Duke Wu of Cheng decides it is time to take over the increasingly powerful kingdom of Hu.
He tells nobody. Then he marries his daughter to Hu's ruler.1
Some time later he calls a council and asks his ministers a question:
"I am considering a military campaign. Which country should we invade?"1
One minister gives the obvious and correct answer: Hu.
The duke seemed angry, and said, "Hu is a sister state now. Why do you suggest invading her?" He had the minister executed for his impolitic remark.1
The ruler of Hu hears about it. Considering that, and the marriage, and "other tokens of Wu's honesty," he takes no precautions to defend himself.
A few weeks later, Cheng forces swept through Hu and took the country, never to relinquish it.1
Look at what the duke actually needed and how he got it.
He needed a demonstration of sincerity that would reach the ruler of Hu indirectly — because anything sent directly would be evaluated as a message, and a message from a man planning to invade you is exactly what a ruler is watching for.
So he manufactured an internal event. A council, a question, an answer, a punishment. All of it inside Cheng, none of it addressed to Hu.
And the news travelled on its own, as gossip about somebody else's court, which is the most credible channel there is. Hu's ruler did not receive a reassurance; he overheard one.
The minister's death is the message, and the minister is the postage.
This is why the case belongs beside Lustig rather than being folded into him.
Every other demonstration of sincerity in Law 12 is self-funded. Lustig risked his own $50,000 for sixty days. Gordon-Gordon paid real handwriting experts out of his own pocket. Borri turned down years of his own fees. Camillus surrendered an advantage that was his to surrender.
Duke Wu spent a minister — a man who answered a direct question correctly, in council, when asked.
That is a completely different cost structure, and it has a specific consequence: the technique is not available to most people. Lustig's version can be run by anyone with capital and nerve. Duke Wu's requires the power to destroy an innocent third party without consequence, which is a property of dukes.
Greene groups them under one law and grades them identically on effectiveness, which conceals the only distinction that reliably sorts this chapter: who bears the cost of the demonstration.
The minister's innocence is not incidental — it is what makes the proof work.
Consider the alternative: the duke punishes someone who had actually offended him. That is ordinary court discipline and it demonstrates nothing about his intentions toward Hu.
The demonstration requires the punishment to be visibly disproportionate, because disproportion is what signals conviction. A ruler who kills a man merely for suggesting an invasion has, apparently, a commitment to the alliance so strong that it overrides his own judgement and his own interest in good counsel.
So the strength of the signal is calibrated by the injustice of the act. A mild rebuke proves little. An execution proves a great deal, precisely because it costs the duke a competent adviser and makes him look unreasonable.
The technique consumes a real asset — the willingness of ministers to answer honestly — and that consumption is what buys the credibility. Every remaining minister in Cheng now knows what happens to accurate answers, which is a permanent tax the duke pays for one invasion.
He was not credulous. He was doing something reasonable, and that is the trap.
He had three pieces of evidence: a marriage alliance, an executed minister, and "other tokens of Wu's honesty." That is a pattern, not a single gesture, and it is exactly the kind of evidence Greene elsewhere tells the reader to trust — effects rather than stated intentions. Nobody had told Hu's ruler that Cheng was friendly; he had watched Cheng behave that way, at cost, repeatedly.
Costly signalling can be faked by anyone willing to pay the cost, and the duke was willing to pay it in someone else's life.
Which means the defence — look at costly actions rather than words — fails precisely against operators who can externalise the cost. Hu's ruler applied the right method and it selected for exactly the wrong kind of counterparty.
Written defensively, because the offensive version requires a power most readers do not have and should not want.
First, ask who paid for the demonstration you are evaluating. A costly signal is only informative about the signaller if the signaller bore the cost. If the price was paid by a third party, it tells you about their willingness to spend others, not about their intentions toward you.
Second, be suspicious of proofs you were not shown. Overheard evidence feels more reliable than presented evidence, which is exactly why it is worth manufacturing. Ask whether the event that reached you could have been staged for the channel it reached you through.
Third, treat disproportionate punishments as data about the punisher, not about the relationship. The signal Hu read was he is committed to us. The available reading was he kills people who answer questions. The second was true and more predictive.
Fourth, notice when a pattern is all recent. The marriage, the execution and the "other tokens" were assembled after the decision to invade. Age of evidence matters — a track record built entirely inside the window of someone's current interest in you is not a track record.
Fifth, if you are the minister: do not answer strategic questions in front of a ruler who has already decided. This is the Law 9 lesson arriving in a new form. The minister's answer was correct, useful, and exactly what he was asked for, and it killed him.
