Abstract

Sunzi's counsel of caution and SAE's Theorem One overlap in conclusion, and the two criteria ask after different objects. Sunzi asks after the properties of the prospective war itself — what losses it will cause, whether they can be recovered, how much remains afterwards, how long it will take. Theorem One asks after the state of the relation before war begins — whether the non-war roads are still open.

From this: any factor that alters the properties of a prospective war may alter the output of Sunzi's criterion, and none alters the output of Theorem One — because such factors reopen no road that has been closed. This volume does not claim that technology makes war reversible, and makes no empirical claim about the effects of technology.

Nine Grounds' precept keep the officers and men in ignorance touches the ground of humanity is an end on one's own side and must be criticised. The verdict is absence rather than dismantling — within the materials examined, no prior self-limitation on one's own side is found available to be dismantled. And on the mechanism of that absence, this volume's original universal proposition is overturned by Mohism; the revised form is a hypothesis about institutional stability, and must be kept apart from the source of norms.

Theorem One's threshold condition — that attacking-the-plan and breaking-alliances be exhausted — depends for its determination on execution-layer knowledge, which SAE places outside its range. This dependency had not previously been registered. And one earlier determination is withdrawn: Sunzi never ranked mou above gui; that ordering was SAE's own taxonomy projected onto his text.

1. The Conclusions Overlap

Attack by Fire, closing passage — the place in the Sunzi that most resembles a self-limitation: no ruler should put troops into the field merely to gratify his own spleen; no general should fight a battle simply out of pique. If it is to your advantage, make a forward move; if not, stay where you are. Anger may in time change to gladness; vexation may be succeeded by content. But a kingdom that has once been destroyed can never come again into being; nor can the dead ever be brought back to life. Hence the enlightened ruler is heedful, and the good general full of caution. Calculations opens in the same direction: the art of war is of vital importance to the state, a matter of life and death, a road either to safety or to ruin.

SAE's Theorem One agrees with this conclusion. War may not be begun out of anger or humiliation — §13 makes this hard, excluding wars of honour and wars of revenge from the cannot-not. And the overlap is real, not a similarity of phrasing: both reject the same class of reason for war, those arising from a subjective state rather than from a situation. Both put it in the most conspicuous place — Sunzi at the close of the book, SAE in the first exclusion of Theorem One.

Where conclusions agree and criteria differ in kind, the two will diverge in identical circumstances — and finding those circumstances is what this volume does.

2. First, a More Convenient Contrast Must Be Excluded

The convenient contrast is that Sunzi's criterion is a mutually compensating weighing and SAE's a threshold. That contrast is unavailable, because it needs something Sunzi does not give. He supplies no weighting structure; a synthetic judgement need not be a compensable weighted sum; an irrecoverable loss need not figure merely as a weight.

So this volume adopts an axis that requires no such premise: what does each criterion ask after. The argument depends on no assumption about compensation, weights, or formulae — only on what three passages each ask after, which a critic can check passage by passage.

3. The Two Criteria Ask After Different Objects

Attack by Fire's reason for caution — anger may in time change to gladness, but a kingdom that has once been destroyed can never come again into being — asks: what losses will this prospective war cause, and can those losses be recovered. Attack by Stratagem's taking a state whole is best asks the other face of the same class: how much remains usable afterwards. Waging War asks the same class again: how long this campaign will take, how much it will consume, whether it will bring in third parties. All three ask after one object: the properties of the prospective war itself.

Theorem One's two conditions are not of that class: has the carrier of cultivation been lost; can cultivation in its mild forms no longer continue. These two ask: before war begins, what state is that relation in — are the non-war roads still open. And the domain must be stated, because this is very easily read as a judgement about the whole Sunzi. It is not. The book contains a great deal asking after other things — form, force and vacuity ask after the balance of strength; Employment of Spies asks after channels of information; and the four rungs from the highest form of generalship is to baulk the enemy's plans downward ask after the available sequence of means, which stands closer to the state of a relation than to the properties of a war. This proposition claims only what each of the three passages asks after.

Not one of the three asks whether other roads are still open. And among Theorem One's two conditions, not one asks how great the cost of this war will be.

