In 1572 the queen mother of France arranged to have a man killed, and the assassin missed.
The target was Gaspard de Coligny, an admiral and one of the leading figures of the Huguenot community, who had become close to her son Charles IX. Catherine de' Médicis feared his growing influence on the young king, so she arranged for a member of the Guise family — one of the most powerful Catholic clans in France — to do it.1
The shot wounded him. It did not kill him. And Coligny, who knew perfectly well who counted him as an enemy, went to the king and named her.2
Greene's account of what followed is four clauses long and covers a civil war: "the failed assassination and the arguments that ensued from it set off a chain of events that led to a bloody civil war between Catholics and Protestants, culminating in the horrifying Massacre of St. Bartholomew's Eve, in which thousands of Protestants were killed."2
This is the chapter's Reversal — the case Greene prints to show what happens when the device fails.
The plan was not simply an assassination. It was an assassination arranged to be misread.
"Secretly, however, Catherine had another plan: She wanted the Huguenots to blame the Guises for killing one of their leaders, and to take revenge. With one blow, she would erase or injure two threatening rivals, Coligny and the Guise family."1
So the Guises are not accomplices being used for convenience. They are the second target, and the killing is the mechanism by which the first target's own faction is turned into the weapon that destroys them.
That is a genuinely elegant construction. Coligny dies. The Huguenots, correctly identifying the Guise assassin, retaliate against the Guises. Two rival power blocs damage each other, and the queen mother is not visible anywhere in it.
It is also, on inspection, three separate operations stacked in sequence, each of which has to succeed for the next to be available.
Count the dependencies, because the failure is not bad luck — it is a structure with no tolerance anywhere.
One: the assassin has to hit. A single shot, at a specific man, in 1572.
Two: the Guise involvement has to be discoverable enough that the Huguenots identify them, but not so discoverable that the arrangement behind it surfaces.
Three: the Huguenots have to retaliate at a scale that damages the Guises without producing a general war — which is to say, a faction inflamed by the murder of its leader has to be precisely as angry as required and no angrier.
The third condition is the one that should have killed the plan on paper, and Greene does not mention it at all. Catherine's design treats a religious faction's grief and fury as a controllable output. Nothing in the design has a governor on it.
What actually failed was the first condition, and the case is usually read as the assassin missed, therefore it went wrong. But conditions two and three were never going to hold either. The plan had one visible failure and two latent ones.
The specific mechanism of exposure is worth isolating, because it is not surveillance, confession, or evidence.
"knowing Catherine as his enemy, he strongly suspected it was she who had set up the attack on him, and he told the king so."2
Coligny had no proof. He had a wound and a list of people who wanted him dead, and he named the most plausible one, to the one person whose opinion mattered.
Which means the screen that Greene says hides the operator's involvement was never load-bearing in the first place. It was only ever protecting Catherine from confirmation, not from suspicion — and a survivor with access to the king does not need confirmation to do damage.
That is the structural lesson the Reversal contains and does not state. The cat's-paw and the scapegoat both assume the target is either dead or without standing. Coligny was alive, wounded, and closer to the king than she was.
Greene's stated moral is about being seen: "if at any moment the screen is lifted and you are seen as the manipulator, the puppet master, the whole dynamic turns around—your hand will be seen everywhere, and you will be blamed for misfortunes you may have had nothing to do with. Once the truth is revealed, events will snowball beyond your control."3
Look at what actually happened against that description.
The thing that went out of control was not Catherine's reputation. It was the mechanism she had built. She had constructed a device for converting one killing into inter-factional violence, and the device worked exactly as designed — a wounded Huguenot leader, an inflamed faction, armed Catholic clans, a court in crisis — except that it ran on past the point where anyone was steering it, and it ran to a massacre.
This is the property that makes an unwitting instrument different in kind from a subordinate. A subordinate can be stopped. A situation cannot. Catherine's cat's-paws were not the Guise assassin alone; they were the Huguenot community, whose predictable response was the engine of the whole plan. You cannot recall a faction.
