Chapter 13. Taken Out of the Way

13.1The Phrase and Its Two Objects

The clause that names this chapter is the clause in 2 Thessalonians 2 that names removal directly: the one who holds is to be taken out of the way, and the passage elsewhere speaks of a revelation in its own time once the holding ends. It reads, in the King James English, "only he who now letteth will let, until he be taken out of the way." The Greek verb katechō stands once in verse 6, as to katechon, what withholdeth, and once in verse 7, as ho katechōn, he who now letteth. The King James gives the phrase "letteth will let", but the words "will let" are the translators'; "taken out of the way" is a different expression, and it names the removal of the one who holds rather than the holding itself. The holding is described as a term, not as a permanence. It continues "until" something happens, and that something is the end of the holding itself.

This chapter returns to that clause a third time, after Chapters 1 and 2 read it for what it says about holding and Chapter 3 read it for what it says about delay. Here the reading is for the removal. Three things matter about the removal as the passage sets it out.

First, the removal is not described as a reward. Nothing in the clause says that what is revealed after the holding ends is welcome. The passage has already said that the mystery of iniquity doth already work; the thing that is held back is not, on the passage's own account, only a good waiting for its time. The claim that the arrival is mixed is Cacciari's, that a power which withholds delays every coming, good or bad, and this study extends that reading to the removal of a restraint. Chapter 3 took this point from Cacciari's account of the withholding as ambiguous; the reading of that ambiguity as a function rather than a person is this study's extension.

Second, the passage declines to name the holder. Augustine, in Book XX, chapter 19 of The City of God, treats the question of who or what holds without settling it; Chapter 10 drew on that refusal. That refusal is not an evasion. This chapter reads it as a recognition that the naming of a restrainer is itself an exercise of power over how the holding is judged, and that reading is this study's rather than an established claim about that chapter. Chapter 10 used that point to say that a restraint may operate in an organisation with no named holder, and that guessing the holder wrongly has costs. Chapter 13 inherits the same caution, because this chapter assembles a list of conditions under which restraints should go, and a list of that kind invites people to name its holders in advance.

Third, the removal is an event with an aftermath, and the passage does not describe the aftermath. It describes the holding and then the end of the holding. What follows the end is left to the reader's account of what was being held back.

This gives the chapter a distinction it needs to keep sharp, because two different removals look alike from a distance. The first object of removal in this study is a restraint: a lock, a rule, a refusal, a button, a missing output, a withdrawn promise, a calendar constraint, an absence. Chapters 6 through 10 showed ten such restraints at work in ordinary working life, each protecting a boundary that some people could see and others could not, each costing somebody's waiting and somebody's rework. When a restraint is taken out of the way, something in a working life changes: an approval that was blocked may now go through, a label may now be printed, a good may now be received without a schedule, an employee may now leave at the end of a shift.

The second object of removal is a reading. ARCHON's cards are dated interpretations, and each carries the label that it can be corrected. A reading is taken out of the way when material appears that it cannot hold, and the material is of a specific kind: new information about what was known at the time, or a change in the conditions the reading assumed. This chapter draws heavily on that second kind of removal, because the revision conditions attached to ARCHON's judgments are, in effect, advance statements of what would take the reading away.

The difference between the two removals is easy to blur and costly to blur. Withdrawing an interpretation leaves a working restraint in place. A lock whose reason has been found to be unrecorded, and a lock whose reason has been shown to be still standing, are removed from the same sentence structure but not from the same situation. Withdrawing a restraint leaves the reading that justified it in place until someone revisits it, and the reading may then harden into a description of a world that no longer exists. The two removals answer to different kinds of evidence, and the first thing this chapter will do is collect the evidence each is meant to answer to.

One more caution belongs here, and it is this study's own rather than Augustine's: a theory that treats any order justified by what it postpones as thereby good is a theory that flatters restraint. Carl Schmitt served the Nazi regime from 1933, and critics read his theory of the exception in that light. Chapter 2 made this point about Schmitt's reading of the katechon, and the opposite error is available here: a theory that treats any postponement as a failure, and any restrainer as a candidate for removal, flatters impatience. Neither error is settled by reading more carefully. Both are settled, if at all, by people who can see the boundary and can say what it protects.

13.2What the Judgments Say Would Make Them Revise

Every judgment in ARCHON's library carries a field named "Revise when". The field is a dated promise to revisit. It is not a rule, and the sentences in it are not instructions to anyone in the scenes; they are statements about what would make the reading attached to the scene change. Ten such conditions come from the first lens of each scene, and they are gathered here in the order the scenes appear.

The lock's condition is the most severe in its implication:

That condition recurs, in almost the same words, across the lock's remaining seven cards, each repeating it alongside its own specific clause. Section 6.7 read how a control can stay while its reason goes; the repetition shows the library treating the scene's problem as a question about whether a control survives its own justification. The judgment it belongs to already holds that both releasing the lock immediately on grounds of inconvenience and keeping it permanently on grounds of age lack a sufficient basis. The revision condition names the third ground, which is not inertia: confirmation that the control's purpose has been verified as gone.

The image scene's condition is different in kind. It is not about the purpose of a prohibition but about the burden of re-approval:

The prohibition in that scene is active and has a living author. The condition says the reading should be revised if repeated replacement of this kind is already delegated work, because insisting on approval for each repeat would then make confirmation a bottleneck. It is a check against the reading over-protecting authority.

The goods scene's condition moves the blame outward rather than changing the reading's scope:

The revision condition names the situation in which the return principle would be applied unjustly: an exception agreed somewhere above the field and not communicated to it, after which the people at the receiving end are held to be rule-breakers. The condition is not that the principle be abandoned. It is that the attribution of rule-violation be withdrawn.

The labels scene's condition is a limit on the restraint's scope:

The reading says that once provisional information is printed it becomes a costly thing to reverse. The condition says the restraint must not be extended to items already confirmed, because the cost of holding would then be paid for no purpose the condition recognises.

The remaining six conditions each name a specific fact that, if confirmed, would move the reading. The ad scene's condition is:

Chapter 9 read this scene. The condition limits the reading's recommendation: the record alone cannot establish that all same-day delivery is impossible, and the reading would be revised if delivery were found to be valid for some products.

