Versailles, and the king is standing at the top of a court filled with noblemen who resent him.
They resent him specifically — the absolute authority he is imposing, the power he has taken from them, the fact that they are here at all. Greene's word is seething.1
And not one of them can show it. In the civilized realm of Versailles, everyone had to be a consummate actor and hide their feelings, particularly toward the king.1
So Louis has a measurement problem, and it is the one every person in a position of any authority has in a smaller form. The higher you sit, the less information reaches you — because the people who would tell you what they think are precisely the people who cannot afford to. Everyone is pleasant. Everyone agrees. And somewhere in that room are the ones who will move against you.
Waiting to observe them will not work, because they are better actors than you are readers.
Louis's answer is the page: he stopped waiting for signals and started producing them.
Greene gives three, and they are worth separating because each is engineered to defeat a different kind of concealment.
The unannounced arrival. He would suddenly appear in their presence, without warning, and look for the immediate expressions on their faces.1
That one is aimed at preparation. A courtier expecting the king has had time to arrange his face, and the arrangement is complete before the king enters the room. Remove the notice and you get the second or two before the mask is fitted. Greene describes the same mechanism elsewhere in this chapter: if you approach someone who harbours negative feelings from an angle, by surprise, you will see displeasure before they have had time to fit on their affable mask — and it shows for a second or two.2
The costly request. He would request that a nobleman move himself and his family to the palace of Versailles — knowing that this was costly and unpleasant — and then carefully observe any signs of annoyance in the face or voice.1
This one is aimed at cheap compliance. Anyone can say yes to something that costs nothing, so agreement carries no information. Louis makes the request expensive, which forces a gap between what the man says and what he feels, and the gap is where the leakage happens. The nobleman will still say yes. The yes is not the measurement.
The insulted ally. He would say something negative about another courtier, an ally of theirs, and notice their immediate reaction.1
The subtlest of the three, because it does not touch the target at all. A man can control his response to an attack on himself — he has rehearsed that, it is the situation he expects. He has not rehearsed a response to an attack on his friend, and the loyalty that shows in that unguarded half-second maps his actual alliances rather than his declared ones.
Greene's decision rule: enough signs of discomfort indicated secret hostility.1
Note enough. Not one sign. The plural is the whole difference between a method and a paranoia, and it is the safeguard the rest of this page depends on.
The obvious objection is that this is just observation with extra steps. It is not, and the difference is worth being precise about, because it is what makes the technique valuable.
Passive observation has two failures that testing does not.
You have to wait, and you do not control the sample. Weeks can pass in which a hostile colleague has no occasion to leak anything, because nothing has come up. Meanwhile you are accumulating pleasant interactions and concluding, reasonably, that things are fine.
And when a signal finally arrives, you cannot attribute it. Someone is short with you on a Tuesday. Is that about you, or about their morning? Greene's own account of decoding failure is that we rush to the first explanation that fits what we want to see — and an uncontrolled observation is exactly the material that error feeds on.3
A test fixes both. You choose the moment, and you know what the stimulus was. When Louis disparages a man's ally and watches the face, whatever appears in the next second has one available cause, because he supplied it.
Which makes this the practical answer to Othello's error rather than another instance of it. Othello interrogated — he applied pressure and read the resulting distress as evidence about something other than his own pressure. Louis applies a stimulus and reads the response as evidence about the stimulus. The difference is whether you know what you did.
Greene extends the method beyond the court and the extension is where it becomes usable.
For envy: if you suspect someone of feeling envy, talk about the latest good news for you without appearing to brag, and look for microexpressions of disappointment on their face.4
The design constraint is in the phrase without appearing to brag, and it is not a courtesy. If you brag, you have supplied a legitimate reason for the flicker of irritation, and the reading is spoiled — you will have measured your own gracelessness. The stimulus has to be clean, which means it must be something a person with no envy would receive with unmixed pleasure.
