The Collateral Order Doctrine

The Collateral Order Doctrine

Every route covered so far depends on a statute that names a category of order. The collateral order doctrine is judge-made, yet it grows out of the same statute that creates the final judgment rule: 28 U.S.C. § 1291, which gives the courts of appeals jurisdiction over “all final decisions of the district courts of the United States.”¹ The doctrine takes its name from Cohen v. Beneficial Industrial Loan Corp., a 1949 decision in which the Supreme Court allowed an immediate appeal from a ruling that did not end the lawsuit.² This section asks when a ruling that leaves the case unfinished can still count as a final decision.

The Supreme Court describes the doctrine not as an exception to the final decision rule but as a “practical construction” of it, which places the doctrine inside the statute’s text.³ Some courts of appeals still call it an “exception” or a “detour,” but the Supreme Court’s description controls.⁴ The difference matters because a collateral order is itself a species of final decision, in the way a Rule 54(b) partial judgment is, so a party appeals it as of right under § 1291 and needs neither the district judge’s certification nor the court of appeals’ permission that § 1292(b) requires.⁵ Because the doctrine hangs on the single word “final,” the Court has kept it narrow: “stringent” conditions must govern its application so that it does not “overpower the substantial finality interests § 1291 is meant to further.”⁶

Three conditions define the class, and an order must satisfy all three. It must conclusively determine the disputed question, resolve an important issue completely separate from the merits of the action, and be effectively unreviewable on appeal from a final judgment.⁷ Each condition answers a different objection to an early appeal. A conclusive ruling gives the court of appeals a settled question to review, not a tentative one the trial judge may soon change. A question separate from the merits means the appellate court will not decide, or have to decide again, what the trial court is still deciding. A ruling that cannot be reviewed effectively after final judgment is one whose harm waiting cannot repair, so the early appeal is the only one that can do any good. Together the conditions describe claims “too important to be denied review and too independent of the cause itself to require that appellate consideration be deferred until the whole case is adjudicated.”⁸

The class is small, and the Supreme Court has warned that the doctrine “must never be allowed to swallow the general rule that a party is entitled to a single appeal, to be deferred until final judgment has been entered.”⁹ The warning rests on the reasoning behind the final judgment rule itself. If every important-seeming pretrial ruling could be appealed at once, piecemeal appeals would become routine, and the delay, the interruption of the trial court, and the burden on the appellate docket would return through a side door. Courts therefore ask whether a kind of order belongs in the narrow class; the weight one ruling carries for one litigant does not decide the question. When an exam stem presents an adverse pretrial order that matches no statutory category, test it against the three conditions. Most orders fail at least one.

The subsections below begin with Cohen itself, take the three conditions in turn, and end with a survey of the orders the Supreme Court has accepted as collateral orders and those it has refused. As a rough sorting rule, rulings that deny an immunity from suit or a bar to prosecution usually qualify, while rulings that a reversal after final judgment can repair usually do not.

Cohen and the Practical Construction

The doctrine takes its name and its logic from Cohen v. Beneficial Industrial Loan Corp., where the Supreme Court could not reach the merits until it decided whether the appeal was allowed at all. Section 1291 provides for appeal only “from all final decisions of the district courts,” and the Court conceded that if Congress had allowed appeals only from final judgments that terminate an action, the order before it would not be appealable.¹⁰ The Court nonetheless read “final” practically, and later cases describe the result as a “practical construction” of § 1291.¹¹ Cohen thus answers a narrower question: which rulings, short of the last one, still count as final decisions?

A shareholder who owned 100 of the corporation’s more than two million shares sued derivatively in federal court, charging its managers and directors with a conspiracy to enrich themselves at the corporation’s expense.¹² While the suit was pending, New Jersey enacted a statute, applicable to pending actions, that made a plaintiff with so small an interest liable for the defendants’ reasonable expenses and attorney’s fees if the suit failed and entitled the corporation to indemnity before the case could be prosecuted.¹³ The district court refused to apply the statute, the Court of Appeals reversed, and the Supreme Court granted certiorari.¹⁴ The Court reached the statute’s validity and its application in federal court only after it settled that the appeal was properly before it, and that threshold ruling is the part of the case this chapter follows.¹⁵

The Court began with what an appeal is. The statute disallows appeal from any decision that is tentative, informal, or incomplete, because an appeal gives the upper court a power of review, not a power of intervention, and so long as a matter remains open, unfinished, or inconclusive, an appeal would intrude on the trial court’s work.¹⁶ The reason follows from the division of labor between the two courts: a ruling the trial judge may still revise gives the court of appeals nothing stable to review, and an appellate opinion on it could be overtaken by the judge’s next order. The Cohen order was different. The district court’s action on the application was concluded and closed before the appeal was taken, so the decision was final in that sense.¹⁷ The order was final on the security question and the trial judge would not alter it, and refusing review might have denied any right of review at all.¹⁸

