A nineteenth-century French officer named Ardant du Picq went back through the accounts of antiquity's most famous battles and found something that could not be explained by fighting.
At Cannae, and at Pharsalus, the losses on the two sides were fantastically disproportionate — a few hundred among the victors, thousands upon thousands among the defeated.1
That ratio is not producible by combat. Two armies of roughly comparable equipment, hacking at each other with edged weapons, do not generate a twenty-to-one casualty differential. Something else was killing people.
Du Picq's answer is the mechanism, and it has three steps.
Through maneuver, the eventually victorious army surprised its enemy and splintered its lines into parts.1
Then:
Seeing their ranks breaking up, losing their sense of solidarity and support, and feeling isolated, soldiers panicked, dropped their weapons, and fled1
And then the arithmetic:
a soldier who turned his back on the enemy was an easy soldier to kill. Thousands were slaughtered this way.1
Greene's conclusion is the sentence the whole thing rests on:
These great victories, then, were essentially psychological.1
Cannae is the proof case. Hannibal was vastly outnumbered. He did not kill fifty thousand Romans in a fight — by making them feel vulnerable and isolated, he made them overreact and retreat in confusion: easy pickings.1
The killing happened after the battle was psychologically over.
The reason a formation matters becomes obvious once you notice what it is doing.
Ancient warfare was hand-to-hand: at every moment, a man was exposed to death from behind and from both sides. Commanders learned early to form soldiers into tight, cohesive ranks — because trusting the men on either side of him not to retreat and leave him exposed, a soldier could fight the man in front of him with more spirit and confidence.1
The line is not primarily a tactical arrangement. It is a device for making a terrifying situation survivable to think inside.
The Romans understood this well enough to engineer it deliberately. They placed the youngest and most impetuous fighters in the front ranks, the most experienced and best in the rear, and everyone else in the centre.
The effect: the weakest soldiers — the ones most prone to panic — were surrounded by the braver and steadier, giving them a powerful sense of security.1
No army went into battle with more cohesion and trust than the Roman legions.1
Note what that arrangement is not. It is not putting your best soldiers where the fighting is hardest. It is positioning your steady people so that they are visible to the ones most likely to break. The formation is optimised for the psychological state of its weakest members, because that is where a collapse starts.
Greene's extension of du Picq to modern life rests on a continuity claim.
Our nature has not changed. Lurking deep in even the most civilized among us is the same basic fear of being alone, unsupported, and exposed to danger.1
And an aggravating factor:
People today are more dispersed and society is less cohesive than ever before, but that only increases our need to belong to a group, to have a strong network of allies—to feel supported and protected on all sides.1
From which:
Take away this feeling and we are returned to that primitive sensation of terror at our own vulnerability. . . . cut people off from their group—make them feel alienated, alone, and unprotected—and you weaken them enormously.1
Greene notes that many commanders turned the terror into an explicit strategy. Genghis Khan was the master — using Mongol cavalry mobility to cut communications and isolate parts of enemy armies so that they felt alone and unprotected. He worked consciously to instil terror.1 Napoleon and Mao's guerrilla forces used the same principle.
The finding has a defensive reading that Greene mostly leaves implicit, and it is the more useful one.
If what breaks an army is the perception that support has gone, then the thing to protect is not strength but the perception. A unit that has taken heavy losses and can still see its flanks holds. A unit at full strength that believes it is alone runs.
Which means the actual vulnerability in any group is not its weakest member's capability. It is that member's visibility of support — and that is a much cheaper thing to attack and a much cheaper thing to defend.
You are part of something under pressure — a team taking criticism, a project being questioned, a group in a fight it might lose.
Watch for the moment the perception of support goes, because that moment is not the same as the moment things actually get worse, and it will arrive first.
The tells are specific and small. Someone stops being copied on things. A meeting happens without a person who would normally be in it. The public defence that used to be automatic becomes a silence. Nobody has withdrawn support; the visibility of support has thinned, and du Picq's finding is that the visibility is what people are actually holding onto.
If you lead the group, spend disproportionate effort on the members most likely to break rather than on the strongest. That is the Roman ordering, and it is counterintuitive because it feels like reinforcing weakness. It isn't — a collapse starts at the least secure point and then propagates, so the least secure point is where reinforcement has the highest return. Concretely: make backing visible rather than merely real. Say it in the room. Copy them on the thing. Be seen to be adjacent.
If you're the one under pressure, understand what you're feeling. The sensation of being exposed is not information about your position. It is information about your view of your position, and the two come apart precisely when someone is working on you. Before acting on it, count actual allies rather than felt ones — a list of names, written down. The gap between those two numbers is usually where the panic lives.
