volley of pila followed by a fierce charge. This chimes exactly with Paddy Griffith's argument that the disciplined British infantry of the Napoleonic Wars beat the French not through winning prolonged firefights but through a single devastating musket volley followed by a charge with the bayonet. 
To understand why this was so you must discard Hollywood notions of close combat. This is hard to do, for the notions are much older than Hollywood. The classical Chinese novels Outlaws of the Marsh and Romance of the Three Kingdoms speak of warriors who exchange five, ten, twenty, and even fifty "rounds" or "clashes" on the battlefield. The long duels of ancient India's great war epic, the Mahabharata, are matched only by the extended contests of its Greek counterpart, the Iliad. All of it is poppycock. Ancient battles did not descend into a series of extended melees when the two front lines collided. The silliness of the Hollywood style of battle becomes immediately apparent when you watch sparring competitions that use pre-modern weapons:
As you can see, most close quarter engagements are decided within seconds. To engage in hand to hand combat is to hang your life on a the balance of a few split second decisions. This is terrifying. It is all the more terrifying if the enemy force is as committed and disciplined as your own. If you survive the first encounter--that is, if you successfully kill the first man who attempts to kill you--there will be another, and then yet another to fill in his place. How long can you keep making instant life-or-death decisions before you make a mistake? The odds are not in your favor. The physical and mental strain of close quarters combat on those in the front lines is simply more than can be borne for any great stretch of time.
Sabin explains why this is important:
What does all this mean for the many cases in Roman infantry battles where neither side broke at the outset, and the combat turned into a prolonged affair? I suggest that close-range sword dueling between steady bodies of infantry must have been a highly unstable state, and one that would require massive injections of physical and psychological energy either to initiate or to sustain for any length of time. It was clearly only the availability of protective armor and shields that made such duels endurable at all, given their apparent intolerability for the unprotected troops of more modern times. I would argue that there must also have been a more physically and psychologically sustainable 'default state' within protracted Roman infantry contests, into which the combatants would naturally relapse if the initial advances by either side failed to trigger an early rout.Sabin does not believe that "pure missile duel" style of battle, decided by one great final charge at its end, accords with the surviving narrative sources:
We can see such 'default states' in a wide variety of other forms of human combat. Anthropological observations of primitive tribes confirm the image in heroic poetry of protracted stand-offs in which individual warriors would move forward to do battle and then retreat into the safety of the supporting mass. Even when lethal weapons are not involved, we can see similar stand-offs between rioting mobs and lines of police, or at an individual level between dueling boxers, who spend much more time circling each other warily and looking for an opening than they do in the actual flurries of blow and counter-blow. I suggest that the default state in protracted Roman infantry combats would have been similar to that between eighteenth- and nineteenth-century infantry, namely a small separation of the two lines so that they could exchange insults and missile fire but were not quite close enough for hand-to-hand dueling. If such a default state existed in Roman infantry clashes, this raises the question of the frequency and duration of actual sword fighting between the opposing lines. Could troops who had closed for such sword play disengage without routing, and re-establish the 'safety distance'? How long a period of sword fighting was physically and psychologically sustainable before the tension had to be broken either by a reversion to the default stand-off or by the flight of one side? What proportion of the overall length of infantry clashes was spent in sword dueling, and what proportion in sporadic missile exchanges from a short distance away? 
Such a radical image seems to me incompatible with the many references in the literature to true hand-to-hand fighting, and it makes it difficult to explain how one side could 'push back' its adversaries during the course of the contest. Hence, unlike in the stalemated firefights of more recent times, I believe that in most Roman battles the lines did sporadically come into contact, as one side or the other surged forward for a brief and localized flurry of hand-to-hand combat. The flurry of combat would end when one side got the worst of the exchange, and its troops would step back to re-impose the 'safety distance' while brandishing their weapons to deter immediate enemy pursuit....But if cold steel was so frightening, why would men engage in close quarters combat at all? Again, Sabin explains:
The model of Roman infantry combat as a dynamic balance of mutual dread fits the overall characteristics of the phenomenon far better than do the alternative images of a protracted othismos [i.e. a group of massed infantry pressing each forward, hopolite style] or continuous sword dueling. It helps to explain why some clashes were decided at the first onset while others dragged on for hours. It accounts for the relatively low casualties suffered by the victorious army, since periods of close range stand-off would be far less bloody than the equivalent firefights in the eighteenth and nineteenth centuries, given the much lower numbers of missiles available and the fact that the great majority would be blocked by the large infantry shields (cf. Livy 28.2, 28.32-3; Caesar, BGI .26; Josephus, BJ 3. I I2-I4). The model also suggests how one side could gradually 'push' another back over distances of hundreds of yards, since if it was always the same side that gave way after the sporadic flurries of hand-to-hand dueling, the accumulation of such small withdrawals would have significant grand tactical impact over time. 