4. What Follows: Technology Can Loosen One and Cannot Loosen the Other

Any factor that alters the properties of a prospective war may alter the output of Sunzi's criterion; and such factors do not alter the output of Theorem One. The argument is analytic: Theorem One asks after the state of the non-war roads, and a technology that makes war cheaper, faster or more precise reopens no road that has been closed, and closes no road that remains open. It alters the side of the war, and Theorem One does not take its value from that side. So advances in military technology can, structurally, loosen Sunzi's class of criterion, and cannot loosen Theorem One.

What this does not claim must be stated. It does not claim that technology makes war reversible — death remains irreversible, and the losses to the attacked party's life, state and relations may remain irreversible; this volume makes no such claim and does not need it. It does not claim that precision and stand-off generally shorten wars — that is an empirical proposition, and this volume has no material for it. It does not claim that Sunzi's criterion will loosen to any specified degree — how much it moves is empirical; this volume states only that the direction of movement is fixed by what the criterion asks after.

And this is not a difference between strict and lax. Sunzi's can be extremely strict: give a kingdom destroyed can never come again into being enough weight and the requirement of caution can rise almost to a prohibition on war. Theorem One's can be extremely lax: read cultivation in its mild forms can no longer continue broadly enough. The difference lies elsewhere — the two criteria change their verdicts when different things change. Sunzi's changes when the properties of the war change; Theorem One's when the state of the relation changes. And in practice these two kinds of change are not synchronised: technology can alter the properties of war enormously while the relation is untouched, and a relation can change utterly while technology stands still. So the two will diverge, and the moment of divergence can be stated in advance.

5. The Strongest Objection, and That Cancelling Operation

The objection: the li — advantage — in if it is to your advantage, make a forward move is not cost. Sunzi's li is a synthetic judgement. Waging War is not about casualties but about the structure of national exhaustion — a thousand pieces of gold a day; long campaigns leave the state's resources insufficient; blunted weapons and dulled ardour, strength exhausted and treasure spent, and then the other feudal lords will take advantage of your extremity to act. Third-party intervention and the international situation are matters at the level of relations — so assigning Sunzi's criterion wholly to the side of the properties of a war is too clean. The objection holds, and it is stronger than the version that merely says li is multi-item.

The response: the other lords will take advantage of your extremity is a judgement about what this war will bring on — what happens if it drags, what happens when strength is spent, whether third parties will move in. It asks after the consequence-structure of the prospective war. Theorem One asks what is already there in the relation before this war occurs. And the distinction can be drawn by one operation: cancel this war from the supposition, and Sunzi's questions have no object — if there is no campaign, there is no long campaign, no blunted weapons, no lords taking advantage. Cancel this war from the supposition, and Theorem One's two questions still have objects — whether those roads are open, whether the carrier of cultivation has been lost, hold equally in the case of not fighting, and it is precisely they that decide whether to fight.

One part of the objection must be granted: Sunzi's criterion does contain inputs at the level of relations. But they enter as components of this war's consequences, not as the prior state of a condition for going to war. The difference is not verbal — one and the same fact, that a certain lord is unfriendly, figures in Sunzi as he will take advantage if fighting starts, and in Theorem One as has the road of peaceful settlement thereby been closed.

6. The Tension Is Already Marked Within the Text

The closing passage of Attack by Fire in fact juxtaposes two different arguments. The irreversibility argument: anger may change to gladness, a kingdom destroyed can never come again into being. The advantage argument: if it is to your advantage, make a forward move; if not, stay where you are. The two are of different character. The former says a certain class of loss cannot be recovered and therefore may not be driven by what can be; the latter says move if it is advantageous. If the advantage is great enough, the latter permits war, while the former confers no veto — it requires caution, not prohibition. The effect of the juxtaposition is that the irreversibility argument no longer constitutes a veto on its own.

This step must be stated precisely, because it very easily slides back into the excluded model. To say that irrecoverability becomes a very weighty consideration is still the language of weighing. All this volume can say is a point about the text: the former sentence requires caution and does not require prohibition; the latter permits movement where there is advantage. Juxtaposed, the former does not acquire the standing of a veto — and that is read off the modality of the two sentences, without assuming any structure of weighing.

