Most of Greene's advice is about what to do to people. This one is about what to stop feeling.
Most important of all, do not take pity. Do not enmesh yourself in trying to help. The infector will remain unchanged, but you will be unhinged.1
Do not take pity. Not do not act on pity — do not have it.
And then, at the end of the chapter, the sentence that removes every exit:
This law admits of no reversal. Its application is universal. There is nothing to be gained by associating with those who infect you with their misery… Ignore this law at your peril.2
Six of the forty-eight laws declare themselves exceptionless. This is the one where an exception is most obviously required, and it is the one Greene defends hardest.
Underneath the coldness is a genuine epistemic instrument, and it deserves separating from the tone:
The answer lies in judging people on the effects they have on the world and not on the reasons they give for their problems.1
That is a rule about what evidence to admit, and it is well-aimed at a real difficulty.
An infector's reasons are usually good ones. Greene concedes it: they "often present themselves as victims, making it difficult, at first, to see their miseries as self-inflicted."3 Each individual misfortune has a coherent, checkable, sympathetic account attached.
So auditing explanations one at a time can never resolve anything, because each audit comes back clean. The rule says: stop scoring the explanations and score the outcomes.
Greene lists what not to do:
When you suspect you are in the presence of an infector, don't argue, don't try to help, don't pass the person on to your friends, or you will become enmeshed.4
Don't argue — because argument is engagement, and this is Law 9's finding arriving in a new context.
Don't try to help — because helping is where identification starts.
Don't pass the person on to your friends — and this is the one nobody obeys, because it does not feel like exposure. Handing someone along is the polite exit. It is also, in the chapter's own metaphor, transmission: you have not ended your contact with the pattern, you have routed it through your network and will meet it again from the other side.
Here is the part worth thinking about, because the instruction is stranger than it looks.
Pity does not cause identification. You can pity someone without taking their problems on as your own — that is roughly what a competent clinician does all day.
So why forbid the feeling rather than the merger?
Because pity argues for the proportionate response. A person who feels pity looks for a way to stay in contact and limit exposure: help with a specific thing, decline the rest, hold a boundary. That is the natural output of caring about someone while seeing clearly.
Greene's instruction is not proportionate. It is removal. And removal cannot survive the presence of pity, because pity keeps generating alternatives to it.
The prohibition on the feeling exists to protect the size of the intervention.
Put the mechanism and the instruction side by side and they do not fit.
The diagnosis, from the chapter's own analysis of Lola Montez's lovers: "Once the lover identified with them, he was lost." The failure point is identification — taking the pattern on as your own.
The instruction: remove the person.
But if the danger is identification, the indicated response is to refuse the identification, not to eliminate the contact. Those are different interventions with very different costs, and the second is only necessary if the first is impossible.
Greene never argues that it is impossible. He asserts it — "Lola Montez could not be helped" — about the person, when the relevant question is whether the lover could have held a boundary.
The chapter has one variable, proximity, and one setting, off. Everything the vault's psychology corpus knows about limited engagement, boundary-setting and sustainable care is outside its range, because it was never modelled.
It would be easy to dismiss this, and that would be wrong.
The rule about effects over reasons is genuinely useful, and it is useful precisely in the case where sympathy is warranted and the pattern is still real. Both can be true. A person can be sincerely suffering, accurate about their circumstances, deserving of help, and reliably leave the people who help them worse off.
Most advice cannot hold those together. It resolves toward they're struggling, help them or they're toxic, leave. Greene's rule keeps the two apart: the reasons may be entirely valid, and the aggregate is still the aggregate.
That is a real contribution. The failure is what he does with it — he treats a diagnostic as a verdict, and jumps from this pattern is real to therefore excise the person without an intervening step.
You have concluded that a pattern around someone is real, and you are deciding what to do.
First, separate the finding from the sentence. Establishing that a pattern exists tells you to change something. It does not tell you what.
Second, ask what you would have to stop doing, before asking who you would have to stop seeing. If the capture mechanism is identification, the smallest effective intervention is to stop taking the problems on — stop explaining them to others, stop planning around them, stop treating their crises as your workload.
Third, run the boundary before the removal, and give it a deadline. Try the proportionate version for a defined period and check the aggregate again. If the pattern continues while you are no longer identified with it, you have learned something the total intervention would never have told you.
Fourth, honour the third prohibition. Whatever you decide, do not solve your own exposure by making an introduction.
Fifth, keep the pity. It is the thing arguing for the smaller intervention, and the smaller intervention is right more often than this chapter allows. If you eventually need the larger one, pity will not have stopped you — it will only have made you check first.