For anger and resentment: use similar tests to probe, eliciting the responses that people cannot suppress so quickly.4
The whole design principle is in that clause. You are not trying to make someone reveal themselves. You are trying to outrun the suppression — every one of these tests works by producing a reaction faster than the mask can be fitted.
And Greene supplies a coarse but genuinely useful classification for what you are measuring toward. In general, people will want to see more of you, want to see less of you, or be rather indifferent. They may fluctuate, but they will tend to veer toward one. And they reveal it in three places: how quickly they respond to your emails or texts, their body language on first seeing you, and the overall tone they take in your presence.4
That is the most immediately checkable thing on this page, and it requires no test at all — response latency is already being recorded, by your inbox, for everyone you know.
It is worth asking why Versailles existed at all, because the tests only work inside it.
Louis brought the French nobility to live at his palace. Greene's framing is that they were seething about the absolute authority he was imposing1 — and the residence requirement was part of imposing it. A nobleman at Versailles is a nobleman not on his own estates, not among his own retainers, and not building a regional power base.
But look at the second function, which is the one this page is about. He had constructed an instrumented environment.
Every one of his three tests requires physical proximity and repeated access. You cannot appear unannounced before a man three hundred miles away. You cannot watch his face when you disparage his ally if he hears about it by letter a fortnight later. The costly-relocation test is itself the mechanism that makes the other tests possible — which is why it comes second in Greene's list and is doing double work: it measures the man's compliance and it delivers him into the range of the instrument.
So Versailles is not a palace with a surveillance problem solved by clever tricks. It is a purpose-built observation facility in which several hundred people who would otherwise be invisible were placed within daily reading distance of the one person they most needed to deceive.
That reframes what the technique costs. Louis's tests are cheap to run and were made possible by an enormous, decades-long structural investment. The tests are the easy part. Anyone reading this page has access to the tests and almost certainly not to the environment — which is the honest limit on how much of this transfers.
Greene is explicit that the point is not the knowledge. The value in detecting possible hostility early is that it increases your strategic options and room to manoeuvre, and he names three, in ascending order of aggression.5
Lay a trap — intentionally stir their hostility and goad them into some aggressive action that will embarrass them in the long run.
A charm offensive — work doubly hard to neutralize their dislike and even win them over.
Create distance — not hiring them, firing them, refusing to interact.
The ordering matters more than the list. Greene puts the aggressive option first and the boring one last, and for almost every reader the ranking should be inverted: distance is the option that is available, cheap, reversible, and almost always sufficient. Goading someone into a self-embarrassing overreach requires that you be able to absorb the overreach, which is a claim about your position rather than your cleverness.
And his own summary of the payoff is the modest one: you will make your path much smoother by avoiding surprise battles and acts of sabotage.5
Which is the real return. Not victory. Not being surprised.
None of this is worth anything unless one claim is true, and Greene states it directly:
People's hostile or resistant actions never come out of the blue. There are always signs before they take any action. It is too much of a strain for them to completely suppress such strong emotions.6
That is a strong claim and the page should be honest that it is doing all the load-bearing. If concealment were perfect, testing would produce nothing.
Greene's account of why we miss the signs is the interesting half, and it is not about perception. The problem is not only that we are not paying attention — it is that we inherently do not like the thought of conflict or disagreement. We prefer to assume people are on our side, or at least neutral. And most often, we do feel something is not quite right and ignore the feeling.6
So the failure is motivated rather than perceptual. Which means the tests have a second function beyond producing a signal: they produce a signal at a moment you have committed in advance to look at, which defeats the avoidance. You cannot decide not to have noticed something you set up in order to notice.
The cue list, for completeness: sudden squinting of the eyes at something you have said, the glare, the pursing of the lips until they nearly disappear, the stiff neck, the torso or feet turning away while you are still in conversation, the folding of the arms as you try to make a point, and an overall tenseness.7 Greene's caveat is important — you will not usually see these unless displeasure has become too strong to conceal, which is why the microexpressions and the subtler signs are what the tests are actually aimed at.
