Line and Unit Replacement

Line replacement is the most difficult aspect of legion warfare to model. It seems to have happened but there is no widespread agreement about just how it might have been done. Connolly gives an excellent description of the process for the maniple legion, Sabin offers good arguments for its importance and for the absolute uniqueness of the Roman method. Neither fully address what the process for the cohort legion might have been. It is the cohort legion that is the focus of this section.


Line replacement begins with Peter Connolly's treatment in Greece and Rome at War. His description is covered in "Alternatives to Gaps: Closing and Re-Opening Them," a page on the Roman Army original site. Wary, Warfare in the Classical World, has much the same description. Sabin seems to favor open gaps with the second line simply moving forward and the first line falling back, the gaps remaining open all the time.

This is for the maniple legion. For the cohort legion there are no similar descriptions. Some would say that line replacement never happened with that legion formation. Yet, something like it must have. The triplex acies formation would have required some mechanism for brining the second and third lines into the battle. Goldsworthy argues that the first two lines routinely supported each other and that the third line was a strategic reserve. At Pharsalus Caesar certainly seems to say that the first two lines were both fully involved in the battle while the third was held back. Furthermore he says that the men in those two lines were tired and that he sent the third line into the fight and that these fresh men helped turn the tide.

Whatever one wishes to call it, the cohort legion had a method of bringing fresh men into the front. If we take Caesar at his word, not only could the second line pass through the first but the third could pass through both.

Eliminating unlikely methods

Sometimes the easiest way to move forward is by eliminating unlikely solutions.

Opening and closing gaps

The solution Connolly and others offer for the maniple legion is unworkable for the cohort legion. In my view, it is unworkable for the maniple legion as well, but it is certainly unworkable for the cohort. A cohort at its smallest possible configuration, with the centuries stacked front to back and each many ranks deep, had to have a front of at least 100 feet. More likely, the cohort had a front of 150 to 200 feet. The cohorts in the illustration below, the formation used for the battle models, are 228 feet wide, including half the gap between cohorts.

It is difficult to imagine a way to replace whole cohorts in a manner similar to that proposed for the maniple legion.

In the first place, would three cohorts adequately replace four? The nearest cohorts of the adjacent legions are shown colored light blue. If all of the legions replaced the first line with the second there would be large gaps between the legions. The gap between the second line cohorts is 258 feet.

Even if the gaps were not a problem, what mechanism could be proposed to get the front line 4 legions out of the way. However one envisions it, the process would entail intricate maneuvers. Cohorts or centuries would have to back out of the line of battle, move sideways, others move up and/or sideways as well. Such complicated maneuvers are unlikely, probably impossible, as a standard method of line replacement. Especially when one keeps in mind that armies were often made up of legions that had not previously fought together and sometimes the legions themselves appear to have been assembled from cohorts recruited for the particular battle.


Philip Sabin proposed a dogtooth formation, one in which the gaps in the front line are left open but are covered by the cohorts of the second line. This is illustrated below.

This seems like an unlikely solution as well. On the right the adjacent legion's cohort is at the same spacing as the other cohorts. In that case the cohorts of the second line would have a gap of 710 feet between them. On the left the two legions have been moved close together. The gap between the cohorts of the second line is only 484 feet. If the second line replaced the first the entire front would be so thinly manned that any enemy could penetrate the gaps.

Nor would such a formation give the Romans a selective advantage. a part of the argument is that the cohorts of the second line would be close enough to cover the gaps so that no enemy would dare penetrate them. Even if this were true, what would the advantage to the Romans be? The enemy troops opposite the gaps would be just as rested and fresh as the Romans of the second line. In fact, without the necessity of making complicated maneuvers, the enemy would actually have the advantage.

However, the real difficulty with this formation is its width. Referring to the example of Pharsalus one more time, the terrain there was considered appropriate for battle by the two greatest generals of the time. There is no mention of any special adjustments made because the plain was too narrow for normal tactics. Therefore I would conclude that the space available there was adequate for a normal deployment. The dogtooth formation would be nearly twice as wide as what the field of Pharsalus could accommodate.