And the tension was not imported by an outside reader; it was felt within the transmission of the text. Giles notes here that if it is to your advantage, make a forward move duplicates Nine Grounds §17 and judges it an interpolation, on the ground that by sense anger may change to gladness should follow immediately on no ruler should put troops into the field to gratify his spleen. This volume need neither accept nor reject that textual judgement. What is worth noting is only that a commentator reading the passage line by line felt the two sentences to be discordant, and felt it strongly enough to propose a textual hypothesis.

7. Acknowledging the Cost: Questionability Is Not Computability

A difference is not a superiority, and one point tells against SAE. Theorem One's practical binding force depends on whether its two conditions can be questioned — and whether the carrier of cultivation has been lost, and whether cultivation in its mild forms can no longer continue, have no settled way of being observed. They require judgement, and judgement can err and can be faked. A criterion that puts checkable items on the surface may in practice bind more tightly than one that is hard to verify.

This must be stated precisely, because it slides easily back into the calculation model. To say Sunzi's li can at least be computed by staff officers, and the resulting figures checked by others presupposes that his criterion has a computable form, and he gives no formula. What can be said is weaker: the items Sunzi's criterion asks after — how long, at what daily cost, in what state of morale, with what attitude among the other lords — several of them have relatively settled ways of being observed, so disputes about them are more readily localised. Whereas can cultivation in its mild forms no longer continue has no comparably settled way of being observed: who determines that a road is closed? What shows that it was open before? This is a difference in questionability, not in computability.

So the difference established here is a difference of object, not of stringency and not of reliability. Reliability is a separate question, turning on mechanisms of verification.

8. A Parallel Conclusion: Taking a State Whole

Attack by Stratagem: in the practical art of war, the best thing of all is to take the enemy's country whole and intact; to shatter and destroy it is not so good. So too it is better to capture an army entire than to destroy it… hence to fight and conquer in all your battles is not supreme excellence; supreme excellence consists in breaking the enemy's resistance without fighting. The sequence runs down to the smallest tactical unit, and the criterion throughout is one: whole is better than shattered.

Sunzi's reason is the usability that comes with wholeness. A country taken whole is still usable — its people, its produce, its walls, its fields are all there; shattered, there is less of it. Waging War has the direct counterpart: the enemy's spoils are used to augment one's own strength. A qualification must be entered, because this is easily read too narrowly: calling it a pure cost argument would be too strong — what quan preserves is not only convertible resources but structures of governance, populations, and the subsequent capacity to control the place. Nor is not-fighting a zero-cost limit: breaking the enemy's resistance without fighting requires investment in attacking-the-plan and breaking-alliances, requires the accumulation of positional advantage, requires the other side to accept. What this volume uses is the weaker step: Sunzi's preference for quan rests on there being a difference in usability between whole and shattered, and that difference can vary.

SAE's reason differs: war is itself an interruption of cultivation, so what is achieved without interruption is better in principle. This has nothing to do with how much remains usable afterwards. So: Sunzi's preference for quan rests on the state of affairs after this war is over; SAE's preference for not fighting rests on the interruption that war itself constitutes. The two ask after different objects, in parallel with the earlier finding — and any factor that alters how much remains afterwards may alter Sunzi's preference, while none alters SAE's: an interruption does not cease to be an interruption because much remains.

A possible objection: breaking the enemy's resistance without fighting exceeds cost calculation — it names a highest good, not an account. That reading has something to it, but the text will not support taking it so far, and the reason lies in the sequence. It does not appear in isolation; it is the apex of the preceding whole/shatter sequence, and every rung of that sequence asks the same thing: how much remains afterwards. If the apex asked something else, it would not be of a kind with the rest and the sequence would break. And the order Sunzi gives immediately after — baulk his plans, prevent the junction of his forces, attack his army, besiege cities — is justified in the same passage by how long a siege takes, what it costs, and that a third of the men are slain while the town still stands. What that reason asks after is still the properties of this war.