Something is off with a colleague and you cannot name it. Nothing has been said. Every interaction is fine.
Start with the free data, before you construct anything. Open your sent folder and check response latency to this person against two others at the same level.4 Not one instance — the pattern over two months. If yours reliably come back last, you have a reading that cost nothing and that nobody performed for you.
Then design one test, not three. The discipline is that a test has to have a clean stimulus and a defined window, or you are just being awkward at someone.
The good-news test is the cleanest and the least aggressive, so start there. You have something genuinely positive coming — a piece of work landed well. Mention it in passing, factually, to a small group including them. The design constraint is that it must not be a brag,4 because a brag gives a decent person a legitimate reason to wince and destroys the measurement.
Then watch the first half-second, and watch it peripherally. Greene is explicit elsewhere that feeling scrutinized makes people freeze up and control their expressions, and that too much direct eye contact will betray you.8 So you are looking at someone else while it registers.
Write down what you saw, immediately, without a conclusion attached. Flat. Congratulated on a half-beat delay. Went back to laptop.
Then do not act on it. One reading is not a finding. Greene's own rule is enough signs,1 and the surest way to convert this method into paranoia is to treat a single flicker as a verdict. Run a second test, a month later, of a different type — mention a piece of criticism of someone they are allied with, mildly, in passing, and watch what happens to their face.1
Two independent tests, two consistent readings, a latency pattern, and the ordinary sense that something has been off — that is a finding. One of those is a mood.
And then take the boring option. Not the trap. Distance: stop routing the thing that matters through them, put the important agreements in writing, and stop being surprised. Greene's own stated payoff is a smoother path and the absence of ambushes,5 and that is entirely achievable without a single confrontation or anyone knowing you ever looked.
One boundary worth keeping. Everything above is aimed at deciding how much to expose yourself to someone. The moment it turns into building a case against a person — collecting flickers as evidence for a conclusion you have already reached — you have stopped running Louis's method and started running Othello's, and the fact that you designed the stimulus makes you more confident and not more correct.
The Versailles court and Louis XIV's use of residence at the palace as an instrument of control over the nobility are historically solid. The three specific tests are Greene's characterization rather than documented episodes — he names no memoir, courtier, or incident — and should be read as a plausible synthesis of how the court operated rather than as recorded practice. 🚩 SINGLE SOURCE; [POPULAR SOURCE].
The load-bearing tension is that this technique's safeguard and its purpose are in conflict. The safeguard is enough signs — multiple readings before a conclusion.1 But the reason you are testing at all is that you already suspect something, and a person running repeated tests on someone they already distrust is in an excellent position to accumulate exactly the confirmations they went looking for. Designing the stimulus removes the attribution problem and does nothing about the prior, and the added confidence from having controlled the setup makes the resulting conviction harder to dislodge.
The only genuine protection is the one this page's workflow specifies and Greene does not: decide in advance what result would clear the person, and check afterward whether any observable response would have. If a warm reaction would have read as performed and a flat one as hostile, you were never testing.
Second tension: the technique is only fully available to the powerful, and Greene presents it as generally applicable. Louis could summon a family to relocate. He could disparage a courtier without consequence. He could appear anywhere unannounced. Every one of the three tests depends on an asymmetry of position — and a peer who tries the unannounced-arrival test on a colleague is not testing, they are being odd. The generalized versions Greene offers are real but weaker, and the strongest of them, response latency, is not a test at all but passive observation of an existing record.
Third, and it should be named: this is a page about deliberately manipulating people's emotional states to extract information they have chosen not to give. The defensive framing is honest — avoiding sabotage is a real interest — and the technique is identical whether the aim is defence or targeting. The vault holds the targeting corpora; this is the same instrument with a different stated purpose.
