Greene opens the mechanism with weather, and the analogy is doing real work:
"Nothing is more terrifying than the sudden and unpredictable. That is why we are so frightened by earthquakes and tornadoes: We do not know when they will strike. After one has occurred, we wait in terror for the next one."1
The important half is the second sentence. The earthquake is over. The terror is not.
What frightens people is not the event but the interval afterwards — a period in which nothing is happening and no posture is safe, because the rule governing when it happens is unavailable. That is a state you can put someone in and leave them in, at no ongoing cost, and it is what the law is actually selling.
The JUDGMENT states it as an exchange: "Humans are creatures of habit with an insatiable need to see familiarity in other people's actions. Your predictability gives them a sense of control. Turn the tables: Be deliberately unpredictable. Behavior that seems to have no consistency or purpose will keep them off-balance, and they will wear themselves out trying to explain your moves."2
They will wear themselves out. That is the yield, and it is worth separating from the intimidation framing that surrounds it.
The target is not primarily frightened. The target is working — generating hypotheses, discarding them, generating more. And the labour is unbounded, because there is no correct answer to converge on. A person trying to model a pattern that has no pattern does not fail and stop; they fail and continue.
Greene supplies a philosophical warrant, and it is more revealing than he probably intends:
"Animals behave in set patterns, which is why we are able to hunt and kill them. Only man has the capacity to consciously alter his behavior, to improvise and overcome the weight of routine and habit. Yet most men do not realize this power. They prefer the comforts of routine, of giving in to the animal nature that has them repeating the same compulsive actions time and time again."3
So predictability is animality, and pattern-breaking is the distinctively human capacity — which most people decline to use.
Two things worth noting about this framing.
First, it does a lot of persuasive work cheaply. Once predictability has been classified as animal, the reader's ordinary reliability is not a virtue with tradeoffs; it is a failure to be human. Nothing argues for this; the categories do it.
Second, the reason Greene gives for why people stay predictable is the interesting part and he moves past it in a clause: "because they mistakenly believe that if they do not unsettle others, they will be left alone."4 That is a real belief, it is often correct, and the chapter dismisses it as an error without examining when it holds.
The chapter's purest illustration is not the famous one. It is Filippo Maria Visconti, last of the Visconti dukes of Milan, who "consciously did the opposite of what everyone expected of him."5
The sequence Greene describes is worth following in full, because it is a machine rather than an anecdote.
The duke showers a courtier with attention. The man comes to expect a promotion. The duke then treats him with the utmost disdain. Confused, the man prepares to leave court — at which point the duke recalls him and treats him well again. Doubly confused, the courtier reasons that his expectation of promotion must have become obvious and offensive, and so begins behaving as though he expects no honour. The duke then rebukes him for his lack of ambition and sends him away.6
Every available posture is punished. Expect promotion: punished. Stop expecting promotion: punished. Leave: recalled. Stay: dismissed.
And Greene gives the only surviving strategy: "Do not presume to know what he wants. Do not try to guess what will please him. Never inject your will; just surrender to his will. Then wait to see what happens."7
The output is not fear. It is the elimination of independent judgement, which is a more useful thing to own in a courtier than fear is. The duke ends the passage "unchallenged and at peace" — and the peace is the point. Nobody can plan against him because nobody can plan at all.
Greene makes a claim that separates this law from most of the corpus: "Unpredictability is most often the tactic of the master, but the underdog too can use it to great effect. If you find yourself outnumbered or cornered, throw in a series of unpredictable moves. Your enemies will be so confused that they will pull back or make a tactical blunder."8
This is unusual and correct. Most of the book's techniques require standing — Law 16's withdrawal, Law 17's own Picasso material, Law 11's dependency. This one does not, and the reason is structural.
Unpredictability's cost to the target is computational, and computation scales with the target's investment in planning rather than with the operator's resources. A large, well-organised opponent has more planning to invalidate, not less. That is why 4,600 men moving in circles in the Shenandoah could freeze a 90,000-man march on Richmond.9
The bigger the machine, the more expensive it is to make it re-plan. Which is the one asymmetry in the book that runs in the underdog's favour by default.
The chapter closes with a caution that is easy to skim and is the law's real boundary: "Unpredictability can work against you sometimes, especially if you are in a subordinate position. There are times when it is better to let people feel comfortable and settled around you than to disturb them. Too much unpredictability will be seen as a sign of indecisiveness, or even of some more serious psychic problem."10
Same behaviour, two readings, and the deciding variable is position.
From above, inexplicable conduct reads as a reason I cannot see. From below, it reads as no reason — which is unreliability, or illness.
That is the same precondition Picasso states in this chapter ("once you have attained a certain level of recognition") and Law 16's Reversal states about absence. Three separate instruments in two chapters, all conditional on a prior standing the reader is assumed to have, and Greene connects none of them.
There is also a cost the chapter never mentions. Everything the Visconti courtier could not do — plan, anticipate, use initiative — is also everything a useful subordinate does. The duke bought peace by destroying the capacity for independent action in the people around him, and then had a court full of people incapable of independent action.
Someone is managing you by being unreadable.
You will know it by a specific feeling: not fear, but the sensation of always having just been wrong about what they wanted. You prepared the thing they asked for and they wanted something else. You held back and they wanted initiative. You are not underperforming; you are failing to hit a target that moves.
The instinct is to try harder to read them. That is the Visconti courtier's move and it is exactly what the mechanism feeds on — every hypothesis you build is more sunk effort to be invalidated, and there is no correct model to arrive at.
So stop modelling. Ask for the criterion in writing, once, plainly, and then act on the criterion rather than on the person. Not as a confrontation — as a request for the standard. If a standard exists, you now have something stable to work against and the unpredictability loses its grip. If no standard is forthcoming after a genuine attempt, you have learned the thing worth knowing: the unreadability is not incidental to how this position works, it is how it works, and no amount of accuracy on your side will resolve it.