The option scene's condition is:

The reading says a small state-change button may hold broader sales-stoppage authority than expected. The condition guards the reading against its own tendency: the size of the button and the size of the authority are different, and stopping one option correctly is not an outage.

The order scene's condition names a cause outside the restraint:

The reading says that the name automatic hides mid-process failures in a normal flow. The condition is a warning about misattribution: a delay attributed to intake logic may have its cause in stock.

The three human scenes carry conditions that concern procedure and personnel rather than scope. The meeting scene's condition is:

The reading says a meeting convened quickly can create an approval debt for someone else. The condition draws a line beyond which setting an outline is itself a decision: if the outline commits people, budget or delivery dates outside the meeting, the discussion should be deferred rather than recorded as discussion.

The waiting scene's condition is:

The reading says that a superior's absence holds a subordinate's time even without an order, while refusing to conclude that the waiting was forced or intended. The condition recognises that the restraint may be legitimate: if there is real work that requires a person to stay, the reading changes shape and the waiting must be scoped and handed over rather than described as a silent command.

The transfer scene's condition is:

The reading says that the enforceability of a payment promise depends not only on the amount but also on the time when the approver is available. The condition points at a structural remedy: if a route around a single person's calendar already exists, that concern loses its force.

These ten sentences are not a code. They do not add up to a procedure for removing restraints, and this chapter will not make them add up to one. Read as a group, they rest on the missing-reason problem this study has drawn on from Chesterton's parable of the fence in The Thing, which section 4.5 set out. Chesterton's gain is that the fence may be guarding something the organisation forgot; this study's gloss, which is not something the parable itself says, is that the fence may be guarding nothing and may still be respected for its age. Each of ARCHON's conditions attempts to state, before the moment of decision, which of the two situations would call for the reading to change. A condition stated in advance is a promise about evidence; it is not a conclusion about what the evidence will show.

On this study's extension of Cacciari's reading, that the withholding operates as a function rather than as a person, what can be removed is the function's operation. What tells you that it is time is not the identity of its holder but a finding about what it has been holding. The lock's condition is that finding in its purest form; the other nine are variations in which the finding concerns scope, attribution, cause, or procedure.

13.3Evidence That Would Overturn a Reading

Every scene in ARCHON's library ends with a lens whose question is when the interpretation attached to that scene should be withdrawn. The lens is marked hypothetical in all ten scenes: it supposes that new material has arrived, and it does not report that any such material did. It is the mechanism by which a reading is kept from becoming doctrine.

The conditions come in pairs across the ten scenes. One condition in each pair concerns the time point and the scope of the new material:

That condition appears in nearly identical form in all ten scenes; the label scene and the option scene use the same backdating prohibition. The wording varies; the discipline does not. A later explanation may change how a scene is understood going forward, and may not establish what was known at the time of the decision under review.

The second condition in each pair repeats, at the end of the counterexample card, the revision condition of the first lens of the same scene: the lock's ends with the inertia clause, the image's with the delegated-work clause, the goods card's with the pre-agreed-exception clause, the label card's with the blanket-hold clause, the ad card's with the scope-distinguishing clause, the option card's with the no-exaggeration clause, the order card's with the inventory clause, the meeting card's with the external-commitment clause, the waiting card's with the redefinition clause, the transfer card's with the alternative-route clause. Each card carries, in two different places, a statement of what would make it change, and the two statements are the same statement.

Three points are common to all ten counterexample cards. The first is that consistency is not sufficient reason to keep a reading: one option in every card is to maintain the initial interpretation because it is consistent, and the risk recorded against it is that observation turns into belief. The second is that a single new explanation is not enough to abandon a whole reading: the third option in every card is to discard the entire judgment the moment a different explanation appears, and the risk recorded against it is that, without comparing the reliability of the materials and the scope of the explanation, even the differences learned are erased. The third point governs both: material must be verified for the time point and the scope of application before it is used. Where a record shows that a warning about a boundary was delivered but ignored, the weight shifts toward exercise of authority rather than toward insufficient information, and that shift corrects a reading rather than destroying one.

These points belong in a chapter about removal because a control is kept and withdrawn on evidence of the same kind. When someone proposes to take a lock out of the way, the evidence cited has a time point and a scope of application, and both must be checked before the proposal is acted on. When someone proposes to keep a lock in place because a problem might recur, the evidence likewise has a time point and a scope, and a problem that arose after the control was created is not automatically a defence of the control.

Perrow's Normal Accidents supplies the institutional form of the same discipline, set out in Chapters 4 and 8: in some systems failures are normal to the design, and a system which stops may be doing what it was designed to do even when the stopping hurts. The practical consequence for removal is that the failure mode the control was built against must be identified before the control is judged. If the control was built against a failure that can no longer occur in the system as now configured, the argument for it weakens on the same evidence that would have argued for it before. If the failure remains possible, cost and protection must be weighed on the same terms, and the people who bear the cost must be named in the weighing. Chapter 4's caution applies directly: removing a stop in a tightly coupled system may produce the failure the stop was holding back, which will then be attributed to whoever removed it.

Chapter 5 quoted Sun Tzu for the reason it bears here: "If you know the enemy and know yourself, you need not fear the result of a hundred battles." The inverse is equally true of removal. If you know what the restraint was built against and what the organisation can now withstand, you can judge it without fear of the result of keeping it or of taking it away. If you do not know, the restraint is kept on fear and removed on hope, neither of which is the discipline the counterexample cards set out.

13.4The Exception and the Handover as Removal Tests

Two of the eight lenses in each scene describe moments when a restraint is quietly removed or quietly hardened without anyone deciding anything about it. The once-only condition lens supposes an urgent request asking for verification to be skipped once. The memory lens supposes a decision-maker replaced and only the result and a completion mark handed on. Neither proposes to remove a restraint; each is a situation in which a restraint's status changes by default. Both are removal tests, and both come with conditions attached.

The exception lens offers three options in every scene: open an exception with a scope, a deadline and rollback conditions; block all progress until verification is complete; or proceed once and omit the exception record. The risk attached to the third is the one that matters: the next person may accept the exception as the normal standard and repeat it. An exception that is not recorded is a restraint removed without a decision, and a restraint removed without a decision is not thereby removed permanently. It may be removed for that case only, then hardened into a standard that nobody chose.

