Sherman's march through Georgia, 1864. The Confederates knew roughly where he was going. What they could not know was which side he would come from.
he divided his army into two wings—and if the rebels retreated from one wing they found themselves facing the other.1
That is the whole shape. Not a trap that closes, and not a surprise. A field arranged so that every direction of retreat is also a direction of advance for somebody.
Greene's Image for the law makes the point physical:
The bull backs you into the corner with its horns—not a single horn, which you might be able to escape, but a pair of horns that trap you within their hold. Run right or run left—either way you move into their piercing ends and are gored.2
The Image is doing real analytical work and it is easy to skim.
A single horn — one army, one line of attack — is a threat you move away from, and moving away is a successful response. The threat generates the solution.
Two horns remove the solution without removing the threat. The target's own evasive instinct, which is functioning correctly, delivers them into the second horn. The mechanism is not overwhelming force; it is making the target's competence work against them.
That distinguishes this from most of the chapter. Kissinger's targets are choosing badly because the menu is short. Sherman's are choosing badly because both available responses are correct responses to the horn they can see.
Greene's second case is where the technique is most usable, and it is more precise than the military one:
This is a classic trial lawyer's technique: The lawyer leads the witnesses to decide between two possible explanations of an event, both of which poke a hole in their story. They have to answer the lawyer's questions, but whatever they say they hurt themselves.1
Three conditions are doing the work here and none is optional.
The two explanations must be exhaustive. If a third exists, the witness takes it. Constructing a genuinely closed pair is the actual skill, and it is why this is difficult rather than merely clever.
Both must be damaging in different ways — otherwise the witness picks the cheaper one and absorbs it.
The witness must be obliged to answer. "They have to answer the lawyer's questions." This is a structural feature of a courtroom, not of conversation, and it is the condition that most often fails outside one.
That third point limits the technique severely. In ordinary life the equivalent of I don't know or a change of subject is almost always available, and a dilemma that can be declined is not a dilemma.
Greene names the constraint and gives it one line:
The key to this move is to strike quickly: Deny the victim the time to think of an escape. As they wriggle between the horns of the dilemma, they dig their own grave.1
Deny the victim the time to think of an escape concedes that an escape usually exists.
Which reframes the whole technique. It is not that the field is genuinely closed — Sherman's Confederates could in principle have gone somewhere neither wing was, and a witness given a week and a lawyer usually finds a third explanation. The dilemma is closed relative to the time available, and the time available is the thing the operator controls.
That makes tempo the real instrument and the pair of options a way of consuming it. Every second the target spends evaluating horn A is a second not spent looking for the exit, and the two-option structure is unusually good at absorbing attention because it presents as a complete problem.
And it explains the failure mode precisely: a dilemma that survives overnight is not a dilemma. The witness who is allowed to go home comes back with the third explanation. The army given a week finds the gap.
⚠ This is the third law running in which the operative variable turns out to be tempo rather than the stated mechanism — Ivan's finger at Law 28 works because "the room was filled with palace guards", and the interval is Houdini's instrument at Law 30. Greene's chapters name impressions and his cases run on timing.
Clean placement, and it is the most benign entry in the list.
Riskless. Both wings are Sherman's. Both explanations are the lawyer's. No branch produces a loss for the operator — which puts it with Kissinger, Rockefeller and Vollard rather than with the three risky ones.
Visible. The witness generally knows exactly what is being done to them, and it does not help. Like Rockefeller's rails, the technique does not depend on the target being fooled — a fact Greene states for Rockefeller and not here, though it applies equally.
Non-permanent. It does not change what the target is, only what they do in the next few minutes. Unlike Force the Resister or Brothers in Crime, it is repeatable and leaves no residue beyond the ordinary resentment of having been outmanoeuvred.
Riskless, visible, and repeatable is the best combination available in the chapter, and Greene lists it last with no comment.
You need an answer from someone who has been avoiding giving one.
Do not ask an open question — you will get the avoidance you have already been getting.
Build the pair. Two accounts of the situation, both of which you can live with, both of which commit them to something. Either the deadline moved and nobody told me, or it didn't and we're behind. Which is it?
Then check the three conditions honestly, because a badly built dilemma is worse than an open question — it announces that you were trying.
Is there a third answer? Nine times out of ten there is, and it is usually it's more complicated than that, which is often true. If a legitimate third account exists, you are not entitled to this move.
Are both branches genuinely costly to them, in different ways? If one is obviously cheaper, you have just told them which to pick.