9. One Passage That Must Be Criticised

Nine Grounds: it is the business of a general to be quiet and thus ensure secrecy; upright and just, and thus maintain order. He must be able to mystify his officers and men by false reports and appearances, and thus keep them in total ignorance. And later in the same chapter: confront them with the deed itself; never let them know your design. The commentators are of one mind and are explicit. Cao Cao: the troops may share in the joy of the outcome, but not in the laying of the plans. Du Mu: the army knows only the orders to advance or retreat; it does not know the advantage to be gained by attack. Zhang Yu: war is the way of deception — and it is applied not only to the enemy, but there are also those who use the way of deception upon their own troops, making them follow without letting them know. Zhang Yu's note is decisive: deception directed at one's own side is not an interpretation brought in from outside but a position the commentarial tradition states of itself.

The practice advocated in Nine Grounds, in so far as it cancels the risk-bearer's standing to know his own situation and the character of his action, touches the ground of humanity is an end on one's own side, and must be criticised. The scope is limited, because this is easily written more broadly than the text supports. What the text plainly supports is the item cancelling the standing to know. What it does not plainly support, and this volume does not claim: that the practice also cancels the position of refusal — Nine Grounds does not discuss whether soldiers may refuse; that it also cancels the position of raising objection; that it covers every rank and every circumstance. Those may be offered as inferences, but must be marked as such and not written into the proposition.

Not falling within this proposition: withholding the timing and route of a march from outsiders; not making a specific point of attack public for a limited period; keeping particular sources of intelligence secret. These do not cancel an executor's knowledge of his own situation or of the character of his action. Falling within it: making an executor unable to know what kind of thing he is doing, or what situation he has been placed in. What Nine Grounds advocates is the latter — keep them in total ignorance, confront them with the deed; never let them know your design — which is not the keeping of one item of intelligence but the cancellation of the standing to know as such. And one inference, marked as an inference: a person who does not know what situation he is in, or what kind of thing he is doing, is in fact also poorly placed to raise a substantive objection or a reasoned refusal.

10. Three Things to Be Said Together

First, this precept is at the level of principle, not of technique. Keep them in total ignorance stands in Nine Grounds alongside quiet and thus secret, upright and just and thus orderly. It is not an incidental tactical suggestion but a specification of what kind of person a general is — calm, inscrutable, upright, orderly, and able to keep his own men in the dark. Four items in parallel, with no shift of register between them.

Second, it cannot be excised from his system. Same chapter: place your army in deadly peril and it will survive; plunge it into desperate straits and it will come off in safety. Placing an army where it cannot retreat, in order to make it fight to the death, is one of the central operations of Sunzi's doctrine of shi — and by the mechanism Sunzi himself gives, it is conditioned on the men not knowing the whole picture. No psychological universal may be added here: to say that an army which knows it is being deliberately placed in deadly ground will mutiny or scatter has no warrant — how an informed unit would react is an empirical question, and history is not short of cases of men who knew and went anyway. What this volume uses is a textual step: Sunzi himself wrote keeping them ignorant as the condition of this practice. Remove it and the mechanism he himself supplies no longer holds.

Third, the verdict is absence, not dismantling. By this series' discipline of determination, dismantling requires first showing that a self-limitation existed in that tradition to be dismantled.

11. The Evidence for Absence

Within the materials examined here — the Sunzi with its commentators, and the Sima Fa's Benevolence as the Root — no prior self-limitation on the side of one's own party is found available to be dismantled. The narrowing matters: to write the Chinese military tradition has no prior self-limitation on its own side would be a universal about a whole tradition, while this volume examined two books. The Wuzi, the Weiliaozi, the Six Secret Teachings, the Three Strategies and the later literature were not examined.

Evidence one: the self-limitations of the Sima Fa all fall on the enemy-facing side. Not to violate the seasons, not to march when the people are afflicted — this is to love one's own people; not to attack a state in mourning, not to exploit a natural disaster — this is to love the enemy's people; not to raise armies in winter or summer — this is to love the people of both. In antiquity, pursuit did not exceed a hundred paces, the following of a withdrawal did not exceed three stages; not to press the incapable, and to pity the wounded and sick; to form ranks and only then to drum. Six items, all of them regulating how the enemy is to be treated. Of these, not to violate the seasons, not to march when the people are afflicted does mention one's own side — but what it regulates is not raising armies in the farming season or during epidemics, a matter of not consuming one's own resources and strength, and it does not touch the soldier's standing to know or to judge.

