Chieh Hsüan starts from a sentence most commanders would rather not say out loud. "The military is not a good affair."1 It requires people to do violent, perverse, inhumane things, which means the men who are best at it are exactly the dangerous ones — the killers, the haters, the liars, the cold strategists. You cannot run an army without them. And here is the trap: "The talents that are advantageous are exactly those that can harm it."2 The very qualities that make a man useful in war are the qualities that make him a threat to his own side. The martial man must kill; the courageous man must hate; the wise man must deceive; the strategist must coldly forbear. Strip those out and you have no army. Keep them in and you have a camp full of people who are, by ordinary standards, bad.
Chieh Hsüan's answer to this is a single word he spends a whole meditation defining: control. "To employ the greedy and deceitful is termed 'control.'"3 Command is not the art of finding good men. It is the art of harnessing dangerous ones — taking the killer, the liar, the greedy thief, and aiming them at the enemy instead of at you.
This is a doctrine of how to lead people you cannot trust toward ends you do choose. Chieh Hsüan, writing late in the Chinese military tradition in the Ping-fa Pai-yen, lays it out as a definition, an original text, and an explanation.4 The core claim: the army cannot cast aside the martial, courageous, wise, or strategists, so it cannot abandon hatred, killing, deceit, and forbearance either. The skill of control is to "employ these capabilities but eliminate the evil aspects, harvest the profits but impede the harms." Do that, and "everything under Heaven will become a resource for the military. Enemies can be summoned, invaders pacified, brigands used, and the distant employed. It all lies in control, that's all."5
The phrase "brigands used" is the heart of it. You do not exile the brigand; you use him — as a brigand, pointed outward. Spy work falls squarely into this frame, because spying is itself one of the dishonorable, deceitful crafts the literati condemned. The reliable spy has to be either a coerced psychopath comfortable in vile circumstances, or an honorable person persuaded to do what they would otherwise abhor — and either way, "the only solution thus resides in motivation and control."6
Chieh Hsüan then names the modes. There are five ordinary ways to control men, each illustrated by a historical army:7
The list is a descending ladder. Virtue is the highest, ruling men by their devotion to a good ruler; authority is the lowest, ruling them by force and fear. "Although their employment differs, they are still unified in being able to manipulate human talent."8 Five different grips on the same dangerous material.
Beyond the five, Chieh Hsüan reaches for something subtler: "control in the realm of the formless."9 Two kinds. First, ch'i: "When the ruler's ch'i is overflowing, the men dare not have any rebellious intentions." Second, strategic power: "When the army's strategic power is flourishing, the warriors dare not have any traitorous thoughts."10 These are not techniques a commander applies; they are conditions of his complete achievement. When the general's ch'i is not flourishing, Chieh Hsüan says, he should not take command at all; when the army's strategic power is not flourishing, it should not be sent out.11 Control at this level is not something you do to the men. It is a state of the whole that makes treachery unthinkable before it is even suppressed.
Picture a stable of half-wild horses — powerful, fast, exactly the animals you need to win a race, and exactly the animals that will trample you if you stand wrong. You do not geld them into docility; a gelded warhorse is a slow one. You learn to ride them. Control is horsemanship for human talent: keep the speed, redirect the kick.
The five modes are five lengths of rein. Virtue rules a man through his love of a good ruler — the longest, lightest rein, requiring almost no force because the man wants what you want. Authority rules him through fear of punishment — the shortest, harshest rein, requiring constant pressure because the man would bolt the moment it slackened. Chieh Hsüan's ordering is a quiet judgment: the lower modes work, but they work because the man is restrained, not because he is aligned. Ts'ao Ts'ao's authority got results; it also produced an army held by force, which is an army that betrays you the instant the force lapses. The higher you climb the ladder, the less the dangerous talent needs watching, because the man's own devotion does the controlling for you.
The explanation carries a sharp warning against the cheapest grip of all. "If one simply gains their allegiance through material incentives but lacks deep plans and farsighted thoughts to control them, then the warriors of the Three Armies will either be arrogant or lax."12 Money alone is not control. Gold buys a man's presence, not his alignment; a man held only by pay grows arrogant when flush and lax when bored, and bolts to a higher bidder. This is why control sits above reward in the doctrine. Reward recruits the dangerous talent; control is what keeps the recruited talent pointed outward instead of inward.
The formless modes are the doctrine's ceiling. Ch'i and strategic power do not negotiate with a man's treachery; they evaporate it. A soldier in an army whose strategic power is "flourishing" does not weigh betrayal and decide against it — betrayal does not occur to him, because the whole shape of the situation makes it absurd. This is control that has become atmospheric. It is also the most demanding, because it cannot be installed; it can only be possessed, and a commander who lacks it, Chieh Hsüan says flatly, should not command.
