Somebody does something that gives you pause. A flash of contempt at a waiter. A commitment quietly dropped. A moment of nastiness toward someone who could not answer back.
And then the sentence arrives, from them or from you: that wasn't like him. He'd had a terrible week.
It may be entirely true. And Greene's corollary says it does not matter:
People never do something just once. They might try to excuse themselves, to say they lost their heads in the moment, but you can be sure they will repeat whatever foolishness they did on another occasion, compelled by their character and habits.1
Then the clause that makes it a hard claim rather than a cynical one: in fact, they will often repeat actions when it is completely against their self-interest, revealing the compulsive nature of their weaknesses.1
That is the load-bearing detail. If people only repeated behaviour that paid, this would be a page about incentives. The evidence for compulsion is repetition at a loss — and that is what makes the corollary predictive rather than merely observational.
Greene's instruction for how to use it is a four-step procedure, and its order matters.
The most significant indicator of people's character comes through their actions over time. Despite what people say about the lessons they have learned and how they have changed, you will inevitably notice the same actions and decisions repeating in the course of their life.2
Step one: notice a salient behaviour. His examples are specific — disappearing when there is too much stress, not completing an important piece of work, turning suddenly belligerent when challenged, or, conversely, suddenly rising to the occasion when given responsibility.2 Note that the last is positive; this is not only a threat-detection instrument.
Step two: research the past with that behaviour fixed in mind.
Step three: re-read what you have already observed. Greene's phrasing is careful — you look at other actions you have observed now in retrospect.2 The data was always there; what was missing was the category that makes it legible as a set.
Step four: watch the present, and see the actions not as isolated incidents but as parts of a compulsive pattern.
And then a sentence that is unusually blunt about where responsibility sits: if you ignore the pattern it is your own fault.2
The Roman case is Greene's proof of the compulsion clause, because every rational consideration pointed the other way and the behaviour continued anyway.
Cassius Severus was a lawyer-orator under Augustus who made his name with fiery speeches attacking high-ranking Romans for their extravagant lifestyles. Bombastic, but full of humour that pleased the public. He gained a following.3
Then the escalation, and note its driver: encouraged by the attention he received, he began insulting other officials, always raising the tone of his attacks.
The authorities warned him to stop.
And here is the first datum that a purely strategic account cannot explain: the novelty wore off and the crowds grew thinner — the thing the behaviour was for had stopped being delivered — but this only made Severus try harder.3
In AD 7 the authorities had had enough. His books were burned and he was banished to Crete.
On Crete he simply continued the campaign, sending copies of his latest diatribes back to Rome.3
Read that soberly. He has been exiled, specifically for this, and the sanction has been demonstrated to be real. There is no audience left in Rome to speak of, no career to build, no faction to lead. And the diatribes go out anyway, by ship, to the people who exiled him.
They warned him again. He ignored it — and then began to harangue and insult the local Cretan officials, who wanted him put to death.3
In AD 24 the Senate moved him to the unpopulated rock of Serifos in the middle of the Aegean, where he spent the last eight years of his life.
Greene's closing image is the one to keep: we can imagine him still concocting more insulting speeches that no one would hear.3
Seventeen years, two exiles, one book-burning, a death sentence sought by an entire island, and a rock with nobody on it — and there is no point in that sequence where the behaviour bends. A tendency that survives the removal of every conceivable reward is not a strategy.
And Greene names the reason we resist the conclusion: it is hard for us to believe that people cannot control tendencies that are so self-destructive, and we want to give them the benefit of the doubt, as the Romans did.4
The Romans gave him three chances. The benefit of the doubt is not a modern softness; it is the default human response to a pattern, and it was wrong every time.
The second half of the method is about scale, and it is the part with the most immediate practical use.
If they are late in finishing simple assignments, they will be late with larger projects. If they become irritated by little inconveniences, they will tend to crumble under larger ones. If they are forgetful on small matters and inattentive to details, they will be so on more important ones.5
The logic is not that small things predict large ones by correlation. It is that the small instance is the unguarded one. Nobody manages their conduct over a trivial assignment, which is precisely why the trivial assignment is a clean sample.
And Greene adds the highest-yield version of this, which requires no waiting at all: look at how they treat employees in everyday settings, and notice if there are discrepancies between the persona they present and their attitude toward underlings.5
The golf carts. In 1969 Jeb Magruder came to San Clemente to interview for a job in the Nixon administration. The interviewer was Bob Haldeman, chief of staff — earnest, completely devoted to the cause, and impressive: honesty, sharpness, intelligence.6
Then they left the interview to take a golf cart on a tour of the grounds, and there were no carts available.
