Law 19 ends on two warnings that Greene explicitly separates and numbers.
"First, in judging and measuring your opponent, never rely on your instincts. You will make the greatest mistakes of all if you rely on such inexact indicators. Nothing can substitute for gathering concrete knowledge. Study and spy on your opponent for however long it takes; this will pay off in the long run."1
*"Second, never trust appearances. Anyone with a serpent's heart can use a show of kindness to cloak it; a person who is blustery on the outside is often really a coward. Learn to see through appearances and their contradictions. Never trust the version that people give of themselves—it is utterly unreliable."*2
The first is the more radical of the two, and it is easy to read past because it sounds like ordinary caution.
It is a direct attack on intuition as a faculty. Not check your instincts, not instincts are a starting point — the greatest mistakes of all. In a book that repeatedly praises reading people, this is the sentence that says the reading cannot be done by feel.
The passage is doing something more precise than be careful, and the three categories are worth pulling apart because the chapter runs them together.
Instinct — your rapid impression, generated by you, from thin evidence.
Appearance — the surface the person presents, which may be unconscious.
The version people give of themselves — the account they actively supply, which is a claim rather than a surface.
Greene rules out all three, and the reasons differ. Instinct fails because it is "inexact." Appearance fails because it can be inverted deliberately — the serpent's show of kindness. Self-report fails because it is a performance with an interest.
What survives is one thing: "concrete knowledge," obtained by "study and spy on your opponent for however long it takes."
Now put that against what Law 19 asks you to assess.
The five types are defined almost entirely by interior states: touchy pride, a fragile ego, a suspicious disposition, a calculating memory. Four of the five are not observable, and Greene says so himself — the Proud Man "may initially disguise it"; the Serpent "will show no anger on the surface."
So the chapter's instruction is: classify people by their interiors, using neither instinct, nor appearance, nor what they tell you — and then it does not say what concrete knowledge of an interior state would consist of.
That is a genuine gap rather than a quibble, because the prohibition is total and the prescription is empty. Study and spy for however long it takes names a duration, not a method.
Greene supplies the answer once, about one type, and does not generalise it.
Of the Serpent: "Recognize this man by his calculation and cunning in the different areas of his life."3
That is the method. Not his demeanour toward you — his conduct in domains you are not part of.
It is not instinct, because it is a record rather than an impression. It is not appearance, because it is behaviour with consequences rather than a surface. And it is not the version people give of themselves, because it was produced before you were a factor and was not addressed to you.
Behaviour toward third parties, observed across contexts, is the only class of evidence about an interior state that satisfies Greene's own constraints — and it is checkable, cheap, and usually available by asking two people.
The chapter mentions it in one subordinate clause and builds its actual prescription (a mild joke at their expense) on a reaction, which is a demeanour, which is exactly what the second caution forbids.
Worth stating flatly, because this is the law's central incoherence.
The mild-joke test is a rapid judgement made from a person's immediate demeanour. "A confident person will laugh; an overly insecure one will react as if personally insulted."4
The closing cautions forbid rapid judgements from demeanour.
There is a defence available — that an involuntary reflex is not an appearance, because it is not presented, and not an instinct, because it is the target's behaviour rather than your feeling. That defence is reasonable and it is the reading this build takes.
But Greene does not make it, and without it the chapter recommends an instrument on one page and prohibits its epistemic basis on the next.
The chapter's own cases support the second warning better than they support anything else in the law, and it is worth saying so rather than only cataloguing the tensions.
Ford's simple house was an accurate appearance — the dealers saw it, registered it, and proceeded with a positional pitch anyway.5 That is not a case of appearances deceiving; it is a case of an appearance being correct and ignored, which is the failure mode the caution should warn about and does not.
Norfleet's appearance was that of an ordinary cattleman, and Furey's ring read him as an ordinary mark. The appearance was accurate about everything except the one variable that destroyed them.6
Together those two suggest a sharper version of the caution than the one Greene writes: appearances are usually accurate and almost always incomplete, and the danger is not that they lie but that they are silent about the specific variable that matters.
You have a strong feeling about someone.
Maybe good — they seem straightforward, you like them, the conversation went well. Maybe bad — something is off and you cannot name it.
Greene's first caution is that this feeling is worth very little, and on the specific question of how will this person behave when I cost them something, he is right. Your impression was formed from a few conversations in which they were, reasonably, presenting themselves.
So do the thing the chapter buries. Find one instance of how they treated someone who crossed them. Not a character reference — a specific past episode and what followed it. Two conversations will usually surface one.
That single data point outperforms months of impression, because it is behaviour, it is historical, it involved someone who is not you, and it was not produced for your benefit.
And take the second caution seriously in the direction the chapter misses. The problem with appearances is not usually that they are false. Ford's simple house was accurate. The problem is that they are silent about the one variable that will decide the outcome — how far this person will go when injured, which almost nothing about ordinary presentation reveals.
So the useful question is not is this appearance genuine. It is what would this appearance look like if the dangerous thing were also true, and if the answer is exactly the same, you have learned that you need a different kind of evidence rather than a longer look at this one.