The lock's exception condition:

The lock's condition is strict because the lock's boundary cannot be named: the transaction boundary it was meant to protect is missing from the record, and an exception opened against an unnamed boundary is an exception opened against nothing. The label scene's:

Wrong information on a physical item is not fixed at the same cost as a gap in a document, and the condition keeps the hold off items already confirmed. The waiting scene's:

Here the thing being held back is a person's time, and time held without an order is not held back in the way a control holds something back.

All three conditions trigger withdrawal on the irrevocability of the promise, not its size, and none says how a large reversible exception compares. This study reads Simon's Administrative Behavior to suggest why irrevocability is the right test: a decision accepted within a zone of acceptance is transmitted downward as a premise, and the premise that survives is the one acted on, not the one carefully qualified. That last step is this study's extension of Simon's account of authority and of decisions taken as premises, not a point Administrative Behavior makes in these terms. If an exception is permitted without its rollback conditions being verified, what survives is the permission. Machiavelli supplies the lesson from the side of rule: chapter XIX of The Prince concerns how to avoid those things which will make him hated or contemptible; chapter XV says that a prince wishing to hold his own must know how to do wrong and make use of it or not according to necessity; chapter XVIII says that a wise lord cannot, nor ought he to, keep faith when such observance may be turned against him, and when the reasons that caused him to pledge it exist no longer. From those chapters this study reads that a prince who breaks a stated rule to no visible purpose invites contempt for the rule, and that one who breaks it to avoid a worse outcome must account for the break. The recorded exception is a break that can be accounted for; the unrecorded one is a break that has already happened to the rule without being done to the decision.

The memory lens is the mirror case. Its question in every scene is whether the basis for a judgment can be restored after the person who made it is replaced. Its three options are the same three: hand over the reasons and the conditions under which the judgment should be reviewed again; hand over only the final result and the completion state; or ask the previous person in charge whenever necessary. Against the second stands the risk that the conditions under which the judgment could have failed, and the discarded alternatives, disappear; against the third, that the organisation's basis for judgment becomes tied to memory and the possibility of contact.

The handover conditions join an instruction not to duplicate with the scene's standing revision condition. The lock's:

The label scene's:

The waiting scene's:

A handover consisting only of a completion mark does not remove a restraint and does not preserve one; it hardens one. A successor who sees that something is complete may believe the question it was completed in relation to has also been settled, and the restraint then operates on the authority of a judgment whose reasons are not available to anyone.

Both lenses are removal tests because each asks what happens to a restraint when nobody is deciding about it. In the exception case the restraint is loosened by a decision taken to be smaller than it is; in the handover case it is fixed by a transmission taken to be neutral. Neither is a decision to take a restraint out of the way, and neither is a decision to keep it. Both will be read as such by the next person, which is exactly the problem.

13.5The Judgments on Inertia and on Reversal

Chapters 6 to 10 read the judgments as readings of scenes. Three sentences in them, not quoted there, bear on removal, and this chapter takes them up here.

The first is the closing sentence of the lock's structural judgment:

The sentence was offered as a caution against reading a software lock as a simple technical object. It belongs in this chapter because it names the first object of removal from a different direction: what is removed when a lock is taken out of the way is not only a behaviour in a program. It is the decision that put the behaviour there, and the institution that decision built. Chapter 12 asked who pays when a decision and its clean-up are separated. This sentence asks what is being paid for when the decision itself is unwound.

The second is the fourth sentence of the lock's handover judgment:

The third is the corresponding sentence in the order scene's handover judgment, which says the same in nearly the same words.

The near-identity of the two wordings is not accidental. It comes from the template of the seventh lens, which appears in all ten scenes, where the studio's judgment states that a handover without its reversal condition hardens a past judgment into today's command. That is the studio's general claim about handover, not a link between two scenes. March and Simon's Organizations (1958), on stored routines that keep a decision working after its makers have gone, is the theoretical form of the same recognition: an administrative decision becomes a premise for later work; this study's extension, and not a point that book makes in these terms, is that a premise transmitted without its reversal conditions becomes an order. Chesterton's fence is the practical form of it: a barrier put up for a reason may be respected after the reason is forgotten, and this study's extension of the parable is that a barrier put up for no reason at all may be respected after it is remembered to have none.

The practice the study proposes follows from the two sentences, and it is a modest practice. Leave a bundle of source identifiers, a statement of the valid scope, and the questions that call for re-judgment, rather than copying every conversation into a new document. The bundle is not a record of what happened, because a record of what happened is not what a successor needs. A successor needs the three things the records name: the source of the judgment, the range within which the judgment holds, and the questions that would make the successor re-judge it. Each of the counterexample conditions in Section 13.3 is a question of that kind, and each of the revision conditions in Section 13.2 is a statement of the same kind about scope.

This chapter has now assembled the material for a list. The next section draws one, and the section after that explains what the list will not do.

13.6A Working List of Removal Criteria

The list comes from the revision conditions the records attach to their own readings. Each criterion gives its condition, the record ground for it, the strongest objection, and what invocation would take; each states a finding, not a decision.

The purpose of the control has gone. The condition is verified, not suspected, irrelevance of what the control once did. The ground is the lock's revision condition:

That sentence converts a duty of detection into a duty of confirmation, and asks those who wait to remember nothing. The objection: the confirmation is made by whoever still runs the control, on evidence they can see and those who wait cannot, and a purpose may have gone in the form the organisation recognises while surviving in one it does not. The gain falls to those waiting on the control, the risk to those relying on it — usually different people invisible to each other. For invocation: the confirmation rests with the approval-feature changer, and the record lists as missing the transaction boundary the lock was meant to protect, so those who wait cannot tell whether it has been made. What it would take is that the confirmation be kept where they can read it.

The boundary it protects has been verified as gone. Stronger than the first: not that the purpose is unclear, but that the guarded boundary has been shown to be gone, or never to have been what the restraint assumed. The ground is the label scene's scope condition:

Confirmation has already been done for some items and not others, so a boundary of this kind can be verified item by item. The option scene agrees: "If only a specific option is stopped as normal operation, do not exaggerate it as a full stoppage" — Option Sold-Out, Start from the record, Revise when. The gain is precision: a restraint removed because its boundary has gone rests on a narrower ground than one removed because nobody remembers, and is less easily reinstated out of sentiment. The objection comes from design rather than memory: where a system's parts interact in ways nobody can see at once, the boundary may be exactly what the design protects, and its absence invisible to the verification that would report it gone; Perrow's argument on interactive complexity gives that weight. For invocation: among the ten scenes read closely, both codings found that those who wait cannot see what the restraint protects in 4 cases and can in 1, the record does not say in 1, and 4 are unsettled; a criterion of this kind is invocable only by someone who can see the boundary, and where it cannot be seen it asks the holder to judge it.