This page hands the rest of the vault the master frame under which spy recruitment makes moral and practical sense. The selection and recruitment pages keep needing dangerous, dishonorable people — the liar, the thief, the curser — and this doctrine explains why that is not a contradiction: the army's whole art is using such people, "brigands used," talent harvested and harm impeded. It gives the reward-and-incentive page its limiting principle: incentives recruit, but control aligns, and money without control breeds arrogance and laxity. And it gives the agent-employment pages a taxonomy — five modes from virtue to authority — for how the deceitful are held once they are recruited.
Chieh Hsüan gives the principle; the Sung-dynasty Hsü Tung gives the exhaustive practice, and the two together complete the doctrine. Hsü Tung makes a point Sawyer calls astonishing for its bureaucratic age: the normal method of selection — prizing learning, literature, and disputation — automatically excludes the men with martial talent and the nerve for dangerous missions.13 The examination system filters out exactly the people the army most needs. So Hsü Tung proposes the opposite: recruit men of every conceivable talent, "however despicable," and assign each to the role his nature fits.14
The assignment list is the doctrine of control made concrete, and it is unsparing. "The fluent and loquacious should be employed to work as spies. Those good at filching things like rats and snatching things like dogs should be dispatched to steal the enemy's signals and investigate the enemy. Those who speak foully and curse a lot should be employed to slander, deprecate, and undermine the enemy."15 Read what is happening here. The thief is not punished for thieving; his thieving is aimed at the enemy's signal-flags. The foul-mouthed curser is not silenced; his cursing is aimed at the enemy's morale. The smooth talker becomes a spy. Each despised trait is matched to the mission where it becomes a weapon.
The matching runs both ways — it is also about not misemploying talent. "Those good at comforting others are not employed in the urgency of combat for fear that they will exhaust themselves but lack courage... The crafty should not be employed in decision-making for fear that they will be indecisive. The recklessly courageous should not be employed in planning against the enemy for fear that they will regard them too lightly."16 Control is precision casting. The crafty man, brilliant at scheming, is useless at decisive command because his craftiness makes him waver; put him where craft helps and decisiveness is not required. Hsü Tung even folds in the diviners and atmospheric officers — the men who read portents and prognosticate — with the instruction that they be kept from confusing each other, and that you esteem only the predictions in which they all concur.17 That last line is the corroboration method smuggled into the talent-assignment system: even your prognosticators are cross-checked.
The payoff is Hsü Tung's whole argument for control. "Although there are both good and evil among the masses in our army, no one will ever be discarded. Since there will thus be no disaffection, when the army moves, it will achieve its objectives."18 No one discarded — the evil talents included, each placed where his evil serves. That is "brigands used" turned into a personnel system.
A commander walks the line of new recruits with an officer beside him, and he is not looking for good men. He stops at a recruit with quick, darting eyes and asks what he was before the war; the answer is a thief, and the commander nods — signals, he tells the officer, send him after the enemy's flags and dispatches, his fingers are already trained. He stops at a loud, foul-mouthed man the other officers wanted to flog for his tongue; the commander listens to him curse and assigns him to the slander detail, to be marched within earshot of the enemy lines. He passes a smooth, fluent talker who could charm a magistrate and marks him for spy work. He passes a brilliant, scheming man and quietly notes him unfit for command — too clever to decide, keep him scheming where craft helps and no order must be given. He is not reforming any of these men. He is casting them, each into the role where the worst thing about him becomes the most useful thing about him, and when he is done not one recruit has been thrown away. That evening he checks his own bearing before he gives the day's orders, because he knows the men read his ch'i, and an army that senses a flagging commander starts, quietly, to think traitorous thoughts.
You are losing control when you have bought the army with money and nothing else. The men hold together while the pay flows and the campaign is easy; the moment funds tighten or boredom sets in they turn arrogant or lax, exactly as Chieh Hsüan warned, because pay was never alignment.19
You are losing control when you have discarded your dangerous talents instead of placing them — when the thief, the curser, the schemer have been flogged or exiled, and you find yourself in the field missing precisely the men who could have stolen the enemy's signals, broken his morale, or out-schemed his staff. A clean army is a weak one.
You are losing control when you have misemployed your talent — put the crafty man in command and watched him dither, put the comforter in the assault and watched him break, put the recklessly brave man in the war council and watched him underrate the enemy into disaster. The trait was good; the casting was wrong.
And you are losing control at the deepest level when your own ch'i has gone flat and the army can feel it. Treacherous thoughts that flourishing strategic power would have made unthinkable begin to surface; the men start to calculate. Chieh Hsüan's diagnosis is blunt — a general whose ch'i is not flourishing should not be in command, because the formless control that suppresses treachery before it forms is simply gone.20
The doctrine carries an optimism Sawyer flags as possibly unearned. He calls Chieh Hsüan's general solution of "control" the "final development of the tradition and perhaps an unfounded optimism."21 The worry is real: control assumes the commander can always harvest the profit and impede the harm, always keep the dangerous talent pointed outward. But the thief who steals the enemy's signals is still a thief; the day the campaign turns, what stops his fingers turning inward? Chieh Hsüan's confidence that evil aspects can be cleanly stripped from useful capabilities is exactly the confidence history keeps puncturing. The doctrine names the danger and then claims to have mastered it — and Sawyer will not fully grant that the mastery is real.