Evidence two: the commentarial tradition reads the Sima Fa as a precedent for Sunzi's precept, not as its opposite. Cao Cao, commenting on Nine Grounds, cites the Sima Fa's on seeing the enemy, make the oath, and Zhang Yu reduces it to: your dispositions should not be divulged in advance. Within this commentarial tradition, the Sima Fa runs in the same direction as Sunzi on this point. Conclusion: within the materials examined, Sunzi is judged zero, not negative. He dismantled nothing; there was nothing on this axis to dismantle.

12. The Mechanism of the Absence, and That Rewriting

Having determined absence, one must ask why there is nothing on this axis. The original proposition: constraints on the enemy-facing side and constraints on one's own side have different mechanisms of generation. Enemy-facing, constraints can be generated by reciprocal equilibrium — that one does not strike a state in mourning, does not exploit disaster, can arise and persist because the other side observes it too and a violator will meet the same treatment next time. This requires no party to hold support at the level of principle: so long as the parties coexist over time, can identify violations, and are able to impose a cost next time, such constraints will arise of themselves and stabilise. On one's own side there is no counterparty to enforce — who redeems the soldier's standing to know, and who makes a commander pay after violating it? The soldier cannot punish the commander's deception, so deception is not reduced by being punished.

A compatible historical observation: constraints of this kind are more conspicuous in the Spring and Autumn period than in the Warring States — the lords of the earlier period coexisted over long periods, fought repeatedly, and knew one another, while the annexationist wars of the later period aimed at the elimination of the adversary, and the premise of repeated play weakened. The character of that paragraph must be stated: it is an explanatory hypothesis, not an established historical mechanism. The collapse of ritual restraint in the Warring States admits several explanations, and this volume has no material for excluding the others.

The original universal form of this proposition has been overturned. The counterexample: the Mohists were a transnational organisation whose internal discipline was enforced by no state's military law, and the Mozi's Lu's Questions records Mozi recalling his student Sheng Chuo from the service of Xiangzi Niu, who had thrice invaded Lu, on the ground that his salary outweighed his righteousness. This is a self-limitation on one's own side, and its support is not reciprocity. Mohism is a counterexample, not an addition. The revised form: self-limitations on the enemy-facing side can be stabilised by reciprocity; self-limitations on one's own side cannot depend on the other party's retaliation, and must be stabilised otherwise — by internal rights, procedures, membership structures, or institutional review. Reciprocity is not the only mechanism by which constraints are generated.

Two qualifications, so that the revision is not read too broadly. First, voluntary membership alone is far from sufficient — a whole bundle is required: rules acknowledged as constitutive of the community; an internal authority or review mechanism; real consequences for violation; exit a real and not merely nominal possibility; and a channel of appeal and reinterpretation. (Five candidate conditions generalised from a single sample, untested for necessity or sufficiency.) Second, why a norm holds and how a norm comes to be stably enforced within an organisation must not be run together. Humanity is an end does not require retaliation to hold, but it will not, by its normative standing alone, produce organisational discipline; conversely an office of Grand Master can stabilise discipline, which does not show that discipline to be grounded.

After the revision, one corollary is unaffected: a constraint hung on reciprocity is effective only on the face where reciprocity obtains; a criterion hung on humanity is an end takes effect without requiring the other party to be capable of retaliation. That is SAE's practical reason for placing its criterion on the ground rather than on custom — custom disappears where reciprocity disappears, and on one's own side there was never any reciprocity available. A constraint that takes effect only when the other party can retaliate is empty in every relation of gross inequality, and the internal relations of one's own party in war are the most unequal of all. And the five conditions all fail in a conscript army — membership is not voluntary, the cost of exit is death, there is no internal review, no channel of appeal, and the rules of military law are not acknowledged by the members but promulgated by the commander. So on the relation Sunzi treats, the conditions of generation remain absent.