Haldeman became frantic. He railed at those in charge of the carts, and his manner was insulting and harsh. He was almost hysterical.6
Greene's assessment: Magruder should have read that as a sign that Haldeman was not what he appeared — that he had control issues and a vicious streak.
He did not. Charmed by the aura of power at San Clemente and wanting the job, he chose to ignore this, much to his later dismay.6
Two things make this the better case study of the two, despite being much smaller than Severus.
The sample size is one, and it was sufficient. Forty minutes of impressive interview conduct told Magruder nothing, because the interview was a managed performance. Ninety seconds at the cart shed told him everything, because nobody manages themselves in front of the cart staff.
And Magruder is not the villain of the story — he is the reader. He saw it. Greene is explicit that he should have seen this incident as a sign, and the failure was not perceptual. He wanted the job. Both men subsequently went to prison over Watergate.
The third instrument is the one Greene handles most carefully, and it corrects a common misreading.
As Lincoln said, "If you want to test a man's character, give him power."7
The reason it works: on the way up, people tend to play the courtier, to seem deferential, to follow the party line, to do what it takes to make it to the top. Once at the top, there are fewer restraints.7
Which means the ascent is a period of near-total measurement failure. Everything you observe about someone who needs something from you is contaminated by their needing it — and that includes almost everyone you evaluate before they arrive.
Lyndon Johnson. Once he reached ultimate security as Senate majority leader, tired of the years he had spent playing the perfect courtier, he relished the power he now had to upset or humiliate those who had crossed him in the past.
The specific behaviours are small and exact. He would approach a senator he disliked and make a point of talking only to his assistant. Or he would get up and leave the floor while a senator he did not like was giving an important speech, making other senators follow him.8
Then Greene's two-part conclusion, and the second part is the correction:
There are always signs of these character traits in the past if you look closely enough — Johnson had revealed such nasty signs in the earliest parts of his political career.
And: so often we think that power has changed people, when in fact it simply reveals more of who they are.8
That is the important half. The folk theory says power corrupts — that the office does something to the person. Greene's claim is that the office removes the constraint that was concealing what was already there. Which has a very different practical implication: if power revealed it, the trait predated the power, and the signs were available to anyone who looked at how the person behaved toward those with nothing to give them.
The instrument has a real limitation and Greene does not name it, so the page will.
It cannot distinguish a pattern from a phase, and it does not try.
Everything above assumes the record is the person. But some of the most useful people in any organization have a bad two years in their history — a period of illness, a marriage ending, one job in which they behaved badly and have not behaved badly since. Applied strictly, this method files them permanently.
The distinguishing feature is not the number of instances but their distribution. Severus's behaviour ran for seventeen years across changing circumstances, escalating rather than decaying, surviving the removal of every reward.3 That is a pattern. Three incidents inside eighteen months, followed by nine clean years, is a period — and the difference is checkable without any psychological insight at all.
The honest form of the corollary, then: people never do something just once, and the useful question is not whether it recurred but whether it recurs across contexts. A behaviour that appears in every setting is character. A behaviour that appeared in one setting and stopped when the setting changed is information about the setting.
Greene closes the chapter with a passage from Cesare Pavese that states the corollary better than his own formulation does, and it is worth having because it identifies the specific illusion involved.
In anything, it is a mistake to think one can perform an action or behave in a certain way once and no more.
(The mistake of those who say: "Let us slave away and save every penny till we are thirty, then we will enjoy ourselves." At thirty they will have a bent for avarice and hard work, and will never enjoy themselves any more.)
What one does, one will do again, indeed has probably already done in the distant past.13
The parenthesis is the sharpest thing in it, and it is a claim this page has not yet made.
Everything above is about reading other people's repetitions. Pavese is describing a mistake people make about themselves, and it is the one nobody sees coming: the belief that a temporary strategy can be adopted and then set down.
The saver at thirty has not failed to enjoy himself through bad luck or continued anxiety. He has spent a decade laying a path, and the path is now what happens when he does not decide otherwise — which is most of the time. He did not acquire a habit. He acquired a character, and he did it by executing a plan that was supposed to be reversible.