Conclusiveness alone did not decide the case, because the statute also does not permit appeals, even from fully consummated decisions, when they are only steps toward final judgment in which they will merge.¹⁹ A ruling merges when the case ends and the ruling folds into the judgment, so the loser can attack it in the single appeal the final judgment rule contemplates. The statute’s purpose is to combine in one review all the stages of a proceeding that can effectively be reviewed and corrected when final judgment results.²⁰ Suppose a trial judge overrules a party’s challenge for cause to a prospective juror. The ruling is conclusive on that question, yet it is one stage of the trial, and the party can attack it after the verdict, so it waits for the final judgment. The Cohen order stood differently, because it made no step toward final disposition of the merits and would not be merged in the final judgment.²¹

That difference mattered because of what the order decided. The state statute entitled the corporation to indemnity before the case could be prosecuted, so a ruling that the corporation had no such right could not be undone by a reversal after the case had been prosecuted to judgment.²² The Court put it this way: when final judgment came, it would be too late effectively to review the order, and the rights the statute conferred, if it applied, would have been lost, probably irreparably.²³ The Court therefore held the order appealable because it was “a final disposition of a claimed right which is not an ingredient of the cause of action and does not require consideration with it.”²⁴ In the words of a later case, such a ruling belongs to the “small class” of rulings that do not conclude the litigation but conclusively resolve claims of right separable from, and collateral to, the rights asserted in the action.²⁵

Later decisions compress this reasoning into the three conditions stated above, and the next three subsections take one condition each. The Court itself restated Cohen in Eisen v. Carlisle & Jacquelin as resting on two grounds: the ruling was not tentative, and it concerned a collateral matter that could not be reviewed effectively on appeal from the final judgment.²⁶ It then applied those grounds to an order placing 90% of the cost of class notice on the defendants, which conclusively rejected their claim that they could not lawfully be made to bear that expense and involved a matter unrelated to the merits.²⁷

The Court fenced in its holding in the same opinion. It did not mean that every order fixing security is appealable, because the right to security there presented a serious and unsettled question. If the right had been admitted or clear, and the order had involved only discretion over the amount of security, a matter the statute makes subject to reconsideration from time to time, “appealability would present a different question.”²⁸ The caution maps onto the first step of the checklist. An order that sets or adjusts an amount stays open to revision as the case develops, so it resembles the tentative rulings the Court excluded, while the Cohen order decided whether the corporation had any right to security at all. The Court left the discretionary-amount question undecided, but its reasoning shows why that order would be a weaker candidate.

Cohen allowed an early appeal because every step of its reasoning pointed the same way: the ruling was final, it would not merge into the merits, and waiting would lose the right. An adverse pretrial ruling that looks important, or even conclusive, has not yet been shown to be a collateral order, because the remaining questions are whether it is separable from the merits and whether review after judgment would come too late.

Conclusive Determination

The first condition comes from the reasoning in Cohen, where the Supreme Court explained that the final-decision statute disallows appeal from any decision that is tentative, informal, or incomplete.³⁰ An appeal gives the upper court a power of review and no power to intervene, so an appeal has to wait while a matter remains open, unfinished, or inconclusive.³¹ The security ruling cleared that bar because the district court’s action on the application was concluded and closed before the appeal was taken.³² Later cases restate the point as the first of three conditions: the order must conclusively determine the disputed question.³³ Eisen v. Carlisle & Jacquelin described Cohen the same way, as a ruling that was not tentative, informal, or incomplete and that settled conclusively the corporation’s claim to require security for costs.³⁴

The condition exists because of what an appellate court does. Review presupposes a finished decision to examine, and a trial judge’s provisional ruling is only a draft of one. If the court of appeals took up a ruling the trial judge remained free to change, the trial court’s next order could overtake its opinion, and the revised ruling would need a second appeal. The early appeal would multiply review in exactly the place where the final judgment rule confines review to a single round. Requiring a settled answer means the appellate court reviews the trial court’s last word on the question.

A literal reading of conclusiveness would make the condition impossible to meet. Rule 54(b) provides that any order or other decision, absent a partial final judgment, “may be revised at any time before the entry of a judgment,” so every collateral order is, in the most rigid sense, not conclusive.³⁵ In Moses H. Cone Memorial Hospital v. Mercury Construction Corp., the Supreme Court met the argument that a stay order was not conclusive by calling it “true only in the technical sense that every order short of a final decree is subject to reopening at the discretion of the district judge.”³⁶ The inquiry therefore looks to the practical situation, not a rigid one.³⁷ As one court of appeals put it, the question is whether “the district court has clearly said its last word on the subject.”³⁸ A judge’s bare power to reopen a ruling does not make it tentative. What makes a ruling tentative is a realistic likelihood of later reconsideration, which arises when the rules invite adjustment or when the ruling belongs to a process that is still developing.³⁹

Class certification shows a ruling that fails the test on the face of the governing rule. Rule 23(c)(1)(C) says that an order granting or denying class certification “may be altered or amended before final judgment,” so the order is tentative by the rule’s own terms.⁴⁰ Before Rule 23(f) allowed interlocutory review of class-action determinations in certain situations, courts held class-certification orders to be mere tentative rulings.⁴¹ In Coopers & Lybrand v. Livesay, the Supreme Court refused to treat a class-certification ruling as a collateral order, finding among other defects that certification is subject to revision by the trial court.⁴² Orders declining to appoint counsel for a plaintiff fail for the same reason, because they may be reconsidered as the trial-court proceeding continues.⁴³ Rule 23(f), a rulemaking route taught later in this chapter, now permits what the collateral order doctrine could not.⁴⁴