The strongest content here is the three-way separation — instinct, appearance, self-report — which is a genuine analytic distinction and rarer in this book than its length suggests.
Tension: the prohibition is total and the prescription is empty. Concrete knowledge of an interior state is demanded and never defined; study and spy for however long it takes specifies a duration rather than a method.
Tension: the cautions forbid the epistemic basis of the chapter's own test. A rapid reading of demeanour is what the mild joke is. The involuntary-reflex defence rescues it, and Greene does not offer that defence.
Tension: the chapter's cases show accurate appearances being ignored, not false ones deceiving. Ford's house, Norfleet's ordinariness. The caution as written aims at the wrong failure.
🚩 SINGLE SOURCE · 🚩 CONSENSUS MISREPRESENTATION — both cautions are asserted with no evidence, no case and no source. "You will make the greatest mistakes of all if you rely on such inexact indicators" is a strong empirical claim about the comparative reliability of intuitive versus deliberate person-judgement, offered with nothing behind it. The research literature on that question is genuinely mixed — rapid judgements are unreliable in some domains and outperform deliberation in others — and the chapter's flat position is not supported here. [POPULAR SOURCE] · [LOW CONFIDENCE]
Open questions. If "never trust the version that people give of themselves" holds generally, what is the status of everything Greene's own practitioners say about themselves — Picasso on the absence of calculation, Lustig on boredom, Jackson on mystifying the enemy? The corpus runs substantially on self-report it elsewhere calls utterly unreliable. And: the cautions are addressed to someone assessing an opponent. What do they imply about assessing an ally, where the same limits apply and the cost of a false negative is completely different?
Within Law 19 these cautions are the chapter's epistemology and they disqualify most of the chapter's own equipment. The typology classifies by interior; the test reads a surface; these cautions forbid surfaces and instincts and supply no alternative. The law contains a taxonomy, an instrument, and a prohibition on the instrument, in that order.
Against Law 14, the convergence is exact and unremarked. Sun-tzu's Authority there rules out spirits, gods, astrology and analogy with past events, leaving only "men who know the enemy situation." These cautions rule out instinct, appearance and self-report, leaving only concrete knowledge. Both are eliminations that arrive at go and find out from the world, five laws apart, and neither references the other. Together they are the corpus's most consistent epistemic position: the future and the interior are both purchased, never inferred.
Against Greene's own later psychology, the tension is sharper still. The Laws of Human Nature is largely a manual for reading people from observable signals — micro-expressions, tone, patterns of speech. That book teaches the skill this passage says will produce "the greatest mistakes of all." Same author, opposite position on whether a person can be read from surfaces, and this build has now logged the same author contradicting himself across books at Mao (L15) and Stonewall Jackson (L17).
Psychology — The Question You Actually Answered
Faced with a hard question you cannot answer — is this machine sound — the mind silently substitutes an easy one it is equipped for: do I trust this person. The swap is invisible and the answer arrives with full confidence.
That is precisely the mechanism Greene's first caution is aiming at, and the pair explains why the caution is so hard to follow.
Will this person retaliate if I cost them something is a hard question requiring evidence nobody has. So the mind answers do I like them, and delivers a confident verdict that feels like a judgement about the original question.
What the pair produces: the instinct Greene warns against is not a weak signal, it is a different signal reported under the wrong name. Your impression is an accurate reading of how pleasant someone is to be around — a question you are genuinely well equipped for — and it is being consumed as an answer about their conduct under injury, which it never addressed.
That is why trust your instincts less is such useless advice, and why the caution needs the buried method to be actionable. You cannot suppress a substitution you cannot detect. What you can do is supply the hard question with its own evidence — a specific past instance of their behaviour toward someone else — at which point there is nothing for the substitution to fill.
Business — Customer-Says-It Truth Asymmetry
That page's principle: when the operator asserts something the other party discounts it; when the other party says it themselves, it registers as true.
Greene's third prohibition is the same asymmetry stated from the receiving end, and the pair produces a useful symmetry the corpus never notices.
The sales page says what they say about their own situation is the load-bearing fact. Greene says "never trust the version that people give of themselves—it is utterly unreliable."
Both are correct, about different things, and the distinction is what makes either usable. Self-report is reliable about preferences and constraints — what someone wants, what they cannot do, what a decision would cost them. It is unreliable about dispositions — what kind of person they are, how they behave under pressure, whether they hold grudges.
The reason is structural rather than moral: people have direct access to their preferences and no privileged access to their own character, and they also have an interest in the second and usually not the first. A prospect saying this would save us money is reporting a fact about their situation. A person saying I'm not the kind to hold a grudge is producing a self-image.
Which gives a clean rule neither page states: believe people about their circumstances and disbelieve them about their character — and Greene's flat utterly unreliable is wrong in the half that makes ordinary cooperation possible.
Sharpest implication. The chapter forbids instinct, appearance and self-report, demands concrete knowledge of an interior state, and never says what that is. The only answer surviving its own constraints — behaviour over time toward third parties — appears in a single clause about one type, and it is the law's most useful sentence.
Generative questions.