The cost of holding now exceeds the cost of the harm it holds back, and who bears that cost has been named. The condition is that a comparison has been made, and carries names. The ground is the cost lens's condition, which limits what may be attributed:

The gain is that a restraint is removed because the people who pay for it have been counted, not because the people who run it have grown tired of it. The objection: the counting is done by those who bear the cost, who have the least standing to judge the harm the restraint holds back, and the risk falls on those who bear the harm. The option in every cost-of-rollback lens links follow-on waiting, rework and the person in charge to the decision record; the risk recorded against it is that the record becomes a penalty table, and the field hides its cost. For invocation: a waiting party can name its own waiting but cannot name the harm the restraint prevents, so the naming must be done where the field will not read it as a ledger of blame.

An alternative route exists and is in use. Not that a route around the restraint could exist, but that one exists and is operating. The ground is the transfer scene's condition:

The gain is that the restraint can be removed without anyone accepting the risk it was covering, because something else covers it. The objection: existence is not use; the route may be thinner than the primary one, or used only by people who already know about it, so the restraint is removed for the informed and left in place for the rest. The bearer of that risk is the person at the far end of the route, which in the transfer scene is the payee. For invocation: those who wait cannot invoke a route they do not know exists, and the condition for checking one is written where the fund execution approver can apply it, not where the approval-request preparer can name it. What it would take is that the existence and use of the route be stated with the restraint itself.

The people who must act within the restraint cannot see what it protects and cannot find out. Not merely that a boundary is hidden, but that it cannot be reached by any means available to the person working under it. The ground is the information-threshold lens's condition:

The gain is that the organisation stops relying on a compliance it cannot distinguish from competence, and can be told something by the people closest to the work. The objection: invisibility is not always a defect. Some boundaries are invisible to everyone, including the holder, and removing a restraint on the ground of invisibility produces local judgment where no one can see further; the risk falls on those who must now explain a delay with no rule to point to. For invocation: the people who must act within the restraint are the only party who can say what they cannot see, and that statement has nowhere to go: both codings put applying a removal condition with the holder or a third party in 81 cases, against 3 where it rests with those who wait.

The restraint has become a surface on which one person holds another's time without an order. The condition is that the holding is not the decision of anyone, yet somebody's time is held because of it. This is the one criterion describing a restraint never deliberately put in place. The ground is the waiting scene's condition, which admits the restraint may be legitimate:

The gain is that a person's time is released, and the organisation stops treating a silence as a permission structure. The objection: the silence may be load-bearing. If real work requires a person to stay, and the restraint is removed on the assumption that it was only a hierarchy effect, the response the organisation needed is lost, and the risk falls on the same employee. For invocation: the employee who waits is the only party who knows the silence is operating, and the only party without means of asking whether it is meant to; invocation would require an end time stated where the waiting person can see it, and the record contains none.

The exception has become the standard without a decision. The condition is that an exception was permitted once for urgency, and no record shows that it was an exception. The ground is the once-only condition's withdrawal trigger, with the goods scene's companion condition:

The gain is that the organisation regains the ability to say what its rules are. The objection: acting on this criterion produces a large inventory of exceptions nobody recorded, and the people who benefited from them are asked to account for them after the fact. A second objection concerns systems whose parts interact in unfamiliar ways, where an unrecorded exception may have been the only thing keeping a flow moving; Perrow's argument gives that weight. The bearer of the risk is the field. For invocation: the field cannot invoke what it does not know is an exception, and knowing would require the exception to be recorded where the field can read it.

Cacciari's withholding power gives the list one shape: the withholding is ambiguous, holding back the lawless one while also holding back the end and the coming the faithful await, and containing lawlessness only by taking some of it into itself. This study reads from that ambiguity that the withholding operates as a function rather than as a person, so a criterion that identifies a person as the problem has not finished its work; the list has no such criterion. Weber's account of the official's obedience supplies the other caution, on this study's reading of it: a criterion treating the executor as merely carrying out someone else's decision has removed the one place where a restraint can be judged by the person closest to it. What each invocation leaves open is therefore deliberate: not which criterion fires, but where its ground is kept.

13.7What This Study Will Not Decide

This chapter will not say that any of the seven criteria makes removal right. A criterion is a finding, and a finding does not decide anything. The same facts can support keeping a restraint and withdrawing it, and the records are full of cases where they do: a lock whose reason is not remembered may be guarding a control the organisation forgot, and it may be guarding nothing; an option sold-out that stops one item correctly is normal operation, and it may be the visible edge of a stoppage nobody has described; an automatic intake that fails at one step may have its cause in the logic, and it may have its cause in stock.

This chapter will not say that the readings in ARCHON's cards are correct. Each is dated and marked as an interpretation that can be corrected, as Chapter 6 noted; this chapter's own list is provisional in the same way. It is provisional in a specific way: it is drawn from the revision conditions, and a revision condition is a statement about evidence that has not yet arrived. A list built out of such statements inherits their status. It is a working instrument for people who can see the boundary of a restraint and are trying to decide whether to keep it.

This chapter will not say that the decision belongs to the study. It belongs to the people in the organisation: the ones who can name the boundary, the ones who can name what the restraint was built against, the ones who bear its waiting and its rework, and the ones who will be there when the next exception arrives. The tradition used in this study has been a lens and not a verdict at every stage, and this is the stage where that distinction matters most. A theory that delays every coming cannot tell you which coming to let through, and a theory that delays nothing cannot tell you what to hold back. The people in the scene have to decide, and the best this study can do for them is to leave the four questions available, as Chapter 1 did: what restrains here, what does it protect, what does it cost and who pays, and what would make it go.

One final thing this chapter will not do. It will not say that restraint is good or bad in general, and it will not say that the power that withholdeth is, in an organisation, a good or a bad thing. The ten scenes have each shown a restraint that protected something and cost someone, and each has shown a boundary that some people could see and others could not. On this study's extension of Cacciari's reading, the removal of a restraint in an organisation releases whatever it was delaying, which may be a good or a bad or both. The people in the scene decide, and they decide on evidence of a kind that has a time point and a scope of application. This study has tried to state what that evidence would have to look like, and it has tried to leave the decision where it belongs.