Cohort by Cohort

If wholesale line replacement does not work, does cohort by cohort replacement work? Could the individual cohorts of the second line simply replace tired cohorts of the first line as needed?

To be clear, this scenario proposes that one entire cohort replace another as the result of some sort of command. It would make no difference for the model if the command originated from the wing commander or came at the request of the lead centurion of the cohort to be replaced.

To better understand the potential dynamics of cohort by cohort replacement it might be beneficial to use another type of diagram. The following illustration is from Frame 7 of the model of legion fighting.

This frame is from early in the battle. But even this early the lines are jumbled. The centuries of alternate cohorts are colored light blue, otherwise it would be difficult to pick out one cohort from the other. Some centuries are more actively engaged than others. In this model the fall back position is around 60 feet. Some, including Sabin, do not think the distance should be that great, that the opposing lines should be only a few feet apart. That would only make the prospect of pulling a cohort out of the line more difficult. In either case, however, the mere fact that the centuries cannot hold their nice tight formations but move around as they encounter the enemy would make cohort replacement tricky. It would not be easy to get the attention of the six centurions, all of whom are tightly focused on the enemy just a few feet away who is trying to kill them.

Ultimately cohort replacement would face the same difficulty that line replacement faced. If successful it would leave the line with three instead of four cohorts. Either there would be a cohort-sized gap someplace or the line would have to be thinned. Either way, it would weaken the front.

Century replacement

Replacement century by century has some merit. It is certainly easier to imagine a small group of men coordinating their replacement by another small group of men. If the stand-off position were on the order of 60 feet there would be adequate separation between the opposing armies for the maneuver to be safely carried out. The process of moving 60 or 80 men out of the line and getting a like number into their place is far easier to organize than it would be for the 400 some men of a cohort.

The three illustrations below show how the three red centuries of the second line could move up to replace the three blue centuries of the first line.

The third frame, bottom left, shows one blue century elongated. This was done to indicate that the men of the century are in a more open formation as they run toward the rear to make space for the red century to move into their position.

The only major difficulty is that there are not enough centuries in the second line to replace those of the first. Either some don't get to have replacements or the model is back to the options of allowing gaps or having thinner lines.

What's left

Individual replacement

When everything else is eliminated there remains the possibility of individual replacement.

Following Goldsworthy, the second line routinely supports the first. At Pharsalus the men of both lines had been in the fighting and were tired by the time the third was committed. Could the first two lines have rotated men to the front on an individual basis? That is, without ever replacing an entire century at once, could fresh men have first been moved forward first within the century and then from the ranks of a supporting century from the second line?

The starting point for this is to visualize how the men from the rear ranks could move forward to take part in the fighting. The next set of illustrations tries to show a possible method.

This is a small century, one of those modeled for Caesar's depleted army. It is used here because it is smaller and therefore easier to get on the screen.

Prior to this illustration the armies charged each other, the first two ranks threw their pila, the armies clashed in hand-to-hand sword fighting. After a few minutes the two sides fell back from each other. That brought the scene to this illustration, the first lull in the fighting.

The first two ranks have swords in their hands, the last three ranks still have their pila.

The enemy line is some distance away; they will be shown in a later illustration. In the model a distance of about 60 feet was proposed. In these illustrations the separation will be reduced to 45 feet to illustrate an alternative to the 60 foot hypothesis.

Behind the 5 ranks of the century are the optio, signifer and tubicen.

In this illustration three men from the third rank are moving forward through the files. The sword fighters in the first two ranks simply step aside a little to make room. Since the enemy line is 45 feet away the momentary disruption of its lines causes no danger to the century. It would take only a matter of a minute or two for all of the men in the third rank to slip forward in this fashion.

