Law 19 contains exactly one usable diagnostic, and it costs almost nothing to run.
"To see if you are dealing with such a type, test them first—make, say, a mild joke at their expense. A confident person will laugh; an overly insecure one will react as if personally insulted. If you suspect you are dealing with this type, find another victim."1
Four features make it a genuinely good instrument, and they are worth naming because most of what this book calls a technique has none of them.
It is cheap. A mild joke costs nothing and can be delivered inside an ordinary conversation.
It is deniable. If it lands badly you were joking, which is a real excuse rather than a transparent one.
It reads a reflex rather than a statement. You are not asking someone how they handle criticism — which produces a rehearsed answer — you are creating a small live instance and watching the response arrive before deliberation can shape it.
And it is run before commitment. Find another victim is the point of the whole thing: the test exists to be run while walking away is still free.
Greene frames the test inside a wider prescription: "All people have insecurities, and often the best way to deceive a sucker is to play upon his insecurities. But in the realm of power, everything is a question of degree, and the person who is decidedly more insecure than the average mortal presents great dangers."2
Note what that concedes. Insecurity is not the disqualifier — insecurity is the lever, and the chapter recommends using it. What disqualifies a target is being unusually far along a distribution everyone is on.
So the test is not screening for a category. It is screening for an outlier, and the mild joke is a cheap way of sampling where on the distribution somebody sits without having to ask.
That is a better description of what the instrument does than identify the insecure man, and it explains why a single small stimulus is adequate: you are not diagnosing a condition, you are checking a magnitude.
The instrument has real coverage limits, and the chapter does not state any of them.
It separates secure from insecure and cannot separate insecure from proud. Both of Greene's first two types — the Arrogant and Proud Man, the Hopelessly Insecure Man — react badly to a joke at their expense. The test tells you to leave, which is the correct instruction for both, so this limitation is survivable. But it means the test cannot tell you which danger you avoided, and the two require different exits: the Proud Man retaliates immediately and the Insecure Man accumulates.
The test is useless against the Serpent by definition. That type "will show no anger on the surface; he will calculate and wait."3 A man who conceals reaction as a matter of practice will pass this test perfectly. The type carrying the harshest prescription in the chapter is invisible to the chapter's only instrument.
It is useless against the Blunt Man, who may simply not follow the joke — and whose non-reaction is indistinguishable from a confident person's equanimity.
So: five types, one test, two covered without being distinguished, and three invisible.
Greene supplies a better method elsewhere in the same law and does not connect it to the test.
Describing the Serpent, he writes: "Recognize this man by his calculation and cunning in the different areas of his life."4
That is a completely different instrument and a much stronger one. It does not read a reaction; it reads a pattern across contexts — how this person operates in domains that have nothing to do with you.
And it is the only method that survives the chapter's own closing prohibitions. Greene forbids relying on instinct and forbids trusting appearances,5 which rules out impression and demeanour. What is left is behaviour, observed over time, toward other people — and behaviour toward third parties is the one class of evidence that cannot be performed at you, because it was produced before you were a factor.
The mild joke and the pattern read trade off exactly. The joke is fast, cheap, available immediately, and shallow. The pattern read is slow, requires access to a history, and is the only thing that reaches the types that matter most.
Greene gives the fast one as the test and mentions the slow one in a clause.
Worth stating plainly, because the chapter does not and the technique's whole ethics live here.
Every reliable assessment of character requires imposing a small cost on someone to see what they do with it. That is not a criticism of the method; it is the method. A mild joke at somebody's expense is, unavoidably, at their expense — a small deliberate unkindness administered to a person who does not know they are being measured.
At the scale of a joke, that is trivial and most people would not object.
But the technique is continuous with much larger versions of itself. Chosroes II, five laws earlier, ran the same operation — construct a situation, read the involuntary response — and paid for it with an innocent courtier per run.6 There is no principled boundary between the two in Greene's corpus; there is only a difference of degree, and the chapter that recommends the small version never marks where the continuum should stop.
The honest formulation: this is a good instrument and it is not a free one, and its cheapness is exactly what makes the escalation easy not to notice.
You are about to rely on someone, and you do not know them well.
Not deceive them — the chapter's frame is a con man's and most people are not running one. But you are about to put something of yours inside their reach: a joint project, a shared client, a piece of work whose failure lands on you.
The mild joke is worth running and it is worth understanding what it tells you. Watch the recovery, not the first reaction. Everyone bristles slightly; the signal is what happens in the following thirty seconds. Does it resolve, or does the conversation acquire a small permanent edge? Does it come back later, reframed, in a context where it does not belong?
That is a real reading and it takes one conversation.
Then run the instrument the chapter buries, because it is the one that reaches the people who matter. Find out how they have treated someone who crossed them, in a situation that has nothing to do with you. Not gossip about their personality — a specific past instance and what followed. This is checkable, it is historical, it cannot be performed at you, and it is the only method that survives Greene's own prohibitions on instinct and appearance.