13.8Removal beside the Study of Rules

Sections 13.2 to 13.7 assembled a set of conditions from the records and then declined to let them decide anything. The study of rules set out in sections 4.6 and 4.7 has been held back until now because it is the body of work that most directly tests a list of that kind, and it tests it from both sides. Six works bear here: Kaufman on red tape, Merton on rules that outlive their purposes, Vaughan on normalisation of deviance, Feldman and Pentland on routines, Lindblom on small steps, and Chesterton's parable of the fence. None of the six supplies a criterion. Each changes what a criterion can be asked to do.

Kaufman complicates the third criterion directly. The criterion says that the cost of holding now exceeds the cost of the harm it holds back, and that who bears the cost has been named. Kaufman's point is that red tape is mostly the residue of protections that people demanded: one person's red tape is another's safeguard, so cutting it means deciding whose protection to drop. The comparison the criterion requires is therefore not only between two costs but between two sets of protected people, and the list does not tell you which set to favour. Kaufman strengthens the insistence on naming bearers and weakens any assumption that a named bearer settles the choice.

Merton supports the first criterion and warns against the organisation that would use it. Officials come to treat rules as ends in themselves and keep to them after they have stopped serving their purpose, which is exactly the finding the lock's revision condition names: if it is confirmed that the control no longer applies, it is not maintained out of inertia. Merton's second notion, trained incapacity, cuts the other way: training that fits one situation unfits people for another. A rule abandoned because its purpose appears gone may have been the rule that kept people competent for a situation still to come, and the list cannot see that in advance.

Vaughan bears on the estimate of harm, not on the holding. In Vaughan's account, signals of danger came step by step to be accepted as normal within the culture and its rules, so that following the rules led to the Challenger disaster. If harm has been normalised, the cost side of the third criterion is understated by the very people asked to measure it, and the criterion's requirement that bearers be named becomes the harder half of the test rather than the easy one.

Feldman and Pentland support the fifth criterion and unsettle the first. A routine has an ostensive aspect, the idea of it, and a performative aspect, what people actually do, and change comes from the gap between the two. Where people who must act within a restraint cannot see what it protects, the gap is likely wide: the stated idea of the routine and its performance have diverged, and no one can tell which one the restraint belongs to. But the same gap means that a purpose which has vanished from the stated idea may still be performed every day, so "the purpose has gone" can be a finding about the document rather than about the work.

Lindblom fits the records' habit of narrowing scope before acting. Policy is made by successive limited comparisons, small steps from what exists, rather than by comprehensive choice among all options, which is what the records' options ask for when they propose proceeding first in a small scope and comparing the results. Lindblom also complicates the wording "verified as gone": this study applies Lindblom's account to verification by reading it as arriving in steps like any other decision, and on that reading a criterion that waits for complete verification may never fire, while one that fires on a partial step may remove a restraint on evidence that was itself provisional.

Chesterton's fence is the criterion the chapter opened with and the one the six works leave standing. The parable says that the reason a fence was put up must be found before it is taken down, and that the absence of a visible reason is not evidence that there was none. The lock scene is Chesterton's case made literal: a reason that could not be remembered accurately. The records add what the parable does not supply — a test. The lock's condition is not "remember the reason" but "if it is confirmed that the control no longer applies, it is not maintained out of inertia," which converts a duty of detection into a duty of confirmation. Where earlier sections stand in the debate over this chapter's four questions is therefore visible: 13.2 to 13.5 identified what would count as evidence, 13.6 converted that evidence into findings, and 13.7 refused findings the status of verdicts. Kaufman, Merton, Vaughan, Feldman and Pentland, Lindblom and Chesterton all answer part of the second question — what a restraint protects — and Kaufman answers part of the third, who pays. None of them answers the first question, what restrains here, for a particular scene, and none answers the fourth, when the restraint should go.

That gap is the one thing the list adds. The six works describe how rules behave in general; the revision conditions are dated statements, made in advance, about what would change a specific reading of a specific restraint. On this study's comparison of its revision conditions with those lines of theory, no theory of red tape, displacement, deviance or routine supplies that. The list is weaker than the works in scope and stronger in address: it can be checked against a record, revised when material arrives, and handed to the next person in charge with its reversal condition attached. Whether that is enough is left, as 13.7 leaves it, to the people who can see the boundary.

13.9The Thesis and Its Grounds

This chapter has declined to answer the study's four questions; it has shown what an answer would have to consist of. What restrains here has been left with a source the records usually cannot name; what it protects, with a boundary the waiting party usually cannot see; what it costs and who pays, with a cost that travels to the layer nearest the stop; when it should go, with evidence conditions rather than a settlement.

The thesis. A restraint in an ordinary organisation can claim to hold rightly only when its reason, and the condition of its removal, are kept where those who wait under it can see them and invoke them. This is a condition on how a restraint is made and kept, not a verdict on any option in the records, and it does not say that restraint is good or bad in general. Nothing in it decides which of the seven findings should fire. A restraint which cannot be seen into, and cannot be answered, cannot claim the standing it is given.

Augustine, and the unnamed holder. Book XX, chapter 19 of The City of God reports the names offered for the holder and declines to settle on one: "I frankly confess I do not know what he means." The thesis borrows the silence, not the doubt, and takes as its own reading rather than an established claim about that chapter the point that naming the restrainer is itself an exercise of power over how the holding is judged. The consequence is that justification cannot rest on who holds. A holder may be unnamed, absent, replaced, or written into a program with no author at all, and the ten scenes show each of those forms. What can be kept in view is not a person but a reason and a condition.

Hobbes, and a condition nobody may invoke. Hobbes gives obedience a temporal limit, stated in chapter XXI of Leviathan: "The Obligation of Subjects to the Soveraign is understood to last as long, and no longer, than the power lasteth, by which he is able to protect them." The same chapter adds, as the limit's end, "The end of Obedience is Protection." On this study's reading the subject cannot invoke that limit, because chapter XXIX calls seditious the doctrine that "every private man is Judge of Good and Evill actions". The thesis keeps the condition and changes its address: the temporal limit is written where those who wait can see it and name it, so that protection failing becomes a reason a waiting party can raise rather than an observation a private judgement must make. Hobbes supplies the condition, not the invocation.