Case-management orders in complex litigation fail for a related reason. In re Recticel Foam Corp., a consolidated action with about two thousand plaintiffs and two hundred defendants, the district court ordered the defendants to share the expense of reproducing certain videotapes.⁴⁵ Recticel sought an interlocutory appeal, and the court of appeals held that the orders were not final. It reasoned that they remained subject to modification under the discovery rules, that the costs in question continued to mount, and that no irreversible legal consequences flowed from them as they stood.⁴⁶ The same features count against conclusiveness for collateral order purposes, because over lengthy discovery a trial court could revisit a cost allocation more than once.⁴⁷ Each order was one move in a continuing process, with further adjustments to come.

Cohen sits on the other side of the line, and the two failing categories show why. The district court there decided whether a state security statute applied to a federal derivative suit at all, a question the trial court would not revisit as the litigation unfolded.⁴⁸ A question of that kind receives one answer, and nothing in discovery or at trial gives the judge cause to reconsider it. A class-certification ruling or a discovery cost allocation depends on how the case develops, so the judge expects to adjust it. The same logic explains why Cohen set apart an order that only fixed the amount of security, a matter the statute left open to reconsideration from time to time.⁴⁹

A yes to question one or two, or a no to question three or four, points against the condition. These questions test only the first of the three conditions. An order that passes them must still be separate from the merits and effectively unreviewable after final judgment, and Coopers & Lybrand found class certification wanting on all three.⁵⁰ Two errors recur on exams. A student who reasons that every pretrial order is revisable, so nothing is ever conclusive, erases the condition and ignores the practical reading the Supreme Court adopted. A student who treats a firm-sounding order as enough has applied only the first condition, and the doctrine requires all three.

Separate and Important

The second condition requires that the order resolve an important issue completely separate from the merits of the action, and it comes from the reasoning of Cohen itself. Once the Court had found the security ruling conclusive, it asked whether the ruling was only a step toward final judgment in which it would merge, and it answered that the matters decided were “not of such an interlocutory nature as to affect, or to be affected by, decision of the merits of this case.”⁵² It then held the order appealable because it was “a final disposition of a claimed right which is not an ingredient of the cause of action and does not require consideration with it.”⁵³ The passage contains two requirements, separateness and importance, and each rests on its own reason.

Separateness protects the single-appeal principle that the final judgment rule serves. An issue intertwined with the merits can be changed or mooted by what happens as the case moves toward judgment, and a court of appeals that decided it early would meet the same facts and law again on appeal from the final judgment.⁵⁴ A truly collateral issue produces no such duplication, because the appellate court decides it once and nothing later in the case gives the trial court reason to look at it again.⁵⁵ Coopers & Lybrand v. Livesay described the orders that fail the test as those involving considerations “enmeshed in the factual and legal issues comprising the plaintiff’s cause of action.”⁵⁶

The order in Cohen sits on the collateral side of that line. The question there was whether a state statute requiring a small shareholder to post security for litigation expenses applied in a federal derivative suit, and the Court treated that question as one that neither affected nor would be affected by the decision of the merits.⁵⁷ Answering it called for reading the statute and deciding how it fit in a federal court. It required no finding on whether the managers had enriched themselves at the corporation’s expense as the complaint charged. The same feature marks the other standard examples. A defense of immunity from suit, whether qualified, absolute, state, or foreign sovereign, is collateral, and so are orders about security or attachment; in Swift & Co. Packers v. Compania Colombiana del Caribe, the Court called a question about attaching a ship “fairly severable from the context of the larger litigious process.”⁵⁸ A denial of a Rule 12(b)(6) motion or a summary-judgment motion is typically not collateral, because those motions ask whether the claim itself can succeed.⁵⁹ A question can be purely legal and still be the merits, so legal character alone does not establish separateness.

The immunity cases show how much depends on the question the appeal actually presents.

The line turns on the question the appeal presents, and the word “immunity” on the motion decides nothing. A legal question about clearly established law stays collateral even when the record contains disputed facts, whereas an appeal that only contests the district court’s view of what happened asks the court of appeals to decide part of the merits.⁶⁰

Class actions put two decisions on opposite sides of the line. In Eisen v. Carlisle & Jacquelin, the district court assigned 90% of the cost of class notice to the defendants, and the Supreme Court held the order appealable because it “involved a collateral matter unrelated to the merits of petitioner’s claims.”⁶¹ Who pays to notify the class can be settled without deciding whether any defendant did what the complaint alleges. Orders granting or refusing class certification fall on the other side, because they are inextricably intertwined with the merits and typically remain open to revision until judgment, and Coopers & Lybrand accordingly refused to treat a class-certification ruling as a collateral order.⁶²

Importance is the second half of the condition, and Cohen supplies its wording too. The Court described the qualifying decisions as those “too important to be denied review,” and it limited its holding with a sentence that shows what it meant: not every order fixing security is appealable, because “it is the right to security that presents a serious and unsettled question.”⁶³ The reason for requiring importance follows from the doctrine’s place in § 1291. If importance were measured by the loser’s stake, every adverse ruling would qualify, since every loser regards the ruling as important, and the doctrine would consume the single-appeal rule. Measuring it by the significance of the question the ruling decides filters out routine rulings, so the early appeal serves the few questions that matter beyond the dispute in which they arise.