And if you're on the other side, know what you're doing. Cutting communications, isolating one part from the rest, making someone feel unsupported — these are effective for the reasons du Picq identified, and they work on the same substrate whether the target is an army or a colleague.
Du Picq's Battle Studies is a genuinely influential text and the observation about disproportionate ancient casualties is his own, well known, and taken seriously in military history. The Roman ordering of ranks by experience is attested. [POPULAR SOURCE] for Greene's framing.
The ancient casualty figures the argument rests on are unreliable. Du Picq was reasoning from Livy, Caesar and Polybius, whose numbers are systematically doubted by modern scholarship — victors' figures for their own losses are routinely understated and enemy losses inflated. The mechanism du Picq proposes may still be correct; but the evidence for it is precisely the kind of data least likely to survive scrutiny, and the argument is presented as though the ratios were measurements.
And the causal direction is assumed rather than shown. Du Picq's reading is: lines break → isolation felt → panic → flight → slaughter. An equally available reading runs: an army already losing, badly commanded or outmaneuvered, has its lines broken because it was losing, and the flight and slaughter follow from tactical defeat rather than from a psychological cascade. The two are hard to separate from a casualty count. What makes du Picq's version persuasive is not the numbers but the corroborating detail — the Roman rank-ordering shows that commanders at the time were already designing around exactly this fear, which is better evidence than the arithmetic.
Greene's evolutionary framing is decoration and should be treated as such. The claim that early humans formed groups out of terror at their own exposure, and that this is the origin of the modern fear of isolation, is a story with no evidence attached. The battlefield mechanism does not need it, and it is the kind of just-so reasoning that this vault flags elsewhere. 🚩
Against Cannae: The Perfect Envelopment, this page supplies the mechanism that page describes the shape of. The double envelopment is the maneuver; du Picq explains why it produced a massacre rather than a defeat. The two readings are complementary and neither states the other — the tactical account explains how the Romans came to be surrounded, and this one explains why being surrounded killed fifty thousand of them.
Against Strategy 19: The Annihilation Strategy, the same mechanism appears again under a different name. Envelopment and division are two ways of producing one psychological state — the conviction that there is no support and no way out. Greene teaches them as separate strategies two chapters apart and never notices they share a single substrate. The unified claim available from both: what destroys people is the removal of perceived options, and the physical arrangement is only the delivery mechanism.
Against Strategy 7: Morale Strategies this is the same subject from the defensive side. That chapter is about building the thing this one is about removing — and the Roman rank-ordering belongs to both, since it is a morale technique implemented as a seating plan.
Behavioral Mechanics → Isolation as the Universal Precondition. Every coercive influence environment — interrogation, high-control groups, an abusive relationship, a hostile workplace — converges on the same first move, which is not persuasion but separation. The specific arguments vary wildly. The isolation does not.
The insight the pairing produces: because isolation is structural rather than rhetorical, it is the one element of coercion that can be detected from outside the compromised system. You cannot reliably evaluate an argument being made to you while you are inside the environment designed to make it persuasive — that assessment uses the faculties under attack. But you can count independent connections, because that is a question with an external answer. Du Picq's soldier does not need to judge whether the retreat is wise; he needs to know whether anyone is still on his flank. The practical translation is that the most reliable self-check available in any high-pressure situation is not "is this reasoning sound" but "how many people who would tell me something inconvenient have I spoken to this month," and the answer to the second is far harder to manipulate than the answer to the first.
Business → The Contagion Point in a Failing Team. Organisations lose teams the way armies lose lines: not gradually, and not from the middle. One person visibly gives up or leaves, and the departure carries information — it says the situation has been assessed by somebody with the same view as you, and found unsurvivable.
The insight neither field gives alone: the first defection is worth vastly more than its headcount, and the value is informational rather than operational. This inverts standard retention thinking, which prices people by contribution and therefore worries most about the strongest performers. Du Picq's Romans put their best fighters where the panic-prone could see them, because the collapse starts at the least secure point regardless of who is most valuable. The corresponding management move is unfashionable and follows directly: in a team under strain, allocate your visible support to whoever is closest to breaking, not to whoever matters most — because the first person to run determines whether anyone else does.
Nobody killed fifty thousand Romans in a fight. They were killed after they turned around — and they turned around because their line broke and they could no longer see anyone beside them. Which means the thing that actually holds people together under pressure is not their strength, their position or even their odds, but a visible flank; and that a group at full strength which believes it is alone will collapse faster than a battered one that can still see its neighbours.