This illustration shows the third rank now moved forward to take its place as the first rank. The old first and second ranks have not stepped back, the new first rank has taken up a position in front of where the first rank was. This was done to show that neither of the first two ranks would have to move backward for this to work. There is sufficient distance between the armies for the 3rd rank to simply take up a new, closer position.

The space where the third rank was is shown empty in this illustration just to make the movements clear. In fact the 4th and 5th ranks could have closed up.

The centurion has kept his position in the first rank, the soldier behind him is not shown moving forward. This was done simply to keep the centurion in his accustomed command position.

This illustration shows various aspects of how the century might initiate an attack against the enemy line. The drawing is complicated and requires some explanation.

The enemy is shown at the top.

At the bottom the first rank is shown in five different phases of the attack. The first figure is in the starting position -- the same place the new first rank was in the previous illustration. The second figure is just starting a short run with his pilum leveled. The third figure is in the act of throwing his pilum. The fourth figure has continued to run and is just drawing his sword. The fifth figure has his sword and shield ready to engage the enemy.

The 4th and 5th figures are taken from the original time-distance studies done for the Roman Army web page which estimated the time it would take to draw a sword after throwing the pilum and the the distance a running man would cover during that time.

The green lines to the side of the illustration are at the distances a running man would cover in one second.

Distances to the enemy line are given for each of the 5 positions.

The first scenario is unworkable if the lines are only 45 feet apart. There is not enough time and space for the soldier to run, throw, draw and get ready for impact -- especially not if the enemy charges in response.

Therefore another sequence is proposed. The sixth figure starts the sequence over again. The 6th and 7th figures are the same as the 2nd and 3rd in the first sequence. But after throwing the soldier does not continue to run but comes virtually to a stop, draws his sword, and then starts to run again. The runner is now 27 feet from the enemy line. Even if the enemy had started to charge there would be time for the soldier to get his sword out.

Finally, in the second file from the left, the illustration shows the swordsmen from the previous first and second ranks following in the charge.

The centurion is not shown participating in the charge though he probably would have.

After a brief period of sword fighting the two sides would fall back away from each other again and the cycle could repeat itself with another pilum-armed rank moving to the front.

The illustration shows only one rank of men involved in the attack. It seems quite feasible for two ranks to come forward and launch the attack. The trade-off would be that the supply of pila for subsequent attacks would be reduced.

This method involves minimal movement within the century. One file at a time slips forward between the men in front. There is little or no shifting of men back to the rear of the formation, this occurs naturally as the rotations continue. The sword fighters are kept at the front, they being the only ones with suitable weapons for close-in fighting.

Second line replacements

After all of the men in the century had rotated to the front, thrown their pila, and taken a turn at hand-to-hand fighting replacements from the second line could be filtered to the front in exactly the same fashion. The only difference would be that there are fewer centuries in the second line. Those centuries may send ranks to reinforce several first line centuries. The centurion could simply call out the instructions for where to go to each rank in turn.

The third line

The third line could send men through both the first and second lines in much the same way.

The alternative would be to invent a mechanism in which the centuries of both the first and second lines could be moved out of the way wholesale to allow the centuries of the third access.

At Pharsalus Caesar reduced the number of cohorts in the third line to create the fourth line of 6 cohorts. Yet he was still able to send this half-strength line into the battle and claim that the fresh men helped turn the tide. For those legions with the reduced third line that would mean two cohorts in the third line replacing seven in the first two lines. It is most unlikely that those 7 cohorts just withdrew from the line and let the two take their place. Instead, there are two possible alternatives.

The first is that the two cohorts sent fresh men in all along the line. The front would have been 4 cohorts wide. The two cohorts could have sent 2 or 3 ranks of fresh men to each of the cohorts. These two or three ranks would have been more than enough to mount a fresh charge along the lines described above.

The second is that the third line sent entire centuries into the line at the trouble spots. Since the original line was only 4 cohorts wide and the third line is 2 cohorts wide this means that a fresh century could have replaced every other one on the front line.