Nixon v. Fitzgerald applied that standard to a question that cleared it. The court of appeals had dismissed the appeal from an order denying a claim of absolute presidential immunity, apparently because its own earlier decision in Halperin v. Kissinger had rejected the claim, and the Supreme Court had affirmed that decision only by an equally divided vote.⁶⁴ The Supreme Court held that the question remained serious and unsettled, because it had never decided the issue itself and the claim invoked the separation of powers.⁶⁵ It treated “serious and unsettled” as an additional requirement that Cohen established, found it satisfied, and let the appeal go forward.⁶⁶

The other end of the range is the order Cohen set outside its own holding. Where a right is admitted or clear and the ruling only exercises discretion over the amount of security, a matter the statute leaves open to reconsideration from time to time, “appealability would present a different question.”⁶⁷ Such a ruling raises no question that matters beyond the case, and the trial court may revisit it, so it falls short of both importance and the conclusiveness the first condition demands. The unsettled half of the inquiry is the easier half to test: is the issue one of first impression, or one that has sharply divided the courts?⁶⁸ A routine ruling that applies settled law to the facts of one case does not fit that description, however much it matters to the party who lost it.

The circuits divide on whether the “serious and unsettled” language states a separate requirement. The First, Second, and Seventh Circuits have applied it as a fourth factor, while the weight of authority, including the Fourth, Fifth, Sixth, and D.C. Circuits, holds that the test has three conditions and that the seriousness of the question bears only on whether the issue is important.⁶⁹ An answer on this point should state the three conditions as the test, point to Cohen’s language, and adjust to the circuit if the problem names one. Where a fourth factor applies, the appellant must show an issue of first impression or a question that has divided the courts. Where only three conditions apply, the same showing still strengthens the argument that the issue is important.

Effectively Unreviewable

The third condition, that the order be effectively unreviewable on appeal from a final judgment, is the key requirement and drives the doctrine.⁷² In Lauro Lines S.R.L. v. Chasser, the Supreme Court left open whether an order denying dismissal on a forum-selection clause was conclusive or separate from the merits, because the order failed this requirement and the appeal could not go forward.⁷³

The idea is usually called now or never. Some rulings carry consequences that arrive at once, so review in the normal course, on appeal after final judgment, comes too late to prevent or undo the harm, and refusing early review then amounts to refusing any meaningful review.⁷⁴ The ordinary wait is tolerable because a reversal after final judgment can usually repair an error: the appellate court can set aside the judgment, order a new trial, or restore money wrongly paid. The condition asks whether that repair will still be available. When it will, the final judgment rule applies as usual and the order stays interlocutory. When it will not, the order is in effect the last word on a right that cannot be recovered later, and the doctrine treats it as a final decision under § 1291.

The cost of waiting cannot satisfy the condition, and the reason lies in how many orders that cost would reach. Every erroneous pretrial ruling puts the loser to trouble and expense that a correct ruling would have spared. If that burden made an order unreviewable, every adverse order would qualify and the final judgment rule would dissolve. The Supreme Court declined to accept that result in Lauro Lines, where the district court had denied motions to dismiss on the strength of a forum-selection clause in passenger tickets, the court of appeals had dismissed the appeal, and the Court affirmed.⁷⁵ The Court conceded that trial in the wrong forum would put Lauro Lines to unnecessary trouble and expense and diminish the value of its contractual right. But there is always value in triumphing before trial rather than after it, and the Court has refused to treat the costs of unnecessary litigation as enough to justify an immediate appeal.⁷⁶ The right must be one that would be essentially destroyed if its vindication waited until trial ended.⁷⁷

A right not to be tried at all is the clearest example of a right that trial itself destroys. The Court has held in criminal prosecutions that deprivation of a right not to be tried is effectively unreviewable after final judgment, and it has carried the same reasoning into civil cases involving immunity from suit.⁷⁸ In Abney v. United States, the defendant claimed that the Double Jeopardy Clause barred his prosecution. Because the clause protects against a second prosecution, not just a second conviction, making him stand trial before appealing would have undermined the right completely.⁷⁹ In Nixon v. Fitzgerald, the President was allowed to appeal an order denying his motion to dismiss on absolute immunity grounds, because absolute immunity spares a defendant not only from ultimate liability but from having to stand trial at all, and a reversal after final judgment could not undo the trial he had already endured.⁸⁰ The Court has explained that “the essence of absolute immunity is its possessor’s entitlement not to have to answer for his conduct in a civil damages action,” and qualified immunity receives the same treatment because it too is an immunity from suit.⁸¹