Open question: the tests work by outrunning suppression.4 But suppression speed is a trainable skill, and the environments where testing is most valuable are also the environments that select for people who are good at it. Does the method therefore work best on the people least likely to be a threat — and worst on exactly the courtier Louis most needed to identify?
Baselining and Behavior Analysis and this page are complementary halves of one method, and neither corpus says so.
Baselining is passive and answers has something changed. It requires no intervention, produces no suspicion, and cannot tell you why. Testing is active and answers what does this specific stimulus produce. It requires you to act, risks being noticed, and delivers an attributable result.
Run alone, each has the other's failure. Baselining without testing waits indefinitely for a signal that may never be occasioned. Testing without a baseline produces a reaction you cannot size — you saw a flicker, and you have no idea whether that is a lot for this person.
Together they produce something neither has: a deviation with a known cause. That is the only form of behavioural reading that is worth much, and it is why Louis's method depended on Versailles. He had years of baseline on every man in that court and the ability to introduce a stimulus at will. Most readers will have one or the other.
These two pages are the problem and its structural solution, and reading them together produces the criterion that neither states alone.
Othello's failure was that he caused the nervousness he then read as guilt — a closed loop in which the observer's own behaviour generates the evidence and is the one variable never considered. Louis also causes the reaction he reads. The two situations are physically identical and epistemically opposite.
The difference is not intention, and it is not skill. It is this: Louis knows what he did and Othello does not. Louis introduced a defined stimulus at a chosen moment and is reading the response to that stimulus. Othello introduced pressure without registering that he had introduced anything, and read the response as data about his wife's fidelity rather than about his own manner.
Which gives a transferable rule for every situation where you are about to draw a conclusion from someone's reaction: name your own contribution to the stimulus before you interpret the response. If you can state it — I mentioned the promotion, neutrally, in front of three people — the reading is usable. If you cannot, or if the honest answer is I've been suspicious of him for a month and he can tell, then whatever you observed is your own prior reflected back and it will feel exactly like evidence.
That criterion is also what separates a test from an interrogation, and it explains why interrogation is such a poor instrument: the interrogator's contribution is enormous, continuous, and invisible to them.
→ Behavioral Economics and Decision Science Hub
Louis is doing revealed-preference measurement, three centuries early, and the pairing sharpens both.
That hub's core distinction is between what people say they prefer and what their costly choices demonstrate — stated preference being cheap, unreliable, and shaped by what the respondent thinks the asker wants to hear. The methodological answer is to observe behaviour under a real cost.
The relocation test is exactly that. Louis does not ask the nobleman whether he is loyal; the answer is worthless because it is free. He imposes a genuine cost and observes the response.1
But Louis adds something the economic version does not have, and it is the more interesting half. Revealed preference reads the choice — did he move or not. Louis reads the half-second before the choice, because in a court every nobleman complies and compliance therefore carries no information.
Which identifies a real gap in the revealed-preference method that the vault can name: in any environment where the cost of refusing is prohibitive, behaviour is fully determined and reveals nothing. Everyone at Versailles obeys. Every employee agrees to the reorganization. The choice data is a constant, and a constant has no variance to measure.
What still varies is the involuntary response, and that is why Louis had to become a reader of faces rather than a keeper of records. The generalization is worth having for anyone running an organization: the more coercive your environment, the less your compliance data means, and the more you are forced back onto exactly the soft, deniable, easily-misread signals this whole chapter warns about. Which is a genuinely uncomfortable finding — it means the leaders with the most power over their people have the worst information, and the tools available to fix it are the least reliable ones in the book.
Sharpest implication: Louis's real achievement was not the tests. It was building the environment in which testing was possible — bringing several hundred people who would otherwise have been invisible into daily reading range of the one person they most needed to deceive. The tests are cheap and portable; the observation facility is not. Which means anyone applying this page's techniques without that structural position is running the easy half of a method whose hard half was decades of statecraft.
Generative questions: