A fourteenth-century Japanese marginal, dropped into Law 47 without comment, contains the best-observed piece of psychology in the chapter.
A famous tree climber is guiding a man up a tall tree and has ordered him to cut the top branches. The man is at a lethal height on branches that might break. The expert watches and says nothing. Then, on the way down, at roof height, he calls out: "Be careful! Watch your step coming down!"1
Asked why — at that height the man could have jumped the rest of the way — the expert explains: "As long as the man was up at a dizzy height and the branches were threatening to break, he himself was so afraid I said nothing. Mistakes are always made when people get to the easy places."2
The silence is the interesting half. It looks like negligence and is the opposite. The expert is not failing to warn during the danger; he is declining to interfere with a regulator that is already working, and correctly identifying that his voice would have been an addition to the load rather than a help with it.
Most safety thinking targets maximum hazard. This targets something else, and getting the distinction right is the whole transferable content.
The quantity the expert is tracking is not danger. It is unattended danger — danger multiplied by the fraction of it the person is not currently watching. High in the tree, hazard is at its peak and attention is also at its peak, and the product is small. At the eaves, hazard has dropped by perhaps two-thirds and attention has dropped to nearly nothing, and the product is larger than it was at the top.
That is a genuinely different optimisation, and it explains a fact people find counterintuitive: that accidents cluster in ordinary conditions rather than extreme ones. Not because ordinary is dangerous, but because ordinary is sub-threshold for fear while remaining above threshold for consequence.
The eaves are still roof height. A fall from there still kills. What has changed is not the physics but whether anyone is looking.
The passage assumes something worth stating explicitly, because it cuts against how fear is usually discussed: that fear is not an impairment to be managed but the primary allocator of attention, and a more reliable one than deliberate intention.
The man in the treetop was not being careful because he had decided to be careful. He was being careful because he was terrified, and terror had narrowed everything down to the branch under his foot. That is not a state you can produce by resolving to concentrate, and it is not a state that survives the removal of its cause.
Which is why the descent is unprotected. He does not decide to stop paying attention. The thing that was paying attention on his behalf switched off, on schedule, at exactly the height where the body stops reading the situation as survivable-if-mishandled.
The practical consequence: you cannot use "I'll be careful" to cover the easy stretch, because the careful you are relying on was never a decision. It was a symptom. Whatever replaces it has to be structural — a rule, a checklist, a person calling out at the eaves — and the expert is exactly that structure, applied at exactly the right moment.
Kenkō then generalises, and the generalisation is sharper than the anecdote: "In football too, they say that after you have kicked out of a difficult place and you think the next one will be easier you are sure to miss the ball."3
Read that closely. The claim is not that easy tasks are dangerous. It is that a task graded easier than the one just completed receives a downgraded allocation — and the grading is comparative.
So the same task, at the same objective difficulty, gets a different amount of attention depending on what preceded it. Follow a hard thing with a medium thing and the medium thing is under-attended. Follow an easy thing with the same medium thing and it is fine.
Attention is allocated by contrast, not by level. That is the mechanism, and it is the one that makes this marginal belong in Law 47 rather than in a chapter about diligence.
Put the pieces together and the chapter's central claim stops being a warning about character and becomes a consequence of how attention works.
A victory is a difficult place successfully exited. Everything after it is graded against it, and therefore under-attended — not by anyone's choice, and not because of arrogance, elation, or intoxication. The very success that ended the hard part is the reference point that downgrades everything following.
That is a better account than the chapter's own. Greene's Interpretation says success "makes you feel invulnerable"4 and treats the resulting failure as a character problem with a moral remedy — be humble, be cautious, recognise the role of luck. If the contrast mechanism is what is actually running, then humility does not touch it. A perfectly humble person coming down from the treetop has the same collapse in attention, because the collapse is not an opinion about themselves.
Which relocates the remedy entirely. Not hold a more accurate view of your own contribution, but install something external at the transition, because the internal regulator has already gone off duty and cannot be argued back on.
Notice when the expert speaks. Not while the man is high. Not at the moment he begins descending. At "the height of the eaves"1 — the specific point at which the situation stops reading as dangerous.
Too early and the warning lands during the danger, where it competes with the fear that is doing the work. Too late and the ground has arrived. The window is narrow and the expert knows precisely where it is, which is what makes him an expert rather than a cautious person.
This is a real design constraint on any intervention of this kind. A warning delivered while someone is still frightened is noise — worse than noise, since it consumes capacity that is fully committed. A warning delivered once they have relaxed is a warning delivered to someone who has already decided the thing is over.
The instrument has to fire at the transition, which means whatever you build has to be triggered by the change in perceived difficulty rather than by the clock or by the objective risk. In practice that means the trigger is a feeling — the moment the thing starts to feel handled — and that feeling is the only reliable signal available.
The passage ends with a remark Greene reprints and does not comment on: "This man belonged to the lowest class, but his words were in perfect accord with the precepts of the sages."2
Worth recording rather than editing out. It is a fourteenth-century Japanese sensibility on display — Kenkō is genuinely surprised that a labourer produced something the sages would endorse, and the surprise is the frame's own.
But there is a substantive point inside the snobbery. The observation came from practice. The tree climber had watched hundreds of people ascend and descend and had noticed where they fell. He was not deducing from a doctrine of vigilance; he had the base rates.