A defense that says the suit is in the wrong place, or was begun in the wrong way, does not have that character. In Van Cauwenberghe v. Biard, a civil defendant claimed immunity from service of process because extradition had compelled his presence in the United States, and the Court held that this was not an immunity from suit, even though service was essential to the trial court’s jurisdiction over him.⁸² The Court framed the critical question as whether the essence of the claimed right is a right not to stand trial.⁸³ Lauro Lines argued that its clause gave it a right not to be haled for trial before tribunals outside the agreed forum, and that appeal after trial in the wrong place could not vindicate that right. The Court has declined to apply the doctrine where the claim is that the suit is not properly before a particular court, as opposed to a claim that the defendant may not be sued at all, and it affirmed the dismissal of the appeal.⁸⁴

Even a right not to be tried does not qualify automatically, because nearly every defendant would prefer to avoid trial. The Court has said that avoiding trial counts only when the trial would imperil a substantial public interest, so a party who claims a right not to be tried must identify “some particular value of high order” in avoiding it.⁸⁵ The values the Court has named are honoring the separation of powers, preserving the efficiency of government and the initiative of its officials, respecting a State’s dignitary interests, and mitigating the government’s advantage over the individual.⁸⁶ The immunity and double jeopardy cases each protect a value of that kind. A forum-selection clause protects a bargain between private parties and implicates no comparable public value.

The modern gloss on the third condition looks at categories of orders and not only at the case in front of the court. In Mohawk Industries, Inc. v. Carpenter, the Court explained that the crucial question is not whether an interest is important in the abstract, but whether deferring review until final judgment so imperils the interest as to justify the cost of allowing immediate appeal of the entire class of relevant orders.⁸⁷ The category approach follows from where the cost of an early appeal falls. Once one kind of order is appealable at once, every litigant who receives that kind of order may file, so the delay, the interruption of the trial court, and the added appellate work multiply across all of them. A single appellant’s unusual hardship therefore does not decide the matter.

Mohawk itself involved an order that a party had waived the attorney-client privilege and must produce documents. The Court recognized the importance of the privilege and still held that disclosure orders adverse to it are not immediately appealable, because postjudgment appeals generally protect the privilege adequately: an appellate court can remedy improper disclosure as it remedies evidentiary errors, by vacating the judgment and ordering a new trial.⁸⁸ The Court added that “[p]ermitting parties to undertake successive, piecemeal appeals of all adverse attorney-client rulings would unduly delay the resolution of district court litigation and needlessly burden the Courts of Appeals.”⁸⁹ The cost-sharing order in In re Recticel Foam Corp. shows the same method on a smaller scale. The court of appeals reasoned there that “[n]othing will be lost: money wrongfully paid by one party to another can be restored by judicial decree.”⁹⁰ A particular co-defendant might lack the funds to repay at the end of the case. But the analysis looks to “the entire category to which a claim belongs,” and in most cases a party can repay if a cost-sharing order is overturned, so no interest in that category justifies an immediate appeal from every such order.⁹¹

Whether a right can be vindicated later also depends on what other routes remain open, and the Court counts them. In Mohawk the Court noted that a party wishing to challenge an adverse privilege ruling still has “standard postjudgment appeals, § 1292(b) appeals, mandamus, and contempt appeals.”⁹² Contempt is the most striking of these. A party ordered to turn over documents can refuse, be held in contempt, and appeal the contempt judgment. The path is risky and harsh, but the Supreme Court has generally treated its availability as showing that a discovery order needs no immediate review as a collateral order.⁹³ Firestone Tire & Rubber Co. v. Risjord applied the same logic to an order denying a motion to disqualify counsel, which the Court held not appealable under Cohen, because whether counsel’s continued representation was prejudicial error is best judged after the litigation is over.⁹⁴ Lauro Lines had not even asked the district court to certify its order under § 1292(b).⁹⁵ Every adequate substitute reduces what waiting costs, and the case for an immediate appeal as of right weakens with it. Mandamus and contempt are taught later in this chapter.

An exam stem that stresses how much the ruling matters to the loser invites the wrong answer. A privilege disclosure order, a cost allocation, and a denied motion to dismiss on a forum clause can each hurt a great deal, and each fails the condition because a reversal after final judgment can repair the error. The stem that earns the appeal describes a claimed immunity from suit or a bar to prosecution, a right that the trial itself would destroy.

Orders That Qualify

A small class of interlocutory orders may be appealed immediately under Cohen: the orders that conclusively determine the disputed question, resolve an important issue completely separate from the merits of the action, and are effectively unreviewable on appeal from a final judgment.⁹⁷ Three conditions become usable only when you know which orders have cleared them, and the orders that have share one feature: each protects a right that a trial, a stretch of custody, or an early round of litigation would use up before any appeal from final judgment could help. The list below collects the categories the cases have accepted.

