Chapter 6 The preservation theorem¶

The "Examination of the Four Truths" opens with the opponent's most comprehensive indictment: "If everything is empty, there is neither arising nor ceasing. Thus there can be no teaching of the Four Noble Truths." From there, the four fruits, the jewels of Saṃgha, Dharma, and Buddha, demerit and merit, cause and effect collapse all the way down.¹ Two thousand years later, the charge is still repeated in plainer words: you have talked tables and chairs out of existence. This chapter proves the theorem's other half, the preservation theorem, in two directions. First, no harm: T and the rejections it authorizes leave E1 and E2 entirely untouched. Second, only thus: E1 and E2 can stand only if T holds. Together, these say that emptiness is no subtraction. It is the condition for the world's functioning. By the end of this chapter's proof, the charge of nihilism will have been wholly reversed.

The preservation theorem, formally stated. Part (a), the direction of no harm. For any assertion A at the level of E1 or E2, T and the rejections it authorizes do not change A's truth conditions. The argument has two steps. The first is classificatory. The scope clause already puts E1 and E2 assertions outside its scope of application. Every rejection carries a type designation and applies only to D3-type propositions. The second is substantive, and it is the crucial one. Ordinary individuation has never used an E3 foundation. Chapter 4's two examples demonstrated this on the spot: the goer is individuated by the activity of going, and a condition is retrospectively identified by its effect. The network of relations suffices for its own operation. Remove the foundation and every knot in the net still bears its load. The foundation is an idle component. Removing it changes the operation of no component in use.

Direction (a) must directly confront a classical opposing principle, the pivot of Sarvāstivāda semantics, later summarized in four Chinese characters, 假必依实: "The nominal must depend on the real." Designated existence can be designated only on the basis of the existence of real dharmas. Chapter 2 cited its criterion, and Essay 15, Chapter 7 devoted a chapter to sorting out its regressive form. If this principle holds, removing the E3 foundation changes precisely the conditions under which E2 assertions hold, and the claim of no harm fails. One step answers it. Grounding is an activity: providing a basis, supporting, sustaining, producing are all jobs performed among conditions. Yet the real thing the opponent offers pledges neither to change with conditions nor to be produced by them. That it cannot enter into causation or serve as a condition follows directly from its definition. Direction (b) below will turn on this very point. What is inert cannot ground anything. Even if such a foundation were posited, it could not take on this job. What has always actually done the individuating and supporting is causes, conditions, parts, and conventions, each dependently arisen, each at the level of E2.

The claim "has never used" thus also receives its universal warrant. We have not counted every case and found no exception; the two examples were only demonstrations in the first place. Rather, as a matter of concepts, it cannot be called upon to do the job. By the contrapositive in direction (b), anything doing work lacks intrinsic nature. "The nominal must depend on the real" is left with only two readings. Read as requiring a real foundation with intrinsic nature, the principle calls for a ground whose own pledge deprives it of every capacity to take on a task. This reading defeats itself. Read as saying that the nominal must be established through various conditions, it is dependent origination, the mechanism this essay itself uses. This step draws its warrant from consequences of the definition D2, not from the main theorem. Direction (a) does not borrow the main theorem's guarantee. Its independence remains intact.

Part (b), the direction of only thus. If a dharma has intrinsic nature, then by D2's second and third conditions it neither arises, ceases, nor changes. By D2's first condition and Chapter 3's difference alternative, it stands in no relation to other dharmas. Taken together, these four points make a bearer of intrinsic nature wholly inert in function: it cannot be produced, undergo change, enter into causation, or serve as a condition. Taking the contrapositive gives this result: whatever functions, whatever arises, ceases, changes, or enters into relations, lacks intrinsic nature. The original states this half in two verses, positive and negative: "Because emptiness holds, all dharmas can be established. If emptiness did not hold, none could be established."² "If there were determinate nature, the world's various forms would neither arise nor cease, but abide forever without decay." Qingmu's commentary adds the empirical aspect: we directly see all things showing change, arising, ceasing, and becoming otherwise, so we know there is no determinate nature. "If there were no emptiness, the unattained could not be attained, afflictions could not be abandoned, nor could suffering be brought to an end." Practice and liberation are among what is preserved. Emptiness upholds supramundane and worldly dharmas alike.³