That is why the observation is better than the doctrine it accords with. A precept about constant alertness is unfalsifiable and unusable — alert when, at what cost, instead of what. The climber's version is a targeting rule, and it is targeted because it was built from the location of actual failures rather than from a principle about how one ought to be.
The hard part is over. It went well. This is the moment.
Say that to yourself as a specific claim, because the signal you are looking for is exactly this feeling — it's handled now — and it is not a report on the situation. It is the sound of your attention being reallocated, and it arrives on schedule whether or not anything has actually become safer.
So do the check now, not later. Whatever the equivalent of the eaves is in your case — the final approval, the send, the last leg of the drive, the paperwork after the deal — that is the part to do slowly, and it will feel absurd to do it slowly, because it is easy. The absurdity is the confirmation.
Better still, do not rely on noticing. The whole lesson is that the noticing apparatus is off. Put something outside yourself at the transition: a person who reviews the final version, a rule that the last step is never done tired, a fixed pause between finishing and sending. It has to be something that fires without your judgment, since your judgment is the thing that just stood down.
And if you are supervising someone through something hard — say nothing while they are frightened. They are already at capacity and your voice is more load. Wait until it starts going well. That is when they need you, and it is the only moment your warning will be heard rather than absorbed.
Evidence quality: the marginal is attributed to Kenkō's Essays in Idleness, fourteenth-century Japan, and is a genuine and well-known passage from that work — better attributed than most of Greene's material. [POPULAR SOURCE] for the surrounding framing; 🚩 SECONDARY WITHOUT PRIMARY in that Greene gives no edition or section.
The psychological claim — that attention allocation is comparative rather than absolute — is [PLAUSIBLE — needs corroboration]. It is consistent with the general finding that judgment operates on contrasts rather than levels, but the specific application to post-difficulty vigilance is asserted here from a fourteenth-century practitioner's observation and no more.
Tension: this passage locates post-victory failure in attention, and the chapter around it locates it in character (arrogance, elation, feeling invulnerable). These imply different remedies — external structure versus internal humility — and Greene runs them together. The attention account is the better-evidenced of the two and appears only in a marginal.
Open questions: the contrast mechanism predicts that the size of the attention drop should scale with the difficulty of what preceded it. Is a small win followed by a routine task under-attended in the same way as a large one? Nothing in the passage or the chapter addresses magnitude.
Greene's marginals are usually decorative — a fable or a quotation running alongside the argument, illustrating its mood. This one contains a competing theory of the chapter's subject, and it is more rigorous than the theory in the body text.
That pattern is worth marking because it recurs across the book. Law 47's other marginal, the Machiavelli passage on the overreaching general, also undercuts the main argument by relocating the danger outside the victor's control. Twice in one chapter, the borrowed material is doing better analytical work than the authored material. [n=2] — two instances in one chapter, not a claim about the book.
The convergence is with the corpus's better moments on state-dependence: Greene is consistently good on the idea that capability varies with condition, that the person who plans is not the person who executes. Law 47's Judgment — set a goal, and when you reach it, stop — is exactly a pre-commitment device built on that insight. The Kenkō marginal supplies the reason such a device is necessary, and Greene never connects them.
Attention as Load-Bearing — the contemplative frame treats attention as a finite structural element rather than as a stance one adopts, and that is precisely what the tree-climber's silence assumes. His refusal to speak during the dangerous phase only makes sense if attention is a load with a ceiling: the man aloft is at capacity, and an incoming instruction does not add to his care, it displaces something already holding weight. The strategic reading of the anecdote misses this — it hears the expert as saying he didn't need warning then, when what he is saying is warning him then would have cost him something. The insight the pairing produces: the descent problem is not a shortage of attention but a release of it, and released capacity does not stay allocated to the task by default — it goes elsewhere, immediately, because the demand that was holding it has ended. Which predicts something the psychological account alone would not. The danger at the eaves is not merely that the man is less careful; it is that his attention has already been recommitted — to the ground, to what comes next, to the conversation he will have when he is down. You are not competing with inattention. You are competing with a live allocation to something else, and that is a harder thing to interrupt than emptiness.
Deep Observation: The Passive Mode — the craft material describes a mode of attention that is receptive rather than directed, and it explains why the tree climber knew what he knew. He was not testing a hypothesis about descent accidents; he had watched, for years, without an agenda, and the pattern surfaced. That is the passive mode's characteristic yield — base rates that nobody set out to collect. What this adds to the psychological account is a claim about where reliable observations of this kind can come from at all. The contrast mechanism is invisible to the person experiencing it, by construction: at the moment your attention downshifts, you are not in a position to notice that it has, and afterwards you will not remember the interval as inattentive. It is only observable from outside, across many instances, by someone with no stake in any one of them. The insight neither frame produces alone: this class of finding — failures that occur precisely where the actor's self-report is least available — is systematically under-supplied by introspective and doctrinal sources and can only be recovered by long undirected watching. Which is a rather good defence of Kenkō's labourer against Kenkō's own surprise, and a reason to weight practitioner observation heavily in exactly the domains where the practitioner cannot explain the mechanism.
The testable claim is the contrast model: that attention allocated to a task is a function of its difficulty relative to the immediately preceding task, not of its absolute difficulty. That predicts a measurable error-rate elevation for routine work performed immediately after demanding work, independent of fatigue, and it should be separable from fatigue by varying the order rather than the total load.
The open edge is magnitude, raised above. If the drop scales with the size of what preceded it, then Law 47's danger is proportional to the size of the victory — which is the chapter's implicit assumption and is nowhere tested.