13. A Dependency Registered

Theorem One's threshold condition — that attacking-the-plan and breaking-alliances be exhausted — depends for its determination on execution-layer knowledge. To determine whether that condition is met, one must know what options those two comprise; and their concrete practice — how to disrupt the other side's designs, how to break up his alliances, what methods are available, under what conditions each is feasible — falls precisely within what this volume's scope note places outside SAE's range.

This does not refute Theorem One, but it stands in tension with §4's the layers do not compete. The threshold condition itself is given by principle — have the other roads been walked is a question of principle; whether it is met must be certified by execution-layer knowledge. A layer of principle may declare that it does not do the execution layer. It may not declare that it does not depend on one.

The practical consequence is not abstract. An agent who does not know what methods attacking-the-plan and breaking-alliances comprise cannot judge whether he has exhausted them. He will conclude there is no road while roads remain — and preventing exactly that is what Theorem One is for. A sincere agent of limited judgement may, in perfect good faith, cross a threshold he should not have crossed, simply because he did not know there was another way. Theorem One cannot prevent this by itself. This dependency had not previously been registered; whether to feed it back into the treatise, this volume does not answer.

14. One Correction, and a Load-Bearing Member

In earlier work this series read Sunzi as ranking mou above gui, and on that basis held that he possessed, in the hierarchy of means, a self-limitation oriented toward 15DD. That reading does not hold and is withdrawn here. The grounds: first, Attack by Stratagem's sequence ranks objects of attack — his plans, his alliances, his army, his cities — and the reason for the ranking is the properties of the war. Second, Calculations' all warfare is based on deception states a mode of operation, and the position in which it appears is decisive: the first chapter, as a general proposition, followed immediately by concrete practices — feign incapacity when able, appear far when near. Third, a general proposition applies to every rung of the sequence, attacking-the-plan included; indeed the commonest way of disrupting an opponent's plan is to make him believe something false. So mou and gui are two orthogonal axes and do not form a ranked sequence. That sequence is SAE's own four-rung taxonomy projected onto his text.

The methodological consequence must be recorded alongside. To project one's own taxonomy onto another's text, and then discover there a tendency oriented toward one's own position, is a self-confirming operation — it guarantees that the observer will always find himself in others. The credibility of such findings is zero, because they cannot fail. The error persisted through several rounds and was found only on returning to the text; and what found it was not finer reasoning but setting out the original position and context of two sentences.

And a downstream consequence: part of a load-bearing member's support has moved. With Sunzi ranked mou above gui withdrawn, the question arises whether SAE's four-rung hierarchy of means has support independent of Sunzi — its legitimacy previously depended in part on the attribution that Sunzi had already drawn a comparable distinction, and that attribution is now withdrawn. Two volumes of this series rest on it: volume three, whose recoil mechanism and first test question issue from §12; and volume seven, whose matching of Gandhi's satyagraha rests on one rung of the hierarchy. Each carries a registration; and the re-argument of the hierarchy itself belongs to §12 of the treatise, not to this series. This series can point out the gap; it cannot fill it.

15. Left Unresolved in This Volume

Whether the registered dependency should be fed back into the treatise is undecided. If it is, the tension with the layers do not compete must be handled; if it is not, an account is owed of how a user of Theorem One is to determine exhausted without commanding execution-layer knowledge — and no such account exists. Related to this is the question of verifiability: Theorem One's binding force depends on whether its two conditions can be questioned, and neither has a settled way of being observed. The two are two faces of one thing — Theorem One needs something it does not itself supply.

The revised form of the proposition about generation has the support of only one strand of the Chinese tradition. Until at least two more strands have been run, it is used as a proposition awaiting verification — German military thought, Islamic siyar jurisprudence and Indian political thought have not been examined.

Finally, an inference this volume cannot verify: the high frequency of borrowed-ground self-limitations on one's own side — those supported by religious command, ritual, or codes of honour — may have a structural explanation: that face has no reciprocity available, and can only borrow a ground from elsewhere. Unverified.