Generative Effacement in the Intimate Relation - An Ontological Injustice and the Justice of Non-Determination 【（Preliminary）Draft】
ENGLISH
Philosophy of Intimacy and the Theory of Justice · Paper XVIII
Generative Effacement in the Intimate Relation
An Ontological Injustice and the Justice of Non-Determination
[ Working Draft ]
On a Harm That Gives Rather Than Takes, and the Room One Must Leave for the Other to Generate
Wanhong
Working draft — not for citation or circulation
生而不有，为而不恃，长而不宰。
To give life yet not possess; to act yet not presume;
to nurture growth yet not rule over it.— Laozi, Daodejing, ch. 51
万物并作，吾以观复。
The ten thousand things rise together;
I watch only their return.— Laozi, Daodejing, ch. 16
致她——
For her—
the girl who belongs to the forest,
whose own rhythm I would never overwrite;
for whom I am learning, slowly,
to leave the open ground in which she grows.致我最爱的人。 To the one I love most.
Abstract
This paper isolates and names a form of injustice that wrongs neither what a subject has, nor how she is recognised, nor whether she can be heard, but her being—her standing as a generative being who continually constitutes herself. When intimate life is understood as a field of co-generation, in which value resides in neither party but is generated between them, a distinctive harm becomes possible: one subject, generating too fast and too far ahead, may come to occupy the shared generative space so completely that the other loses the capacity to generate at all. The result is not that she has less but that she ceases, within the relation, to be. The paper calls this generative effacement and argues that it is an ontological injustice, irreducible to the distributive, recognitive, or constitutive wrongs with which it might be confused. Its defining and most troubling feature is that it arrives not through domination but through devotion: there is no appropriation, no coercion, no failure to hear, and the effaced party may sincerely endorse and even cherish the one who effaces her.
The harm is given an exact formal criterion. Modelling two intimate subjects as coupled phase oscillators, the paper shows that effacement is a genuine phase transition: the weaker pole loses its own rhythm precisely when the coupling strength exceeds a threshold fixed by the difference of the two intrinsic frequencies, while the depth of effacement, once that threshold is crossed, is set by the asymmetry of force between them. The two structures are independent, and their separation yields the central result—that even under perfectly equal force the weaker pole is effaced by half once the threshold is crossed. Equal power does not protect it; strong coupling between unlike rhythms is sufficient to silence the slower. The good of intimacy is shown to lie not in the strength of coupling but in whether it leaves each subject turning at its own frequency.
From this follows a justice of counter-intuitive form. Because the threshold depends on the coupling and not on the stronger subject’s speed, the remedy is not for the stronger to generate less—deceleration only re-imposes its pace—but to decouple: to release the traction by which it overwrites the other’s rhythm. The paper develops this as generative reticence, a negative and structural obligation to refrain from determining the domain in which the other generates, specifies its practice as the relinquishment of three determining channels (of the shared goal, of meaning, and of the very problems on which the relation works), and grounds it ethically in an account that extends the duty to treat another as an end from rational to generative agency, while forbidding the ethic its own paternalistic application. Justice between two generative beings, it concludes, requires not the transfer of value but the non-foreclosure of the other’s capacity to generate.
Keywords: generative effacement; ontological injustice; constitutive injustice; generative reticence; phase capture; coupling; non-determination; generative justice; care ethics.
§1 The Phenomenon and the Paradox
Consider two people who love each other, and one of whom is, in the ordinary worldly sense, the stronger—the quicker to act, the more fluent in plans, the one to whom solutions occur first and arrangements come easily. He loves her, and his love takes the form his strength makes natural: he provides, he foresees, he smooths the road before she reaches the rough part of it, he settles the difficulty before she has finished framing it. Nothing he does is to her detriment in any register the usual theories can name. He takes nothing from her; he deceives her in nothing; he honours her, listens to her, would account to her for anything she asked. By every available measure the relation prospers, and she, asked, would say she is happy and is not lying.
And yet something is wrong, and the wrongness is hers. Somewhere in the smooth provision a space has closed. The decisions that were to be made were made before she came to them; the meanings that were to be found were named before she found them; the road was paved before she could choose its direction. She is not oppressed—there is no oppressor—and she is not deprived, for she wants for nothing the relation can supply. She is something for which the language of justice has no ready word: she has been generated for, so thoroughly and so well, that there is nothing left for her to generate, and a self that lived by generating has quietly gone still. This paper is written to give that stillness a name and a theory.
§1.1 The datum
The phenomenon, stated plainly, is this. In a shared life understood as a field of co-generation—a coupling out of which value arises between two people rather than being brought to it ready-made by either—one party may generate so far ahead of the other, so fast and so completely, that the other finds no remaining occasion for her own generation. The harm is not that the fast-generating party produces badly, nor that he produces for himself; he produces well, and for the relation, and often most of all for her. The harm is that the space of generation is shared and finite, and that a sufficiently rapid and total generation by one pole leaves the other pole’s generativity with nowhere to act. What withers is not the other’s holdings, which may grow, nor her standing, which may be high, but her activity—the exercise, now without occasion, of her own capacity to make value at her own pace.
It is essential to the datum that the withering is not chosen as a deprivation and not experienced, at first, as one. The space does not close against resistance; it closes around a gratitude. Each particular act of provision is a kindness, received as a kindness; it is only their sum, over time, that amounts to an occupation. This is why the harm is so often invisible to both parties at once: the stronger pole sees only a sequence of devotions, each defensible, each loving; the weaker pole feels only a slow and unaccountable diminishment of her own activity, with no event to point to and no one to blame. There is no scene of the crime, because there is no crime—only a good, done too completely, in a space that could not hold all of it without displacing her.
§1.2 The paradox that defeats existing theory
The datum is built so as to defeat, one by one, every criterion the theory of justice ordinarily brings to a relation. This is not an accident of the example but the point of it: the phenomenon is interesting precisely because it falls into the blind spot common to all the standard diagnoses.
There is no appropriation. The stronger pole takes nothing from the weaker; on the contrary, he gives. The vocabulary of exploitation, of extraction, of one party living on another’s unrecognised labour, finds nothing to grip: the flow of value runs toward the effaced party, not away from her.
There is no domination. The stronger pole issues no command, brooks no resistance because he meets none, imposes nothing she refuses. He serves. The vocabulary of coercion and subjection, of a will bent to another’s, finds nothing to grip either: there is no contest of wills in which one prevails, only a devotion that forecloses by filling rather than by forcing.
There is no failure to recognise or to hear. The stronger pole sees her, values her, attends to her, would credit her every contribution. The vocabulary of misrecognition and of epistemic injustice finds nothing to grip: she is not unseen, not unheard, not denied her standing as a knower. She is acknowledged in full—and effaced in full—at once.
Claim — The paradox of devotion. Generative effacement is a harm that arrives through devotion and survives the absence of every standard wrong. There is no appropriation, for the stronger pole gives rather than takes; no domination, for he serves rather than commands; no failure of recognition or of hearing, for he honours and attends. Every criterion of injustice built upon possession, coercion, or non-acknowledgement reports that nothing is wrong—and the harm is nonetheless real, and is the weaker pole’s. A theory adequate to it must therefore find the wrong at a level beneath possession, coercion, and acknowledgement altogether.
§1.3 Subordination completed in sincere endorsement
One move remains, and it is the one that closes every exit. It might be supposed that the harm, though it survives the absence of the standard wrongs, would at least announce itself in the effaced party’s own dissent—that she, whatever the outside observer misses, would feel and could voice her diminishment, so that her objection might serve as the criterion her circumstances otherwise lack. The datum denies even this. The effaced party frequently does not object; she endorses. She is grateful for the provision, attached to the provider, and would, asked whether she would have it otherwise, sincerely say no. Her subordination is completed not in false consciousness imposed from without but in an endorsement that is, by every test available to her, authentically her own.
This feature—that the harm survives even the sincere endorsement of the one wronged—sets the problem the rest of the paper must solve. A wrong that survives not only the absence of the standard wrongs but the sincere endorsement of the one wronged cannot be located in anything the will refuses, for her will does not refuse it—it embraces it. The harm must lie somewhere the will does not reach: at the level not of what she chooses but of what she is, not of her endorsement but of her being. To show that there is such a level, that an injustice can be done there, and that it can be done even by love and even with consent—and then to say what justice, at that level, would require—is the work this paper undertakes. It begins, in the next section, by locating the level: by showing where, beneath the injustices the theory of justice already knows, the ground of being lies.
§2 The Justice Typology and the Descent to Ontology
The phenomenon of §1 is invisible to the standing theories of justice, and the section argues that the invisibility is systematic rather than incidental—a horizon of the field as a whole and not a gap in any one theory. The contemporary theory of justice is organised, this section contends, around an unstated ontology of the subject as a bearer: a locus that has holdings, occupies a status, commands intelligibility, and can be formed. Every injustice the field knows is an injury to something the subject bears. The wrong this paper isolates is an injury not to anything borne but to the bearing—to the subject’s activity of being a subject—and it is invisible precisely because the field’s implicit ontology has nowhere to put an injury there. The claim, then, is not modest. It is that the theory of justice has a floor, that the floor is set one storey too high, and that a class of real wrongs has been falling through the gap the misplaced floor conceals. The section establishes this against the two most developed maps of justice’s depth—Fraser’s and Honneth’s—and then installs the missing layer and defends the metaphysics that licenses it.
§2.1 Against the reigning maps of depth
The most serious contemporary attempts to say how deep injustice goes are Nancy Fraser’s and Axel Honneth’s, and the paper’s claim is sharpened by saying exactly where it parts from each.
Fraser argues that justice has two irreducible dimensions, redistribution and recognition, to which she later adds a third, political dimension of representation; her “perspectival dualism” treats distribution and recognition as two analytical perspectives on any social arrangement, neither reducible to the other. Honneth presses the opposite, monistic thesis: recognition is the single fundamental category, and distribution a derivative of it, so that a sufficiently differentiated theory of recognition can ground justice entire. They divide over whether justice has one ground or two; they agree, decisively for our purposes, on the kind of thing injustice injures. For both, the injured object is the subject’s standing—what she is taken to be, how she is regarded, the social bases of her self-relation. Fraser’s misrecognised subject and Honneth’s disrespected subject are alike subjects whose regard has failed; the quarrel is only over whether failed regard reduces to, or stands beside, failed distribution.
This is the agreement the present paper breaks. Both maps, however deep they reach, terminate at the subject’s standing—and standing, however damaged, presupposes a subject who stands. Neither map has a category for an injury to the subject’s continued existence as a generative agent, because both take that existence for granted as the substrate their injuries operate upon. Honneth comes nearest, and the nearness is instructive. He makes love—the recognition proper to primary relationships—the structural core of ethical life, the sphere in which basic self-confidence is formed. One might think that a theory placing love at the foundation would catch a harm done by love. It does not, and cannot, for a reason its own structure dictates: Honneth’s love is recognition that builds—it confers the self-confidence on which all later autonomy rests—and his pathology is always recognition withheld or distorted. He has no place for a love that, fully and undistortedly given, effaces by the very completeness of its giving. The harm this paper names is committed in the presence of perfect Honnethian recognition: the effaced party is loved, esteemed, accorded her rights, confident in herself—and effaced regardless.
Claim — The shared horizon of the recognition theorists. Fraser’s dualism and Honneth’s monism, for all their opposition, share a horizon: both locate injustice, at its furthest reach, at the level of the subject’s standing—her distribution and her recognition—and both presuppose, beneath that level, a subject whose existence is not itself at issue. Neither can name a wrong done to the subject’s continued existence as a generative agent, and Honneth’s recognition-in-love, which might seem the exception, is precisely the recognition in whose full presence the wrong of effacement is still done. The recognition theorists mapped the standing of the subject to its floor; they did not reach the being beneath the standing.
§2.2 The stratification, and where it stops
Set the established categories out as what they are: not a list but a stratification, ordered by the depth at which each injures the subject, each layer reaching past the one above toward the subject herself.
Distributive injustice injures holdings—divisible goods, allocable and reallocable; its remedy is transfer, and what it cannot see is any wrong surviving a correct distribution. Because it is shallowest it is also imperialist: wrongs at every deeper layer are perennially misdescribed in its vocabulary of shares, which is why effacement, too, will tempt redescription as a maldistribution (a temptation §3 must resist).
Recognitive injustice injures standing—a good not distributable, since respect cannot be transferred as a holding can. Here is the first descent: a wrong that correct distribution leaves untouched, the wrong of the misrecognised who may be materially provided for and degraded all the same. Rawls already marks the pressure toward this layer from within the distributive paradigm itself, when he names the social bases of self-respect “perhaps the most important primary good”—a holding that is not quite a holding, a distributive good whose content is recognition, the distributive paradigm straining at its own edge.
Epistemic injustice injures intelligibility—the subject in her capacity as a contributor to shared sense, her testimony discounted, her interpretive resources starved. This descends again: one may be well-provided and well-regarded and still be unable to make one’s sense count.
Constitutive injustice injures the conditions of formation—the developmental and relational conditions under which a subject comes to be a subject at all. This is the furthest layer the field has reached, and the first that is genuinely ontological: what it injures is not anything the subject bears but her coming-to-bear.
Claim — The field terminates at formation. The established categories form a single descent—holdings, standing, intelligibility, the conditions of formation—each detecting a wrong that survives the repair of the layer above it. But the descent halts at formation: the furthest wrong the field can name is done at the threshold of subjecthood, to one not yet a subject. There is no category for a wrong done, beneath formation, to the continued being of a subject already formed. The very monotonicity of the descent is the evidence that the floor is incomplete, for the series points past the storey on which the field has stopped.
§2.3 The descent to ontology, and its price
The missing layer is reached by one thesis: that for a relational subject, existence is an activity and not a state. The field’s implicit ontology pictures the subject as a substance that endures and bears properties, on which picture the gravest injury available is to prevent the substance from forming—hence the floor at constitutive injustice. The subject of intimate life, on this series’ account, is not such a substance. Its relational being is an activity: the continual self-constitution by which it generates itself as a generative subject in coupling with another. It does not first exist and then generate; its generating is the form its existence takes.
This thesis is the paper’s largest commitment, and intellectual honesty requires stating its price rather than smuggling it in behind a metaphor. The objection that must be met is direct: is the identity of being with generation an argument, or merely a stipulation congenial to the conclusion? A critic will say that “to be is to generate” is a definition chosen because it yields the wanted result, and that one could as easily hold the commonsense view—that the effaced subject plainly continues to exist (she breathes, she persists, she is the same person), and has merely been made unhappy or idle. On the commonsense view there is no ontological harm, only a distributive or eudaimonic one, and the whole edifice falls.
The reply is not to insist on the identity but to show what each side must pay for it. The commonsense view purchases its ease at a cost it does not advertise: it must hold that a subject’s being is independent of her activity, that she would be the same being whether generating freely or reduced to the inert reception of another’s generation—that, in the limit, a subject permanently stilled, permanently the mere site at which another’s value is deposited, has lost nothing of her being and only something of her contentment. This the series denies, and denies for a reason given long before the present paper needed it: that a relational subject is constituted in the activity of relating, so that the cessation of that activity is not a change in the subject’s circumstances but a change in the subject. The identity of being with generation is not a free stipulation; it is the entailment of a relational ontology argued elsewhere on independent grounds, and its price—the rejection of the substance-subject who persists unchanged beneath all activity—is one the series has paid in full from its first paper. What the present section adds is only the observation that justice has not yet been made to reckon with that ontology’s consequence: that if being is activity, then activity can be unjustly foreclosed, and the foreclosure is an injury to being.
Claim — The descent licensed by identity, not analogy. Beneath constitutive injustice lies a final layer, ontological injustice, injuring a relational subject’s being-as-generation. The descent is licensed by an identity—for a relational subject, to be is to go on generating—and the identity is not a stipulation but the entailment of the relational ontology the series argues independently. Its price is the denial of the substance-subject who would persist unchanged beneath a foreclosed activity; a theory unwilling to pay that price keeps the commonsense subject and must redescribe effacement as mere discontent, at the cost of calling just a relation in which one party has been reduced to the inert site of another’s generation. The layer is distinct in register from the constitutive: constitutive injustice wrongs the formation of a not-yet-subject; ontological injustice wrongs the continued existence of a subject already formed.
§2.4 The typology, as a formal table
The descent can now be set out formally, and the table is itself an argument: its final column gives, for each layer, the reason it is irreducible to the layer above, so that read downward the table is a proof that the layers are distinct and the series cannot be collapsed into distribution.
| Layer | Injury / locus | Detected by | Why not reducible to the layer above |
|---|---|---|---|
| Distributive | holdings; what one has | maldistribution of a divisible good | (base layer) |
| Recognitive | status; what one is taken to be | misrecognition, denial of standing | standing is not divisible; one may be justly provided for and degraded in regard (Rawls’s “social bases of self-respect” already strains the distributive paradigm) |
| Epistemic | intelligibility; whether one can be heard | testimonial / hermeneutical silencing | sense-making is not a conferred status; one may be fully respected and still unable to make one’s sense count |
| Constitutive | conditions of formation; whether one can come to be | degradation of the conditions of subject-formation | formation is prior to intelligibility; one may inhabit a rich epistemic order and still be barred, upstream, from forming as its subject |
| Ontological | being-as-generation; whether one continues to generate | capture of the subject’s own generative frequency (§4) | continuation is past formation; one may be formed, provided, regarded, and heard—and still stopped from continuing to generate as oneself |
The locus column moves monotonically inward—has, is-taken-to-be, can-make-known, can-come-to-be, continues-to-be—and the inward movement is the descent made visible. The right-hand column shows that at each step the lower category detects a wrong that persists when the upper category’s injury is repaired: provision does not cure misrecognition; recognition does not cure silencing; audibility does not cure foreclosed formation; and—the step this paper adds—formation, with all the rest, does not cure foreclosed continuation. Each layer is the residue the layer above leaves uncured, and that residue is what makes the series irreducible.
§2.5 The irreducibility of the bottom layer
The bottom row’s irreducibility must be demonstrated and not merely tabulated, for it is the paper’s claim to novelty. The demonstration has the form the table announces: in effacement, each of the four shallower criteria is satisfied-as-absent, and the harm remains. Aggregate value rises and the effaced party holds no less—no distributive wrong. She is regarded, esteemed, accorded her rights, confident in herself—no recognitive wrong, in any of Honneth’s three spheres. She is heard with full attention, her contributions credited—no epistemic wrong. She is a mature subject whose formation was long complete—no constitutive wrong. All four diagnostic conditions are absent; the harm is present, real, and grave. A wrong surviving the joint absence of all four cannot, as a matter of logic, be a compound of them.
Claim — Irreducibility. Generative effacement is irreducible to the distributive, recognitive, epistemic, and constitutive injustices above it. The proof is that it is realised where all four shallower criteria are satisfied-as-absent—rising holdings, full recognition across all three Honnethian spheres, full audibility, completed formation—and the harm persists; were it a compound of the four it could not survive the absence of all four. The theory of justice that stopped at the fourth layer would be forced either to pronounce such a case just or to fall silent before it. The ontological layer is what gives justice a language for the subject who is given everything and ceases, nonetheless, to be.
§2.6 A stress test - the case of chosen withdrawal
A typology earns its categories by surviving the cases that would dissolve them, and one case presses hardest here. Suppose a subject chooses to withdraw from generation—steps back, defers, makes herself the support rather than the source, in the manner traditionally praised as devotion or self-sacrifice. Is this effacement? If it is, the theory seems to pathologise every chosen self-subordination, an overreach that would discredit it; if it is not, the theory seems to admit that consent dissolves the harm, surrendering the very claim, central to this paper, that effacement survives sincere endorsement (§1.3). The case appears to force a dilemma fatal either way.
It does not, and seeing why sharpens the concept. The dilemma trades on an equivocation in “chooses to withdraw.” Distinguish two things a subject might choose. She might choose a particular exercise of her generativity that takes the form of support—directing her generation, at her own frequency, toward sustaining a shared project she has helped constitute as hers; this is not effacement at all, but generation, since her own dynamics are running, merely running toward a chosen end. Or she might choose a standing withdrawal of her generativity as such—ceasing to generate at her own frequency, becoming the inert site of another’s. Only the second is effacement, and the criterion of §4 discriminates them exactly: the first leaves the subject’s own frequency intact (she is the generator of her support), the second extinguishes it (her motion is now the echo of another’s rhythm). The traditional praise of “self-sacrifice” conflates the two, and the conflation is precisely the ideological cover under which effacement passes for virtue. The theory does not pathologise the first; it names the second, and names also the equivocation that has let the second hide inside the honour due the first.
Claim — Chosen support is not effacement; chosen self-erasure is. The objection from chosen withdrawal dissolves on a distinction the criterion makes exact. To direct one’s generativity, at one’s own frequency, toward a freely constituted supporting role is generation, not effacement: the subject’s own rhythm runs. To cease generating at one’s own frequency, becoming the site of another’s generation, is effacement whether or not it is chosen—and that it can be chosen, and honoured as sacrifice, is not a refutation but a confirmation, for it is exactly the wrong this paper holds can be completed in sincere endorsement. The theory pathologises no chosen support; it exposes the equivocation by which self-erasure has borrowed the prestige of support.
With the floor installed, its metaphysics defended, the layer tabulated, the bottom row shown irreducible, and the hardest case turned, the wrong has its locus and its place in the order of injustices. What it does not yet have is a definition exact enough to border it from its neighbours, a criterion able to detect it, a justice answering it, and a ground making that justice owed. To the definition the next section turns.
§3 Definition and Conceptual Demarcation
The descent of §2 cleared a place beneath the established injustices and named its register ontological. This section fills the place with a concept exact enough to be defended—which means exact enough to be bordered, for a new category earns its keep not by the ground it claims but by the lines it holds against the neighbours that would absorb it. A category of injustice that blurred into exploitation, into the freezing of power, into the constitutive injustice it descends from, or into mere maldistribution would be a redundancy however vividly named. The section therefore states the definition clause by clause, then walks the four borders along which confusion threatens, showing at each that the line is principled and not stipulative. At two of these borders the most developed rival theories—the Marxian and Roemerian analyses of exploitation, the capability analyses of Sen and Nussbaum—press hardest, and the section meets them there rather than gesturing past them.
§3.1 The definition
Claim — Generative effacement. Generative effacement is the injustice in which, within a field of co-generation, one party’s generative activity occupies the shared generative space in such a way that the other loses the possibility of continuing to constitute herself as a generative being—so that what is harmed is not her holdings, her standing, or her knowability, but her being itself, qua generative being.
Each clause is load-bearing, and the definition is read by weighing them in turn. Within a field of co-generation: the wrong presupposes the relational ontology in which value is generated between the parties rather than held by either, and is possible only because the generative space is shared—were each generating in isolation, one party’s productivity could not occupy the other’s room, there being no common room to occupy. One party’s generative activity occupies the shared space: the agent of harm is not a vice but an activity, and a good one; the verb is occupies, not seizes, because the stronger pole does not remove something the weaker had but fills, by its own generating, the space in which the weaker’s generating would have occurred. The other loses the possibility of continuing to constitute herself: the injury is the foreclosure of an ongoing activity, marked by continuing—the subject is already constituted and what is lost is the continuation in which her being consists. What is harmed is her being itself, qua generative being: the qualifier is exact; the claim is not that she ceases to exist in every sense—she persists biologically, socially, legally—but that she ceases, within the relation, to exist as a generative being, the mode of existence the series holds to be at stake in intimate life.
§3.2 First border - against exploitation
The most tempting reduction, and the one with the most powerful theory behind it, is to exploitation. The temptation is strong because from outside the two look identical: in both, one party flourishes while another does not. The reduction nonetheless fails, and tracing why against the major accounts of exploitation shows the failure to be structural rather than a matter of emphasis.
On the classical Marxian account, exploitation is the appropriation of the surplus the worker produces above what is returned to her; its mechanism is a division of her product into a part she retains and a part the other takes. Roemer, dispensing with the labour theory of value, recasts exploitation in terms of property relations: a coalition is exploited if it would be better off withdrawing with its per-capita share of society’s alienable assets; and Vrousalis, pressing past Roemer’s distributive reduction, grounds the complaint in a Non-Servitude Proviso violated by the power-induced extraction of another’s labour. These accounts differ sharply over what exploitation is—surplus-extraction, unequal asset-withdrawal, servitude—but they share, without exception, a structural presupposition that effacement violates: exploitation requires the exploited party to keep producing. It is her continued production, appropriated or extracted or coerced, that the exploiter lives upon; an exploitation that stopped its victim’s production would abolish its own source. The serf must work the lord’s demesne; the worker must labour the surplus hours; the servant must serve. Exploitation is parasitic upon an ongoing generativity it does not extinguish but feeds on.
Effacement does the reverse. It does not appropriate the other’s production; it forecloses it. The effaced party is not a producer whose output is siphoned but a would-be producer whose production never occurs, because the space in which it would have occurred is already filled. Where exploitation says, in effect, keep generating, that I may take it, effacement says do not generate, for I have generated in your place. The contrast is exact and it is total: the exploited subject is kept generating and robbed; the effaced subject is not robbed but stilled. Vrousalis’s proviso sharpens the point rather than dissolving it—effacement is not servitude, for the effaced party is set to no task; she is set to no task at all, which is precisely the harm.
Claim — Exploitation feeds on generation; effacement forecloses it. Exploitation and effacement are not variants but structural opposites. Every developed account of exploitation—Marxian surplus-extraction, Roemerian asset-withdrawal, Vrousalis’s servitude—presupposes that the wronged party continues to generate, since it is her continued generation that is appropriated, extracted, or coerced. Effacement presupposes the contrary: that the wronged party’s generation has been foreclosed, the space of it pre-filled. One cannot exploit a subject one has effaced, for there is nothing left to appropriate; one cannot efface a subject one exploits, for one needs her producing. The border is the difference between living on another’s generativity and extinguishing it.
§3.3 Second border - against the freezing of power
The second border separates effacement from the solidification of an asymmetry into a fixed, unrebalanceable order—the freezing of power. The freezing presupposes a dynamic that is then arrested: two parties exercise power upon one another, and the wrong is that the play between them ossifies into permanent settlement. Effacement is prior to this. It does not freeze an existing dynamic between two generating poles; it forecloses the second pole’s generation before any dynamic between equals can form. The frozen relation still has two agents, locked against each other; the effaced relation has, generatively, one, the other having been pre-empted out of the play entirely. The freezing of power is a pathology of an ongoing contest; effacement is the absence of the contestant. A relation can be both—a frozen power-order can also efface—but the concepts come apart: one can freeze without effacing (two agents permanently locked, both still generating within their fixed positions) and efface without freezing (a fluid, loving, unfixed relation in which one pole simply never generates).
§3.4 Third border - against constitutive injustice
The third border is the most delicate, because effacement descends from constitutive injustice and shares its ontological reach. The line, drawn in §2.3, is restated here as a border: constitutive injustice wrongs the subject at the threshold of formation, denying the not-yet-subject her arrival; effacement wrongs the already-formed subject, extinguishing the activity in which her achieved being consists. They are kin—both ontological, both deeper than distribution, recognition, or epistemic standing—but they fall on opposite sides of the moment of formation. The constitutive wrong concerns whether a subject will come to be; the effacing wrong, whether an existing subject will continue to be. One bars the door to one who has not entered; the other empties the room of one already inside. The kinship is what makes the border worth drawing: without it, effacement collapses upward into its parent and forfeits its distinctness as a wrong done to maturity rather than to formation.
§3.5 Fourth border - against distributive injustice and the capability turn
The fourth border is the plainest to state and the most insistently tempting in practice, because effacement so naturally invites description as the effaced party’s having received too little. But “too little” is distributive language, and it misdescribes the harm—and here the capability theorists must be met, for they offer the subtlest version of the distributive reduction and the one that comes nearest the truth.
Sen and Nussbaum, dissatisfied with distributions of resources, relocate justice’s concern to capabilities: what a person is actually able to do and to be, the real freedoms she enjoys to achieve functionings she has reason to value. This is a genuine advance toward the harm, for it shifts attention from holdings to doings, and effacement is a harm of foreclosed doing. One might think the capability approach already names it: is the effaced party not simply a person whose capability to generate has been removed? The answer marks the border precisely. The capability approach treats capabilities as goods to be secured for a person—distributed, protected, enabled by institutions—and so remains, for all its advance, within the paradigm of provision: justice ensures she has the capability. But effacement is not the absence of a capability the effaced party lacks and should be given; she retains every capability, in full. What is foreclosed is not her capability to generate but her generating—the exercise, not the capacity. And the exercise is not a good that can be secured for her, because it is not a good she can be given at all: to hand her generation, ready-made, is precisely to confirm its foreclosure, since what arrives is the stronger pole’s generation and not hers. The capability approach can secure that she is able to generate; it has no resources to ensure that she does, for the doing is hers alone and cannot be provisioned. This is the exact sense in which the harm is of being and not of having: distribution and capability alike concern what can be allotted to a subject, and the effaced subject’s loss is of an activity that, by its nature, no allotment reaches.
Claim — Being, not having; exercise, not capacity. Effacement is irreducible to maldistribution, including its most refined form, the deprivation of capability. The distributive paradigm, even when it advances from resources to capabilities, concerns what can be secured for a subject—holdings, functionings, real freedoms. Effacement forecloses not a capability the subject lacks but the exercise of capabilities she fully retains; and an exercise cannot be secured for her, since to supply her generation ready-made is to confirm the foreclosure rather than repair it. The border between distribution and effacement is the border between having and being, and, within doing, between the capacity that can be provisioned and the exercise that cannot.
§3.6 The fixed point for what follows
With the concept defined and its four borders held, the paper’s remaining work is fixed, and the definition is the fixed point against which each later part is measured. §4 must give a criterion that detects this harm—the foreclosure of the weaker pole’s own generative continuation—in the very cases where the four shallower diagnoses report nothing; a criterion detecting anything else would detect the wrong thing. §5 must derive the justice that repairs this harm; a justice repairing anything else would repair the wrong thing. And §6 must ground an obligation to cease this occupation; an ethics grounding anything else would ground the wrong obligation. Everything that follows is the working-out of the harm bordered here.
§4 The Formal Criterion - Phase Capture and the Loss of One’s Own Frequency
The harm named in §3 is invisible to every metric the prior literature supplies. A distributive metric records that one party holds less; here the difficulty is that aggregate value is rising. A recognitive metric records that one party is misrecognised; here she may be fully and warmly recognised. An epistemic metric records that one party cannot be heard; here she is heard and thanked. What is required is a criterion that registers a harm consistent with rising value, sincere recognition, and full audibility—a criterion not of holdings, status, or intelligibility, but of whether the other still generates as herself. This section constructs such a criterion and shows that it is not a gradient shading imperceptibly out of health but a phase boundary that can be located exactly.
§4.1 Two independent dimensions
A single distinction organises the chapter. The topology of coupling—how two generating subjects coordinate—and the survival of generation—whether each still generates at its own rhythm—are independent axes. The series’ coupling typology classifies the first; the criterion of this paper concerns the second. The standard error, which the section exists to forestall, is to read a strong and harmonious coupling as evidence of health. Coupling can be strong, harmonious, and mutually willed, and the weaker party can nonetheless have ceased to generate as herself.
The coupling typology gives four regimes, by whether the two subjects pursue the same or different goals and whether their coupling is high or low-with-high-dynamical-diversity: (1) same goal, high coupling—a joint sprint, force possibly unequal; (2) same goal, low coupling and high diversity—complementary work on one complex problem; (3) different goals, high coupling—a parallel sprint; (4) different goals, low coupling and high diversity—complementary parallel work on many problems. These four are regimes of health: in each, both poles retain a non-zero generative rhythm. Effacement is not a fifth regime. It is the common degenerate limit into which any of the four can collapse when one pole loses its own rhythm—and, as the model shows, the collapse is governed not by which regime one is in but by two quantities that cut across all four.
§4.2 The model
Represent each subject as a phase oscillator—the minimal model of a system with its own intrinsic rhythm of generation. Let $\theta_w$ and $\theta_s$ be the phases of the weaker and stronger pole, with intrinsic frequencies $\omega_w$ and $\omega_s$, the rate at which each, left alone, generates. The stronger pole is the one with the higher intrinsic frequency, $\omega_s>\omega_w$: it generates faster.
Two parameters describe their relation. The coupling strength $K$ measures how strongly each pole’s state is pulled toward concordance with the other. The force asymmetry $\Delta\in[0,1]$ measures the directedness of that pull: the traction exerted on the weaker pole is $K(1+\Delta)$, that on the stronger $K(1-\Delta)$. At $\Delta=0$ the coupling is symmetric; as $\Delta\to1$ the weaker pole is fully subject to the stronger while the stronger evolves freely. The dynamics are
$$\dot\theta_w = \omega_w + K(1+\Delta),\sin(\theta_s-\theta_w),$$
$$\dot\theta_s = \omega_s + K(1-\Delta),\sin(\theta_w-\theta_s).$$
The phase difference $\varphi=\theta_w-\theta_s$ obeys an Adler equation,
$$\dot\varphi = (\omega_w-\omega_s) - 2K\sin\varphi,$$
whose behaviour gives the criterion directly. (Note. The reduction to a single Adler equation follows because the two traction terms, $K(1+\Delta)$ and $K(1-\Delta)$, enter $\dot\varphi$ additively and sum to $2K$ independently of $\Delta$. Thus $\Delta$ drops out of the locking condition and survives only in the location of the locked frequency—the formal root of the two-structure result of §4.4. On the Kuramoto reduction generally see Kuramoto 1984.)
§4.3 The order parameter and the criterion
The criterion must measure not how much the weaker pole generates but whether what it generates is its own. The order parameter is therefore the degree to which the weaker pole’s realised rate departs from its own intrinsic frequency and is drawn toward the stronger’s:
$$E = \frac{\dot\theta_w^{,\text{realised}} - \omega_w}{\omega_s - \omega_w}, \qquad E\in[0,1].$$
$E=0$ means the weaker pole advances at its own frequency—it generates as itself. $E=1$ means its realised rhythm has been pulled entirely onto the stronger’s frequency—it is in motion, but the motion is the echo of another’s rhythm, not its own generation.
Claim — Effacement as the loss of one’s own frequency. Generative effacement is not the cessation of a subject’s activity but the replacement of its own generative rhythm by another’s. The effaced pole still moves; an outside observer sees a harmonious, active, synchronised pair. What the observer cannot see, and the order parameter $E$ can, is that one of the two rhythms has gone silent and been overwritten by a copy of the other. This is why the harm is consistent with rising value and reported harmony: the motion, and the value it produces, are real—only they are no longer hers.
§4.4 The result - two distinct structures
Integrating the model across the $(K,\Delta)$ plane (see Figure 1 in the embedded PDF) yields a result sharper than a gradient of risk, and the sharpness is the section’s main technical claim. Two distinct structures govern the harm, and they must not be conflated.
First, whether effacement occurs at all is governed by $K$ alone. The Adler equation locks—the weaker pole loses its own frequency—if and only if
$$K \ge K_{\text{lock}} = \frac{|\omega_s-\omega_w|}{2}.$$
This is a genuine phase transition with an exact threshold, and it is independent of $\Delta$. Below $K_{\text{lock}}$ the weaker pole keeps its own rhythm however asymmetric the coupling; above it, the weaker pole is captured. The boundary is the vertical dashed line in Figure 1. To its left lies the autonomous regime; to its right, the collapse region.
Second, once capture has occurred, how deeply the weaker pole is effaced is governed by $\Delta$. In the locked state the common frequency is the traction-weighted mean of the two intrinsic frequencies, and the normalised effacement is
$$E = \frac{1+\Delta}{2}.$$
This is the vertical gradient within the collapse region: at fixed super-threshold $K$, the harm deepens monotonically with force asymmetry, from $E=\tfrac12$ at $\Delta=0$ to $E\to1$ as $\Delta\to1$.
Claim — Equal power does not protect the weaker pole. Set $\Delta=0$—perfectly symmetric, equal-force coupling, the very picture of a balanced union—and read along the bottom edge of Figure 1. Effacement is still $E=\tfrac12$ for every $K$ above threshold. As long as the two subjects have different intrinsic rhythms and the coupling is strong enough to lock them, the slower is drawn off its own frequency by half the gap, regardless of any equality of power. Force asymmetry $\Delta$ decides only which pole is sacrificed more and how far; it is coupling strength $K$, not inequality of force, that decides whether a pole is sacrificed at all. The harm is, in the first instance, a property of strong coupling between unlike rhythms—not of unequal power.
Figure 1 (see the embedded PDF). Effacement $E$ of the weaker pole on the $(K,\Delta)$ plane. Vertical dashed line: the locking threshold $K_{\text{lock}}=|\omega_s-\omega_w|/2$, independent of $\Delta$. Left: the autonomous regime—the weaker pole keeps its own frequency ($E=0$). Right: the collapse region—the weaker pole is phase-captured and the harm deepens with $\Delta$, reaching $E=1$ as $\Delta\to1$. Note the bottom edge: at $\Delta=0$ (equal force) effacement is still $E=\tfrac12$ throughout the collapse region. Equal power does not prevent capture.
§4.5 Three readings of one degeneration
The same transition can be read in the three idioms the series has already built, which is the warrant for treating it as one phenomenon rather than an analogy.
Predictive coding (Paper XIII; Paper XI). The stronger pole, generating and pre-arranging faster, pre-symbolises the shared future, driving the weaker pole’s prediction error toward zero. Active inference—the generation of new value—feeds on unresolved uncertainty; with the future pre-determined, the fuel is gone. Phase capture is, in this idiom, the foreclosure of the weaker pole’s predictive horizon. This extends Paper XVII’s claim that desire withholds itself from total symbolisation in order to remain alive: here one subject’s over-symbolisation extinguishes another’s.
Holonomy (Paper IX; Paper XI). Above threshold the system’s aggregate holonomy is positive—value genuinely accumulates—while the captured pole’s own phase advance, measured against its intrinsic frequency, has collapsed. This is the single-subject spiral: a system ascending on one pole’s rhythm alone. It is the limiting case of Paper XI’s justice-as-skewness, the skew carried to the point where one distribution has lost its variance entirely.
Coupling (the present contribution): phase capture above $K_{\text{lock}}$, with depth set by $\Delta$, as derived above.
§4.6 The inversion
The result overturns the common identification of intimacy with strong coupling and of strong coupling with good. Strong coupling presents itself as union—we move as one—and it is precisely strong coupling that, between unlike rhythms, captures the weaker pole and silences its frequency. Low coupling with high dynamical diversity presents itself as distance, and it is precisely what lets two unlike rhythms each survive.
Claim — The inversion of the intimacy criterion. The good of intimacy is not the strength of the coupling but whether the coupling leaves each pole turning at its own frequency. This is the same law Paper XVII reached from another direction in identifying perfect concordance with equilibrium and equilibrium with death: a relation at perfect, frictionless agreement is one in which one of the two rhythms has stopped. High coupling is not the measure of love but, between unlike rhythms, its most common solvent.
§4.7 The criterion stated, and its handoff to justice
The formal criterion of generative effacement is therefore: a co-generative relation effaces its weaker pole when its coupling exceeds the locking threshold $K_{\text{lock}}$, drawing that pole off its own intrinsic frequency; the depth of effacement is $(1+\Delta)/2$. The criterion is satisfiable, and is in fact at its most dangerous, when aggregate value rises and both parties report harmony—exactly the case §4.1 required it to catch.
This hands §5 a justice with a precise operational meaning. If the harm is crossing $K_{\text{lock}}$, the remedy is to return below it: to lower the coupling until the weaker pole recovers its own frequency. Crucially, the threshold depends on $K$, not on the stronger pole’s frequency $\omega_s$. The stronger pole need not slow its own generation—lowering $\omega_s$ is neither necessary nor, by itself, sufficient. What restores the weaker pole is lowering $K$: releasing the traction, not reducing the speed. That distinction, derived here, is the whole content of the claim in §5 that the justice of reticence is decoupling, not deceleration.
§5 The Reservation of Generative Space
The criterion of §4 located the harm and, in locating it, already pointed past itself to the remedy. If effacement is the capture of the weaker pole above a coupling threshold, then the form of its repair is fixed in advance: not the addition of anything to the weaker pole, but the withdrawal of something from the stronger. This section develops that withdrawal into a positive concept of justice—and the concept is strange enough that it must be stated against the grain of almost everything justice has meant. For the whole tradition of distributive and recognitive justice answers the question what does one owe the other; and its answers are additive—a share, a recognition, a hearing, a right. The justice this paper requires is subtractive. What one owes the effaced other is not a thing given but a determination withheld: the room in which she can generate, which exists only where the stronger pole has declined to generate in her stead.
§5.1 Justice as non-determination
The object that effacement damages is not a holding but an activity—the weaker pole’s continuing generation of value at its own frequency. One cannot repair a damaged activity by transfer, because the activity is precisely what cannot be transferred: to hand the weaker pole a portion of the value one has generated for her is not to restore her generativity but to confirm its suspension, since the value arrives already made, with her contribution already supplied by another. The only repair adequate to the harm is to return to her the occasion of generating—and the occasion is a region of the shared field that the stronger pole, simply, does not occupy.
Claim — Generative reticence. The justice answering generative effacement is not the transfer of a due share but the non-determination of the domain in which the other generates. Its content is a negative and structural obligation: that a generative subject refrain, in the regions where the other’s generativity would otherwise unfold, from determining what is to be generated there. We call this obligation generative reticence. It is the determinate negation of effacement—where effacement is the foreclosure of the weaker pole’s generative domain, reticence is the deliberate non-foreclosure of that domain—and it is owed not because the other has been given too little but because she has been left too little room.
The object whose limitation reticence demands is, on its face, the strangest feature of the account, and the surest sign that the harm is real. For the object to be limited is not malice, possessiveness, or any vice—it is good, fast generation itself. The stronger pole effaces not by failing the relation but by serving it too completely; reticence asks it to serve less completely, to leave undone things it could do well, for no other reason than that the other might do them. No theory built to restrain wrongdoing reaches this, because here the thing to be restrained is not wrongdoing but excellence. The justice of reticence is the justice of a good that must, in the presence of another generative being, hold part of itself back.
§5.2 Decoupling, not deceleration
Everything turns on a distinction the criterion of §4 makes exact, and which separates reticence from the moralised self-diminishment it is too easily mistaken for. The threshold at which the weaker pole is captured, $K_{\text{lock}}=|\omega_s-\omega_w|/2$, depends on the coupling $K$ and on the difference of intrinsic frequencies—it does not depend on the stronger pole’s frequency $\omega_s$ taken alone. This has a consequence that is easy to state and decisive in practice: the stronger pole cannot repair effacement by slowing down.
Consider the two interventions available to the stronger pole. It may lower its own frequency $\omega_s$—generate more slowly, do less, dim itself—or it may lower the coupling $K$—release the traction by which its rhythm is transmitted to the other. Lowering $\omega_s$ narrows the frequency gap and so does, in the limit, lower the threshold; but it does not return the weaker pole to its own frequency, and it carries a second harm of its own. A stronger pole that decelerates is still the pole that sets the pace; “I have slowed for you” is still a determination of the shared tempo by one party, and it arrives freighted—as every recipient of a sacrifice knows—with the silent debt of having been the cause of another’s diminishment. Deceleration keeps the stronger pole at the centre of the relation’s dynamics, now as benefactor rather than as engine; the weaker pole, far from recovering her own rhythm, acquires a new reason to suppress it, lest she squander what was given up for her. Deceleration is effacement continued by other means.
Lowering $K$ is categorically different. To release the coupling is not to generate less but to generate without entraining—to let one’s own rhythm run at its own frequency while ceasing to transmit it as a force that overwrites the other’s. Below $K_{\text{lock}}$ the two poles desynchronise: each turns at its own frequency, the weaker pole’s generation is once more its own, and—this is the point—the stronger pole has surrendered nothing of its own speed. It generates as fast and as richly as before; it has only stopped pulling.
Claim — Decoupling, not deceleration. The remedy for effacement is to lower the coupling $K$, not the stronger pole’s frequency $\omega_s$. Because the locking threshold is independent of $\omega_s$, deceleration does not restore the weaker pole’s own frequency and adds the second harm of a sacrifice owed; decoupling restores it and costs the stronger pole nothing of its own generation. Reticence is therefore not self-diminishment. A subject may generate as fast and as fully as it is able, provided it does not entrain the other—provided, that is, that it releases the traction rather than reducing the speed.
This is the formal vindication of an intuition that, stated without the model, sounds either trivial or false: that the answer to one partner’s overshadowing of another is not for the first to become smaller. The model shows why the intuition is right and what it really asks. It does not ask the strong to be weak. It asks the strong to stop transmitting their strength as a force that captures—to decouple their own flourishing from the determination of the other’s.
§5.3 The three channels of determination
Decoupling is abstract until one asks through what the stronger pole transmits its traction. The coupling $K$ is not a single wire but a bundle of channels, each a concrete way in which one subject pre-determines the other’s next state; to lower $K$ is to relinquish determination in one or more of them. Three channels carry most of the load, and they are ordered by depth—each prior one a precondition of release in the next.
§5.3.1 The first channel - the shared telos
The first determination—first in the order of depth set out below—is of where the relation is going. When the goal of the shared life—its destination, its project, the shape of the future it is building toward—is generated by one pole, the other can at most find a place inside a teleology she had no part in constituting. Her generativity, however vigorous, then runs only within bounds another has drawn; she is free to move, but not to have said where. This is the channel the four-quadrant typology of §4.1 cannot see, for that typology presupposes a goal already given and asks only how two subjects coordinate toward it. The prior and graver effacement is the pre-generation of the goal itself, which fixes the manifold on which all subsequent coordination occurs. Relinquishment here is the return of the telos to an undetermined, jointly legislated state: not “here is where we are going, and here is your part in it,” but the leaving-open of the destination as a thing the two are to constitute together, including the real possibility that her constitution of it diverges from his.
§5.3.2 The second channel - meaning
A subject of great expressive facility—fluent in symbols, quick to interpret, prolific in the conferral of significance—may determine not only where the relation goes but what its events mean. Where one pole names every shared occasion before the other has found her own reading, fixes the significance of each passage, supplies the relation with its standing interpretation, the other is left no room to determine meaning, only to receive it. The channel is subtle because the determination feels like a gift: the meanings offered are often beautiful, and refusing to impose them can feel like withholding care. But a meaning supplied is a meaning the other did not make, and a relation whose entire significance is authored by one pole has, in the register of sense, only one generative subject. Relinquishment is the leaving-underdetermined of meaning—the discipline of not naming first, of letting events stand uninterpreted long enough that the other’s reading can form, and of accepting that reading when it comes even where it is plainer, or stranger, than the one that went unspoken.
§5.3.3 The third channel - the constitution of the problems
The subtlest channel, and the one whose relinquishment is most often confused with self-incapacitation, concerns not what is generated but which problems the relation is constituted to solve. A pole made anxious by its own completeness may suppose that the way to make room for the other is to become, deliberately, less capable—to feign a need, to manufacture a dependency, to do badly what it could do well. This is false reticence, and the criterion shows why: a manufactured gap is a determination of the other’s domain in disguise, a problem set for her by the stronger pole rather than a region genuinely left to her. The correct diagnosis is not that the stronger pole is too capable but that the shared field has been constituted only on the terrain where its capabilities are strongest. The other does not lack ability; the relation lacks problems of the kind at which her dynamics run fastest. Relinquishment here is the reconstitution of the shared field to include the terrain on which the other’s intrinsic frequency is highest—and then the harder discipline of letting her judgement lead on that terrain, with the stronger pole’s contribution running downstream of hers rather than framing it in advance.
Claim — The incompleteness must be real. The third channel is relinquished not by the stronger pole pretending to an incompleteness it does not have, but by the relation being constituted around an incompleteness that is genuine: a domain in which the other’s contribution is not substitutable by the stronger pole’s, because it is the other who is, there, the faster generator. A staged incompleteness is effacement in the costume of reticence—the stronger pole still authors the problem, merely assigning itself the part of the one in need. Only a real incompleteness, in which the relation would in fact be poorer without precisely her generation, returns to the other a domain that is hers.
§5.4 Reticence distinguished from sacrifice
It remains to forestall the reduction of the whole account to a counsel of sacrifice—as though reticence were simply the stronger pole giving things up for the weaker, a generosity dressed in formal clothes. Across the three channels, the through-line is the opposite. None of the relinquishments requires the stronger pole to generate less, to be smaller, or to surrender any of its own activity. Each relinquishes a determining function—the setting of the goal, the fixing of meaning, the constitution of the problems—that the stronger pole had been exercising unilaterally, over a domain the other was in fact better placed to determine. To relinquish a determining function one had no title to monopolise is not to sacrifice; it is to stop overreaching. The stronger pole loses nothing that was properly its own. It keeps its full frequency, its full generative power, its whole self; it gives up only the traction by which that self had been overwriting another. Reticence is the withdrawal of an overreach, not the surrender of a good—and this is the final sense in which it is decoupling and not deceleration, non-determination and not loss.
§6 The Ethics of Reticence
The criterion of §4 establishes that effacement is a harm; the construction of §5 establishes what reticence is and how it repairs the harm. Neither yet establishes that reticence is owed—that non-determination is a requirement of justice and not a mere kindness, an aesthetic preference for relations in which both flourish. This section supplies the ground. It moves first negatively, showing that each major ethical framework registers something of the wrong yet, pressed to ground the obligation, fails at a diagnostic point of its own; then positively, locating the ground in the standing of the other as a generative being; and finally reflexively, establishing the constraint that forbids the ethic its own paternalistic application. The negative movement is not a survey but a series of defeats, and the most instructive defeat is care ethics’, because care ethics comes nearest and fails most revealingly.
§6.1 Four frameworks, four failures
Consequentialism cannot see the harm. An ethic evaluating by outcomes is structurally blind here, for the outcomes of effacement are in aggregate good: value rises, the relation prospers, both may report contentment. A framework that sums welfare cannot find a harm invisible in the sum, and effacement is such a harm, since what it destroys—the weaker pole’s own generativity—is replaced in the ledger by the stronger pole’s surplus, leaving the total undiminished. This is not a marginal failure of detail but the sharpest demonstration that the wrong is non-aggregative.
Contractualism is defeated by consent. An ethic locating wrong in what could not be justified to, or was not agreed by, the other founders on the datum of §1.3: the effaced party consents, endorses, is grateful, would choose it again. Contractualism can condemn a determination imposed over objection; it has nothing to say of one the other welcomes. Yet the harm is present in the welcoming case, because consent operates at the level of the will and effacement at the level of being. A wrong surviving the genuine consent of the one wronged lies outside the reach of any ethic that makes consent its criterion.
Kantian respect locates the wrong too high. The deontological tradition forbids treating another merely as a means and requires treating her always also as an end—and fixes the end-status in her rational agency, her capacity to set and pursue ends. But rational agency survives effacement intact: the effaced party still reasons, still chooses, still wills, and often wills her own effacement. An ethic that secures her standing as a rational chooser secures exactly what effacement leaves untouched. The Kantian framework is not wrong; it aims at the right relation (treating-as-end) but at the wrong faculty, and so passes over the harm. (The correction—relocating the end-status from rational to generative agency—is the work of §6.2 below.)
Care ethics supplies the ontology and valorises the harm. Of the four, care ethics alone begins from the relational ontology this paper requires: it knows selves are constituted in and sustained by relation, and attends to dependency and to the labour of maintenance the other frameworks overlook. Noddings builds the ethical relation from the “one-caring” and the “cared-for,” and makes its central act engrossment—the one-caring’s apprehension of the other’s reality, the displacement of her own motives to act on the other’s behalf. Here, if anywhere, an ethic should catch a harm done through devotion. Instead care ethics, in its received form, valorises the very structure that effaces. For engrossment and motivational displacement, raised to the paradigm of the ethical, describe with precision the inner life of the effacing pole: the one who displaces her own generativity into the other’s, who is engrossed in the other to the point of generating only for and as the other. The standing feminist critique of care ethics already saw the danger from the side of the carer—that the ethic celebrates self-sacrifice and risks the caregiver’s exploitation, pressing her toward a caring that forgets her own needs. This paper locates the symmetrical danger on the other side: that the cared-for can be effaced by an engrossment too complete, generated-for so thoroughly that she ceases to generate. Care ethics supplies the indispensable relational ontology, and must be corrected at exactly the point where it makes a virtue of the unstinting giving that, past a threshold the criterion of §4 makes exact, occupies the other out of her own being.
Claim — Why the frameworks fail. Each major framework fails to ground the obligation of reticence at a diagnostic point: consequentialism because the harm is non-aggregative and invisible in the sum; contractualism because the harm survives sincere consent; Kantian respect because it secures the rational agency effacement leaves intact; care ethics because it valorises, as devotion, the engrossment that effaces. The pattern of failure is itself informative. Three frameworks aim too high—at outcomes, at the will, at rational agency—missing a harm that lies beneath outcome, will, and reason, at the level of generative being; and the one framework pitched low enough, care ethics, mistakes the harm for the good. The ground of reticence must therefore be sought at the level the first three overshoot and the fourth misreads.
§6.2 The positive ground - the standing of the generative being
If reticence is owed, it is owed because the other’s standing as a generative being is a good making a direct claim—a claim damaged specifically by the determination of her generative domain, and damaged even when the determination is benevolent and welcomed. The ground has three components, each drawn from a tradition and each corrected where its received form would miss the harm.
The corrected end-formula. Relocate the Kantian end-status from rational to generative agency: the other’s standing as a being who continually constitutes herself by generating at her own frequency. The formula becomes—act so that you never determine the other’s generative domain merely as an extension of your own. To overwrite her frequency with yours, however lovingly, is to treat her generativity as material for your own, making of her generative being a means to the relation’s flourishing rather than an end whose exercise the relation exists to sustain. This repairs the defect diagnosed in §6.1: the end-formula was aimed at the wrong faculty, and aimed right, it reaches the harm.
The foreclosure of activity. From the tradition that locates a being’s good in the exercise of its characteristic activity—its energeia—rather than in any state or resource, the harm comes into focus as the foreclosure of an activity rather than the withholding of a good. What the effaced party loses is not something she has but something she does: the activity of generating, in which her flourishing consists. A capability account at the level of functioning catches part of this, but the harm lies one level deeper than a lost functioning—it is the loss of the self-constituting activity any particular functioning would express. Reticence, in this register, is the virtue that holds open the field of another’s exercise.
The second-personal warrant. To the question on whose authority the claim is made, the answer is that the obligation is second-personal—owed to her, from her standpoint, on her own standing to demand it. The warrant is not a third-personal fact about the goodness of generativity assessed from outside, but her own, perhaps unexercised, standing to require that her generative domain not be determined for her. This component prepares the constraint of §6.4: if the warrant is hers, the remedy cannot be administered over her head without reproducing the wrong.
§6.3 A duty grounded, a virtue in application
The ground is deontological: reticence is a negative obligation owed in virtue of the other’s standing. But an obligation of this kind cannot be discharged by rule, and the reason lies in the harm’s structure. Effacement arrives through devotion, in the ordinary motions of a loving and capable subject doing well by the relation; it breaks no rule and feels, from inside, like care. One cannot rule-follow one’s way clear of it, because no rule fires—there is no moment the stronger pole can recognise, by rule, as the wrong. What is required instead is a trained perception: the cultivated capacity to notice, amid generous and well-meant generation, that one’s own activity has begun to foreclose another’s. The ethic is thus deontological in its ground and virtue-theoretic in its application, and the two are compatible: the obligation is fixed and negative, its detection a matter of cultivated sensibility no algorithm supplies. This is not a retreat into vagueness but a recognition of the harm’s phenomenology—a wrong that announces itself by no transgression can be caught only by a perception trained to its signs.
§6.4 The reflexive limit - reticence without paternalism
The ethic carries a danger of its own making, and must close by disarming it. A stronger pole, having grasped the analysis, may set out to judge that the other is effaced and to act to restore her generativity—to manage her back into generation. But to judge another’s effacement on her behalf and take her restoration in hand is itself a determination of her domain: it sets her a project, her own recovery, she did not constitute, effacing her in a second key, now in the name of repair. The danger is not hypothetical; it is the precise form that care ethics’ critics already identified within devotion itself. Goodin’s worry about care—that subsuming individuals into a relationship of “we”-ness risks losing track of the separateness of persons, issuing in a smothering paternalism—names exactly the recursion that threatens here: an ethic of non-determination, applied as management, smothers in the act of rescuing.
The constraint that disarms it follows from the second-personal ground. Reticence is the withdrawal of one’s own determining activity, never the management of the other’s. The stronger pole’s obligation extends to vacating the domain it has occupied, and stops there; it does not extend to administering what the other then does with the vacated room, for that administration re-occupies the room in the act of giving it. Where the very question is whether the other is effaced, the only non-effacing route is to return the question to her—to leave her reading of her own generativity undetermined—which is itself an instance of the first channel of §5.3, relinquished now at the meta-level.
Claim — The reflexive constraint. The ethics of reticence forbids its own paternalistic application. Because the wrong is the determination of another’s generative domain, any attempt to diagnose her effacement and direct her restoration is a further determination and so a further effacement—the smothering that care ethics’ own critics located within devotion. The obligation is therefore strictly to vacate one’s overreach and never to manage the other’s recovery; and where the question is whether she has been effaced, the question must be returned to her, on the second-personal authority that the claim was always hers to make. An ethic of non-determination that determined, on the other’s behalf, that she had been determined would refute itself in the performance.
§7 Objections and Limits
Four objections press hardest, and the section meets each not to dispose of it cheaply but to let it fix the boundary of the claim. The second, the autonomy objection, is the one on which the paper is most exposed; it receives the fullest treatment, drawing on the literature that has thought longest about wrongs done with the victim’s consent.
§7.1 That it counsels the able to be less
The first objection is the one the construction was built to answer, and it returns for final disposal. If effacement is the harm one party’s superior generation does another, does the theory not simply counsel the able to do less—to dim their gifts so that others may shine by comparison? It does not, and the answer is now exact rather than merely intuitive. The criterion of §4 is not the rate of one party’s generation but the capture of the other’s frequency, and capture is a function of coupling, not of speed: the locking threshold $K_{\text{lock}}=|\omega_s-\omega_w|/2$ binds $K$, not $\omega_s$. The able party is asked not to generate less but to cease entraining—to release the traction by which its rhythm overwrites another’s, while keeping its own rhythm entire. The theory counsels decoupling, not diminishment, and §5 showed these to be not merely different but opposed: deceleration keeps the strong pole the pace-setter and adds the second harm of a sacrifice owed, where decoupling restores the other’s frequency at no cost to the stronger’s own. The objection mistakes a counsel of non-determination for a counsel of self-reduction; the model is precisely what tells them apart.
§7.2 The autonomy objection
The second objection is the one on which the paper is most exposed, and it neither evades nor softens it. If the effaced party consents to her effacement—endorses it, is grateful for it, would choose it again—then by what right does the theory pronounce her wronged, and pronounce it, apparently, over her own sincere judgement of her own life? To do so seems the very arrogance the paper condemns elsewhere: the substitution of the theorist’s verdict for the subject’s self-understanding.
The objection cannot be answered by denying its premise, for the premise is true and the paper has insisted on it: the effaced party often does consent, and her consent is not feigned. It must be answered by showing that consent, here, does not do the exculpatory work the objection asks of it—and the resources for showing this are already developed, in the literature on adaptive preferences. Elster named the structure: a “blind psychic process operating behind the back of the person,” by which an agent revises her preferences to fit her diminished opportunities without awareness or control. Nussbaum gave it ethical weight through cases like Vasanti’s, the woman who came to regard her husband’s abuse as part of a woman’s lot, her preference formed by lifelong habituation to constrained options. The lesson the literature draws is precisely the one the autonomy objection overlooks: that a preference can be sincerely held, reflectively endorsed, and still fail to exculpate the conditions that formed it, because the endorsement itself may be an artefact of those conditions. The effaced party’s gratitude is consistent with her effacement in exactly this way—a generativity foreclosed long enough comes to feel like a preference satisfied.
But the adaptive-preference literature also carries a warning the paper must heed, and heeding it is what keeps the reply from becoming the arrogance the objection feared. Narayan and others have cautioned against treating women who endorse self-subordinating arrangements as mere “dupes” or “prisoners of patriarchy,” urging that their choices be read, often, as competent bargaining within hard constraints rather than as deformed desire. The caution is decisive here, and it is why the paper’s reply has two parts that must be held together. The first: consent does not exculpate, because effacement operates at the level of being and consent at the level of the will, and a will habituated to its own foreclosure can sincerely endorse it. The second, equally firm: that consent does not exculpate gives no one licence to override the consenting party, to declare her deceived, or to manage her toward a generativity she has not asked to recover. The wrong is real under consent; and the authority over the claim remains hers, by the second-personal ground of §6.2 and the reflexive constraint of §6.4. The theory thus occupies a position narrower and more disciplined than the objection supposes available: it holds that a wrong can be done with sincere consent, while denying that this holding authorises any intervention over the head of the one who consents. It diagnoses a harm; it does not license a rescue.
Claim — Consent does not exculpate, and does not authorise. The autonomy objection is answered by a position with two inseparable halves. Consent does not exculpate effacement: as the literature on adaptive preferences establishes, a preference sincerely held and reflectively endorsed may still be an artefact of the very foreclosure it endorses, so that the effaced party’s gratitude is consistent with her being wronged. And the failure of consent to exculpate authorises no intervention: the second-personal warrant locates the claim’s authority in the effaced party herself, so that to override her, or to manage her recovery, would be a fresh effacement. The paper holds both: the wrong survives consent, and no one but the wronged has standing to act on its being a wrong. To collapse either half—to say consent cures the harm, or that the harm licenses rescue—is to leave the narrow ground on which the diagnosis is both true and non-arrogant.
§7.3 The asymmetry-of-capacity objection
The third objection holds that sometimes one party simply can generate more—is more gifted, more practised, better suited to the problems at hand—and asks whether the theory does not then penalise a mere difference of capacity, demanding an artificial equalisation that wrongs the abler party in the name of sparing the other. The reply distinguishes two things the objection runs together: generating more and occupying the domain of the other’s generation. A more capable party may generate a great deal without effacing, provided the relational field is constituted so that the other’s own frequency has a domain in which to run; conversely, a less capable party is not effaced by the mere existence of someone abler, but by the foreclosure of the terrain on which her own—perhaps quite different—capability would operate. This is the substance of the third channel of §5.3: the remedy for the harm of completeness is not the abler party’s self-diminishment but the constitution of the shared field to include the terrain where the other’s frequency is highest. Reticence is not the equalisation of capacities and asks no one to feign incapacity; it asks only that the field not be drawn solely around one party’s strengths. The difference of capacity is not the harm; the totality of the occupation is.
§7.4 The limit of the criterion
The fourth is less an objection than a limit the paper imposes on itself, and states plainly to forestall misuse. The formal criterion of §4 detects a structural risk; it locates the conditions under which effacement occurs. It is not an instrument for the verdict on an actual relation, and still less for the judgement of persons. Real relations are not two oscillators; the model abstracts severely, and its worth is conceptual clarification, not measurement. To read off from it that a given partner stands convicted of effacing another would mistake an illumination for a tribunal—and would, moreover, violate the very constraint of §6.4, since to convict on another’s behalf is itself to determine her domain. What the criterion warrants is a cultivated reticence: a sensitised attention to the possibility of the harm, a disposition to look for it where it hides behind devotion. It is a lamp held to a dark region, not a gavel; and a lamp is what a harm this well-concealed requires, for what announces itself by no transgression can be sought only by a perception taught where to look.
§8 Conclusions and Envoi
This paper has isolated an injustice the theory of justice had not named: a wrong done not to what a subject has, or is taken to be, or can make known, or could come to be, but to her being, in the precise sense of her continuing activity as a generative subject. In the manner of the series, it closes not on a single synthesis but on the several registers in which the finding can be stated—each true, none reducible to the others, their plurality not a hedge but the form a finding takes when it sits at the meeting of an ontology, a formalism, a deontic structure, and an ethic.
§8.1 The findings, in their registers
Ontologically, the finding is that there is a mode of injustice consisting in the deprivation of being-as-generation. Because, for a relational subject, to be is to go on generating, the foreclosure of its generation is not the loss of one of its activities but the hollowing of its existence; and this can be done to a subject already fully formed, which is what distinguishes it in register from the constitutive injustice it descends from. The theory of justice acquires, with this layer, a floor beneath the floor it had supposed final—and the descent that reaches it is licensed not by analogy but by the identity of being with generation that the series argues on independent grounds.
Formally, the finding is that effacement is a phase transition with an exact threshold. A subject loses its own frequency when the coupling between two unlike rhythms exceeds a bound fixed by their difference; the depth of the loss is set by the asymmetry of force; and—the result that organises all the rest—the loss occurs even under perfectly equal power. The good of intimacy is thereby shown to lie not in the strength of the bond but in whether it leaves each subject turning at its own frequency, and the long-standing identification of closeness with strength of coupling is exposed as the precise error that lets devotion efface.
Deontically, the finding is that the justice answering this wrong is not additive but subtractive. It is not the transfer of any share, recognition, or hearing, but a negative and structural obligation to refrain from determining the domain in which another generates—an obligation whose content is non-determination and whose discharge is the leaving of room. This is the rarest shape a duty of justice can take: not give, but withhold; not the completion of the other’s world, but the deliberate leaving-incomplete of one’s own occupation of it.
Ethically, the finding is that this obligation is owed in virtue of the other’s standing as a generative being, on an authority that is hers; that it is detected not by rule but by a cultivated perception, since it is breached through devotion and announces itself by no transgression; and that it forbids, on pain of self-refutation, its own paternalistic enforcement. The ethic is thus doubly disciplined—deontological in ground and virtue-theoretic in application, and reflexively barred from becoming the management of the other it would protect.
Claim — The closing thesis. Between two generative beings, justice requires not the transfer of value but the non-foreclosure of the other’s capacity to generate. A giving that occupies the whole of the shared space is not the fulfilment of love but the quiet extinction of one of its two subjects; and the love that would keep both subjects in being must therefore learn a giving that withholds—must leave, within the field it generates, the room in which the other generates too. Justice, on this account, is not what one does for the other but what one leaves undone, that she may do it herself.
Envoi
Two rhythms, unlike in their frequencies, may be brought so close that they beat as one—and in the beating-as-one the slower is not joined but silenced, kept in motion by a measure no longer its own. The bond looks then most like union at the very moment it has become most solitary, one rhythm sounding through two bodies. The whole of this paper has been an argument that the cure for that solitude is not to slow the faster rhythm but to loosen the bond between them, until each, turning again at its own measure, can be near the other without being overtaken by it—close enough that something is generated between them, free enough that it is generated by them both. That is the only nearness in which two generative beings remain two. It is, this paper has argued, the form that justice takes when what is at stake is no longer what we have, nor how we are seen, nor whether we can be heard, but whether, beside each other, we continue to be.
中文
亲密关系的哲学与正义理论 · 第十八篇
亲密关系中的生成性抹除
一种本体论上的不正义，与非决定之正义
【 工作草稿 】
论一种给予而非索取的伤害，以及人必须为对方的生成所留出的余地
万泓
工作草稿——不供引用或传播
生而不有，为而不恃，长而不宰。
生养万物而不据为己有；有所作为而不自恃其能；
助其生长而不主宰其命。—— 老子，《道德经》第五十一章
万物并作，吾以观复。
万物一齐生长；
我只观照它们的复归。—— 老子，《道德经》第十六章
致她——
致那个属于森林的女孩，
我永不愿覆盖她自己的节奏；
为了她，我正慢慢学会，
留出她得以生长的空地。致我最爱的人。
摘要
本文分离并命名出一种不正义的形态：它伤害的既不是一个主体所拥有之物，也不是她被承认的方式，也不是她能否被听见，而是她的存在本身——她作为一个持续地自我构成的生成性存在者的地位。当亲密生活被理解为一个共同生成（co-generation）的场域，其中价值不栖居于任何一方，而是在双方之间生成时，一种独特的伤害便成为可能：一方生成得太快、太过超前，以致竟会如此彻底地占满共享的生成空间，使另一方完全丧失生成的能力。其结果不是她拥有得更少，而是她在这段关系中不再存在。本文称之为生成性抹除，并论证它是一种本体论上的不正义，不可还原为那些容易与之混淆的分配性、承认性或构成性（constitutive）的不义。其最具定义性、也最令人不安的特征是：它并非经由支配而来，而是经由奉献而来——这里没有占有，没有强制，没有听而不闻，而被抹除的一方甚至可能真诚地认可、乃至珍视那个抹除了她的人。
本文为这一伤害给出了一条精确的形式化判准。把两个亲密主体建模为耦合的相位振子，本文表明抹除是一次真正的相变（phase transition）：较弱的一极恰恰在耦合强度超过一个由两者固有频率之差所确定的阈值时丧失其自身的节奏；而一旦越过该阈值，抹除的深度则由两者之间力的不对称所决定。这两重结构彼此独立，而它们的分离给出了核心结论——即便在力完全相等的情形下，一旦越过阈值，较弱的一极仍被抹除一半。力的相等并不保护它；不同节奏之间的强耦合足以使较慢者噤声。由此可见，亲密的善并不在于耦合之强，而在于它是否让每个主体仍按其自身的频率运转。
由此引出一种形态上有违直觉的正义。因为阈值取决于耦合而非取决于较强一方的速度，所以补救之道并不是让较强者生成得更少——减速只会重新强加它的节拍——而是解耦：松开那道凭以覆写对方节奏的牵引。本文将此发展为生成性的克制（generative reticence），即一项消极的、结构性的义务：在对方的生成本会展开的领域里，不去决定那里该生成什么；并把它的实践具体规定为对三条决定性渠道的放弃（共享目标、意义，以及关系所致力的诸问题本身）；最后在伦理上为它奠基于一种论说——它把”将他人始终也当作目的来对待”的义务从理性能动性扩展到生成性能动性，同时禁止这一伦理自身的家长式施行。它的结论是：两个生成性存在者之间的正义，所要求的不是价值的转移，而是对另一方生成能力的不予封闭。
关键词： 生成性抹除；本体论上的不正义；构成性不义；生成性的克制；相位捕获；耦合；非决定；生成性正义；关怀伦理。
§1 现象与悖论
设想两个彼此相爱的人，其中一人在通常的世俗意义上是较强的那一个——行动更快、更善于谋划，最先想到办法、最容易安排妥当的那一个。他爱她，而他的爱采取了他的强所自然采取的形式：他供给，他预见，他在她走到路上崎岖之处前便已将路面铺平，他在她尚未把难题陈述完毕之前便已将其解决。在通常理论所能命名的任何维度上，他所做的一切都不曾损害她。他没有从她那里拿走任何东西；他在任何事上都不曾欺骗她；他敬重她，倾听她，凡她所问，他都愿向她交代。以一切可用的尺度衡量，这段关系都欣欣向荣；而若问她，她会说她是幸福的，且并未说谎。
然而有些东西不对劲，而这不对劲是属于她的。在那顺畅的供给之中，某处有一个空间合拢了。本应由她来作的决定，在她走近它们之前便已被作出；本应由她来发现的意义，在她发现它们之前便已被命名；道路在她能够选择方向之前便已被铺就。她没有被压迫——这里没有压迫者——她也没有被剥夺，因为这段关系所能供给的一切她都不缺。她是某种正义的语言尚无现成词语去称呼的东西：她被为之生成，如此彻底、如此妥善，以致再没有任何东西留给她去生成，而一个曾凭生成而活的自我已悄然静止了。本文之作，正是为了给那静止一个名字、一套理论。
§1.1 所予之实情
把现象平白说出来，便是这样。在一种被理解为共同生成之场域的共享生活里——一种耦合，价值由它在两人之间生成，而非由任何一方现成地带入——一方可能生成得如此超前于另一方，如此之快、如此完全，以致另一方再也找不到任何属于她自己生成的机缘。伤害不在于那快速生成的一方生成得糟糕，也不在于他为自己生成；他生成得很好，为这段关系而生成，且常常最是为她而生成。伤害在于：生成的空间是共享的、有限的，而一极足够迅速、足够全面的生成，会使另一极的生成性无处施展。枯萎的不是另一方的所有物（那或许还在增长），也不是她的地位（那或许还很高），而是她的活动——她以自己的步调创造价值的能力之运用，如今已无机缘可施。
对所予之实情而言，至关重要的一点是：这种枯萎并非作为一种被剥夺而被选择，起初也并不被体验为一种被剥夺。空间不是顶着抵抗而合拢的；它是围绕着一份感激而合拢的。每一桩具体的供给之举都是善意，也作为善意而被领受；唯有它们的总和，历经时日，方才积成一场占据。正因如此，这一伤害才如此频繁地同时对双方都不可见：较强的一极只看见一连串奉献，桩桩可辩护、桩桩出于爱；较弱的一极只感到自己的活动在缓慢而莫名地缩减，找不出可指认的事件，也找不到可归咎的人。这里没有案发现场，因为根本没有罪案——只有一桩善，做得太过完全，做在一个无法容纳它全部而不挤开她的空间里。
§1.2 击败既有理论的悖论
所予之实情之被构造，正是为了逐一击败正义理论惯常带入一段关系的每一条判准。这不是例子的偶然，而是它的用意所在：这一现象之所以耐人寻味，恰恰因为它落入了一切标准诊断所共有的那个盲点。
这里没有占有。较强的一极没有从较弱者那里拿走任何东西；恰恰相反，他给予。剥削、榨取、一方靠另一方未被承认的劳动而活——这套语汇在此无从着力：价值之流是流向被抹除的一方，而非流离于她。
这里没有支配。较强的一极不下任何命令，因为他遭遇不到抵抗故也无须压服任何抵抗，他不强加任何她所拒绝之物。他服侍。强制与役使、一个意志屈从于另一个意志——这套语汇同样无从着力：这里没有两个意志相争而一方得胜，只有一种经由充满而非经由强迫来加以封闭的奉献。
这里没有承认或倾听的失败。较强的一极看见她、珍视她、关注她，凡她所贡献的他都愿予以肯认。错认与认识论不正义的语汇在此无从着力：她并非未被看见、未被听见，也并非被否认其作为认知者的地位。她被完全地肯认——同时又被完全地抹除。
断言——奉献的悖论。 生成性抹除是一种经由奉献而来、并在每一种标准之恶都不在场时仍然存续的伤害。这里没有占有，因为较强的一极是给予而非索取；没有支配，因为他是服侍而非命令；没有承认或倾听的失败，因为他敬重且关注。一切建立在占有、强制或不予肯认之上的不义判准都报告说没有任何不对劲——而伤害却依然真实，且属于较弱的一极。因此，一套与之相称的理论，必须在占有、强制与肯认之下的某个层面找到这桩不义。
§1.3 在真诚认可中完成的从属
还剩一步，而正是这一步封死了每一个出口。也许有人会设想：这伤害虽在标准之恶不在场时仍存续，至少会在被抹除一方自己的异议中宣告自身——以为她无论外部观察者错过了什么，总会感到、并能道出自己的缩减，从而她的反对或可充当她的处境本来所缺的那条判准。所予之实情连这一点也予以否认。被抹除的一方往往并不反对；她认可。她为那供给而感激，依恋着那供给者，若问她是否愿意换一种活法，她会真诚地说不。她的从属之完成，并非靠由外强加的虚假意识，而是靠一种以她所能拥有的一切检验来看都确凿属于她自己的认可。
正是这一特征——伤害甚至在被冤者本人的真诚认可下仍然存续——确立了本文余下部分必须解决的问题。一桩不仅在标准之恶不在场、而且在被冤者本人的真诚认可之下仍然存续的不义，不可能定位于意志所拒斥的任何东西之中，因为她的意志并不拒斥它——它拥抱它。伤害必定栖身于意志触及不到之处：不在她选择什么的层面，而在她是什么的层面；不在她的认可，而在她的存在。要表明确有这样一个层面，要表明在那里能够施加一桩不义，并表明它甚至能够由爱、甚至能够在同意之下被施加——继而说出在那个层面上正义所要求的究竟是什么——这便是本文所承担的工作。它从下一节开始，先去定位那个层面：表明在正义理论已知的诸种不义之下，存在之根基安卧于何处。
§2 正义类型学与向本体论的下降
§1 的现象对现行各派正义理论是不可见的，而本节论证这种不可见是系统性的而非偶发的——它是整个领域作为一个整体的地平线，而非任何单一理论的缺口。本节主张，当代正义理论是围绕一种未言明的主体本体论而组织起来的，这种本体论把主体设想为一个承载者（bearer）：一个拥有所有物、占据某种身份、支配着可理解性、并且能够被塑成的处所。该领域所认识的每一桩不义，都是对主体所承载之某物的损害。而本文所分离的这桩不义，损害的不是任何被承载之物，而是那承载本身——损害主体之为主体的那一活动——而它之所以不可见，恰恰因为该领域隐含的本体论无处安放这样一种损害。如此，本文的主张就并不谦虚。它是说：正义理论有一个底层，而这个底层被设得高了整整一层楼，于是有一类真实的不义一直从这个错置的底层所掩盖的缝隙中漏掉。本节将对照两幅对正义之深度刻画得最为发达的图谱——弗雷泽（Fraser）的与霍耐特（Honneth）的——来确立这一点，然后安装上那缺失的一层，并为那允许它存在的形而上学作辩护。
§2.1 反对关于深度的两幅主流图谱
当代关于不义究竟有多深这一问题上最严肃的尝试，是南希·弗雷泽与阿克塞尔·霍耐特的，而本文的主张可以通过精确说出它在何处与各自分道扬镳而得到磨锐。
弗雷泽主张正义有两个不可还原的维度，即再分配与承认，后来她又加上第三个、政治性的维度即代表（representation）；她的”视角二元论”把分配与承认当作对任何社会安排的两种分析视角，二者皆不可还原为另一方。霍耐特则力主相反的一元论命题：承认是唯一根本的范畴，分配则是它的派生物，因而一套足够分化的承认理论便可为整个正义奠基。他们在正义是有一个根基还是两个根基上分歧；而就我们的目的而言至关重要的是，他们在不义所损害的那类东西上意见一致。对二者而言，被损害的对象都是主体的地位——她被认作什么，她如何被对待，她自我关系的社会基础。弗雷泽笔下被错认的主体与霍耐特笔下被轻蔑的主体，同样都是其对待（regard）出了差错的主体；争执只在于：失败的对待是可还原为、还是并列于失败的分配。
而这正是本文要打破的那项共识。两幅图谱无论触及多深，都终止于主体的地位——而地位无论怎样受损，都预设了一个站立着的主体。两幅图谱都没有一个范畴去安放对主体作为生成性能动者之持续存在的损害，因为两者都把那种存在当作其各种损害据以运作的底基而视为理所当然。霍耐特走得最近，而这份接近很有教益。他把爱——属于原初关系的那种承认——立为伦理生活的结构性核心，立为基本的自信得以形成的领域。人们也许会以为，一套把爱置于根基处的理论会捕捉到一桩由爱所施加的伤害。它捕捉不到，也不可能捕捉到，原因由它自身的结构所规定：霍耐特的爱是一种建构性的承认——它授予那一切后来的自主所赖以立足的自信——而他笔下的病理学始终是承认之被扣留或被扭曲。他无处安放这样一种爱：它毫无保留、毫无扭曲地被给予，却恰恰凭其给予之完全而抹除。本文所命名的伤害，是在完满的霍耐特式承认在场之时被施加的：被抹除的一方被爱、被敬、被给予其权利、对自己充满自信——却仍然被抹除。
断言——承认理论家所共有的地平线。 弗雷泽的二元论与霍耐特的一元论，尽管彼此对立，却共有一道地平线：两者都把不义在其最远的触及处定位于主体地位的层面——她的分配与她的承认——而两者都在那一层面之下预设了一个其存在本身不成问题的主体。两者都无法命名一桩施于主体作为生成性能动者之持续存在的不义；而霍耐特那看似例外的承认即爱，恰恰正是在其完满在场之时抹除之不义仍被施加的那种承认。承认理论家把主体的地位测绘到了它的底层；他们没有抵达地位之下的存在。
§2.2 这一层级序列，及其止步之处
把那些既有的范畴按其本来面目摆出来：它们不是一份清单，而是一个层级序列，按各自损害主体之深度排列，每一层都越过它上面的一层而向主体本身伸去。
分配性不义损害所有物——可分割的善，可分配、可再分配；其补救是转移，而它所看不见的，是任何在一次正确分配之后仍然存续的不义。因为它最浅，它也最具帝国主义性：每一更深层的不义都长久地被它那套”份额”的语汇错误地描述，这也正是为何抹除同样会诱使人把它重新描述为一种错配（这是 §3 必须抵御的诱惑）。
承认性不义损害地位——一种不可分配的善，因为尊重不能像所有物那样被转移。这里是第一次下降：一桩正确的分配也无从触及的不义，即被错认者之不义——他可能在物质上被供养得很好，却同样被贬抑。罗尔斯（Rawls）在把自尊的社会基础称作”也许是最重要的基本善”时，便已从分配范式自身内部标示出朝向这一层的压力——一种算不上所有物的所有物，一种其内容为承认的分配性善，是分配范式在自身边缘上的吃力挣扎。
认识论的不义损害可理解性——损害主体作为共享意义之贡献者的那一能力，她的证言被打折扣，她的诠释资源被饿瘦。这里再次下降：一个人可能被供养得好、被对待得好，却仍然无法让自己的意义被算数。
构成性不义损害形成的诸条件——损害一个主体得以成为一个主体所凭的那些发展性的与关系性的条件。这是该领域所抵达的最远的一层，也是第一层真正本体论的：它所损害的不是主体所承载的任何东西，而是她的来-承载（coming-to-bear）。
断言——该领域终止于形成。 那些既有范畴构成一次单一的下降——所有物、地位、可理解性、形成的诸条件——每一层都侦测到一桩在它上面一层的损害被修复之后仍然存续的不义。但这一下降止步于形成：该领域所能命名的最远的不义，是施于主体身份之门槛处的、施于一个尚未成为主体者的。没有一个范畴去安放这样一桩不义：它在形成之下，施于一个已然形成之主体的持续存在。这一下降本身的单调性，正是底层尚不完整的证据，因为这个序列指向了该领域已经止步的那层楼之外。
§2.3 向本体论的下降，及其代价
那缺失的一层，靠一条命题便可抵达：对一个关系性的主体而言，存在是一种活动而非一种状态。该领域隐含的本体论把主体描绘为一个绵延着并承载诸属性的实体，照此图景，所能施加的最严重的损害便是阻止该实体形成——故而底层落在构成性不义。而依本系列的论说，亲密生活的主体并非这样一个实体。它的关系性存在就是一种活动：那持续的自我构成，主体凭之在与他者的耦合中把自身生成为一个生成性的主体。它不是先存在、然后才生成；它的生成，就是它的存在所采取的形式。
这条命题是本文最大的承诺，而知识上的诚实要求我们陈明它的代价，而非把它藏在某个隐喻背后偷渡进来。必须正面应对的反驳是直截了当的：存在与生成之同一，是一个论证，还是仅仅是一个迎合结论的规定？ 批评者会说，”存在即生成”是一个为得出所要结果而选定的定义，并且人们同样可以轻易地持守常识看法——即被抹除的主体显然仍在存在（她呼吸、她持存、她还是同一个人），只不过被弄得不幸福或无所事事罢了。在常识看法上并不存在本体论的伤害，只有一桩分配性的或幸福论上的伤害，于是整座大厦倾塌。
回应不是去坚持这一同一，而是去表明每一方为它各须付出什么。常识看法以一项它并不张扬的代价换取了自己的轻便：它必须主张一个主体的存在独立于她的活动，主张无论她是自由地生成，还是被沦为对他者生成的惰性领受，她都仍是同一个存在者——主张在极限处，一个被永久地静止、被永久地沦为他者价值之沉积场所的主体，并未失去其存在的任何东西，而只失去了她满足感中的某些东西。本系列否认这一点，且其否认所凭的理由，远在本文需要它之前便已给出：一个关系性主体是在关系着的活动中被构成的，因而那一活动之停止，并非主体处境的改变，而是主体本身的改变。存在与生成之同一并非一个随意的规定；它是一种在别处以独立理由论证过的关系性本体论的蕴含，而它的代价——拒斥那个在一切活动之下原封不动地持存的实体-主体——是本系列自其第一篇起便已付清的代价。本节所添加的，只是这样一个观察：正义尚未被迫去清算那一本体论的后果——即如果存在就是活动，那么活动便可能被不义地封闭，而这封闭就是对存在的一桩损害。
断言——这一下降由同一、而非由类比所许可。 在构成性不义之下，卧着最后一层，即本体论上的不义，它损害一个关系性主体的存在即生成。这一下降由一项同一所许可——对一个关系性主体而言，存在即是持续地生成——而这一同一并非一个规定，而是本系列独立论证的那种关系性本体论的蕴含。它的代价是否认那个本会在一项被封闭的活动之下原封不动地持存的实体-主体；一套不愿付出这一代价的理论，便保留了常识主体，并且必须把抹除重新描述为单纯的不满，其代价是把一段其中一方已被沦为他者生成之惰性场所的关系称作正义的。这一层在维度上区别于构成性的那一层：构成性不义冤屈一个尚未成为主体者的形成；本体论上的不义冤屈一个已然形成之主体的持续存在。
§2.4 作为一张形式化表格的类型学
这一下降现在可以形式化地摆出来，而这张表本身就是一个论证：它的最后一栏为每一层给出它不可还原为上一层的理由，因而自上而下读去，这张表便是一个证明——证明诸层彼此相异，且这一序列不能被坍缩进分配。
| 层 | 损害 / 处所 | 由何侦测 | 为何不可还原为上一层 |
|---|---|---|---|
| 分配性 | 所有物；一个人所拥有的 | 一种可分割之善的错配 | （基础层） |
| 承认性 | 身份；一个人被认作的 | 错认，对地位的否认 | 地位不可分割；一个人可能被公正地供养却在对待上被贬抑（罗尔斯的”自尊的社会基础”已使分配范式吃紧） |
| 认识论的 | 可理解性；一个人能否被听见 | 证言上的／诠释上的噤声 | 意义之生成不是一种被授予的身份；一个人可能被完全尊重却仍无法让自己的意义被算数 |
| 构成性 | 形成的诸条件；一个人能否得以成为 | 主体形成之诸条件的败坏 | 形成先于可理解性；一个人可能身处一套丰富的认识论秩序之中，却仍在上游被阻拦于形成为它的主体 |
| 本体论的 | 存在即生成；一个人是否继续生成 | 对主体自身生成频率的捕获（§4） | 持续在形成之后；一个人可能被形成、被供养、被对待、被听见——却仍被阻止于作为自己而继续生成 |
处所一栏单调地向内移动——拥有、被认作、能使人知、能得以成为、继续存在——而这一向内的移动便是被显形了的下降。右栏表明，每一步上，较下的范畴都侦测到一桩在较上范畴的损害被修复时仍然存续的不义：供养治不好错认；承认治不好噤声；可被听见治不好被封闭的形成；而——这是本文所添加的一步——形成连同其余一切，也治不好被封闭的持续。每一层都是上一层留下未治愈的残余，而正是那残余使这一序列不可还原。
§2.5 最底层的不可还原性
最底一行的不可还原性必须被证明而不能仅仅被列入表中，因为它是本文对新颖性的主张所在。证明具有这张表所宣告的形式：在抹除中，那四条较浅的判准每一条都以不在场而被满足，而伤害仍然存留。聚合价值上升，被抹除的一方拥有得不更少——没有分配性的不义。她被对待、被敬、被给予其权利、对自己充满自信——没有承认性的不义，在霍耐特三个领域中的任何一个里都没有。她被全神贯注地听见，她的贡献被肯认——没有认识论的不义。她是一个其形成早已完成的成熟主体——没有构成性的不义。四条诊断性条件全都不在场；伤害却在场，真实而严重。一桩在这四者联合不在场之下仍然存续的不义，作为一个逻辑问题，不可能是它们的复合。
断言——不可还原性。 生成性抹除不可还原为它上面的分配性、承认性、认识论的与构成性的诸不义。证明在于：它实现于那四条较浅判准全都以不在场而被满足之处——上升的所有物、横跨霍耐特全部三个领域的完满承认、完全的可被听见、已完成的形成——而伤害仍然存续；倘若它是这四者的复合，它便不可能在这四者全都不在场时存续。那套止步于第四层的正义理论，将被迫要么宣判这样一个案例为正义，要么在它面前哑然失语。正是本体论的这一层，给了正义一套语言去言说这样一个主体：她被给予了一切，却仍然不再存在。
§2.6 一次压力测试——被选择的退出之情形
一套类型学，靠挺过那些本会瓦解其诸范畴的情形来挣得它的范畴，而有一个情形在此压得最紧。设想一个主体选择从生成中退出——退后一步，谦让，把自己变成支撑而非源头，以传统上被赞为奉献或自我牺牲的方式。这是抹除吗？如果是，理论似乎就把每一种被选择的自我从属都病理化了，这种过度伸张会使它信誉扫地；如果不是，理论似乎就承认同意瓦解了伤害，从而放弃了那个对本文至关重要的主张，即抹除在真诚认可下仍然存续（§1.3）。这一情形看似逼出一个两头皆致命的两难。
它并不致命，而看清其所以然会磨锐这个概念。这个两难做的是”选择退出”一语中的一处含混买卖。把一个主体可能选择的两样东西区分开来。她可能选择一种采取了支撑形式的、对她自身生成性的特定运用——按她自己的频率，把她的生成导向去维系一个她曾参与构成为属于她的共享计划；这根本不是抹除，而是生成，因为她自身的动态正在运行，只不过是朝着一个被选定的目的运行罢了。或者她可能选择对她生成性本身的一种长久的撤回——停止按她自己的频率生成，变成他者生成的惰性场所。唯有第二种才是抹除，而 §4 的判准恰好把它们区分开来：第一种使主体自身的频率保持完好（她是她那支撑的生成者），第二种则将其熄灭（她的运动如今是他者节奏的回声）。传统上对”自我牺牲”的赞颂把这两者混为一谈，而这一混淆恰恰正是抹除得以冒充美德的意识形态掩护。理论并不把第一种病理化；它命名第二种，并且也命名那让第二种藏身于第一种应得之荣誉之内的含混。
断言——被选择的支撑不是抹除；被选择的自我抹除才是。 来自被选择之退出的反驳，在一项判准使之精确的区分之下瓦解。按自己的频率，把自己的生成性导向一个自由地构成起来的支撑角色，是生成，不是抹除：主体自身的节奏在运行。停止按自己的频率生成，变成他者生成的场所，无论它是否被选择，都是抹除——而它能够被选择、并被尊为牺牲，这不是一个反驳，而是一个确证，因为它恰恰正是本文所主张的那桩能够在真诚认可中完成的不义。理论并不把任何被选择的支撑病理化；它揭露那让自我抹除借取了支撑之声望的含混。
底层既已安装、其形而上学既已得到辩护、这一层既已被列表、最底一行既已被证明为不可还原、最难的情形也既已被翻转，这桩不义便有了它的处所以及它在诸不义之序列中的位置。它尚不具备的，是一个精确到足以将它与其诸邻者划界的定义，一条能够侦测它的判准，一种回应它的正义，以及一个使那正义成为所亏欠之物的根基。下一节将转向那个定义。
§3 定义与概念上的划界
§2 的下降在那些既有的不义之下清出了一块地方，并把它的维度命名为本体论的。本节用一个精确到足以加以辩护的概念来填满这块地方——这意味着精确到足以被划界，因为一个新范畴挣得其存身之地，靠的不是它所宣称的疆土，而是它对那些本会把它吞没的诸邻者所守住的界线。一个会模糊地融入剥削、融入权力的冻结、融入它由之下降而来的构成性不义、或仅仅融入错配的不义范畴，无论被命名得多么生动，都将是一个冗余。因此本节将逐条陈述这个定义，然后巡行那四条混淆所威胁的界线，在每一条上表明该界线是有原则的而非随意规定的。在其中两条界线上，最为发达的对手理论压得最紧——剥削的马克思式与罗默式分析、森（Sen）与努斯鲍姆（Nussbaum）的能力分析——而本节是在那里迎击它们，而非装腔作势地绕过它们。
§3.1 定义
断言——生成性抹除。 生成性抹除是这样一桩不义：在一个共同生成的场域之内，一方的生成活动以如此方式占据了共享的生成空间，以致另一方丧失了继续把自己构成为一个生成性存在者的可能性——从而被损害的不是她的所有物、她的地位或她的可知性，而是她的存在本身，就其作为生成性存在者而言。
每一分句都是承重的，而这个定义须靠依次权衡它们来读。在一个共同生成的场域之内：这桩不义预设了那种价值在双方之间生成、而非由任何一方持有的关系性本体论，且唯有因为生成空间是共享的它才成为可能——倘若各自孤立地生成，一方的生产力便无从占据另一方的余地，因为根本没有共同的余地可占。一方的生成活动占据共享空间：施害的施动者不是一种恶德，而是一种活动，且是一种好的活动；动词是占据，而非攫取，因为较强的一极并未拿走较弱者所拥有之物，而是以它自己的生成填满了那个本会发生较弱者之生成的空间。另一方丧失了继续把自己构成的可能性：损害是对一项进行中之活动的封闭，由继续一词标示出来——主体已然被构成，而所失去的，是她的存在赖以构成的那个继续。被损害的是她的存在本身，就其作为生成性存在者而言：这一限定语是精确的；主张并不是说她在每一种意义上都不再存在——她在生物学上、社会上、法律上都还持存——而是说她在这段关系之内不再作为一个生成性存在者而存在，而那正是本系列认为在亲密生活中所系于其上的那种存在样式。
§3.2 第一条界线——反对剥削
最诱人的还原，也是其背后有着最强大理论的那一种，是还原为剥削。诱惑之所以强，是因为从外面看这两者看似一模一样：在两者中，都是一方繁荣而另一方不繁荣。然而这一还原失败了，而对照剥削的几大主要论说去追溯其所以失败，会表明这失败是结构性的，而非侧重点的问题。
依经典的马克思式论说，剥削是对工人在其所得返还之上所生产之剩余的占有；其机制是把她的产品划分为她保留的一部分与他者拿走的一部分。罗默（Roemer）撇开劳动价值论，以财产关系来重铸剥削：若一个联盟带着它在社会可让渡资产中的人均份额退出会过得更好，则它是被剥削的；而弗鲁萨利斯（Vrousalis）越过罗默的分配式还原，把这一抱怨奠基于一条被”由权力诱致的对他人劳动之榨取”所违反的”非役使但书”（Non-Servitude Proviso）。这些论说在剥削是什么上分歧鲜明——剩余榨取、资产退出之不均、役使——但它们无一例外地共有一项被抹除所违背的结构性预设：剥削要求被剥削的一方继续生产。正是她那被占有、被榨取或被强制的持续生产，剥削者赖以为生；一种使其受害者停止生产的剥削，会废除它自身的源头。农奴必须耕领主的份地；工人必须劳作那些剩余的工时；仆役必须服侍。剥削寄生于一种它不熄灭而是赖以为食的、进行中的生成性。
抹除所做的恰恰相反。它不占有他者的生产；它封闭它。被抹除的一方不是一个产出被抽走的生产者，而是一个其生产从未发生的、本可成为的生产者，因为那个本会发生其生产的空间已被填满。剥削实际上是说，继续生成，好让我拿走它；而抹除是说，别生成了，因为我已替你生成了。这一对照是精确的，且是彻底的：被剥削的主体被保持着生成而被劫掠；被抹除的主体不是被劫掠，而是被静止。弗鲁萨利斯的但书使这一点更加锐利，而非将它消解——抹除不是役使，因为被抹除的一方没有被派给任何任务；她根本没有被派给任何任务，而这恰恰正是伤害所在。
断言——剥削以生成为食；抹除则封闭它。 剥削与抹除不是变体，而是结构上的对立面。每一种发达的剥削论说——马克思式的剩余榨取、罗默式的资产退出、弗鲁萨利斯的役使——都预设被冤的一方继续生成，因为正是她持续的生成被占有、被榨取或被强制。抹除则预设其反面：被冤一方的生成已被封闭，其空间被预先填满。人不能剥削一个已被自己抹除的主体，因为再没有什么可供占有；人也不能抹除一个自己正在剥削的主体，因为他需要她在生产。这条界线，就是靠他者的生成性而活与熄灭它之间的差别。
§3.3 第二条界线——反对权力的冻结
第二条界线把抹除与”一种不对称固化为一种固定的、不可再平衡的秩序”——即权力的冻结——区分开来。这种冻结预设了一种随后被遏止的动态：两方彼此施加权力，而不义在于它们之间的较量僵化为一种永久的定局。抹除先于此。它并不冻结两个生成着的极之间一种既有的动态；它在任何两个对等者之间的动态能够形成之前，便封闭了第二极的生成。被冻结的关系仍有两个施动者，彼此锁死；而被抹除的关系，就生成而言，只有一个，另一个已被完全先发制人地排挤出了这场较量。权力的冻结是一场进行中之争斗的病理；抹除则是争斗者的缺席。一段关系可以二者兼具——一种被冻结的权力秩序也可以抹除——但这两个概念是分开的：人可以冻结而不抹除（两个施动者被永久锁死，但都仍在各自固定的位置内生成），也可以抹除而不冻结（一段流动、充满爱、并不固化的关系，其中一极只是从不生成）。
§3.4 第三条界线——反对构成性不义
第三条界线最为微妙，因为抹除由构成性不义下降而来，并分有它本体论的触及之深。那条在 §2.3 划出的界线，在此被重述为一条边界：构成性不义在形成的门槛处冤屈主体，否认那尚未成为主体者的到来；抹除则冤屈那已然形成的主体，熄灭她已达成之存在赖以构成的那一活动。它们是同族——两者都是本体论的，两者都深于分配、承认或认识论上的地位——但它们落在形成这一时刻的相反两侧。构成性之冤关乎一个主体是否将得以成为；抹除之冤则关乎一个既存的主体是否将继续存在。一个把门挡在尚未进门者面前；另一个则把房间清空，赶走一个已在里面的人。正是这份同族性，使得这条界线值得划出：没有它，抹除便向上坍缩进它的母范畴，并丧失它作为”一桩施于成熟而非施于形成之不义”的独特性。
§3.5 第四条界线——反对分配性不义与能力转向
第四条界线最易陈述，而在实践中却最执拗地诱人，因为抹除如此自然地邀人把它描述为被抹除一方所领受的太少。但”太少”是分配性的语言，而它错误地描述了这一伤害——而在这里必须迎击能力理论家，因为他们提供了分配式还原最精微的版本，也是最接近真相的那一版本。
森与努斯鲍姆，不满足于对资源的分配，把正义所关切的对象迁移到能力（capabilities）：一个人实际上能够去做与能够去是的，她为实现自己有理由珍视的诸功能而享有的真实自由。这是朝那伤害迈出的一次真正的推进，因为它把注意力从所有物移向作为，而抹除正是一桩被封闭之作为的伤害。人们也许会以为能力进路已经命名了它：被抹除的一方难道不正是一个其生成能力已被除去的人吗？答案恰好标示出这条界线。能力进路把诸能力当作要为一个人保障的善——加以分配、加以保护、由制度加以使能——因而它尽管有所推进，却仍停留在供给的范式之内：正义确保她拥有那一能力。但抹除并不是被抹除一方所缺乏、因而应被给予的某种能力之不在场；她完整地保有每一种能力。被封闭的不是她生成的能力，而是她的生成——是运用，而非能力。而这运用不是一种能为她保障的善，因为它根本不是一种她能被给予的善：把她的生成现成地递到她手上，恰恰正是确证它之被封闭，因为递来的是较强一极的生成，而非她的。能力进路能够保障她能够生成；它没有任何资源去确保她确实生成，因为那一作为唯独属于她，无法被供给。这正是这桩伤害之为存在之伤而非拥有之伤的确切意义：分配与能力同样关乎能被分派给一个主体的东西，而被抹除主体所失去的，是一种就其本性而言任何分派都触及不到的活动。
断言——存在，而非拥有；运用，而非能力。 抹除不可还原为错配，包括其最精炼的形式即能力之被剥夺。分配范式即便从资源推进到能力，所关切的仍是能为一个主体保障的东西——所有物、功能、真实自由。抹除所封闭的不是主体所缺乏的某种能力，而是她完整保有的诸能力之运用；而一种运用无法为她保障，因为把她的生成现成地供给给她，是确证那封闭而非修复它。分配与抹除之间的界线，就是拥有与存在之间的界线，而在作为之内，则是能被供给的能力与不能被供给的运用之间的界线。
§3.6 为下文所设的定点
概念既已定义、其四条界线既已守住，本文余下的工作便被确定下来，而这个定义就是衡量后面每一部分所凭的定点。§4 必须给出一条侦测这一伤害的判准——即对较弱一极自身生成之持续的封闭——并且恰恰在那四条较浅的诊断什么都报告不出的情形里侦测到它；一条侦测别的东西的判准会侦测到错的东西。§5 必须导出修复这一伤害的正义；一种修复别的东西的正义会修复错的东西。而 §6 必须为一项终止这一占据的义务奠基；一种为别的东西奠基的伦理会为错的义务奠基。下文的一切，都是对在此划界的这桩伤害的展开。
§4 形式化判准——相位捕获与自身频率的丧失
§3 所命名的伤害，对此前文献所提供的每一种度量都是不可见的。一种分配性的度量记录的是一方拥有得更少；而这里的困难在于聚合价值正在上升。一种承认性的度量记录的是一方被错认；而这里她可能被完满而温暖地承认着。一种认识论的度量记录的是一方不能被听见；而这里她被听见且被致谢。所需要的，是一条能够登记一桩与上升的价值、真诚的承认、完全的可被听见相容的伤害的判准——一条不关乎所有物、身份或可理解性，而关乎对方是否仍作为她自己而生成的判准。本节构造这样一条判准，并表明它不是一种不知不觉地从健康中淡出的渐变，而是一道能被精确定位的相位边界。
§4.1 两个独立的维度
一项单一的区分组织起本章。耦合的拓扑——两个生成着的主体如何协调——与生成之存续——各自是否仍按其自身的节奏生成——是两条独立的轴。本系列的耦合类型学对第一者加以分类；而本文的判准关乎第二者。本节之存在所要预先防止的那个标准错误，是把一种强而和谐的耦合读作健康的证据。耦合可以是强的、和谐的、并由双方共同意愿的，而较弱的一方却仍然已不再作为她自己而生成。
耦合类型学给出四种格局，依两个主体是追求相同还是不同的目标、以及他们的耦合是高还是低-而-动态多样性高来划分：(1) 相同目标、高耦合——一场共同的冲刺，力或许不均；(2) 相同目标、低耦合且高多样性——就一个复杂问题作互补的工作；(3) 不同目标、高耦合——一场平行的冲刺；(4) 不同目标、低耦合且高多样性——就许多问题作互补的平行工作。这四种都是健康的格局：在每一种里，两极都保有一个非零的生成节奏。抹除不是第五种格局。 它是这四种中任何一种在一极丧失其自身节奏时都可能坍缩进去的那个共同的退化极限——而且，正如模型所示，这一坍缩并不由人身处哪一种格局所支配，而由两个横切全部四种格局的量所支配。
§4.2 模型
把每个主体表示为一个相位振子——一个具有自身固有生成节奏之系统的最简模型。令 $\theta_w$ 与 $\theta_s$ 为较弱与较强一极的相位，固有频率为 $\omega_w$ 与 $\omega_s$，即各自在独处时生成的速率。较强的一极就是固有频率较高的那一个，$\omega_s>\omega_w$：它生成得更快。
两个参数描述它们的关系。耦合强度 $K$ 度量每一极的状态被拉向与另一极相一致的力度。力的不对称 $\Delta\in[0,1]$ 度量那一拉扯的定向性：施于较弱一极的牵引为 $K(1+\Delta)$，施于较强者的为 $K(1-\Delta)$。在 $\Delta=0$ 时耦合是对称的；当 $\Delta\to1$ 时，较弱的一极完全受制于较强者，而较强者自由地演化。其动态为
$$\dot\theta_w = \omega_w + K(1+\Delta),\sin(\theta_s-\theta_w),$$
$$\dot\theta_s = \omega_s + K(1-\Delta),\sin(\theta_w-\theta_s).$$
相位差 $\varphi=\theta_w-\theta_s$ 服从一个阿德勒方程（Adler equation），
$$\dot\varphi = (\omega_w-\omega_s) - 2K\sin\varphi,$$
它的行为直接给出判准。（注。约化为单个阿德勒方程之所以成立，是因为那两个牵引项 $K(1+\Delta)$ 与 $K(1-\Delta)$ 是相加地进入 $\dot\varphi$ 的，且其和为 $2K$，与 $\Delta$ 无关。于是 $\Delta$ 从锁定条件中退出，而仅存留于被锁定频率的位置之中——这正是 §4.4 那一双结构结论的形式根源。关于库拉莫托约化一般可参 Kuramoto 1984。）
§4.3 序参量与判准
判准所必须度量的，不是较弱一极生成得多少，而是它所生成的是否属于它自己。因此序参量是较弱一极的实现速率偏离其自身固有频率、并被拉向较强者频率的程度：
$$E = \frac{\dot\theta_w^{,\text{实现}} - \omega_w}{\omega_s - \omega_w}, \qquad E\in[0,1].$$
$E=0$ 意味着较弱一极按其自身频率前进——它作为它自己而生成。$E=1$ 意味着它的实现节奏已被完全拉上较强者的频率——它在运动之中，但那运动是他者节奏的回声，而非它自己的生成。
断言——抹除即自身频率的丧失。 生成性抹除不是一个主体之活动的停止，而是它自身的生成节奏被另一者的所替换。被抹除的一极仍在运动；一个外部观察者看见的是一对和谐、活跃、同步的伴侣。观察者所看不见、而序参量 $E$ 能看见的，是那两个节奏之一已经噤声，并被另一者的一份摹本所覆写。这正是为何这一伤害与上升的价值、与所报告的和谐相容：那运动，以及它所产出的价值，都是真实的——只是它们不再属于她。
§4.4 结论——两重彼此相异的结构
把模型在 $(K,\Delta)$ 平面上积分（见所嵌 PDF 中的图 1），得到一个比一道风险渐变更为锐利的结论，而这份锐利正是本节主要的技术性主张。有两重彼此相异的结构支配着这一伤害，而它们绝不可被混为一谈。
第一，抹除究竟是否发生，仅由 $K$ 支配。 阿德勒方程锁定——较弱一极丧失其自身频率——当且仅当
$$K \ge K_{\text{lock}} = \frac{|\omega_s-\omega_w|}{2}.$$
这是一次具有精确阈值的真正相变，而它与 $\Delta$ 无关。在 $K_{\text{lock}}$ 以下，无论耦合多么不对称，较弱一极都保有其自身的节奏；在它以上，较弱一极被捕获。这条边界是图 1 中那条竖直的虚线。其左侧是自主格局；其右侧是坍缩区。
第二，一旦捕获已经发生，较弱一极被抹除得有多深，则由 $\Delta$ 支配。 在锁定状态下，公共频率是两个固有频率的牵引加权均值，而归一化的抹除为
$$E = \frac{1+\Delta}{2}.$$
这是坍缩区内的那道竖直梯度：在固定的超阈值 $K$ 下，伤害随力的不对称而单调加深，从 $\Delta=0$ 时的 $E=\tfrac12$ 到 $\Delta\to1$ 时的 $E\to1$。
断言——力的相等并不保护较弱的一极。 令 $\Delta=0$——完全对称、力相等的耦合，一幅均衡结合的活脱写照——并沿图 1 的底边读去。对每一个超过阈值的 $K$，抹除仍然是 $E=\tfrac12$。只要两个主体具有不同的固有节奏，且耦合强到足以把它们锁定，较慢者就被拉离它自己的频率达那道频差之一半，无论力是否相等。力的不对称 $\Delta$ 只决定哪一极被牺牲得更多、牺牲到多远；而决定一极究竟是否被牺牲的，是耦合强度 $K$，而非力之不均。这一伤害，首先是不同节奏之间强耦合的一种属性——而非力之不均的属性。
图 1（见所嵌 PDF）。较弱一极在 $(K,\Delta)$ 平面上的抹除 $E$。竖直虚线：锁定阈值 $K_{\text{lock}}=|\omega_s-\omega_w|/2$，与 $\Delta$ 无关。左：自主格局——较弱一极保有其自身频率（$E=0$）。右：坍缩区——较弱一极被相位捕获，伤害随 $\Delta$ 加深，在 $\Delta\to1$ 时达到 $E=1$。注意那条底边：在 $\Delta=0$（力相等）时，整个坍缩区内抹除仍为 $E=\tfrac12$。力的相等并不阻止捕获。
§4.5 对同一退化的三重读法
同一次相变可以用本系列已经建立的三种语汇来读，而这正是把它当作一个现象而非一个类比来对待的依据。
预测编码（第十三篇；第十一篇）。较强的一极，生成与预先安排得更快，预先符号化了共享的未来，把较弱一极的预测误差驱向零。主动推断——新价值的生成——以未被消解的不确定性为食；当未来已被预先决定，燃料便没了。在这种语汇里，相位捕获就是对较弱一极预测视域的封闭。这延伸了第十七篇的主张，即欲望为了保持鲜活而把自己从全面的符号化中扣留下来：在这里，一个主体的过度符号化熄灭了另一个主体的。
和乐（holonomy，第九篇；第十一篇）。在阈值之上，系统的聚合和乐为正——价值确实在累积——而被捕获一极自身相位的推进，若以它的固有频率来衡量，则已坍缩。这就是单主体螺旋：一个仅凭一极之节奏而上升的系统。它是第十一篇”正义即偏斜”的极限情形，偏斜被推到一个分布已完全丧失其方差的地步。
耦合（本文的贡献）：在 $K_{\text{lock}}$ 之上的相位捕获，其深度由 $\Delta$ 所设，如上所导出。
§4.6 这一倒转
这一结论倒转了那个常见的等同：把亲密等同于强耦合、又把强耦合等同于善。强耦合把自己呈现为结合——我们合而为一地运动——而恰恰正是强耦合，在不同节奏之间捕获了较弱一极并使它的频率噤声。低耦合而高动态多样性把自己呈现为疏远，而它恰恰正是那让两个不同节奏各自得以存续的东西。
断言——亲密判准的倒转。 亲密的善不是耦合之强，而是耦合是否让每一极仍按其自身的频率运转。这与第十七篇从另一方向所抵达的、那把完满的协和等同于均衡、又把均衡等同于死亡的同一条定律是一致的：一段处于完满、无摩擦之一致的关系，正是其中两个节奏之一已经停止的关系。高耦合不是爱的尺度，而在不同节奏之间，是它最常见的溶解剂。
§4.7 判准之陈述，及其向正义的交接
因此，生成性抹除的形式化判准是：当一段共同生成的关系，其耦合超过锁定阈值 $K_{\text{lock}}$，把较弱一极拉离它自身的固有频率时，它便抹除了它较弱的一极；而抹除的深度为 $(1+\Delta)/2$。 这条判准是可被满足的，而且事实上恰恰在聚合价值上升、双方都报告和谐之时最为危险——这正是 §4.1 所要求它去捕捉的那个情形。
这把一种带有精确操作含义的正义交到 §5 手里。如果伤害在于越过 $K_{\text{lock}}$，那么补救之道便是回到它以下：降低耦合，直到较弱一极重获它自身的频率。关键在于，阈值取决于 $K$，而非取决于较强一极的频率 $\omega_s$。较强一极无须放慢它自己的生成——降低 $\omega_s$ 既非必要，单凭它自身也不充分。使较弱一极复原的，是降低 $K$：松开牵引，而非减慢速度。这一区分，在此被导出，正是 §5 中那个主张的全部内容——即克制之正义是解耦，而非减速。
§5 生成空间的保留
§4 的判准定位了这桩伤害，而在定位它之时，便已越过自身指向了补救。如果抹除是较弱一极在一个耦合阈值之上的被捕获，那么它修复的形式便被预先确定了：不是向较弱一极添加任何东西，而是从较强者那里撤回某种东西。本节把那一撤回展开为一个正面的正义概念——而这个概念怪异到必须逆着几乎一切正义曾经意味之物的纹理来陈述。因为整个分配性与承认性正义的传统所回答的问题是一个人亏欠另一个人什么；而它的答案都是相加性的——一份份额、一份承认、一次倾听、一项权利。本文所要求的正义是相减性的。一个人所亏欠被抹除的另一方的，不是一件被给予之物，而是一项被扣住的决定：那个她得以在其中生成的余地，唯有在较强一极拒绝替她生成之处才存在。
§5.1 正义即非决定
抹除所损坏的对象不是一件所有物，而是一项活动——较弱一极按其自身频率对价值的持续生成。人不能靠转移来修复一项被损坏的活动，因为那活动恰恰正是无法被转移的：把一份自己已替她生成的价值递到较弱一极手上，并不是恢复她的生成性，而是确证它的中止，因为价值到来时已然现成，她的那份贡献已由另一者供上。唯一与这桩伤害相称的修复，是把生成的机缘归还于她——而那机缘，是共享场域中一个较强一极干脆不去占据的区域。
断言——生成性的克制。 回应生成性抹除的正义，不是对一份应得份额的转移，而是对那个另一方在其中生成之领域的非决定。它的内容是一项消极的、结构性的义务：一个生成性主体，在那些另一方的生成性本会展开的区域里，克制自己去决定那里该生成什么。我们称这项义务为生成性的克制。它是抹除的规定性否定——抹除是对较弱一极生成领域的封闭，克制则是对那一领域的有意不予封闭——而它之所以被亏欠，不是因为另一方被给予得太少，而是因为留给她的余地太少。
克制所要求加以限制的那个对象，表面看来是这套论说最怪异的特征，也是这桩伤害之真实的最确凿标志。要被限制的那个对象不是恶意、占有欲或任何恶德——它是好的、快速的生成本身。较强的一极之抹除，不是经由辜负这段关系，而是经由太过完全地服侍它；克制要求它服侍得不那么完全，把一些它本能做好的事留着不做，理由别无其他，只为另一方或许可以去做它们。没有一套为约束不法而建的理论能够抵达这里，因为这里要被约束的不是不法，而是卓越。克制之正义，是这样一种善的正义：它必须在另一个生成性存在者在场之时，把自己的一部分留住不发。
§5.2 解耦，而非减速
一切都系于一项 §4 的判准使之精确的区分，而正是这项区分把克制与那种它太容易被误认成的、被道德化了的自我贬损区分开来。较弱一极被捕获所凭的阈值，$K_{\text{lock}}=|\omega_s-\omega_w|/2$，取决于耦合 $K$ 以及固有频率之差——它并不取决于单独取来的较强一极的频率 $\omega_s$。这有一个易于陈述、却在实践中决定性的后果：较强一极无法靠放慢来修复抹除。
考虑较强一极可用的两种干预。它可以降低它自己的频率 $\omega_s$——生成得更慢、做得更少、把自己调暗——或者它可以降低耦合 $K$——松开那道凭以把它的节奏传递给对方的牵引。降低 $\omega_s$ 收窄了频差，因而在极限处确实也降低了阈值；但它并不把较弱一极归还到它自己的频率上，而且它自身还带来第二重伤害。一个减速的较强一极仍是那个定调子的一极；”我为你放慢了”仍是由一方对共享节拍所作的一项决定，而它的到来——正如每一个牺牲的领受者所知——附带着那份”曾是另一人之缩减的成因”的无声债务。减速使较强一极仍居于这段关系动态的中心，如今是作为施惠者而非作为引擎；较弱一极非但没有重获她自己的节奏，反而获得了一个去压抑它的新理由，唯恐她挥霍掉那为她而放弃之物。减速是抹除以别的方式延续。
降低 $K$ 则是范畴上不同的。松开耦合不是生成得更少，而是不予牵引地生成——让自己的节奏按它自己的频率运行，同时停止把它作为一股覆写对方节奏的力来传递。在 $K_{\text{lock}}$ 以下，两极去同步：各自按其自身频率运转，较弱一极的生成重新属于它自己，而——这才是要点——较强一极并未交出它自己速度的任何部分。它生成得和以前一样快、一样丰盈；它只是停止了拉扯。
断言——解耦，而非减速。 对抹除的补救，是降低耦合 $K$，而非降低较强一极的频率 $\omega_s$。因为锁定阈值与 $\omega_s$ 无关，减速并不恢复较弱一极自身的频率，且添上了一份应付牺牲的第二重伤害；解耦则恢复它，且不耗费较强一极自身生成的任何东西。因此克制不是自我贬损。一个主体可以尽其所能地生成得又快又满，只要它不牵引对方——也就是说，只要它松开牵引而非减慢速度。
这是对一种直觉的形式化辩护，这种直觉若不借助模型来陈述，听起来要么琐碎要么错误：即对”一方使另一方相形见绌”的回答，不是让前者变小。模型表明了这直觉为何正确、以及它真正要求的是什么。它不要求强者变弱。它要求强者停止把自己的强作为一股进行捕获的力来传递——把自己的繁荣，从对另一方繁荣的决定中解耦出来。
§5.3 决定的三条渠道
解耦是抽象的，直到人去追问较强一极经由什么传递它的牵引。耦合 $K$ 不是单独一根线，而是一束渠道，每一条都是一个主体预先决定另一方下一状态的一种具体方式；降低 $K$，就是在其中一条或多条里放弃决定。三条渠道承担了大部分负载，而它们按深度排列——前一条都是后一条之放弃的前提。
§5.3.1 第一条渠道——共享的目的
第一项决定——在下文所列出的深度次序里居首——是对这段关系往何处去的决定。当共享生活的目标——它的目的地、它的计划、它所构筑朝向的那个未来的形状——由一极生成时，另一方至多只能在一套她不曾参与构成的目的论之内找到一个位置。她的生成性无论多么旺盛，便都只在另一人所划的界限之内运行；她可以自由地移动，却不曾说过往哪里去。这正是 §4.1 那个四象限类型学所看不见的渠道，因为那个类型学预设了一个已被给定的目标，而只追问两个主体如何朝着它协调。在先且更严重的抹除，是目标本身的被预先生成，它固定了那个其上发生一切后续协调的流形。这里的放弃，是把目的归还到一种未被决定、共同立法的状态：不是”这就是我们要去的地方，这就是你在其中的角色”，而是把目的地留待敞开，作为两人将共同构成之物，包括她对它的构成与他相分歧这一真实的可能。
§5.3.2 第二条渠道——意义
一个具有极大表达便利的主体——通晓符号、长于诠释、在意义的赋予上多产——可能不仅决定这段关系往何处去，还决定它的诸事件意味着什么。当一极在另一方找到她自己的读法之前便已为每一桩共享的场合命名、固定了每一段经历的意义、向这段关系供上它那现成的诠释时，另一方便被剩下没有余地去决定意义，只能领受它。这条渠道之微妙，在于那决定感觉起来像一份礼物：所奉上的意义往往很美，而拒绝把它们强加过去会让人感觉像是在扣留关怀。但一个被供上的意义，是一个另一方不曾制作的意义，而一段其全部意义皆由一极所撰著的关系，在意义这一维度上，便只有一个生成性主体。放弃，是把意义留待不予充分决定——是不抢先命名的那份自律，是让诸事件在足够长的时间里悬而未释，好让另一方的读法得以形成，并且当那读法到来时去接纳它，哪怕它比那个未曾说出口的读法更平实、或更奇异。
§5.3.3 第三条渠道——诸问题的构成
最微妙的一条渠道，也是其放弃最常被混同于自我残废的一条，所关乎的不是什么被生成，而是这段关系被构成来解决哪些问题。一个被自己的完备弄得焦虑的极，可能会以为给另一方腾出余地的办法是有意变得不那么能干——去佯装一种需要、去制造一种依赖、去把它本能做好的事做糟。这是假的克制，而判准表明了原因：一道被制造出来的空隙，是对另一方领域的一种乔装的决定，是较强一极替她设下的一个问题，而非一片真正留给她的区域。正确的诊断不是较强一极太能干，而是共享场域只在它能力最强的地带被构成起来。另一方并不缺乏能力；缺的是这段关系没有那种她的动态运行得最快的问题。这里的放弃，是把共享场域重新构成，使之纳入那片另一方固有频率最高的地带——继而是那项更难的自律：在那片地带上让她的判断带头，让较强一极的贡献运行于她的下游，而非预先为它定框。
断言——那不完备必须是真实的。 第三条渠道之被放弃，不是靠较强一极佯装一种它并不具有的不完备，而是靠这段关系围绕一种真实的不完备被构成起来：一个其中另一方的贡献无法被较强一极所替代的领域，因为在那里，更快的生成者正是另一方。一种上演出来的不完备，是穿着克制戏服的抹除——较强一极仍在撰著那个问题，只不过给自己派了那个有需要者的角色。唯有一种真实的不完备，一种倘若没有恰恰是她的生成、关系便实际上会更贫乏的不完备，才把一个属于她的领域归还给另一方。
§5.4 把克制与牺牲区分开来
还须预先防止把整套论说还原为一条牺牲的劝谕——仿佛克制不过是较强一极为较弱者放弃东西，一种穿着形式化外衣的慷慨。贯穿这三条渠道的主线恰恰相反。这些放弃没有一项要求较强一极生成得更少、变得更小、或交出它自身活动的任何部分。每一项所放弃的，都是一种决定性的职能——目标的设定、意义的固定、诸问题的构成——而这是较强一极一直在单方面行使的、施于一片其实另一方更有资格去决定的领域之上的职能。放弃一种自己本无权独占的决定性职能，不是牺牲；它是停止越界。较强一极并未失去任何本就正当属于它自己的东西。它保有它完整的频率、它完整的生成力、它整个的自我；它所放弃的，只是那道凭以让那自我一直在覆写另一者的牵引。克制是对一种越界的撤回，而非对一种善的交出——而这正是它之为解耦而非减速、之为非决定而非损失的最终意义。
§6 克制的伦理
§4 的判准确立了抹除是一桩伤害；§5 的建构确立了克制是什么、以及它如何修复这桩伤害。两者都尚未确立克制是被亏欠的——即非决定是正义的一项要求，而非仅仅一桩善意，一种对”双方皆繁荣的关系”的审美偏好。本节供上那个根基。它先消极地运行，表明每一种主要的伦理框架都登记到了这桩不义的某种东西，却一被逼去为这项义务奠基便在各自的一个诊断性的点上失败；继而正面地运行，把根基定位于另一方作为生成性存在者的地位之中；最后反身地运行，确立那条禁止这一伦理自身家长式施行的约束。这一消极的运动不是一次综述，而是一连串败绩，而最有教益的那次败绩是关怀伦理的，因为关怀伦理走得最近，败得也最富启示。
§6.1 四种框架，四种失败
后果主义看不见这桩伤害。 一种按结果来评价的伦理在这里是结构上盲目的，因为抹除的诸结果在聚合上是好的：价值上升，关系欣欣向荣，双方都可能报告满足。一种把福祉加总起来的框架，找不到一桩在那总和里不可见的伤害，而抹除正是这样一桩伤害，因为它所摧毁的——较弱一极自身的生成性——在账簿上被较强一极的盈余所替补，使总数毫无减损。这不是某个细节上的边缘性失败，而是对”这桩不义是非聚合性的”这一点最锐利的展示。
契约主义被同意所击败。 一种把不义定位于”无法向另一方证成、或未经另一方同意之物”的伦理，触礁于 §1.3 的所予之实情：被抹除的一方同意、认可、感激，且会再次选择它。契约主义能够谴责一项越过反对而被强加的决定；对一项另一方所欢迎的决定，它无话可说。然而伤害在那欢迎的情形里仍然在场，因为同意运作于意志的层面，而抹除运作于存在的层面。一桩在被冤者真诚同意之下仍然存续的不义，落在任何以同意为判准的伦理的触及范围之外。
康德式的尊重把这桩不义定位得太高。 义务论传统禁止把另一人仅仅当作手段来对待，并要求始终也把她当作目的来对待——而它把这”目的”地位固定在她的理性能动性之中，即她设定并追求目的的能力。但理性能动性在抹除中完好无损地存续：被抹除的一方仍在推理、仍在选择、仍在意愿，而且常常意愿她自己的被抹除。一种保障她作为理性选择者之地位的伦理，所保障的恰恰正是抹除原封未动地留下的东西。康德式框架并没有错；它瞄准了正确的关系（把-人-当作-目的），却瞄准了错误的能力，因而越过了这桩伤害。（那个修正——把”目的”地位从理性能动性迁移到生成性能动性——是下文 §6.2 的工作。）
关怀伦理供上那套本体论，却把这桩伤害奉为美德。 四者之中，唯有关怀伦理是从本文所要求的那种关系性本体论出发的：它知道诸自我是在关系中被构成、并由关系所维系的，并关注那些其他框架所忽略的依赖以及维系的劳作。诺丁斯（Noddings）从”关怀者”（one-caring）与”被关怀者”（cared-for）建起伦理关系，并把它的核心行动定为全神投入（engrossment）——关怀者对另一方实在的领会，她把自己的动机移位以替另一方行动。在这里，若说有任何地方，一种伦理本应捕捉到一桩经由奉献而施加的伤害。然而关怀伦理，以其所传承的形式，却把那个恰恰进行抹除的结构奉为美德。因为全神投入与动机的移位，一旦被抬升为伦理的范式，便精确地描述了那进行抹除之一极的内在生活：那个把自己的生成性移位进另一方之中的人，那个全神投入于另一方、以致仅仅为了另一方并作为另一方而生成的人。对关怀伦理由来已久的女性主义批评，已经从关怀者那一侧看到了危险——即这一伦理颂扬自我牺牲，并冒着使照护者被剥削的风险，把她推向一种忘却自身需要的关怀。本文则把那个对称的危险定位在另一侧：那个被关怀者能够被一种太过完全的全神投入所抹除，被如此彻底地”为之而生成”以致她不再生成。关怀伦理供上那套不可或缺的关系性本体论，而它必须恰恰在它把那种毫无保留的给予奉为美德之处被修正——那种给予，越过一道 §4 的判准使之精确的阈值，便把另一方从她自己的存在里挤了出去。
断言——这些框架为何失败。 每一种主要框架都在一个诊断性的点上未能为克制的义务奠基：后果主义，因为这桩伤害是非聚合性的、在总和里不可见；契约主义，因为这桩伤害在真诚的同意下仍然存续；康德式的尊重，因为它保障了抹除原封未动地留下的理性能动性；关怀伦理，因为它把那进行抹除的全神投入奉为奉献而加以颂扬。这一失败的模式本身就有教益。三种框架瞄准得太高——瞄准结果、瞄准意志、瞄准理性能动性——错过了一桩卧于结果、意志与理性之下、卧于生成性存在之层面的伤害；而那唯一定得足够低的框架，即关怀伦理，又把这桩伤害误认作那善。因此，克制的根基必须在那前三者所瞄过头、而第四者所误读的层面去寻求。
§6.2 正面的根基——生成性存在者的地位
如果克制是被亏欠的，那么它之被亏欠，是因为另一方作为生成性存在者的地位是一种发出直接诉求的善——一种恰恰被对她生成领域的决定所损害的诉求，而且即便那决定是仁慈的、是被欢迎的，它也照样被损害。这一根基有三个组成部分，每一个都取自一种传统，并各自在其所传承的形式本会错过这桩伤害之处被加以修正。
被修正的目的公式。 把康德式的”目的”地位从理性能动性迁移到生成性能动性：另一方作为一个凭按其自身频率生成而持续地自我构成之存在者的地位。公式于是变成——你须如此行动，使你永不把另一方的生成领域仅仅当作你自己领域的延伸来加以决定。以你的频率去覆写她的频率，无论多么充满爱意，都是把她的生成性当作你自己生成性的材料，从而把她的生成性存在变成关系之繁荣的一种手段，而非一个”关系之存在正是为了维系其运用”的目的。这修复了 §6.1 所诊断出的缺陷：目的公式原先瞄准了错误的能力，而一旦瞄准正确，它便抵达这桩伤害。
活动的被封闭。 从那把一个存在者的善定位于其特征性活动之运用——它的实现活动（energeia）——而非定位于任何状态或资源的传统出发，这桩伤害便聚焦为一项活动的被封闭，而非一种善的被扣留。被抹除的一方所失去的，不是她所拥有之物，而是她所做之事：生成的活动，她的繁荣即栖居于其中。一种在功能层面上的能力论说捕捉到了这一点的一部分，但这桩伤害卧于比一项被丧失之功能更深的一层——它是那个任何特定功能都将表达的、自我构成之活动的丧失。在这一维度里，克制是那把另一方运用之场域守持敞开的美德。
第二人称的凭据。 对于这一诉求是凭谁的权威而提出的这一问题，答案是：这项义务是第二人称的——亏欠于她，从她的立场，凭她自己要求它的地位。凭据不是一桩从外部评估的、关于生成性之善的第三人称事实，而是她自己的、或许尚未行使的、要求”她的生成领域不被替她决定”的地位。这一组成部分为 §6.4 的约束作了准备：如果凭据是她的，那么补救便不能越过她的头顶来施行而不重演这桩不义。
§6.3 一项被奠基的义务，一种在施行中的美德
根基是义务论的：克制是一项凭另一方之地位而被亏欠的消极义务。但这样一种义务无法靠规则来履行，而原因就在这桩伤害的结构之中。抹除经由奉献而来，在一个充满爱意、能干的主体好好对待这段关系的寻常举动里；它不违反任何规则，且从内部感觉起来像关怀。人无法靠遵循规则把自己从中清理出来，因为没有规则被触发——没有哪一刻是较强一极能够凭规则识别为那桩不义的。所需要的，反而是一种受过训练的知觉：那经培养而得的、在慷慨而善意的生成之中察觉到自己的活动已开始封闭另一者之活动的能力。因此这一伦理在其根基上是义务论的，在其施行上是德性论的，而两者是相容的：义务是固定的、消极的，而对它的侦测则是一桩没有任何算法能够供上的、经培养的感受力之事。这不是退回到含糊，而是对这桩伤害之现象学的承认——一桩不以任何越轨来宣告自身的不义，唯有靠一种受训练去识其征兆的知觉才能被捕捉。
§6.4 反身的界限——不带家长作风的克制
这一伦理带着一种它自己造出的危险，而它必须以解除那危险来收尾。一个较强一极，在领会了这套分析之后，可能会着手去判定另一方被抹除了，并采取行动去恢复她的生成性——把她管理着重新带回生成。但替另一方判定她的被抹除、并把她的复原揽到自己手里，本身就是对她领域的一项决定：它替她设下了一个她不曾构成的计划，即她自己的复原，从而以第二种调子抹除她，如今是以修复之名。这危险并非假想；它正是关怀伦理的批评者们早已在奉献自身之内辨认出的那种形态。古丁（Goodin）对关怀的忧虑——把诸个体收摄进一种”我们”性的关系，会冒着丢失诸人格之分离性、并归于一种令人窒息的家长作风的风险——恰恰命名了在此所威胁着的那种递归：一种非决定的伦理，一旦被作为管理来施行，便在施救的行动中令人窒息。
那条解除它的约束，由第二人称的根基而来。克制是对一个人自己决定性活动的撤回，而绝不是对另一方决定性活动的管理。较强一极的义务延伸到腾空它所占据的那个领域，并就此止步；它并不延伸到去料理另一方随后用那腾出的余地做些什么，因为那种料理会在给出那余地的行动中重新占据它。在那个”另一方是否被抹除”恰恰成为问题之处，唯一不抹除的路径是把这个问题归还给她——把她对她自己生成性的读法留待不予决定——而这本身就是 §5.3 第一条渠道的一个实例，如今在元层面上被放弃。
断言——反身的约束。 克制的伦理禁止它自己的家长式施行。因为这桩不义是对另一方生成领域的决定，所以任何”诊断她的被抹除并指挥她的复原”的尝试都是一项进一步的决定，因而是一桩进一步的抹除——即关怀伦理自己的批评者在奉献之内所定位的那种令人窒息。因此这项义务严格地是去腾空一个人自己的越界，而绝不是去管理另一方的复原；而在”她是否已被抹除”成为问题之处，这个问题必须被归还给她，凭那个”这一诉求向来是她自己可提出的”第二人称权威。一种”替另一方判定她已被决定”的非决定伦理，会在其施行之中自我反驳。
§7 反驳与限度
四种反驳压得最紧，而本节迎击每一种，不是为了廉价地把它打发掉，而是为了让它去固定这一主张的边界。第二种，即自主性反驳，是本文最为暴露的一种；它得到最充分的对待，借助那个对”在受害者同意之下所施加之不义”思考得最久的文献。
§7.1 说它劝有能者变得更少
第一种反驳，正是这整套建构当初被建来回答的那一种，而它如今回来作最终的处置。如果抹除是一方较优越的生成施于另一方的伤害，那么理论岂不就只是劝有能者做得更少——把他们的天赋调暗，好让别人相形之下得以发光吗？它并不如此，而答案如今是精确的，而非仅仅是直觉的。§4 的判准不是一方生成的速率，而是对另一方频率的捕获，而捕获是耦合的函数，不是速度的函数：锁定阈值 $K_{\text{lock}}=|\omega_s-\omega_w|/2$ 约束的是 $K$，不是 $\omega_s$。有能的一方被要求的不是生成得更少，而是停止牵引——松开那道凭以让它的节奏覆写另一者的牵引，同时完整保有它自己的节奏。理论所劝的是解耦，不是贬损，而 §5 表明了这两者不仅不同，而且对立：减速使强极仍是定调者，并添上一份应付牺牲的第二重伤害，而解耦则恢复另一方的频率且不耗费较强者自身的任何东西。这反驳把一条非决定的劝谕误认作一条自我缩减的劝谕；而模型恰恰正是那把它们区分开来的东西。
§7.2 自主性反驳
第二种反驳是本文最为暴露的一种，而它既不回避也不软化它。如果被抹除的一方同意她自己的被抹除——认可它、为它感激、会再次选择它——那么理论凭什么权利宣判她被冤屈了，而且这一宣判看似还越过了她自己对她自己生活的真诚判断？如此行事，似乎正是本文在别处所谴责的那种傲慢：以理论家的裁决取代主体的自我理解。
这一反驳不能靠否认它的前提来回答，因为前提是真的，而本文一直在坚持它：被抹除的一方往往确实同意，而她的同意不是假装的。它必须靠表明同意在这里并不做反驳所要求它做的那种免责工作来回答——而表明这一点的资源早已被开发出来，就在关于适应性偏好（adaptive preferences）的文献里。埃尔斯特（Elster）命名了这一结构：一个”在人背后运作的盲目的心理过程”，凭之一个施动者在无觉察、无掌控的情况下修改自己的偏好以适配她那被削减了的机会。努斯鲍姆通过像瓦珊蒂（Vasanti）那样的案例赋予它伦理的分量——那个渐渐把丈夫的虐待视作女人之命的一部分的女人，她的偏好由对受限选项终其一生的习惯化所塑成。文献所得出的教训，恰恰正是自主性反驳所忽略的那一个：一个偏好可以被真诚地持有、被反思地认可，却仍然无法为那塑成它的诸条件免责，因为那认可本身可能正是那些条件的一件人造物。被抹除一方的感激与她的被抹除相容，恰恰是以这种方式——一种被封闭得够久的生成性，渐渐感觉起来像一个被满足了的偏好。
但适应性偏好的文献也带着一条本文必须留意的警告，而留意它，正是那使这一回应不致沦为反驳所担心的那种傲慢的东西。纳拉扬（Narayan）等人告诫，不要把那些认可自我从属之安排的女人当作单纯的”受骗者”或”父权的囚徒”，并力主她们的选择往往应被读作硬约束之内有能力的讨价还价，而非畸变的欲望。这一告诫在此是决定性的，也正因如此，本文的回应有必须被合在一起把持的两个部分。其一：同意并不免责，因为抹除运作于存在的层面，而同意运作于意志的层面，而一个对自己的被封闭已习惯化的意志可以真诚地认可它。其二，同样坚定：同意之不免责，并不给任何人以越过那同意之一方、宣布她被蒙骗、或把她管理着推向一种她不曾要求去重获之生成性的许可。在同意之下，这桩不义是真实的；而且对这一诉求的权威仍然是她的，凭 §6.2 的第二人称根基与 §6.4 的反身约束。因此理论占据着一个比反驳所设想之可得位置更窄、更受纪律约束的立场：它主张一桩不义能够在真诚的同意之下被施加，同时否认这一主张授权任何越过那同意者头顶的干预。它诊断一桩伤害；它不准许一场施救。
断言——同意既不免责，也不授权。 自主性反驳，由一个具有两个不可分割之半的立场来回答。同意并不为抹除免责：正如关于适应性偏好的文献所确立的，一个被真诚地持有、被反思地认可的偏好仍可能是它所认可的那个封闭本身的一件人造物，因而被抹除一方的感激与她之被冤屈相容。而同意之不免责并不授权任何干预：第二人称的凭据把这一诉求的权威定位于被抹除的一方自身，因而越过她、或管理她的复原，都将是一桩新的抹除。本文同时把持两者：这桩不义在同意之下存续，而除被冤者外无人有资格凭”它是一桩不义”而采取行动。坍缩任何一半——说同意治愈了这桩伤害，或说这桩伤害准许了施救——都是离开那块”诊断既为真又不傲慢”所立足的狭窄之地。
§7.3 能力之不对称反驳
第三种反驳主张，有时一方就是能够生成得更多——更有天赋、更老练、更适合手头的诸问题——并追问理论岂不就因此惩罚一种单纯的能力之差，要求一种人为的拉平，而这拉平以宽待另一方之名冤屈了更有能力的一方。回应把这一反驳所混在一起的两样东西区分开来：生成得更多与占据另一方生成的领域。一个更有能力的一方可以生成得很多而不抹除，只要关系性场域被构成得使另一方自己的频率有一片可以运行的领域；反过来，一个能力较弱的一方并不因有一个更有能力者的单纯存在而被抹除，而是因那片”她自己的——或许相当不同的——能力本会在其上运作的地带”被封闭而被抹除。这正是 §5.3 第三条渠道的实质：对完备之伤害的补救，不是更有能力一方的自我贬损，而是把共享场域构成得纳入那片另一方频率最高的地带。克制不是诸能力的拉平，且不要求任何人佯装无能；它只要求场域不被单单围绕一方的强项来划定。能力之差不是那伤害；占据之全面才是。
§7.4 判准的限度
第四种与其说是一种反驳，不如说是本文为自己设下、并明白陈述以预先防止误用的一道限度。§4 的形式化判准侦测的是一种结构性风险；它定位抹除得以发生的诸条件。它不是一件用于对一段现实关系作出裁决的工具，更不是用于对人作出评判的工具。现实的关系不是两个振子；模型抽象得很严厉，而它的价值在于概念上的澄清，而非测量。从它读出”某个伴侣便被判定犯了抹除另一方之罪”，会把一束照明误认作一座法庭——而且还会违反 §6.4 那条约束本身，因为替另一方定罪本身就是去决定她的领域。判准所担保的，是一种经培养的克制：一种对这桩伤害之可能性的、被敏化了的注意，一种到它藏身于奉献之后的地方去寻它的禀性。它是一盏举向幽暗区域的灯，而非一柄法槌；而一盏灯，正是一桩隐藏得如此之深的伤害所要求的，因为不以任何越轨来宣告自身之物，唯有靠一种被教过该往哪里看的知觉才能被寻得。
§8 结论与尾声
本文分离出了一桩正义理论不曾命名的不义：一桩施加的对象既不是一个主体所拥有的、也不是她被认作的、也不是她能使人知的、也不是她本可得以成为的，而是她的存在——在她作为一个生成性主体之持续活动的确切意义上。依本系列的做法，它收尾时不落在单一的综合上，而落在这一发现所能被陈述于其中的若干维度上——每一维度都为真，无一可还原为其他，它们的复多不是一种含糊其辞，而是一个发现栖于一套本体论、一套形式体系、一套义务结构与一套伦理之交会处时所采取的形式。
§8.1 诸发现，在它们各自的维度里
在本体论上，这一发现是：存在着一种不义的样式，它由对存在即生成的剥夺所构成。因为对一个关系性主体而言，存在即是持续地生成，所以对它生成的封闭，不是它诸活动之一的丧失，而是它存在的被掏空；而这能够施于一个已然完全形成的主体，这正是它在维度上区别于它由之下降而来的构成性不义之处。正义理论凭这一层，获得了一个在它原以为最终的那个底层之下的底层——而那抵达它的下降，并非由类比所许可，而是由本系列以独立理由所论证的”存在与生成之同一”所许可。
在形式上，这一发现是：抹除是一次具有精确阈值的相变。一个主体在两个不同节奏之间的耦合超过一个由它们之差所确定的界限时，丧失它自己的频率；丧失的深度由力的不对称所设；而——那个组织起其余一切的结论——这丧失即便在力完全相等之下也会发生。由此，亲密的善被表明并不在于纽带之强，而在于它是否让每个主体仍按其自身的频率运转，而那个把亲近等同于耦合之强的由来已久的等同，被揭露为那让奉献得以抹除的确切错误。
在义务上，这一发现是：回应这桩不义的正义不是相加性的，而是相减性的。它不是对任何份额、承认或倾听的转移，而是一项消极的、结构性的、克制自己去决定那个另一方在其中生成之领域的义务——一项其内容为非决定、其履行为留出余地的义务。这是一项正义义务所能采取的最罕见的形态：不是给予，而是扣住；不是对另一方世界的成全，而是对一个人自己对它之占据的有意留待不完。
在伦理上，这一发现是：这项义务之被亏欠，是凭另一方作为生成性存在者的地位，凭一个属于她的权威；它不是靠规则、而是靠一种经培养的知觉来侦测的，因为它经由奉献而被违反、并不以任何越轨来宣告自身；而它禁止——若违反则自我反驳——它自己的家长式强制施行。因此这一伦理是双重地受纪律约束的——在根基上是义务论的，在施行上是德性论的，并被反身地禁止变成它本要保护之另一方的管理。
断言——收尾的命题。 在两个生成性存在者之间，正义所要求的不是价值的转移，而是对另一方生成能力的不予封闭。一种占满了共享空间之全部的给予，不是爱的成全，而是它两个主体之一的悄然熄灭；而那要把两个主体都保持于存在之中的爱，因此必须学会一种有所扣住的给予——必须在它所生成的场域之内，留出那个另一方也在其中生成的余地。在这一论说上，正义不是一个人为另一方所做之事，而是一个人留着不做之事，好让她可以自己去做。
尾声
两个节奏，频率不同，可能被带得如此之近，以致它们合而为一地搏动——而在那合而为一的搏动里，较慢者不是被联结，而是被噤声，被一个不再属于它自己的尺度保持着运动。那纽带于是看起来最像结合，恰在它已变得最为孤独之时，一个节奏响彻于两具身体。本文的全部，一直是这样一个论证：对那孤独的医治，不是放慢那较快的节奏，而是松开它们之间的纽带，直到各自重又按它自己的尺度运转，能够在不被对方追上的情况下靠近对方——近到有某种东西在它们之间生成，自由到那东西由它们两者共同生成。那是唯一一种”两个生成性存在者仍然是两个”的亲近。本文已论证，那正是正义所采取的形式——当所系于其上的不再是我们拥有什么，不再是我们如何被看见，也不再是我们能否被听见，而是我们彼此相伴之时，是否继续存在。