• Orders denying absolute immunity or qualified immunity, to the extent the denial turns on a question of law.⁹⁸
• Orders denying state sovereign immunity, immunity under the Foreign Sovereign Immunities Act, or Eleventh Amendment immunity from suit in federal court.⁹⁹
• Orders rejecting a criminal defendant’s double jeopardy defense.¹⁰⁰
• Orders denying bail or denying in forma pauperis status.¹⁰¹
• Orders allocating most of the cost of class notice to the defendants, as in Eisen v. Carlisle & Jacquelin.¹⁰²
• Orders denying an attorney’s motion to withdraw, in the decisions of some courts of appeals.¹⁰³
• Orders vacating a maritime arrest or attachment, in the decisions of some courts of appeals.¹⁰⁴

Immunity from suit explains the first two entries and sets the pattern for the rest. A defense to liability is fully vindicated by a reversal after final judgment, because the court of appeals can set the judgment aside. Qualified immunity, which the Court placed in the class in Mitchell v. Forsyth, is a defense to liability and something more: it is also a limited “entitlement not to stand trial or face the other burdens of litigation.”¹⁰⁵ That second feature controls appealability, because discovery and trial are the burdens the immunity exists to prevent, and both are finished before an appeal from final judgment is heard. Absolute immunity rests on the same footing, since it entitles its holder not to have to answer for conduct in a civil damages action.¹⁰⁶ A right of that kind is destroyed by the delay the final judgment rule imposes, so waiting for final judgment eliminates the protection instead of postponing review of it.

Nixon v. Fitzgerald shows the doctrine at work on absolute immunity. After discovery, the district court denied summary judgment, held that the action must proceed to trial, and ruled that Nixon could not claim absolute presidential immunity.¹⁰⁷ Nixon took a collateral appeal to the D.C. Circuit, which dismissed it summarily, apparently because its decision in Halperin v. Kissinger had already rejected the immunity claim.¹⁰⁸ The Supreme Court held that the appeal belonged in the court of appeals, relying on earlier holdings that orders rejecting claims under the Speech or Debate Clause and the Double Jeopardy Clause are appealable under the Cohen criteria.¹⁰⁹ The remaining question was Cohen’s added requirement of a serious and unsettled question, which the Court found met because it had never ruled on presidential immunity.¹¹⁰ The dismissal was therefore wrong. A court of appeals’ own precedent does not settle a question that remains open in the Supreme Court.

Ashcroft v. Iqbal applies the same reasoning to qualified immunity at the pleading stage. Government officials moved to dismiss on qualified immunity grounds, the district court denied the motion, and the court of appeals affirmed. The Supreme Court took up the court of appeals’ jurisdiction first, because subject-matter jurisdiction cannot be forfeited or waived and must be considered when fairly in doubt.¹¹¹ A denial of qualified immunity qualifies so long as it turns on an issue of law. The Court described such an order in terms that track Cohen’s three conditions in order: it conclusively determines that the defendant must bear the burdens of discovery, it is “conceptually distinct from the merits of the plaintiff’s claim,” and it would prove “effectively unreviewable on appeal from a final judgment.”¹¹² The Court was candid about the doctrine’s limits. It observed that the doctrine “may have expanded beyond the limits dictated by its internal logic,” yet called its application to qualified immunity “well established,” and it confirmed that an order rejecting qualified immunity at the motion-to-dismiss stage is a “final decision” under § 1291.¹¹³

The respondent’s jurisdictional argument produced the holding specific to Iqbal. He conceded that the court of appeals could decide whether the complaint alleged a violation of clearly established law but denied that it could decide whether the complaint was sufficiently pleaded, so that an appeal resting on failure to state a claim would fall outside the doctrine.¹¹⁴ The Court rejected that narrowing, because appellate jurisdiction in these appeals is “not so strictly confined,” and it pointed to earlier decisions reviewing interlocutory denials of qualified immunity on the elements of a claim.¹¹⁵ The reason is structural. Whether a complaint sufficiently alleges a clearly established violation cannot be decided in isolation from the facts pleaded, so the sufficiency of the pleading is “inextricably intertwined with” and “directly implicated by” the qualified immunity defense.¹¹⁶ An official who loses a motion to dismiss on qualified immunity may therefore appeal at once, and the court of appeals may ask whether the facts pleaded establish a violation of clearly established law. What the complaint must contain is a pleading question governed by Rule 8 and Rule 12(b)(6), and Iqbal decides only that the court of appeals has power to ask it on this appeal.

A limit appears when the denial rests on the record and not on the complaint. In Johnson v. Jones, five police officers were accused of beating the plaintiff, and the summary judgment order determined only that a genuine issue of material fact existed about whether three of them took part.¹¹⁷ The Court held that such an order is not a final decision, because only decisions turning on an issue of law are subject to immediate appeal, and deciding whether the evidence creates a genuine dispute is a “fact-related” inquiry.¹¹⁸ The concerns are the ones that support the final judgment rule itself. To answer the question, the court of appeals might have to consult a “vast pretrial record, with numerous conflicting affidavits, depositions, and other discovery materials,” a task that falls more within the district court’s ken and that may inefficiently duplicate questions arising on appeal after final judgment.¹¹⁹ Iqbal fell on the abstract side of that line, because the court of appeals read only the four corners of the complaint, and judging whether a complaint has the “heft” to state a claim is a task well within an appellate court’s core competency.¹²⁰ The two cases differ in what the court of appeals would have to read, a pleading in one and a contested record in the other.

The line does not depend on whether the facts are disputed. Courts have held that legal issues such as qualified immunity may remain appealable even when the underlying facts are disputed and the denial rested on the presence of material factual disputes.¹²¹ What controls is the question the appellant asks the court of appeals to answer.

