Digital Equipment Corp. v. Desktop Direct, Inc.

delivered the opinion of the Court.

Section 1291 of the Judicial Code confines appeals as of right to those from “final decisions of the district courts.” 28 U. S. C. § 1291. This case raises the question whether an order vacating a dismissal predicated on the parties’ settlement agreement is final as a collateral order even without a district court’s resolution of the underlying cause of action. See Cohen v. Beneficial Industrial Loan Corp., 337 U. S. 541, 546 (1949). We hold that an order denying effect to a settlement agreement does not come within the narrow ambit of collateral orders.

I

Respondent, Desktop Direct, Inc. (Desktop), sells computers and like equipment under the trade name “Desktop Direct.” Petitioner, Digital Equipment Corporation, is engaged in a similar business and in late 1991 began using that trade name to market a new service it called “Desktop Direct from Digital.” In response, Desktop filed this action in the *866 United States District Court for the District of Utah, charging Digital with unlawful use of the Desktop Direct name. Desktop sent Digital a copy of the complaint, and negotiations between officers of the two corporations ensued. Under a confidential settlement reached on March 25, 1992, Digital agreed to pay Desktop a sum of money for the right to use the “Desktop Direct” trade name and corresponding trademark, and for waiver of all damages and dismissal of the suit. That same day, Desktop filed a notice of dismissal in the District Court.

Several months later, Desktop moved to vacate the dismissal and rescind the settlement agreement, alleging misrepresentation of material facts during settlement negotiations. The District Court granted the motion, concluding “that a fact finder could determine that [Digital] failed to disclose material facts to [Desktop] during settlement negotiations which would have resulted in rejection of the settlement offer.” App. to Pet. for Cert. 13a. After the District Court declined to reconsider that ruling or stay its order vacating dismissal, Digital appealed.

The Court of Appeals for the Tenth Circuit dismissed the appeal for lack of jurisdiction, holding that the District Court order was not appealable under §1291, because it neither “end[ed] the litigation on the merits” nor “[fell] within the long-recognized ‘collateral order’ exception to the final judgment requirement.” 993 F. 2d 755, 757 (1993). Applying the three-pronged test for determining when “collateral order” appeal is allowed, see Cohen, supra; Coopers & Lybrand v. Livesay, 437 U. S. 463 (1978), the Court of Appeals concluded that any benefits claimed under the settlement agreement were insufficiently “important” to warrant the immediate appeal as of right. Although Digital claimed what it styled a “right not to go to trial,” the court reasoned that any such privately negotiated right as Digital sought to vindicate was different in kind from an immunity rooted in an explicit constitutional or statutory provision or “compel *867 ling public policy rationale,” the denial of which has been held to be immediately appealable. 993 F. 2d, at 758-760. 1

The Tenth Circuit recognized that it was thus deviating from the rule followed in some other Courts of Appeals, see Forbus v. Sears, Roebuck & Co., 958 F. 2d 1036 (CA11 1992); Grillet v. Sears, Roebuck & Co., 927 F. 2d 217 (CA5 1991); Janneh v. GAF Corp., 887 F. 2d 432 (CA2 1989); but see Transtech Industries, Inc. v. A & Z Septic Clean, 5 F. 3d 51 (CA3 1993), cert. pending, No. 93-960. We granted certiorari, 510 U. S. 942 (1993), to resolve this conflict and now affirm.

II

A

The collateral order doctrine is best understood not as an exception to the “final decision” rule laid down by Congress in § 1291, but as a “practical construction” of it, Cohen, supra, at 546; see, e. g., Coopers & Lybrand, supra, at 468. We have repeatedly held that the statute entitles a party to appeal not only from a district court decision that “ends the litigation on the merits and leaves nothing more for the court to do but execute the judgment,” Catlin v. United States, 324 U. S. 229, 233 (1945), but also from a narrow class of decisions that do not terminate the litigation, but must, in the interest of “achieving a healthy legal system,” cf. Cobbledick v. United States, 309 U. S. 323, 326 (1940), nonetheless be treated as “final.” The latter category comprises only those district court decisions that are conclusive, that resolve important questions completely separate from the merits, and that would render such important questions effectively unreviewable on appeal from final judgment in the underlying action. See generally Coopers & Lybrand, supra. Immediate appeals from such orders, we have ex *868 plained, do not go against the grain of § 1291, with its object of efficient administration of justice in the federal courts, see generally Richardson-Merrell Inc. v. Roller, 472 U. S. 424 (1985).