And the plain record: a competent official was executed to make a neighbouring king relax, and Greene reports it as a technique for building a reputation for honesty.
Strongest support. The mechanism is complete and economical — a stated intention, a manufactured internal event, an indirect channel, an observable outcome — and it isolates something no other case in the chapter does: sincerity demonstrated to a third party who is never addressed.
Tension — it is filed under selective honesty and contains none. The duke's council question was a lie, his anger was performed, and the marriage was a device. What is demonstrated is not honesty but commitment, which is a different signal with different mechanics.
Tension — the cost is externalised and Greene does not remark on it. The minister is described only as having made an "impolitic remark." Greene's narration adopts the court's framing of a man who answered correctly.
Tension — the technique taxes the duke permanently and the case does not follow through. A ruler who executes ministers for accurate counsel has degraded his own information supply. That cost lands after the anecdote ends.
🚩 SINGLE SOURCE · 🚩 SECONDARY WITHOUT PRIMARY — no dynasty, no dates, no text named. The story is a well-known one from the Chinese strategic tradition (it appears in Han Fei's writings, which Greene quotes elsewhere in this same chapter as an Authority) and he cites nothing here. [POPULAR SOURCE]
Open questions.
Greene builds The Arguer at Law 9 from an almost identical scene, and the two together produce something neither states.
At Law 9, an Athenian engineer contradicts a Roman consul about a battering ram and is flogged to death. Greene's reading: the engineer was a type, he impugned a superior's intelligence, and his error was arguing at all.
Here, a minister answers a question he was directly asked, correctly, and is executed. Greene's reading: the duke was clever.
Same event, and the analytical attention moves entirely depending on whose page it is. At Law 9 the dead man is the subject and is diagnosed at length; the consul is weather. At Law 12 the dead man is a mechanism and the killer is the subject.
The insight the pair produces: this corpus's unit of analysis is whoever is being taught, and the other party is always scenery — which means the same act generates opposite lessons depending on which chapter you are in. A reader who takes Law 9 seriously learns not to answer honestly upward. A reader who takes Law 12 seriously learns that honest subordinates are a consumable.
Both readings are in the book, neither is cross-referenced, and together they describe a court in which candour has been eliminated from both directions.
→ behavioral-mechanics · Lustig and Al Capone
Lustig returns $50,000 to Capone to the penny, apologises for failing, and is handed $5,000 unasked.
Both are demonstrations of sincerity engineered to disarm, and they are the chapter's two poles on the question of who pays.
Lustig's demonstration is self-funded and reversible: he risks his own capital, and if Capone had simply pocketed the money and thrown him in the river, the loss would have been his alone.
Duke Wu's is externally funded and irreversible: the cost lands on a man who is not party to the transaction and cannot be compensated afterwards.
The insight neither page produces alone: self-funded demonstrations are a technique and externally-funded ones are a privilege, and the distinction determines who can run the play at all. Greene's chapter reads as a set of transferable skills; roughly half its cases require the power to spend other people, which is not a skill.
That also explains the scale difference in what each buys. Lustig's own money bought him one disarmed gangster and $5,000. A minister's life bought a kingdom. The returns track the externality, which is an uncomfortable finding to leave standing and this chapter never states it.
→ psychology · Mi Tzu-hsia and the Half-Eaten Peach
At Law 5, the same two acts — taking the ruler's coach, offering him a half-eaten peach — are read first as devotion and later as insolence. Only the ruler's regard changed between the readings.
The minister's answer runs the same mechanism, compressed into a single moment.
"Hu should be invaded" is either excellent counsel or an impolitic remark, and which one it is was decided before he opened his mouth — by a plan he did not know existed.
The insight the pair produces: Mi Tzu-hsia's page shows meaning supplied retroactively by a ruler's changed affection. This shows it supplied in advance, by a ruler's undisclosed intention — and the second is worse, because there is no period during which the subordinate was safe and could have noticed the shift.
The minister was never in the situation he thought he was in. He was answering a question about strategy; he was actually delivering a line in a performance staged for an audience in another country. And that is the general hazard of serving someone with private plans: your conduct is being evaluated against a script you have not read, and correctness is no protection because correctness was never the criterion.
Sharpest implication. Hu's ruler did the right thing. He ignored words, watched costly actions, and read a pattern rather than a single gesture — which is precisely the epistemic discipline this book recommends everywhere else. Costly signalling only tells you about the signaller if the signaller bore the cost, and the duke paid in a third party's life. So the recommended defence selects specifically against operators willing to externalise, and the more rigorously Hu's ruler applied it, the more thoroughly he was taken.
Generative questions.