A second tension: the five modes are presented as a ranked ladder, virtue highest and authority lowest, yet the lowest mode produced two of history's most effective armies, Ssu-ma Yi's and Ts'ao Ts'ao's.22 If authority works, in what sense is virtue "higher"? The doctrine implies virtue produces a more durable alignment while authority produces only restraint — but it does not prove the durability claim, and the historical record of effective tyrannical armies presses against it. The ranking is moral as much as practical, and the source does not fully separate the two.
An open question the formless modes raise and never answer: ch'i and strategic power are said to make treachery unthinkable, but they are conditions you either possess or lack, not techniques you can build. What does a commander do when his ch'i is not flourishing, besides Chieh Hsüan's counsel of despair — don't take command? The doctrine's highest mode is also the one it offers no method for acquiring.
Sawyer reads Chieh Hsüan and Hsü Tung as a convergent pair across dynasties — Sung-era Hsü Tung supplying the exhaustive practice for the principle Chieh Hsüan would later distill.23 The convergence is genuine in substance: both insist the army must use despicable talent rather than purge it, and both treat command as casting rather than character-screening. But the framing as a continuous "tradition" reaching its "final development" in Chieh Hsüan is Sawyer's synthesis across a long span, and it carries his characteristic continuity instinct. Hsü Tung's concrete role-assignment and Chieh Hsüan's abstract five-modes-plus-formless meditation are doing different things — one is a personnel manual, the other a philosophy of command — and reading them as one maturing doctrine smooths a gap that may reflect distinct projects rather than a single line of descent. The shared core (use the deceitful, place the dangerous) is well-attested in both; the developmental arc that makes Chieh Hsüan the tradition's culmination is Sawyer's reading, and his judgment that it represents "perhaps an unfounded optimism" is itself a tell that he does not fully trust the synthesis he has assembled. The continuity should be checked against the texts' actual dates and aims before the ladder from Hsü Tung's practice to Chieh Hsüan's theory is taken as a real historical sequence.
Handshake with Compliance and Influence Architecture. Chieh Hsüan's five modes of control are a complete map of how one person bends another's behavior to a chosen end, and they line up against the behavioral-mechanics question — how do you actually get compliance? — almost mode for mode. Virtue and public spirit correspond to alignment-based influence, where the target wants what you want and needs no pressure; administration corresponds to systems-based compliance, rules and procedures that channel behavior; authority corresponds to coercive influence, compliance held only by fear of cost. The structural parallel is exact, but the insight emerges from the tension: behavioral-mechanics typically studies influence on a single target who can walk away, while Chieh Hsüan is controlling dangerous talent that is inside your own organization and armed. That changes the calculus. The behavioral-mechanics practitioner can afford to lose a non-compliant target; the commander cannot, because the uncontrolled killer in his own camp turns the weapon around. The insight neither domain generates alone is that the lowest, most coercive influence mode is the most dangerous precisely when the target is most useful — a held-by-fear elite soldier is exactly the man whose defection is catastrophic. Control doctrine thus supplies behavioral-mechanics with a missing variable: the cost of the target's defection, which should govern how high up the alignment ladder you must climb. You can rule a stranger by fear; you cannot safely rule your own best assassin that way.
Handshake with Agent Qualification and Selection. Selection asks who is fit to be an agent; control asks how the unfit-but-useful are held once chosen. Hsü Tung dissolves the boundary between them — his thief and his curser would fail any virtue-based selection and yet are the very men he assigns to steal signals and break morale. Holding the two pages together exposes a tension in the idea of "qualification" itself: the agent doctrine elsewhere prizes honorable, intelligent recruits, while Hsü Tung's system explicitly wants the despicable, "no one ever discarded." The insight is that selection and control are inverse operations on the same population. Selection screens out the dangerous; control screens them in and aims them. A doctrine that only selects produces a clean, weak roster missing the rat-fingered thief who alone can take the enemy's signals; a doctrine that only controls produces a camp of dangerous talent with no filter at all. The complete practice is the two in tension — select for the missions where character matters, control for the missions where only the despicable trait will do — and neither page, alone, sees that the army needs both gates running at once.
Sharpest implication. Command is not the art of assembling good men; it is the art of using bad ones — harvesting the profit of the killer, the liar, the thief while impeding the harm — and the entire competence of a commander reduces to whether he can keep dangerous talent pointed outward. The clean army is the weak army. The question is never whether your men are good; it is whether you control them.
Generative questions.