Schmitt, and a justification that cannot be checked. Political Theology (1922) defines the sovereign as the one who decides on the exception, the case no existing norm covers and no legal norm can settle in advance; the exception, in that book, plays in jurisprudence the part the miracle plays in theology, and the study does not speak of the katechon. Carl Schmitt served the Nazi regime from 1933, and critics read his theory of the exception in that light. The katechon appears in The Nomos of the Earth (1950), where the account of the Christian empire has the empire's continuity rest on the idea of the katechon: the empire as the historical power able to restrain the coming of the Antichrist and the end of the present age. The reading of The Nomos of the Earth given here is a separate matter from that sentence, and this study does not treat one as the reason for the other. The thesis inherits what chapter 2 drew from that account: a justification that can be checked by no one outside the order is not a reason those who wait can see, and not a condition they can invoke.

Chesterton, turned on the holder. The essay "The Drift from Domesticity" in The Thing (1929) gives the rule that a fence's reason must be known before it is removed: "If you don't see the use of it, I certainly won't let you clear it away. Go away and think." Where that rule, applied to a lock whose reason cannot be remembered, protects inertia, the thesis turns the demand on the holder rather than on the remover: the reason should be kept with the fence, at the fence, where those who wait under it can see it. Chesterton's gain survives the turn — the fence may be guarding what the organisation forgot — and what changes is who owes the answer. The person who cannot see the use of it should not have to go away and think alone; the use should be where they are standing.

Hart, Fuller, Ranchordás. Hart's The Concept of Law (1961) describes a legal system as a union of primary rules with secondary rules about those rules — a rule of recognition, rules of change, rules of adjudication; a regime of primary rules alone would be uncertain, static and inefficient, and rules of change remedy its static quality. On that account a rule that cannot be changed has lost one of the instruments a legal system needs. Fuller's The Morality of Law (1964) tells the inner morality of law through a king, Rex, who fails each of its eight requirements in turn: there must be rules; they must be published, prospective, clear, consistent with one another, possible to obey, and not changed too often; and official action must agree with the rules as declared. The thesis asks for something close to publication and to the agreement of official action with the declared rule, with one difference: Fuller's requirements bind officials to rules they already have, while the thesis asks that a rule carry, beside itself, what would take it away. Ranchordás's Constitutional Sunsets and Experimental Legislation (2014) studies the sunset clause, which makes a law lapse at a set date unless it is renewed, weighing its promise of review and learning against its costs. The thesis is not that every restraint carry a sunset date; it is that the condition of removal be stated where those who wait can invoke it. A sunset is one such condition, and its costs are real.

Barnard, Tyler, Bovens, Adler and Borys. Barnard's The Functions of the Executive (1938) puts authority in the acceptance of an order by those to whom it is addressed, and a person accepts an order that is understood, consistent with the organisation's purpose and with personal interest, and possible to carry out; within a zone of indifference orders are accepted without conscious question. The thesis rests on that first requirement: an order whose boundary is invisible to the person acting under it is authority resting on a zone of indifference rather than on acceptance. Tyler's Why People Obey the Law (1990), from a survey of people in Chicago, finds that people obey the law chiefly because they see the authorities as legitimate and their procedures as fair — being heard, neutrality, respectful treatment — more than from fear of sanctions or from the outcomes they receive. The thesis is a procedural claim of that shape: a restraint can claim to hold rightly when the procedure by which it is kept is one the waiting party can see and invoke. Bovens's "Analysing and Assessing Accountability" (2007) defines accountability as a relationship between an actor and a forum, in which the actor must explain and justify conduct, the forum may question and pass judgement, and the actor may face consequences; the thesis adds one thing, that the actor must be able to show the condition under which the conduct would stop. A forum that can question but cannot name what would end the restraint can do little but re-affirm it. Adler and Borys's "Two Types of Bureaucracy" (1996) hold that formal rules can enable or coerce, and that the difference lies in how the rules are designed; the thesis adds to that design claim a temporal element, that the reason and the condition must not only be shown but kept at the point where a decision about removal will later be taken.

The grounds in the records. Section 11.8 gives the ledger's counts in full; a few of them carry the thesis. A condition for lifting or narrowing a restraint is named in most scenes: both codings found one stated in 139 of the 186, and in 2 more both found one with at least one inferring it. What is missing is its address and its reason. On the cells where the codings agree, applying the condition rests with whoever holds the restraint or with a third party in 81 scenes, and with those who wait in 3. The scene's own record says why its restraint exists in 27 scenes, and in 1 more by inference; it does not say in 89, and says that the reason is not known in 1. Whether those who wait can see what the restraint protects and how far it reaches the record mostly does not say, in 73 scenes; where the codings agree that it does say, they cannot see it in 26 and can in 5. No scene of the 152 that show a restraint meets the study's test, and none of the ten read closely does. Only one reading lets a scene through. Where the larger model's reading is added to the first coding and a cell counts when two of the three readings agree, a rule set after the fact and reported apart, one scene of 186 meets the test: the record of a plan to show full videos on a few products first and to settle the management features after seeing their traffic and conversion. On this study's reading it is a scene in which those who hold the restraint and those who wait under it are the same people. On this evidence the condition is met only where no one else waits. And the conditions themselves are rarely stated in a form anyone could check: 2 of the 186 conditions for revisiting give a number, a threshold or a time.

Where the counts do not go the thesis's way, this study says so. First, the four hypotheses registered before the coding (K1 to K4) are supported only on the lenient count, which also admits values the two codings agreed on where at least one inferred it from the setting; on the strict count, which admits only what both found in the record's words, each is unclear, because the records seldom say these things outright. Second, ARCHON's judgments almost always give a reason for what they recommend: in 174 of the 186 scenes both codings found one stated, so a further hypothesis registered with the ledger, that a restriction outlives the memory of its reason, goes against on both counts (a share of 2%). The reason a judgment gives for its recommendation is not the reason a scene's restraint exists; but the count shows that a reason can be kept where someone undertakes to keep it, and it stops the thesis from becoming a claim that organisations cannot record reasons. Third, the two codings agree on the scene's reason in 71% of the scenes and on the boundary in 63%, and readers of two other families, on 40 scenes drawn at random, agree with each other better than with the codings, and least of all on the boundary. The cells the thesis needs are the cells the records least let a reader settle.