Reversing the reductio. Now settle the charge of nihilism. Define nihilism as denying that the ordinary level holds. Test both parties. The theory of emptiness: direction (a) is proved, and not a single item of E1 or E2 is lost. The farmer sows, the traveller goes on a journey, and the practitioner abandons afflictions, each in place. The theory of intrinsic nature: direction (b) is proved, and everything freezes. Without arising and ceasing, no farming; without change, no practice; without relations, no cause and effect. Who destroys the ordinary? Derivation gives the answer, not attitude. At this point in the original exchange, an exact structural symmetry deserves to be set out as it stands. The opponent opens with "If everything is empty, there is neither arising nor ceasing. Thus there can be no teaching of the Four Noble Truths." Nāgārjuna returns it in kind: "If everything is nonempty, there is no arising or ceasing. Thus there can be no teaching of the Four Noble Truths."⁴ The same verse, one character changed and thrown back. The opponent's inference holds only on the opponent's own reading: only if emptiness is read as annihilating nonexistence does it entail no arising or ceasing. On the reading distinguished by types, the antecedent that really entails no arising or ceasing is intrinsic nature. The indictment has the right contents and the wrong addressee.

The verdict too comes with the original: "You say I cling to emptiness and find fault with me for it. The fault you now allege is absent from emptiness." "The fault is now your own, yet you turn it upon me, like a person riding a horse who forgets what he is riding."⁵ The horse-rider analogy follows the same line as Chapter 5's exhibit. Every step the accuser takes, advancing a thesis, inferring, speaking, listening, rides on the horse of dependent origination. He uses everything that emptiness preserves for him to bring a case against emptiness.

The verse of three identifications, read three ways. "Dharmas born of the various causes and conditions I declare to be nonexistence. They are also nominal designation and also the meaning of the middle way. Never has there been a dharma not born of causes and conditions. Hence there is no dharma that is not empty."⁶ One verse gives three clauses of the preservation theorem. Reading one, emptiness: whatever is dependently arisen lacks intrinsic nature. This rejects every E3 reading and is the theorem's rejecting side. Reading two, nominal designation: what is dependently arisen is established through individuation within a network of language and names. This is the positive mechanism of E2. "Nominal designation" is no term of disparagement but the name of a manner of individuation. The sense of designation is E2, and it was already waiting here when Chapter 1 set out the definitions. Reading three, the middle way: free of the two extremes of existence and nonexistence, it rejects both real existence and annihilating nonexistence. What stands between these two rejections is E2 itself. "The middle way" names two rejections together with one level preserved, not a third metaphysical position. Qingmu's commentary ends with just this point: "Since these dharmas have no nature, they cannot be said to exist; since there is no emptiness either, they cannot be said not to exist."

The place of the two truths. "The Buddhas rely on two truths to teach the Dharma to living beings: first, conventional truth; second, ultimate truth." "Without relying on conventional truth, one cannot attain the ultimate; without attaining the ultimate, one cannot attain nirvāṇa."⁷ This essay assigns the two truths their formal places. Conventional truth is the legitimacy of assertions at the levels of E1 and E2; ultimate truth is T itself. The two truths distinguish levels of assertion, not two worlds. There is one world and two levels of discourse. One convergence across schools deserves recording here. The Sarvāstivāda's own definition of conventional truth says that "to say, in accordance with conventional principles, that pots and the like exist is true, not false," as Chapter 2 quoted. Neither party will pronounce the ordinary level false. They differ only over what remains at the ultimate level. The Sarvāstivāda keep an inventory of what survives decomposition and analysis; this essay's ultimate level retains only a metatheorem.

One more sentence from Qingmu's commentary must be quoted separately, since it brings Chapter 5's inventory of resources down to earth: "The ultimate always depends on speech. Speech is conventional. Therefore, without relying on convention, the ultimate cannot be spoken of." The meta level is expressed in conventional language. Conceptual grammar and logic are themselves conventional tools. This is no flaw but a design: the ultimate must be expressed starting from convention. Disputes among schools over readings of the two truths are the subject of the next essay. Here we only mark the point without developing it.