Then Pavese's account of why this feels like fate from inside:
The agonizing thing in life is that it is our own decisions that throw us into this rut, under the wheels that crush us. (The truth is that, even before making those decisions, we were going in that direction.)13
Both halves at once. The decisions are genuinely ours, and they were already downstream of a direction we were already travelling. Which is exactly the position Greene's whole chapter occupies — neither pure choice nor pure fate, but a choice made along a groove.
And the closing sentence supplies the mechanism, with the mysticism explicitly refused:
A decision, an action, are infallible omens of what we shall do another time, not for any vague, mystic, astrological reason but because they result from an automatic reaction that will repeat itself.13
Not for any vague, mystic reason. Pavese is doing the same relocation Greene does — taking the ancient intuition that actions foretell a life, and grounding it in automaticity rather than in the stars.
The practical residue: any arrangement you are entering "just for now" should be costed as though it were permanent, because the thing that will make it permanent is not circumstance but the path it lays while you are in it.
You are about to make a decision about somebody — hire, promote, partner, marry — and something small has been nagging at you.
Name the specific behaviour, not the feeling. Not something's off. One sentence with a verb: he was contemptuous to the person at reception. She agreed to the deadline and then renegotiated it twice. If you cannot produce the sentence, you have an impression and not a datum, and the exercise stops here.
Then go looking for it backwards. Greene's third step is the one everybody skips: re-read what you have already observed, in retrospect, with the category in hand.2 You will usually find two or three more instances you had individually explained away — and this is the whole method, because the pattern was never hidden, it was disaggregated.
Weight the unguarded samples much more heavily than the managed ones. The interview is a performance and tells you about their competence at performing. The cart shed is not.6 So: how did they treat the assistant, the driver, the person who got the order wrong? That thirty seconds outranks the forty-minute conversation, and it is nearly always available if you arrange to be present for it.
Check whether the behaviour has ever cost them anything. This is the compulsion test, and it is what separates a strategy from a pattern.1 If the conduct has persisted through consequences — a lost client, a dissolved partnership, a warning taken seriously by them at the time — you are looking at something that will not be argued out of them by you either. If it has always been rewarded, you may simply be looking at a rational actor in a permissive environment, and a different environment may produce different behaviour.
Then run the distribution check before you conclude. Across how many years, and how many separate settings? Escalating, steady, or fading? A pattern that fades across contexts is a phase, and treating it as character is the error this method is most prone to.
And check yourself, in one sentence, honestly. Magruder saw it and wanted the job.6 So: what do you want here, and would you be reading this evidence the same way if you did not want it? That question has saved more people than any amount of behavioural observation, and it takes ten seconds.
Finally, if you proceed anyway — which is often the correct decision, since most patterns are tolerable if you know about them — proceed with the pattern priced in rather than hoped away. Do not hire the person who disappears under stress and then build a plan that requires them under stress. Greene's whole point is that the behaviour is not a risk that might materialize. It is a specification.
Cassius Severus is a real figure and the outline — the invective career, the AD 7 banishment to Crete with his books burned, the later removal to Serifos — is attested in Tacitus and Seneca the Elder. The Magruder–Haldeman golf cart episode comes from Magruder's own memoir, written after Watergate by a man with obvious reasons to portray his own judgment as having been overwhelmed by circumstances. 🚩 MOTIVATED REASONING on that source. Johnson's conduct as majority leader is well documented. [POPULAR SOURCE] throughout.
The load-bearing tension is that this page and Greene's own closing section contradict each other. Here, people never do something just once, the pattern is compulsive, and ignoring it is your own fault.1 Forty pages later, the reader is told that with self-examination they can catch themselves in time and step back and mitigate the effects of their patterns.9
Both cannot be general. And the resolution is uncomfortable, because it is the same asymmetry that appears throughout this book: the reader is granted plasticity and everyone the reader evaluates is denied it. That is precisely the superiority bias Greene documents in Law 1 — we ascribe our own conduct to circumstance and other people's to character — reproduced in the structure of his own advice.
The defensible version is a base rate rather than a law. Most people do not undertake the sustained work that mitigation requires, so past behaviour is your best available predictor — which justifies using the record while leaving the exception real. That is a smaller claim, and it is the one the evidence supports.
Second tension: the Severus case proves less than it appears to. Greene reads seventeen years of escalating self-destruction as compulsion. It is equally consistent with a man who had correctly identified that invective was the only thing he was good at, whose identity was entirely constituted by it, and for whom stopping would have been a form of death. That is not the absence of choice; it is a choice made under a constraint Greene does not model — and the difference matters, because one reading says the man could not stop and the other says he would not.