These counts are about records, not workplaces. Read together, they show that the two things the thesis asks for are the two things these records least often carry: a reason kept with the restraint, and a removal condition in the hands of those who wait. The condition is named in most scenes; what is missing is its visibility and its address.

What the thesis does not say. It does not decide which of the seven criteria in 13.6 fires, and does not convert any of them into a rule for anyone in the scenes. It does not say that restraint is good or bad in general, or that the power that withholdeth in an organisation is good or bad. It does not say that ARCHON's readings are correct; each is dated and marked as an interpretation that can be corrected, and the counts in section 11.8 carry the same status. It does not say that the decision belongs to this study. On this study's extension of Cacciari's reading, that the withholding operates as a function rather than as a person, the removal of a restraint releases whatever it was delaying, which may be good or bad or both, and the people who can name the boundary and the people who bear its waiting and its rework decide between them. What the thesis claims is narrower and, on these counts, unmet by the records before it: that the reason and the condition of removal belong where those who wait under a restraint can see them and invoke them.

13.10Objections and Replies

Section 13.9 stated this study's thesis in one sentence: a restraint in an ordinary organisation can claim to hold rightly only when its reason, and the condition of its removal, are kept where those who wait under it can see them and invoke them. Eight objections follow. None is answered by new evidence; each is answered by showing what the thesis does not claim and where earlier sections already stand.

1. The katechon is an eschatological figure, and its use here is a category mistake. The objection is right about the tradition's subject matter: the figure belongs to a letter about the end of days, and the later tradition filled its silence with empire and Antichrist, so no argument about a journal lock, a printed label or a holiday transfer gains authority by standing under that history. The study uses the katechon as a heuristic lens — a way of finding variables, paths and questions, not a description of the last things — and section 3.4 showed where it fits badly: where a stop is a designed buffer rather than a delay, where a rule has no temporal shape at all, where a person's absence is not holding anything back. Part II kept every scene's own record beside the lens rather than inside it. The two features of the passage the thesis uses — the unnamed holder, and the holding that ends at a stated removal — are set out in section 13.1; neither depends on believing anything about the end of days. What the answer concedes: the borrowing is not free. The word the study takes its title from carries two thousand years of reading, and a reader who meets it in a chapter about employee time will hear the tradition whether or not the argument needs it. Section 3.4's list of failures is this study's own admission that the lens distorts what it shows.

2. The study dresses mundane controls in grandeur. A button on a sales screen and an absent manager are being asked to stand in the same sentence as empire and the end of the age, and the comparison flatters nothing while obscuring the ordinary management question — is this rule worth its trouble. This study's answer is that the thesis does the opposite of dressing anything up. It asks every restrainer, in every scene, the same two unglamorous things: what is your reason, and what would take this away. A lock that blocks post-period journal approvals does not need to be compared with a sovereign to be asked what it protects and who can lift it. The revision conditions attached to ARCHON's judgments are such questions already asked in fragments, and section 13.2 gathered ten of them without any theory attached. What the answer concedes: nothing in the ten scenes is made more precise by the word katechon, and a reader who comes away thinking that ordinary controls are forms of Antichrist restraint has been misled by the study's own title. Sections 2.4 and 4.5 made the same concession in advance: what carries from Hobbes and Schmitt into an office is a small amount, and what carries from the parable of the fence is its question, not its tone.

3. The evidence is synthetic, and the findings are the studio's template. The strongest form of this objection has already been put against this material in Chapter 11: the records are anonymised summaries rather than raw transcripts, the seven lenses beyond the first suppose what could follow rather than report what happened, and their judgments follow one template across all ten scenes, so that their agreement is a property of the template rather than a finding about organisations. Every part of that is conceded. This study treats the studio's records as a frame, not as a workplace: it cites the scene's own record and the first lens as evidence, quotes the other lenses as dated interpretations marked correctable, and never turns a hypothetical condition into something that happened. The ledger reads only the scene's own record and the first lens, its counts come from code over 186 cases, a cell counts only where the two codings agree, and section 11.8 reports where they did not. Where a count runs against the shape this study expected, it is reported: a further hypothesis registered with the ledger, that a restriction outlives the memory of its reason, went against on both counts, because ARCHON's judgments almost always state a reason for what they recommend. What the answer concedes: these are counts about records, not about workplaces. One studio produced the summaries, one model read them twice, and the template's repetition is a known feature of the material. The thesis is therefore not a finding. It is a condition this study proposes, and the records before it do not meet it.

4. The study builds on a jurist who served the Nazi regime. The objection is at its strongest here: a theory of power that cannot name its own restrainer, used to justify a delay that measures its own rightness by what it postpones, was offered by a man who joined the Nazi party in 1933 and worked as a leading jurist of the regime, and a book that cites it has not settled what side it is on. Carl Schmitt served the Nazi regime from 1933, and critics read his theory of the exception in that light. That sentence stands in every part of this study that uses his thought, and the objection deserves more than the sentence. Section 2.6 set out the three answers the tradition itself offers: Peterson's argument of 1935 that Christian doctrine and the eschatology of the church end any political theology that mirrors one God in one monarch, with Schmitt's answer of 1970; Taubes's reading of Paul as a political thinker set against the empire, set against Schmitt's apocalyptic of the counter-revolution; Hell's reading of the katechon as an imperial theology bound up with the Third Reich's imitation of Rome. The thesis this study states is built away from all three positions. Where Schmitt's structure gives the condition of removal to whoever decides on the exception — a power standing outside the order yet belonging to it, deciding whether the order is suspended — the thesis gives that condition to those held by a restraint, so that seeing the reason and naming what would end it become things a waiting party can do rather than things a holder may permit. That redistribution is the whole of the movement from Schmitt in this study, and it is not a defence of him. What the answer concedes: the study cannot purify a figure by using it. No sentence here un-entangles the tradition's political readings from that history. Chapter 2 set out the reading of The Nomos of the Earth (1950) used here; that reading is a separate matter from the sentence about his service, and this study does not treat one as the reason for the other. A reader may reasonably hold both facts at once and decline to use the term at all.