Greene reads the outcome as exposure and its consequences. The case shows something more specific and more useful: the failure mode of a cat's-paw is not that you are discovered. It is that the thing you started does not stop.
There is a situation you want changed, and you have found a way to change it that involves nobody knowing you were involved.
The plan has several steps and you can see all of them. That clarity is the dangerous part, because a plan you can see the whole of feels like a plan you control.
Go back and mark each step by who executes it. Some steps are yours. Some are somebody else's, done for their own reasons. Now cross out every step of the second kind and read what is left.
What is left is the entirety of your control. Everything else is prediction.
Then ask the question the case is built around, and ask it about the last step rather than the first. If the person or group at the end of the chain reacts twice as strongly as you expect, what happens? Not will they — assume they do. Catherine's design needed the Huguenots angry enough to attack the Guises and not angry enough to start a war, and there was no version of the plan in which she could adjust that dial after the shot was fired.
And the third check, which is the one people skip because it feels like pessimism. Where is the off switch? If new information arrives on day four, what specifically do you do? If the honest answer is nothing, it's running, then you have not built a plan. You have started something, and the only question left is whether its natural end is somewhere you can live.
Greene's own advice here is unusually sound and unusually meek: "It is often wiser to use such dupes in more innocent endeavors, where mistakes or miscalculations will cause no serious harm."4 Which is to say: use this only where you can afford to have no control at all.
The chapter draws a warning about visibility. The case supports a warning about irreversibility, and the two are not the same.
If exposure were the real risk, then a better-concealed version of the same plan would have been safe. It would not have been. A plan whose middle steps are executed by an inflamed religious faction is not made safer by the queen being harder to identify — the massacre did not require anyone to know who arranged the shot.
There is a second lesson in the Reversal's position. Greene puts this case at the end of a chapter that has spent four pages recommending scapegoats and instruments, and the case he chooses to close on is one where the technique produces a mass killing. He then advises using it for small things.
The chapter's demonstrations are all large and its advice is small, and it never accounts for the gap.
There is one more thing the case exposes about the whole law.
Every successful instance in Law 26 — the commissariat chief, de Orco, Louis Howe, Cleopatra's Romans — involves an instrument that is a person or a defined body. One man, one lieutenant, one secretary, one general with an army under discipline.
Catherine's instrument was a population. The Guise family was the visible hand, but the plan's actual engine was the Huguenot community's predicted reaction, and communities do not have the property that makes instruments usable: bounded, legible, terminating response.
Greene does not distinguish these. The chapter treats someone who does the dirty work for you as a single category, and the category silently spans from a beheaded quartermaster to a religious faction of tens of thousands.
The technique's reliability collapses somewhere between those two, and the book contains no method for locating the boundary.
🚩 SINGLE SOURCE · 🚩 SECONDARY WITHOUT PRIMARY. The attribution of the Coligny assassination attempt to Catherine is one of the most contested questions in sixteenth-century French history, argued over for four hundred years and unresolved. Greene states her authorship, her secret second plan, and her intent as fact, without citation or qualification.1
🚩 The causal chain is compressed to the point of distortion. "The failed assassination and the arguments that ensued from it set off a chain of events" covers weeks of court crisis, competing factions, and decisions by multiple parties, culminating in a massacre whose responsibility is itself disputed.2 The compression makes a contested political catastrophe read as the consequence of one woman's failed scheme.
Tension with the chapter's own success cases. Cleopatra ran an almost identical structure — position a powerful third party so their own interests destroy your rivals — and it worked twice, across thirteen years, with two different Romans.5 Greene offers no variable distinguishing her success from Catherine's failure except the outcome itself. This is the sorting-by-outcome pattern the build has now logged at half a dozen laws.
Open question. If the real risk is irreversibility rather than exposure, then concealment is not the operator's main safeguard and the chapter's whole emphasis is misplaced. What would the safeguard be, and does any case in the book demonstrate one?
Set this beside the law's opening case and the chapter's two halves disagree about what the technique is for.