And the honest note: sometimes the answer really is to leave, and no amount of boundary-setting fixes it. This law is right about that case. It is wrong to present that case as all of them, and wrong to declare the matter closed.
Strongest support. The effects-over-reasons rule is a genuine instrument, correctly aimed at a situation where explanation-auditing structurally fails. The three prohibitions are specific, and the third is a real insight about transmission through networks.
Tension — the instruction does not follow from the diagnosis. Identification is named as the capture point; removal is prescribed. Documented above.
Tension — the exemption is unusable. Greene's own opening exempts those "brought down by circumstances beyond their control," who "deserve all the help and sympathy we can give them" — and then supplies a rule that admits no reasons, which is the only evidence by which that group could be identified. The exemption and the method are incompatible, four paragraphs apart.
Tension — the no-reversal claim is refuted in the same chapter's own margin. Musashi, printed inside Law 10, describes infecting an opponent deliberately — "you can infect them with a carefree, drunklike spirit, with boredom, or even weakness." A law that has an offensive form printed in its own gutter is not a law without a reverse.
Tension — against the vault's psychology corpus. Do not take pity sits in direct opposition to most of the vault's material on repair, boundaries and sustainable care, which treats limited engagement as the skill and total withdrawal as the failure mode. Preserved, not resolved.
🚩 SINGLE SOURCE — the clinical framing (quarantine, disease, virus) is metaphor performing the function of a mechanism, with no clinical literature cited anywhere in the chapter. [POPULAR SOURCE]
Open questions.
The instruction sits against The Enemy as Untapped Resource at Law 2, and the pair exposes what Greene's real sorting criterion is.
Law 2 has Talleyrand recruiting Fouché, the man who had tried to have him assassinated, and Kissinger holding a private Saturday meeting with three of his own would-be kidnappers. The instruction there is that a proven enemy is workable material — that hostility is a resource if you are cold enough to use it.
Law 10 says that a person who is merely unlucky and unstable must be excised entirely, with no attempt at engagement, and that the law has no exceptions.
So the book will keep an assassin and cut off a depressive.
The variable is not danger — the would-be kidnappers were literally more dangerous. It is usefulness. An enemy has capabilities, a network, and motives you can trade with. An infector, in Greene's account, offers nothing.
Which makes the coldness of Law 10 less a moral position than an accounting one, and clarifies the whole corpus: people in this book are sorted by what can be extracted from them, and the only ones who get cut off completely are the ones with nothing to extract. Greene never states that criterion, and it explains the two chapters better than either explains itself.
→ psychology · The Infector
The psychology page carries the mechanism — identification as the capture point — and the five diagnostic signs.
Setting the two together shows a gap that neither shows alone: the diagnosis is about a process and the instruction is about a person.
Identification is something you do. It has stages, it is partly voluntary, and it is the kind of thing that can be interrupted. The five signs, by contrast, classify them — a fixed property, assessed once, producing a permanent verdict.
The insight neither page produces alone: Law 10 diagnoses a behaviour and sentences an identity, and the switch happens silently between the Interpretation and the Keys to Power. That is why the chapter has no graduated response available — you cannot partially excise a category. Had it stayed with the process it named, the natural output would have been a set of practices, and the vault would be filing this under boundary-setting rather than under avoidance.
→ history · Lola Montez and the Chain of Ruin
The case page lays out the five men and the pattern — and, crucially, the confounds: Dujarier died of a duel he provoked while drunk, Ludwig abdicated in 1848 along with much of Europe, Heald died in a boating accident.
That matters here more than on the case page itself, because this is the page that turns the pattern into an instruction.
The evidence supports there is a real pattern around this person. The instruction is therefore excise, feel nothing, and treat the matter as closed. The gap between those is where the confounds live — and an instruction with no exceptions has no room to be wrong about one of the five links.
The insight the pair produces: the strength of aggregate evidence and the strength of the response it licenses are not the same quantity, and this law conflates them. Aggregation is exactly the right tool for detecting the pattern; it is a poor basis for a total, irreversible, exceptionless action, because it cannot tell you which links were real. Greene's method correctly identifies that something is happening around Lola Montez and provides no way to establish how much of it is her — and then forbids the pity that would have made you ask.
Sharpest implication. The prohibition on pity is the most revealing line in the chapter, because pity is not the mechanism of capture — identification is. Greene has to forbid the feeling because the feeling reliably produces a better strategy than the one he is recommending: a person who pities looks for the smallest effective intervention, and this law has only the largest one in stock. Whenever a piece of advice needs you to stop feeling something in order to work, the thing to examine is not the feeling.
Generative questions.