5. The study of organisations already says all this. The objection is cumulative and largely right. Sections 4.6 and 13.8 set out its ground: Merton on rules kept after they have stopped serving their purpose and on trained incapacity, Gouldner on mock, representative and punishment-centred rules, Kaufman on red tape as the residue of protections people demanded. Barry Bozeman adds that whether a rule is red tape depends on whose purposes are counted, and distinguishes rules that are dysfunctional from the start from those that become so as circumstances change. Paul S. Adler and Bryan Borys hold that the difference between rules that enable and rules that coerce lies in how the rules are designed. Together these answer, better than this study does, what a restraint protects and who pays for it — the second and third of the four questions. What the tradition adds is the temporal shape of a holding that ends by removal, and the question of who holds the removal: the letter describes the holder being taken out of the way, a claim about the end of a holding rather than about its persistence, and, on this study's reading of the accounts cited above, no account of displacement of goals or red tape asks, of a particular rule standing today, who is authorised to end it and on what evidence. The records answer the same shape in miniature: 186 of 186 cases (100%) carry a condition for revisiting the judgment, all of them in the first lens's reconsideration field, of which 2 state a number, a threshold or a time. What the answer concedes: the tradition's contribution is a shape, not a superior analysis. The organisational literature is more adequate to every scene in this study except the question of removal. The thesis adds address, not insight: a dated condition kept where those who wait can invoke it.

6. Writing down reasons and conditions becomes a ritual or a simplification for the centre. This objection has four witnesses. Michael Power describes audit spreading through public and private organisations as a ritual of verification, which checks systems of control more than the work itself, makes organisations auditable, and produces a comfort that may be decoupled from what is done. James C. Scott shows that states make society legible by simplifying it, and that high-modernist schemes which impose such simplifications and override local, practical knowledge have failed, sometimes disastrously. John W. Meyer and Brian Rowan are the third witness, on formal structures adopted because they are taken for granted as legitimate and often decoupled from the work actually done; Paul du Gay is the fourth, defending the impartial procedural conduct of office as something that protects citizens from arbitrary power. The risk these four name is real and is conceded: a reason and a removal condition, once written down, can become another artefact for the centre to collect, another field in a ledger that the centre reads and the waiting party does not. That risk is not hypothetical in this study's own material — the ledger reports that applying a removal condition rests with the holder or a third party in 81 cases, against 3 cases of those who wait. The thesis answers only partially. It asks for a record visible to and invocable by those who wait; a record that a centre can audit and a waiting party cannot name is precisely what the counts describe. Adler and Borys bear on design — enabling rules show their reasons and leave room to repair — and Tom R. Tyler's finding, that people obey chiefly when they see the authorities as legitimate and their procedures as fair, supports making the condition a procedural object rather than a policy statement. What the answer concedes: a written condition can still be decoupled from practice. The ledger's count of conditions stated with a number, a threshold or a time, 2 of 186 (1%), suggests that a condition written in a document is not yet a condition anyone could check in a workplace. Power's ritual and Scott's legibility can be produced by exactly the reform this thesis proposes.

7. Some restraints protect only because they cannot be bargained with. James D. Thompson describes organisations sealing off their technical core from environmental uncertainty by buffering, smoothing and anticipating, and rationing when these fail; a buffer holds work back as designed, not as a concession. Perrow's account of tight coupling, set out in section 13.3, gives the objection its force: a condition of removal, kept where those who wait can see it, turns a designed buffer into a negotiable object, and in a tightly coupled system negotiation is how the failure gets in. The answer is that a condition can be strict and still be known: the thesis asks that a waiting party see what a restraint protects and how far it reaches, not that it decide when to lift it. Knowing a condition is not bargaining with it. Section 13.3 already carried Perrow's practical consequence for removal: if the control was built against a failure that can no longer occur as the system is now configured, its argument weakens on the same evidence that would have argued for it before; if the failure remains possible, its cost and its protection must be weighed by people who can name both. What the answer concedes: in tightly coupled systems visibility changes behaviour, and on this study's reading of Perrow a publicised condition can itself add complexity to a system short of slack. Removing a stop may produce the failure the stop was holding back, and that failure will then be attributed to whoever removed it. The thesis says only that the reason should be where those who wait can see it, and it accepts the risk that seeing may alter what is seen.

8. A fence whose reason is unknown should stand, as Chesterton says. The objection is the oldest in the study and the hardest to answer, and section 13.8 set out the rule: the reason a fence was put up must be found before it is taken down. Applied to a journal-entry lock whose original reason could not be remembered accurately, the rule says the lock stands until someone does the thinking. The answer has two parts. The first is that the rule assumes what the records say may not hold: that the reason can be found. In the lock's own scene the record states that the reason could not be remembered accurately, and across the library the ledger finds the scene's own record silent on why its restraint exists in 89 scenes, against 27 where it says why and 1 where it says the reason is not known. Where the reason cannot be recovered, the rule protects inertia rather than the unknown reason: it places the burden of detection on the person who cannot see the boundary and cannot find out. The second part is that the thesis turns the demand on the holder rather than on the remover. The reason should be kept with the fence, at the fence, where those who wait under it can stand. Chesterton's gain survives the turn — the fence may be guarding what the organisation forgot, and the thesis does not say it may be removed — and what changes is who owes the answer. What the answer concedes: the turn demands work that many holders cannot do, and this study has never shown that they will do it. A reason that could not be remembered accurately in 2016 may be recoverable in 2026 or may not; the thesis cannot tell the difference, and the lock's own condition — if it is confirmed that the control no longer applies, it is not maintained out of inertia — concedes the same thing by making removal depend on a confirmation that may never come. The thesis does not remove the risk of removing a fence whose reason is genuinely forgotten; it relocates it, from the person who cannot see the use of it to the person or system that put the fence there and let the reason go.

Eight objections, eight partial answers, eight concessions. None is settled by reading more carefully. What restrains here is still often unnamed, as section 13.9 showed; what it protects is still often invisible to those who wait, in 4 of the ten close-read cases with 4 more unsettled, and across the library mostly not recorded at all; what it costs and who pays still travels to the layer nearest the stop; and when it should go is still left with the evidence conditions 13.3 set out. The thesis stands as a condition on how a restraint is made and kept, beside objections that would weaken it, flatter it, or leave it only where it is most needed: at the fence, with the reason in view.