Why would parts of each line sporadically surge forward into contact? The key individuals would surely be the 'natural fighters' and junior leaders, who would encourage a concerted lunge forward to overcome the understandable reluctance among their comrades to be the first to advance into the wall of enemy blades. Roman sub- units such as centuries, maniples, and cohorts offered an ideal basis for such localized charges, whereas tribal warriors would mount less disciplined attacks led by the bolder spirits among them. The many accounts of Roman standard-bearers carrying or flinging their standards towards the enemy to embolden the onslaught of their comrades (as at Pydna and in Caesar's invasion of Britain) are of obvious relevance in this connection (Plutarch, Aem. 20; Caesar, BG 4.25). Across an overall infantry battlefront many hundreds of yards wide, the back and forth movement of individual sub-units or warrior bands just the crucial few yards to engage in or disengage from hand-to-hand combat would not prejudice the maintenance of the overall line. If such flurries of sword fighting were not quickly decisive, then sheer physical and nervous exhaustion, coupled with the killing or wounding of the key junior leaders who were inspiring their men to engage, would lead the two sides to separate back to the default stand-off. The fact that even phalangites could step back facing the enemy (as at Sellasia) indicates that there was usually sufficient 'give' within infantry formations to allow front-rankers to shy away from their adversaries without bumping immediately into the man behind. Indeed, when this flexibility was removed and troops became too closely packed together, thereby hindering their ability to use their weapons properly or to step back from clashes which were not going well, they risked exposing themselves to one-sided slaughter. Something like this clearly happened at Cannae, and it could well be that a key reason why flank and rear attacks were so devastating was not just the psychological shock they caused but the fact that they crowded the victims in on one another, removing their ability to re-establish the 'safety distance' and so to recover their cohesion and fighting effectiveness. Sabin goes on to describe how this model of legionary activity makes sense of the ambiguous descriptions of Rome's famous maniple system, and why the maniples would be so effective in this style of combat. I encourage those interested in Roman history to read the entire thing. But I hope readers can see how easily Sabin's insights transfer to the wars of men who lived far from Rome. Not every army in the pre-modern world had maniples, but many had large infantry contingents intended to destroy their enemies in close quarters. The tempo of their battles would have been decided by fear and terror, as it was with the Romans. Sabin's model of periodic surges of courage temporarily hurling front lines together should be the default image of every mass infantry battle waged in the pre-modern era.
Other posts on The Scholar's Stage about pre-modern warfare:
"The Radical Sunzi"
T. Greer. The Scholar's Stage (2 January 2015).
"ISIS, the Mongols, and the "Return of Ancient Challenges"
T. Greer. The Scholar's Stage (18 December 2014).
"What Edward Luttwak Doesn't Know About Ancient China (Or a Short History of the Han-Xiongnu Wars)"
T. Greer. The Scholar's Stage (4 September 2014).
"Whence Springs a Strategic Canon?"
T. Greer. The Scholar's Stage (9 April 2013).
EDIT (27/10/2015 11:00 AM): A reader has pointed out to me that Philip Sabin has recently published a book that fleshes out this model and uses it to analyze the narrative accounts of famous Roman and Greek battles. I have not read it yet, but it looks interesting: Lost Battles: Reconstructing the Great Clashes of the Ancient World.
EDIT (27/10/2015 4:00 PM): Also see my short follow up post to this one: "A Few More Thoughts on the Terrors of Pre-Modern Battle."
 Philip Sabin, "The Roman Face of Battle," Journal of Roman Studies 90 (2000), p. 5
 Ibid, 13
 Ibid, 14
 Ibid, 15
 Ibid, 16