The first argument is abstract in the sense Johnson used the word, and the second is fact-based. An exam stem that reports a denial of summary judgment in a qualified immunity case therefore calls for a further question about what the official is asking the court of appeals to decide.

The other immunity entries follow the same pattern. A defense of immunity from suit, whether qualified, absolute, state, or foreign sovereign, is the textbook collateral issue, because it protects the defendant from the burdens of suit itself as well as from an adverse judgment.¹²² The criminal-side entries have the same structure. An order rejecting a double jeopardy defense qualifies, because the defense protects a defendant against facing a second trial, and a reversal after that trial comes too late to prevent it.¹²³ Orders denying bail were likewise held effectively unreviewable on appeal from a final judgment, because the question is whether the defendant is held in custody before trial, and that time cannot be returned once the trial has ended.¹²⁴

Two further categories rest on lower-court decisions, and an answer should say so. Some courts of appeals treat an order denying an attorney’s motion to withdraw as a collateral order, and the Ninth Circuit has treated an order vacating a maritime arrest or attachment the same way.¹²⁵ The arrest ruling fits the second condition, since a question about attaching a ship is “fairly severable from the context of the larger litigious process.”¹²⁶

The notice-cost ruling in Eisen stands apart from the others because it involves no immunity and no custody. The district court imposed 90% of the notice costs on the defendants, and the Supreme Court held that the order fell within “that small class” described in Cohen, because it “conclusively rejected respondents’ contention that they could not lawfully be required to bear the expense of notice” and “involved a collateral matter unrelated to the merits of petitioner’s claims.”¹²⁷ The Court reached that result by applying the two grounds on which Cohen was decided: the ruling was not tentative, informal, or incomplete, and it concerned a collateral matter that could not be reviewed effectively on appeal from the final judgment.¹²⁸ Who bears the expense of telling the class about the suit can be decided without deciding whether any defendant did what the complaint alleges, which is the separateness the second condition demands. The same reasoning does not reach the ruling on whether a class exists at all, because class-certification orders are enmeshed in the merits.¹²⁹ Eisen is therefore authority for class-notice allocation specifically. The cost-sharing order in Recticel, discussed above, failed the conclusiveness and unreviewability conditions, so no general rule makes every cost order appealable.

Orders That Do Not Qualify

The collateral order doctrine reaches an interlocutory order only if the order meets all three conditions, and the Supreme Court has described the class that meets them as small.¹³¹ The clearest proof is how often the Court has refused to enlarge it. Those refusals follow one pattern: in nearly every one, the party who lost the pretrial ruling can still obtain full review on appeal from the final judgment, so the order fails the third condition even when it is conclusive and even when it matters a great deal to the loser.¹³²

The third condition does most of the refusing because a reversal after final judgment can repair most pretrial errors. The court of appeals can set aside the judgment, order a new trial, or restore money wrongly paid, and the party who lost the ruling then stands where a correct ruling would have left it, less the expense of the first trial.¹³³ That expense cannot supply the missing condition, because every erroneous pretrial order costs the loser something, and the Court has declined to treat the costs of unnecessary litigation as enough to justify an immediate appeal.¹³⁴ The inquiry also looks at the whole category to which a ruling belongs. A single party’s unusual hardship, such as a co-defendant who may lack the funds to be repaid after a cost allocation is reversed, does not move an entire class of orders into the doctrine.¹³⁵

The refusals sort into five groups, and the list below names the orders in each so that you can match an adverse ruling to its group:

• Privilege, discovery, and case management: orders compelling disclosure of material claimed to be privileged, orders allowing or excluding evidence, discovery and cost-sharing orders, and sanctions orders.¹³⁶
• Forum and venue: orders denying dismissal based on a contractual forum-selection clause or on forum non conveniens.¹³⁷
• Counsel: orders granting or denying a motion to disqualify counsel because of a conflict.¹³⁸
• Class actions: orders granting or denying class certification.¹³⁹
• Ordinary dispositive motions: orders denying a motion to dismiss, whether for failure to state a claim or for lack of personal jurisdiction, and orders denying summary judgment.¹⁴⁰

Privilege and discovery orders show the pattern most plainly. In Mohawk Industries, Inc. v. Carpenter, the district court ruled that a party had waived the attorney-client privilege and must hand over documents, and the Supreme Court held that the ruling could not be appealed at once. A postjudgment appeal generally protects the privilege adequately, because the court of appeals can undo an improper disclosure as it undoes an evidentiary error, by vacating the judgment and ordering a new trial.¹⁴¹ Other discovery orders fare the same way. An order to turn over documents or answer a discovery demand is often final and collateral in the ordinary sense, yet it typically cannot be appealed under the doctrine, because the party can disobey it, be held in contempt, and appeal the contempt judgment.¹⁴² That path is risky and harsh, but the Court has generally treated its existence as showing that the discovery order needs no immediate review. The contempt route is taught later in this chapter.¹⁴³

Case-management orders in large litigation add a second reason for refusal. In In re Recticel Foam Corp., the court of appeals found that an order requiring the defendants to share the cost of reproducing videotapes remained subject to modification under the discovery rules and that no irreversible legal consequences flowed from it as it stood.¹⁴⁴ Money wrongly paid by one party to another can be restored by judicial decree, so waiting costs the payer nothing that a later judgment cannot return.¹⁴⁵ The Supreme Court has also placed sanctions orders, as in Cunningham v. Hamilton County, and orders rescinding settlement agreements, as in Digital Equipment Corp. v. Desktop Direct, Inc., outside the doctrine.¹⁴⁶

Orders about the place of trial fall into the same pattern. In Lauro Lines S.R.L. v. Chasser, the district court denied motions to dismiss that relied on a forum-selection clause in passenger tickets, and the Supreme Court affirmed the dismissal of the appeal, because the order failed the third condition and the Court had no need to decide whether it was conclusive or separate from the merits.¹⁴⁷ Trial in the wrong forum would put the defendant to unnecessary trouble and expense and would diminish the value of its contractual right. But the Court insisted that the asserted right be one essentially destroyed if its vindication waits until trial ends, and added expense does not meet that standard.¹⁴⁸ Van Cauwenberghe v. Biard reached the same result for a defendant who claimed immunity from civil process because extradition had brought him into the country, and for an order denying dismissal on forum non conveniens grounds.¹⁴⁹

The contrast with the immunity cases explains why forum rulings fail. A defendant who says that this court may not hear the suit claims a right a reversal can restore, because the court of appeals can vacate a judgment entered in the wrong forum.¹⁵⁰ A defendant who says that no court may make it answer a civil damages action at all, as the President did in Nixon v. Fitzgerald, asserts a right that trial itself consumes, and the denial of absolute immunity is therefore appealable at once.¹⁵¹

Disqualification of counsel presents a related question about timing. In Firestone Tire & Rubber Co. v. Risjord, the Court held that an order denying a motion to disqualify counsel could not be appealed under the collateral order doctrine, because there was no sufficient showing that the ruling would be effectively unreviewable after judgment on the merits.¹⁵² The reason is practical. The effect of counsel’s continued participation is hard to assess until the litigation is over, and at that point the appellate court can decide whether letting the representation continue was prejudicial error, so no immediate review is needed.¹⁵³ The refusal runs in both directions, since orders granting a disqualification motion fail as well as orders denying one.¹⁵⁴ Some courts of appeals do accept an order denying an attorney’s own motion to withdraw as a collateral order, so the presence of a lawyer in the dispute does not settle the question, and each order is tested against the conditions.¹⁵⁵

Class certification rulings fail all three conditions at once. In Coopers & Lybrand v. Livesay, the Supreme Court rejected the collateral order argument because certification is subject to revision by the trial court, the class issues are intertwined with the merits, and the ruling can be reviewed on appeal from a final judgment.¹⁵⁶ An order certifying a class fares no better, because it too is inextricably intertwined with the merits and typically open to revision until judgment.¹⁵⁷ The label class action therefore decides nothing by itself, because the order in Eisen v. Carlisle & Jacquelin placing 90% of the cost of class notice on the defendants was held appealable as a final determination of a subject separate from the merits.¹⁵⁸ A plaintiff whose individual claims are too small to pursue alone may argue that a denial of certification ends the case. But the Court criticized that death knell theory as resting on policy considerations better left to Congress and as letting plaintiffs, but not defendants, appeal early.¹⁵⁹ The death knell theory and the rule-based route that followed it are taken up in the next sections.

Denials of ordinary dispositive motions are the largest group and the easiest to place. A denial of a Rule 12(b)(6) motion or a summary-judgment motion is typically not collateral, because those motions ask whether the claim itself can succeed, and that question is the merits.¹⁶⁰ Such a denial may be conclusive and may matter greatly to the defendant, yet it is classic error that can be raised on appeal from the final judgment, and nothing is lost by waiting except the ordinary costs of litigation.

The comparison to draw is with a governmental entity that moves to dismiss on Eleventh Amendment immunity. The order refusing to dismiss such an action qualifies as a collateral order, because the defense claims freedom from suit in federal court itself.¹⁶¹ Government status alone does not carry the appeal, though. In Will v. Hallock, government agents tried to appeal a ruling that the judgment bar of the Federal Tort Claims Act did not defeat the suit against them, and the Court refused. It reasoned that the judgment bar rests on avoiding litigation for its own sake, and that treating that interest as enough would make collateral order appeal a matter of right whenever the government lost a motion to dismiss under the Act.¹⁶² What earns the appeal is the kind of right the defendant asserts, whoever the defendant happens to be.

When an exam stem describes an adverse pretrial order, the first step is to place it in one of these five groups. An order that fits a group goes back to the final judgment rule or to another route in this chapter, and the Court counts those routes as real alternatives, since a party facing an adverse privilege ruling still has postjudgment appeal, certification under § 1292(b), mandamus, and contempt.¹⁶³ An order that fits no group must still pass all three conditions, and the third is where these orders fail. Expense, the abstract importance of the right, and the conclusiveness of the ruling cannot make up for a right that a reversal after final judgment would restore.