Ts'ao Ts'ao's scapegoat was defensive: an accusation was travelling toward him and he redirected it. The operation was complete within a day, involved one death, and terminated.
Catherine's was offensive and open-ended: nothing was travelling toward her, she initiated in order to remove two rivals, and the operation's completion depended on other people continuing to act for an indefinite period.
Greene's Reversal warns about a screen being lifted, which is a defensive concern — it is about being seen. But the case he uses to illustrate it is an offensive operation whose disaster had nothing to do with concealment. The warning is written for the first half of the chapter and demonstrated with the second, and the mismatch is why the Reversal's advice (use it for small things) reads as a non sequitur after the case it follows.
The vault's Law 5 material on the same territory sits at The Conspiracy Problem, and its finding — that the natural instinct to recruit co-conspirators is where most plots begin and why most fail — bears directly on what Catherine actually built.
Behavioral mechanics — The Conspiracy Problem
Greene's own argument there is about arithmetic. Someone inside your organisation needs to go, you cannot do it alone, and so the natural instinct is to find others who feel the same way — and that instinct is where most plots begin and most plots end, because every additional participant is another point of failure, another person who can talk, hesitate, or be turned.
Catherine's plan looks at first like the opposite of a conspiracy. She recruits almost nobody: a Guise, an assassin. The Huguenots are never told anything. That should make it more robust by the conspiracy page's own logic.
It is less robust, and understanding why produces the insight neither page reaches.
A conspirator is a point of failure who can be managed. An unwitting participant is a point of failure who cannot. You can watch a co-conspirator, reassure him, pay him, or remove him. You can do none of those things to a faction that does not know it is in your plan, because any contact with it would reveal the plan's existence.
So the conspiracy page's arithmetic needs a second term. The risk of a plot is not simply the number of people who know; it is the number of people whose behaviour the plan depends on — and unwitting dependants are strictly worse than witting ones, because they carry the same failure probability with none of the manageability.
Which inverts the usual reading of Catherine's design. It looks tight because so few people were told. It was in fact the loosest plan in the chapter, because more of it was being executed by people she could not speak to than in any other case in the law.
Psychology — Paranoia as Institutionalized Anxiety Displacement: The Root Mechanism
That page describes a trade a population makes with itself. Diffuse, nameless dread about an uncertain and threatening world is intolerable. Structured fear — of a specific enemy, with a face and a strategy and an intention — is more tolerable. So institutions organise the anxiety by supplying the enemy, and a psychological problem gets solved at the level of the nation.
France in 1572 was already carrying that load. Two confessional communities, a decade of intermittent war, a young king, and a court in which every faction believed the others were plotting — which they were.
Catherine's plan assumed she was introducing a stimulus into a system that would respond proportionately. What she was actually doing was supplying a named enemy to a population already saturated with unstructured dread, and the displacement page explains why the response was not proportionate to the stimulus at all.
The wounded Coligny did not add a grievance to a stable situation. He gave an existing, intolerable, free-floating fear a face and a date — and the relief of that conversion is enormous, which is exactly why the reaction outran anything the original injury could justify. Thousands died for an assassination attempt that failed.
Here is what the pair yields, and it generalises well past this case. The size of a reaction is set by the accumulated dread available to be discharged, not by the size of the provocation. Which means the operator's estimate — Catherine's, and every operator's — is systematically wrong in one direction, because he calibrates against the injury he is inflicting and the environment calibrates against everything it was already carrying. The more anxious the system, the more violently it converts a small trigger, and the less any of that is visible from the inside of a plan.
Sharpest implication. Catherine's disaster is filed under being discovered and is really about starting something with no off switch. The screen was never her protection — Coligny suspected her without evidence and told the king, which is what any survivor does. What destroyed her was that her plan's middle steps were assigned to a religious faction, and a faction is an instrument you can aim once and never recall. The property that makes an unwitting agent deniable is the same property that makes him uncorrectable, and the chapter recommends the first without pricing the second.
Generative questions