Corollary: a verdict does not transmit identity. Once the intrinsic-nature assertions about two terms a and b have each been rejected, exactly one thing follows: at the level of intrinsic nature, both their identity and their difference are rejected. The difference set up through intrinsic nature, a ≠ b, requires each side to stand without depending on the other, and it has been rejected along with π; the merging of two bearers of intrinsic nature into one is rejected as well, by the identity horn of the dilemma in Chapter 3. Two things do not follow. First, that the two are without difference at the level of designation: difference at that level is ≠, not ≠, and direction (a) of this chapter does not allow the rejections to alter it. Second, that a = b: Leibniz's law carries properties from identity, and the reverse inference, from a shared property to identity, is not valid. Besides, no property is shared here. Pure negation does not commit to a complementary class, so what the rejections give to a and b is only two verdicts. If one insists on reading "lacking intrinsic nature" as a predicate, then by T-strong it is true of every term in the domain; its extension is the whole domain, and it distinguishes nothing. To pass from joint rejection to nondifference takes two substitutions. The first writes the rejection as an internal negation, ¬(a ≠ b), which brings the rejected presupposition back onto the scene and violates Lemma 1. The second erases the star and reads ≠ as ≠; at the level of designation, ¬(a ≠ b) is equivalent in classical logic to a = b, and this step violates direction (a). Neither step belongs to the derivation. The eight negations in the treatise's verse of homage, "neither identical nor different," state this corollary in the original: at the ultimate level, "not different" is paired with "not identical" from the first verse.⁸

What this chapter delivers. Both directions of the preservation theorem are proved. The record of the reversal of the charge of nihilism includes the original's symmetrical verses and horse-rider analogy. The verse of three identifications combines the theorem's three clauses in a classical formulation. The two truths receive their formal places, with one instance of convergence across schools. One corollary follows: a verdict does not transmit identity, and the opening verse's "neither identical nor different" states it in the original. The theorem's full form has now emerged: it rejects every combination of real existence and intrinsic nature while upholding the whole of ordinary life and practice. Chapter 7 conducts the self-audit; Chapter 8 takes the theorem beyond its domain to test it; Chapter 9 gives the verdict.

A restatement in plain language follows the same rule as before: change only the language, add no content, and let the first half govern any discrepancy.

This time the opponent's indictment is complete. If you say everything is empty, there is no arising or ceasing; without arising and ceasing, no Four Truths; without the Four Truths, no practice or attainment of its fruits. Then there is no Saṃgha, no Dharma, no Buddha, and all good and evil, cause and effect are scrapped. In today's words: you have talked the world out of existence.

The answer has two steps. First, this judging machine processes only one kind of claim, the claim that something is solidly real and established on its own. Tables, chairs, electrons, crops, and practice are all left untouched. Besides, Chapter 4 showed that these things never stood on that mysterious foundation in the first place. Walking identifies the walker; an effect retrospectively identifies its condition. The network of relations suffices on its own. Remove the idle foundation and nothing at work changes.

There is an old doctrine to answer directly here, though. The opponent's semantics has an ancestral rule: the nominal must rest on the real. Things registered under names need something solidly real underneath to prop them up. Chapter 2 quoted the criterion behind this rule, and Essay 15, Chapter 7 spent a whole chapter on its version of questions that keep chasing one another. If it holds, pulling out the foundation changes the rules for registration at the ordinary level, and the claim of no harm cannot stand. One reply suffices. Propping things up is work: providing a basis, supporting, connecting, producing are all jobs. Yet the definition of the real thing the opponent puts forward fixes it as neither changing with conditions nor being produced by them. It follows that it neither enters into causation nor serves as a condition. The second step below will turn on this very point. Something that does no work cannot prop anything up. Even if such a foundation were really posited, it could not take the job. What has actually been doing the work underneath has always been causes, conditions, parts, and conventions, all dependently arisen, all at the ordinary level.