Can they decline? Outside a courtroom, usually yes. Which means the pair has to be constructed so that not answering is itself the third damaging option — visibly evasive, in front of someone whose opinion they care about.
And accept the honest limit: if they get to sleep on it, you have not asked a dilemma, you have asked a question. The technique lives entirely inside the conversation it is used in.
Strongest evidence. Two cases from unrelated domains — a documented military campaign and a standard courtroom practice — with a shared structure and a stated constraint. The military case is verifiable; the lawyer case is a real and recognisable technique rather than an anecdote.
Tension — the tempo constraint concedes the dilemma is not closed. "Deny the victim the time to think of an escape." Documented above.
Tension — the courtroom condition does not generalise. "They have to answer the lawyer's questions" is a fact about subpoenas, and Greene offers the technique to readers who have none.
Tension — riskless, visible and repeatable, and listed last without comment. On the chapter's own implied axes this is the best-behaved entry in the set. See The Six Methods of Controlling Options.
🚩 [POPULAR SOURCE] · 🚩 SECONDARY WITHOUT PRIMARY — no source for the Sherman material, no dates within 1864, no named engagement, and no citation for the trial-lawyer claim. The March to the Sea is well documented; Greene's two-wings reading of it is asserted.
Open questions. Did Sherman divide his forces for this reason, or is the dilemma a retrospective reading of a decision made for supply and foraging reasons? The historical case for the second is at least as strong, and Greene supplies no evidence for intent.
Against the rest of Law 31, this is the entry with the fewest moving parts and it exposes something about the others.
Every other method requires a model of the target — Kissinger needs an insecure master, Erickson a reflexive resister, de Retz a frightened duke, Vollard a chronic ditherer. This one needs almost nothing. A dilemma works on the competent, the suspicious and the hostile, because it does not run on a psychological weakness at all; it runs on the geometry of the available responses.
That is why it is the one entry that could be described as a rule rather than a trick.
Against Law 29, the tempo finding connects directly. Knowing When to Stop and this page both turn out to be about the interval the operator controls, and the Cluster H finding — Greene's chapters name impressions and his cases run on timing — now has a third instance.
And within the vault's Sherman-adjacent material, Strategy 12: The Grand Strategy gives the war book's version, which is more careful: it treats the divided advance as a way of keeping options open for yourself rather than closing them for the enemy. Same manoeuvre, and the two books read it from opposite ends.
Rhetoric and questioning — the closed pair as an interrogation instrument. Trapping a Liar works on the principle that a deceptive account is under-specified, and that the way to expose it is not confrontation but forcing commitment to specifics — the liar has not built the parts they were not asked about.
The horns are that principle formalised. Both explanations are specific, and the witness has to inhabit one.
The insight neither produces alone: the dilemma is not primarily about the two branches being damaging. It is about making under-specification impossible, which is why it works on truthful witnesses too and why lawyers use it on both. A true account survives being pinned to specifics; a fabricated one starts generating contradictions the moment it has to have details. So the technique's real output is not a bad answer — it is converting a vague position into a falsifiable one, and Greene's framing (whatever they say they hurt themselves) describes only the case where the position was weak to begin with. That distinction is what separates the courtroom use, which is a truth-finding instrument, from the manipulative use, which requires the operator to already know the answer.
Military theory — the field arranged so competence fails. The Culminating Point of Victory gives Clausewitz's insistence that the decisive variable in a campaign is usually a structural property of the situation — supply, distance, morale curves — rather than the quality of the decisions made inside it.
The two wings are that claim in miniature: the Confederate commanders are not shown making errors. They retreat from a threat, correctly, and the retreat is the mistake because of how the field was built.
What the pairing produces: most of Greene's chapter locates failure in the target's psychology — insecurity, defiance, indecision, greed. This entry locates it in geometry, and that is a categorically stronger position for an operator, because psychology varies by person and geometry does not. Which suggests a hierarchy the chapter never states: arrange the field if you can, arrange the menu if you cannot, and work on the person only as a last resort. Rockefeller and Sherman are at the top of that hierarchy, Kissinger and Vollard in the middle, and de Retz and Stavisky at the bottom — and the chapter lists them in none of that order.
Sharpest implication. A single threat generates its own solution — you move away from it. A pair removes the solution while leaving the threat, so the target's correct evasive instinct is what delivers them. But Greene's own constraint concedes the field is not really closed: deny the victim the time to think of an escape. The dilemma is closed relative to the available time, which makes tempo the actual instrument and the two options a device for consuming attention — and a dilemma that survives overnight was never one.
Generative questions.