And run the honest check on yourself before you do either. You are administering a small unkindness to gather information from a person who has not consented to being measured. At the scale of a joke that is ordinary social behaviour and not worth agonising over. The thing to watch is the drift — the point at which the test gets larger because the small one was inconclusive, and the person paying for your certainty has stopped being a party to a joke.
The instrument's strength is that it reads an involuntary response rather than a statement, which is the property that makes it resistant to preparation — and it is one of very few techniques in this book that a reader could actually execute this week.
Tension: the test covers two of five types and cannot distinguish them, and the three it misses include the Serpent, which carries the chapter's harshest prescription.
Tension: the chapter's better instrument is in a clause. "Recognize this man by his calculation and cunning in the different areas of his life" is the only method that survives Greene's own closing prohibitions, and it is mentioned once, about one type, and never generalised.
Tension: the technique is continuous with Chosroes' loyalty test five laws earlier, and the corpus offers no principle for where on that continuum to stop.
🚩 SINGLE SOURCE — the test is asserted with no case, no source, and no evidence. It is presented as practitioner knowledge alongside the typology's uncited "artists—con and otherwise—of the past." [POPULAR SOURCE] · [LOW CONFIDENCE] on the claim that the reaction reliably discriminates.
Open questions. What is the false-positive rate? A person having a bad day, or one who has recently been the target of real hostility, will read as insecure — and the prescription is find another victim, so the cost of a false positive is a foregone relationship you will never learn was a mistake. And: does the test work on someone who has read this book? The reflex it reads is fast, which suggests yes — but a person expecting to be tested may substitute a performed laugh, and the instrument has no way to detect a rehearsed reflex.
Within Law 19 this is the operational half of the typology, and the mismatch between them is the chapter's central weakness: five categories defined by interior states, and one test that reads a surface reaction.
Against the closing cautions in the same law, there is a direct tension the chapter never registers. Greene says never trust appearances and never rely on your instincts — and then the only instrument he supplies is a rapid judgement made from a person's immediate demeanour, which is close to the definition of both. The test survives only on the argument that an involuntary reflex is not an appearance, and Greene does not make that argument.
Against Chosroes II at Law 14, this is the same instrument at the opposite end of the scale — construct a situation, read the involuntary response, act on the reading. Chosroes' version destroys an innocent person per run by design; this one costs a moment's discomfort. The corpus contains both and no criterion, which is worth stating as a corpus-level observation rather than a complaint about either page.
And against Ford in this same law, the test would have failed anyway. Ford's immunity was not insecurity or its absence — it was that he did not hold the distinction the pitch required, and no reaction test detects a missing valuation. The chapter's best-fitting case is one its own instrument could not have caught.
Psychology — False Confession Psychology
That page's uncomfortable finding is that the conditions which most reliably produce a confession are the same conditions that most reliably degrade the truth of what it contains — pressure produces output, and the output's reliability falls as the pressure rises.
The mild joke is the same instrument at minimum dose, and the pair makes the dose-response visible.
At a very low stimulus, what you get is close to a genuine reading: a small unexpected provocation, an unrehearsed reaction, and a signal that reflects the person rather than the situation.
But the instrument has the same curve. Increase the stimulus to get a clearer signal — a sharper joke, a more public one, a real slight — and at some point you stop measuring a disposition and start manufacturing one. A person who reacts badly to a serious insult has told you nothing about their baseline.
What the pair produces: the test's validity depends on the stimulus staying small, and the operator's incentive runs the other way. An ambiguous result invites a stronger test, and a stronger test is less valid, not more. The instrument degrades precisely when you most want an answer — which is the same structure the interrogation literature documents, and it means mild is not a stylistic recommendation but the load-bearing condition.
Business — Secret Shop Your Own Business
That page's method is to experience your own operation as a customer would — unannounced, without identifying yourself — because what people do when they are not performing for you is different from what they report.
That is the same epistemics as the mild joke, aimed at an organisation instead of a person: create an ordinary situation and observe the unprepared response.
What the pair produces is the general principle both are instances of, and it is more useful than either. Self-report is unreliable not because people lie but because being asked changes the thing being measured — an employee who knows they are being assessed performs assessment-behaviour, and a person asked how they take criticism produces an account of how they would like to take it.
So the only reliable observations are of behaviour that was not produced for the observer. That gives a clean hierarchy of evidence quality, which Law 19 gropes toward and never states: what they did before you were a factor (best — Greene's buried pattern-read), what they do when they do not know they are being watched (the secret shop), what they do in a situation you created (the mild joke), and what they tell you about themselves (worst — the thing Greene explicitly warns is "utterly unreliable").
The chapter recommends the third and dismisses the fourth. The first two are better and it names neither as a method.
Sharpest implication. The test's validity is inversely related to how badly you want a clear answer — an ambiguous result invites a stronger stimulus, and a stronger stimulus measures the situation rather than the person. Mild is not a matter of taste; it is the condition on which the whole instrument works.
Generative questions.