So the assurance behind "has never used" does not come from counting. We have not counted all the cases in the world and failed to see it clock in. Chapter 4's two examples were only demonstrations in the first place. By definition it cannot take this job. Whatever is doing the work, by the reversed inference in the second step below, lacks intrinsic nature. The ancestral rule has only two readings left. If the nominal must rest on something real with intrinsic nature, the prop summoned to do the work is dismissed by its own definition. This reading collapses on its own. If the nominal must stand through various conditions, that is dependent origination, exactly the mechanism this chapter describes. This step uses only consequences already contained in the definition of intrinsic nature. It makes no use of the main theorem, so the independence of the no-harm half remains.

The second step is harsher. Give everything intrinsic nature and the world freezes on the spot. Something with intrinsic nature neither arises, ceases, changes, nor relates to anything else. This follows directly from the definition. A seed is forever a seed and cannot sprout; a sick person is forever sick and cannot recover; an ordinary person is forever ordinary and cannot attain the path. The original's two verses, positive and negative, settle the matter: because there is emptiness, everything can be established; without emptiness, nothing can. If all things really had determinate nature, gods, humans, animals, everything would neither arise nor cease, abiding forever without decay. Yet the world before our eyes is plainly arising, ceasing, and changing.

Then who, exactly, is the nihilist? Those who speak of emptiness preserve every item of ordinary life. Those who speak of real existence freeze it all to death. The original has a fine exchange here. The opponent begins by reciting that if everything is empty, there is no arising or ceasing and no Four Truths. Nāgārjuna changes one character in the same verse and recites it back: if everything is nonempty, then there is no arising or ceasing and no Four Truths. The indictment has the right contents and the wrong addressee. The verdict is ready too: the fault is your own, yet you turn around and pin it on me, like someone riding a horse who forgets he is on horseback. Our accuser needs inference to advance a thesis, language to speak, and progress to practise. Each of these rides on the horse of dependent origination. He uses all the possessions that emptiness has preserved for him to prosecute emptiness.

The most famous verse has three statements, and they are this chapter's three clauses. Of something put together by causes and conditions, calling it empty means it lacks a foundation that establishes itself. Calling it nominally designated means it stands through registration in a network of names. Nominal designation is no counterfeit; it is a way of registering things. Calling this the middle way means taking neither end, neither solidly real existence nor utter nonexistence. Reject both ends, and what stands between them is the ordinary world itself.

Finally comes the arrangement of two levels of discourse. At the ordinary level, keep speaking, and what is true stays true. A pot is a pot, truly and not falsely. Even the Sarvāstivāda say this themselves, so the two sides have no case against each other here. At the ultimate level, only one statement remains, the theorem. Two levels are not two worlds. They are two levels of speaking about the same world. And what is said upstairs must use the language downstairs. The original makes this plain: the ultimate always depends on speech, and speech is conventional. The ladder must stand downstairs. This is no defect but a design. Other readings of these two levels and the history of their disputes are the subject of the next essay. Here we stop at marking the point.

There is one more corollary. Suppose the claims that two things each have an entity of their own are both rejected. Exactly one thing follows: at the level of entities, saying that they are one thing and saying that they are two things each standing on its own are equally untenable. Two things do not follow. One is that they have no difference at the ordinary level: this judging machine does not touch differences at the ordinary level at all. The other is that they are one and the same thing: the rejections hand these two things only two verdicts and put no shared label on them. If one insists on treating "having no entity" as a label, it sticks to everything, and using it to tell things apart tells nothing apart. To get from both claims rejected to no difference, one has to swap things twice. First, the rejection is restated as a negation, that they do not differ in entity, which invites the two things back to be compared. Then this negation is moved to the ordinary level, as if they did not differ in everyday terms either, and having no difference at the ordinary level is just being the same thing. Neither swap is derived. The eight negations in the verse with which Nāgārjuna opens the treatise, "neither identical nor different," say exactly this.

The theorem's full form is now visible. On one side it rejects every claim of something solidly real that establishes itself as a thing in its own right. On the other, it upholds the ordinary world together with practice and liberation. There is also a corollary: rejecting both sides does not make the two sides one. The next chapter checks for flaws; Chapter 8 tests the theorem beyond its domain; Chapter 9 gives the verdict.

References¶