Lulls, some pros and cons

Sabin's notion of lulls in the battle is supported by some and challenged by others. I have made it a part of the model because it solves several problems that no other model proposed has.

The pesky pila

If only two ranks can throw their pila during the initial charge then there are a lot of pila to account for. In the century formation used above there are 5 ranks per century, 3 lines of centuries -- 15 ranks altogether. Only two have used their pila. The Romans did not arm 13 ranks with a weapon for which they had no use. Whatever model is proposed should describe just how those 13 ranks found a way to use their pila.

If there is continuous close-in fighting then it is difficult to explain how those pila get used. They are almost useless to the men in the front rank. The pilum is not a thrusting weapon and cannot be thrown with full effect from a distance of two feet. It cannot effectively be lofted over the heads of the first two or three ranks to fall among the enemy ranks because, to be lofted overhead and fall back at a distance of only a few yards it cannot be thrown with any velocity. And there would not be either enough time for the missile to rotate to a point-down orientation, it would simply fall flat. For full effect the pilum has to be thrown hard, not lofted.

The scenario suggested in the models does at least give a plausible way for those 13 pila to be used. The lulls provide the opportunity. If the armies fall back to a psychologically safe zone of 30 to 60 feet there is enough space for an effective throw. If the ranks can rotate to the front then each local charge can begin with a fresh volley of pila.

The sources do not describe any such action. But neither do the sources tell us how those other pila were used.


A small piece of the battle scene would be the need for water for the soldiers. Most ancient battles were summer affairs and Roman wars tended to be in the Mediterranean sun. Add hot metal helmets and armor. The soldiers would be subjected to several hours of pre-battle maneuvers, sometimes to a lengthy wait in ranks while the other army stalls, and then to anything from one to 4 hours of fighting. Even if each man carried the equivalent of a canteen, about a quart, a half-liter, of water into battle, it seems very likely that there would be a need for additional water during a long battle. Lulls and stand-offs provide an opportunity to get water to even the front lines.

If fighting were continuous and close-in it is hard to imagine how a man with a sword in one hand and a shield in the other could take so much as a sip of water. In that situation even the soldier in the last rank would be taking a risk to sheath his sword, perhaps even set his scutum down so that he could get at his water flask.

If the standards were kept at the back of the century it they could have served as a marker for water delivery.

Rest -- physical and psychological

Fighting of any sort is exhausting, both physically and mentally. If the fighting were continuous and close-in then the men in the first line would be under extreme stress for the duration of the battle. Even those in the 5th rank would be so close to the enemy that their stress levels would not be relieved. If there were lulls in the fighting but the two sides stayed within a few feet of each other there would be the opportunity for some physical rest but very little. With the enemy only feet away each man would have to maintain a constant heightened sense of alertness, watching the enemy carefully for signs of attack, keeping both sword and shield at the ready at all times.

Lulls and distance would provide a safety zone within which the individuals could relax their guard at least momentarily.


The rank rotation scheme proposed above is intended to suggest a possible way it might have been done. It is offered as food for thought. It does serve as a vehicle for highlighting some of the considerations that any complete model of Roman fighting should address. It does not solve all of the problems. For example, it presumes that the enemy is willing to have lulls and stand-offs as well. But, one must ask, were the German tribal armies, for example, at that level of sophistication or did they just keep pouring on the men until the battle was over? If that were the case then the model offered above would not work. The new model should, however, still address the basic questions -- how the rear ranks got to the front, how the second and third lines got into the fight, how all those rear-rank pila were used, even how men managed to get a drink of water during the battle.

A complete model would go beyond the Roman side and address just how the enemy force solved its similar problems. It would consider how Roman tactics might have changed depending on what type of enemy it faced. Would Roman fighting have been different when facing another Roman army than when facing Celts, Germans, Greeks or Persians? Almost certainly.

Next Page: Line Relief by Centuries

Site Map

© 2003, Gary Brueggeman. All rights reserved world wide. No part of this work may be reproduced in part or whole, in any form or by any means, without permission from the author.