Third, the practical hazard: this method is a machine for confirming first impressions. Its instruction is to form a hypothesis from one salient behaviour and then search the past for instances of it — which is a description of confirmation bias with steps. Greene's own chapter on decoding warns that interpreting cues through existing bias makes observations not only useless but also dangerous.10 The distribution check and the honest self-check in the workflow are the only safeguards, and neither is in the source.
Open question: every instrument on this page measures behaviour toward people who cannot retaliate — the cart staff, the assistant, the subordinate after the promotion. That is treated as revealing the true self. But it is equally a measure of something narrower: how someone behaves when unobserved by anyone who matters to them. Is a person who is reliably decent to those who can help them and vicious to those who cannot a person with a vicious character, or a person with an unusually accurate model of consequences? Greene assumes the first, and the second predicts the same behaviour.
Howard Hughes and the Pattern is this page's longest-running proof and also its sharpest complication. Four identical failures across five decades, in four industries — and the man had stated the correct diagnosis of his own defect, publicly and precisely, in 1930.11
Which means the corollary held despite insight rather than in its absence, and that is a stronger version of the claim than Greene makes anywhere. He uses Hughes as an illustration of the pattern. The case actually establishes something more specific: articulate self-knowledge is not among the variables that predict whether the pattern will recur.
Character as Engraving supplies the mechanism, and the two pages need each other. Without the laid-path account, this page is a cynical heuristic; with it, the repetition is what the model predicts — the default route requires no decision and therefore appears whenever no decision is made.
Erickson and the Second Language converges on the small-signs principle from a completely different direction. Erickson's readings all came from unmanaged behaviour — how someone walked toward a desk, how they sat down, how lint was picked off a sleeve. Both accounts locate the information in exactly the material that nobody thinks to curate, and the convergence across a clinician and a strategist is the best available evidence that the principle is sound.
The relationship is mechanism and instrument, and putting them together yields a correction that neither states.
The engraving model says character is laid down as paths, that the deepest layers are unavailable to awareness, and that repetition occurs because the path already exists.12 This page converts that into a prediction about other people.
But the conversion silently upgrades a tendency into a certainty. The mechanism is probabilistic: a laid path is the route of least resistance, not the only available route — which is exactly why the engraving page can also promise that a person who examines themselves may catch a pattern in time.
So the honest instrument is not they will do it again. It is: they will do it again unless something has been built that makes the alternative cheaper than the default — and you can check for that, because it leaves traces. Has anything in their environment changed that would raise the cost of the behaviour? Have they built a structure, a commitment, a relationship that runs the correction for them?
That gives the method a discriminator it otherwise lacks, and it is the one the distribution check is really testing. A behaviour that stopped when the setting changed did not stop because the person reformed. It stopped because the new setting made the default expensive — which also tells you it will return the moment the setting does.
→ Power, Authority and the Social Hub
Greene's claim that power reveals rather than changes8 is a genuine minority position, and setting it against that hub's material shows what it costs him.
The dominant tradition holds the opposite: that positions of authority do something to their occupants — that the absence of accountability, the deference of subordinates, the volume of decisions, and the flattery all reliably alter judgment over time, in people who were not previously like that.
Both accounts fit Johnson. Greene says the nastiness was always there and majority leader removed the restraint.8 The institutional reading says years of courtiership followed by unchecked authority would produce that conduct in a wide range of people.
And the case cannot adjudicate between them, because Greene's own escape clause is unfalsifiable: there are always signs in the past if you look closely enough.8 Retrospective sign-finding will succeed on anyone.
What the pairing produces is a test neither side proposes, and it is the practically important one: the two theories differ on whether removing the power reverses the behaviour. If power revealed a fixed trait, a demoted person should stay the same. If power produced it, they should partially revert.
That is checkable, it happens constantly — executives leave, politicians lose office, founders get replaced — and it is almost never examined. The answer would settle a great deal about how much weight to put on this page's core instrument, and about whether the standard remedy for bad leadership is finding better people or building tighter constraints.
Sharpest implication: the pattern was never hidden — it was disaggregated, and each instance came with a plausible local excuse that was probably true. Which means the work is not perception but assembly, and the reason nobody does it is that assembling the instances requires you to have written them down at a time when each one individually did not seem worth writing down.
Generative questions: