Rout
A Socratic walk-through of the rout — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why does a unit break and run after losing only a small fraction of its strength?
A formation that has lost a modest fraction of its number is, by any arithmetic of strength, still overwhelmingly intact. Most of its people are unhurt, most of its weapons work, and the enemy has not physically removed its ability to fight.
And then it dissolves. Not withdraws in order — dissolves, individually, at speed, discarding equipment, in a way that is very hard to stop and often costlier than the fighting was. Something has collapsed that is not the unit's material capacity, and the arithmetic of losses plainly does not capture it.
Reasoning it through
REASONING #Start by asking what is holding an individual in place at all. Standing in danger is not the default behaviour of a person; it is sustained by a structure. Several things do that work at once: the belief that the line will hold, the presence of specific people either side who would see you leave, the expectation that orders will keep arriving, and the calculation — usually not conscious — that staying is safer than the alternatives.
That last one is the load-bearing element and it is easy to miss. In close combat, a formation that holds together is genuinely safer per person than the same people running: fleeing turns your back, breaks mutual support, and historically was when most casualties in a defeat occurred. So staying is not merely brave; under the assumption that everyone else stays, it is correct.
Now notice the structure of that reasoning. Under the assumption that everyone else stays. If the formation is going to dissolve anyway, then being among the last to leave is the worst possible position. So each individual's best action depends entirely on what they expect the others to do — which means there are two self-consistent outcomes: everyone stays and staying is right, or everyone runs and running is right. The unit is sitting at one of two stable states, not on a sliding scale of morale.
That immediately explains the suddenness. A system with two stable states does not move gradually between them. It sits in one until something pushes it past a tipping point, and then it goes to the other quickly, because each person leaving makes leaving more attractive for those remaining. The transition is fast for the same reason a queue collapses into a crowd.
And it explains why the trigger is so often trivial relative to the losses. What matters is not the size of a shock but whether it changes what people expect others to believe. A visible event that everyone can see everyone else has seen is far more destabilising than a larger loss nobody witnesses — the flank being turned, the commander falling, a cry taken up, the unit next door starting to move. These are shared signals, and they update everybody's estimate of everybody's estimate at once. A quiet twenty per cent loss can leave a unit steady; a visible small one at the wrong moment can break it.
Now the countermeasures follow, recognisable across very different armies and eras, and nearly all work on expectations and observability rather than capacity. Keeping units together so people fight beside those they know. Standards, colours, music — shared visible signals that the formation still exists. Officers where they can be seen. Reserves committed at the moment of wavering, whose value is partly the visible fact of their arrival. Depth, so those in front cannot see an empty space behind. None adds firepower; all stabilise the belief that everyone else is staying.
The analogy
THE ANALOGY #Think of a crowded footbridge where everyone is walking calmly. Each person is safe because everyone else is moving steadily, and each keeps their pace partly because the others do.
Someone stumbles at the front and a few people surge. Nobody has been hurt and the bridge is sound. But now the calculation has changed for everyone: if a crush is coming, you want to be out before it, and the worst place is at the back of one that has started. The stumble did not remove anyone's ability to walk calmly. It removed the reason to believe everyone else would.
A crowd's panic has no adversary, whereas a rout has an enemy actively trying to produce it and to exploit it once begun — so the military case has a party working to trigger the transition and to make the flight as costly as possible, which is why the losses of a rout are concentrated after the collapse rather than before.
Clarifying the model
THE MODEL #This is not cowardice, and framing it that way both misdescribes it and predicts badly. Units that break are frequently the same units that fought well minutes earlier, and the individuals are largely the same people. A model based on the character of the soldiers cannot explain why a unit's behaviour changes discontinuously without its composition changing at all, nor why the countermeasures that work are structural rather than exhortative. The threshold belongs to the formation, not to the person.
"Losses cause breaking" gets the causation roughly backwards. Losses matter mostly as signals — as evidence about how the fight is going and about what others will do — rather than as attrition of capacity. This is why the same casualty figure has such different effects depending on whether it was visible, whether it fell on the flank, and whether the unit could see reserves behind it. Any account keying on a percentage will fit poorly.
The literatures here are contested, and some famous claims are weak. The mid-century work arguing that only a minority of soldiers fired their weapons is disputed on methodological grounds and should not be leaned on. That primary-group cohesion matters more than ideological commitment is better supported but has also been qualified. I am describing a structure that fits the observed pattern, not claiming the empirical weightings are settled.
Withdrawal is not rout, and the distinction is the practical one. An ordered withdrawal keeps the structure intact and can be resumed from; a rout destroys it, which is why rallying a broken unit takes it out of the fight for far longer than its losses would suggest. The whole art of the countermeasures is to keep a unit that must give ground on the ordered side of the line.
The falsification test. If the mechanism is a coordination threshold rather than accumulated attrition, then units at identical loss levels should break at very different rates depending on whether the losses were visible and whether the unit can see support behind it — and measures that add no firepower but increase mutual observability should measurably reduce breaking. If breaking tracked casualty percentage regardless of visibility, formation depth or unit cohesion, the account here would be wrong and simple attrition would explain it.
A picture of it
THE PICTURE #How to readRead the boxes as states of the formation, not of any individual. The two self-loops are the argument: the loop on Holding absorbs losses without moving, while the loop on Breaking is self-accelerating, and both are stable in their own way. Wavering is the only state with a genuine choice of exits, which is why every countermeasure in the piece is aimed at that box. Note how far it is from Routed back to Holding — the return path runs through rallying and reconstitution, which is why a rout costs so much more than its casualties.
What became clearer
WHAT CLEARED #A unit does not break because it has run out of capacity. It breaks because holding is a coordinated act that only makes sense while everyone expects it to hold, so the formation has two stable states and a tipping point between them. That is why collapse is sudden rather than gradual, why the trigger is often small but always visible, and why the practices armies use to prevent it — keeping people beside those they know, showing the standard, committing a reserve where it can be seen — work on what people believe about each other rather than on what the unit can do.