And run the check in the other direction, because it is the one with teeth. If people around you seem oddly tentative — waiting rather than proposing, checking rather than deciding — ask what they would have to believe about you to behave that way. The Visconti mechanism does not require intent. Being inconsistent by accident produces the same courtier, and it buys you the same peace, and costs you the same capacity for independent action in everyone you work with.
The strongest material is Visconti, because Greene describes a procedure with all four branches — expect, stop expecting, leave, stay — rather than an outcome. That structure is what makes it a mechanism rather than an anecdote.
Tension: the chapter's own flagship case is not about unpredictability. Spassky's diagnosis at Reykjavik — "a man who is willing to commit suicide has the initiative" — names a stakes asymmetry, not a pattern problem, and it explains the outcome better than the chapter's thesis does. Documented on its own page; not resolved here.
Tension: predictability is classified as animality, and the reason people stay predictable is dismissed in a clause. They mistakenly believe that if they do not unsettle others, they will be left alone — a belief that is frequently correct and never examined.
Tension: the law is conditional on standing and says so only at the end. The closing warning inverts the entire chapter for subordinates, which is most readers, and arrives after the technique has been recommended for four pages.
🚩 SINGLE SOURCE · 🚩 SECONDARY WITHOUT PRIMARY — no source for the Visconti material, which is reported as habitual practice ("he might suddenly shower a courtier with attention") rather than as a documented episode, with the courtier's internal reasoning narrated as fact. The Han-fei-tzu Authority and La Bruyère marginal carry names and centuries but no works. [POPULAR SOURCE]
Open questions. Does the mechanism require the target to believe there is a pattern to find — and if so, would a target who correctly concluded "this is simply arbitrary" be immune, or does the arbitrariness itself do the work? And: the Visconti court was at peace. What was the cost of that peace in the quality of the decisions his court produced, and does the chapter's success criterion (unchallenged) have any room for it?
🚩 The Authority endorses the chapter's goal by the opposite method, and this is the seventh instance in the build of a marginal or Authority quotation supplying something the body lacks.
Han-fei-tzu: "The enlightened ruler is so mysterious that he seems to dwell nowhere, so inexplicable that no one can seek him. He reposes in nonaction above, and his ministers tremble below."11
Every case in Greene's body text scrambles patterns actively — Fischer's deliberate blunders, Visconti's oscillations, Jackson's marches in circles, Picasso's withdrawal. Han-fei-tzu's ruler is unreadable through nonaction: he is not erratic, he is absent, and the ministers tremble at a vacancy rather than at a pattern.
Those are different instruments with different costs. Active scrambling requires continuous performance and manufactures the events it is read from. Nonaction requires nothing and cannot be over-read, because there is nothing to read. The Authority is describing Law 16's withdrawal and Greene has printed it as an endorsement of Law 17's disruption.
Within the chapter, the Reversal concedes that a stable pattern is itself an asset — "it sets up a smoke screen… it allows you on rare occasions to do something completely against the pattern." Which means predictability is not the animal condition the body text calls it; it is a stored resource that unpredictability spends. The chapter argues both and marks neither.
Psychology — The Paradox of Mental Control
That page's finding is the white-bear problem: suppressing a thought requires holding it in mind to check whether it has gone, so the monitoring process keeps the thought alive. The effort to stop thinking about something is what sustains it.
Visconti's courtier is in the same trap with a different content.
He is trying to stop guessing what the duke wants — Greene's own surviving instruction is "do not presume to know what he wants, do not try to guess what will please him." But not-guessing requires continuously checking whether you are guessing, which is a form of attending to the duke's preferences at all times.
What the pair produces: the instruction Greene gives as the survival strategy is unexecutable, and its unexecutability is why the mechanism holds. You cannot deliberately not-model someone you depend on; the attempt is itself a model with a monitoring process attached. Which explains the observed outcome better than fear does — the courtier is not terrified, he is occupied, permanently, by an operation he has been told to stop performing.
And it locates the real defence, which is not psychological at all. You cannot stop attending to an unpredictable superior by choosing to. You can only reduce your dependence on predicting them, which is a change in exposure rather than in attention.
Business — Consistent, Persistent, Predictable Actions — The Trust Builder
That page's claim is the exact inverse of this law's: trust is constructed by being consistent (same stance each time), persistent (over time), and predictable (they can forecast your next move). Predictability sounds like a negative and is a positive, because the other party no longer has to spend energy guessing.
Set the two side by side and they agree completely about the mechanism and disagree about which side of it you want to be on.
Both say: an unpredictable person imposes forecasting costs on everyone around them. The sales page treats that cost as the thing destroying the relationship. Greene treats the identical cost as the thing producing the control.
The insight neither generates alone: they are describing the same tax and differ only on who is paying and what is being bought. Which resolves the apparent contradiction and produces a usable rule. Forecasting costs are worth imposing when you want the other party occupied and compliant and do not need anything from their initiative — Visconti's court, which was peaceful and produced nothing. They are worth eliminating when you need the other party to act on your behalf without supervision, which requires them to be able to predict you.
So the choice is not between two theories of influence. It is a question about what you need from the person: compliance is bought with unpredictability and initiative is bought with reliability, and no one can supply both to the same counterparty.
Sharpest implication. Unpredictability does not primarily frighten people — it occupies them, permanently, with an unsolvable modelling problem, and the occupation consumes exactly the capacity for independent judgement that makes them useful. Visconti bought an unchallenged court by destroying its ability to act without him, and the chapter counts that as a win.
Generative questions.