But we have also repeatedly stressed that the “narrow” exception should stay that way and never be allowed to swallow the general rule, id., at 436, that a party is entitled to a single appeal, to be deferred until final judgment has been entered, in which claims of district court error at any stage of the litigation may be ventilated, see United States v. Hollywood Motor Car Co., 458 U. S. 263, 270 (1982). We have accordingly described the conditions for collateral order appeal as stringent, see, e. g., Midland Asphalt Corp. v. United States, 489 U. S. 794, 799 (1989), and have warned that the issue of appealability under §1291 is to be determined for the entire category to which a claim belongs, without regard to the chance that the litigation at hand might be speeded, or a “particular injustice]” averted, Van Cauwenberghe v. Biard, 486 U. S. 517, 529 (1988), by a prompt appellate court decision. See also Richardson-Merrell, supra, at 439 (this Court “has expressly rejected efforts to reduce the finality requirement of § 1291 to a case-by-case [appealability] determination”); Carroll v. United States, 354 U. S. 394, 405 (1957).

B

Here, the Court of Appeals accepted Digital’s claim that the order vacating dismissal (and so rescinding the settlement agreement) was the “final word on the subject addressed,” 993 F. 2d, at 757 (citation omitted), and held the second Cohen condition, separability, to be satisfied, as well. Neither conclusion is beyond question, 2 but each is best left *869 untouched here, both because Desktop has made no serious effort to defend the Court of Appeals’ judgment on those points and because the failure to meet the third condition of the Cohen test, that the decision on an “important” question be “effectively unreviewable” upon final judgment, would in itself suffice to foreclose immediate appeal under §1291. 3 Turning to these dispositive factors, we conclude, despite Digital’s position that it holds a “right not to stand trial” requiring protection by way of immediate appeal, that rights under private settlement agreements can be adequately vindicated on appeal from final judgment.

C

The roots of Digital’s argument that the settlement with Desktop gave it a “right not to stand trial altogether” (and that such a right per se satisfies the third Cohen requirement) are readily traced to Abney v. United States, 431 U. S. 651 (1977), where we held that § 1291 entitles a criminal defendant to appeal an adverse ruling on a double jeopardy claim, without waiting for the conclusion of his trial. After holding the second Cohen requirement satisfied by the distinction between the former jeopardy claim and the question of guilt to be resolved at trial, we emphasized that the Fifth Amendment not only secures the right to be free from multi *870 pie punishments, but by its very terms embodies the broader principle, “ ‘deeply ingrained in ... the Anglo-American system of jurisprudence,’ ” that it is intolerable for “ ‘the State, with all its resources ... to make repeated attempts to convict an individual [defendant], thereby subjecting him to embarrassment, expense and ordeal and compelling him to live in a continuing state of anxiety and insecurity.’” 431 U. S., at 661-662 (quoting Green v. United States, 355 U. S. 184, 187-188 (1957)). We found that immediate appeal was the only way to give “full protection” to this constitutional right “not to face trial at all.” 431 U. S., at 662, and n. 7; see also Helstoski v. Meanor, 442 U. S. 500 (1979) (decision denying immunity under the Speech and Debate Clause would be appealable under § 1291).

Abney’s rationale was applied in Nixon v. Fitzgerald, 457 U. S. 731, 742 (1982), where we held to be similarly appeal-able an order denying the petitioner absolute immunity from suit for civil damages arising from actions taken while petitioner was President of the United States. Seeing this immunity as a “functionally mandated incident of the President’s unique office, rooted in the . . . separation of powers and supported by our history,” id., at 749, we stressed that it served “compelling public ends,” id., at 758, and would be irretrievably lost if the former President were not allowed an immediate appeal to vindicate this right to be free from the rigors of trial, see id., at 752, n. 32.

Next, in Mitchell v. Forsyth, 472 U. S. 511 (1985), we held that similar considerations supported appeal under §1291 from decisions denying government officials qualified immunity from damages suits. An “essential attribute,” id., at 525, of this freedom from suit for past conduct not violative of clearly established law, we explained, is the “entitlement not to stand trial or face the other burdens of litigation,” id., at 526, one which would be “effectively lost if a case [were] erroneously permitted to go to trial,” ibid. Echoing the reasoning of Nixon v. Fitzgerald, supra (and Harlow v. Fitz *871 gemid, 457 U. S. 800 (1982)), we explained that requiring an official with a colorable immunity claim to defend a suit for damages would be “peculiarly disruptive of effective government,” and would work the very “distraction . . . from . . . dut[y], inhibition of discretionary action, and deterrence of able people from public service” that qualified immunity was meant to avoid. See 472 U. S., at 526 (internal quotation marks omitted); see also Puerto Rico Aqueduct and Sewer Authority v. Metcalf & Eddy, Inc., 506 U. S. 139,147 (1993) (State’s Eleventh Amendment immunity from suit in federal court may be vindicated by immediate appeal under § 1291).

D

Digital puts this case on all fours with Mitchell. It maintains that it obtained dual rights under the settlement agreement with Desktop, not only a broad defense to liability but the “right not to stand trial,” the latter being just like the qualified immunity held immediately appealable in Mitchell. As in Mitchell, that right must be enforceable on collateral order appeal, Digital asserts, or an adverse trial ruling will destroy it forever.

While Digital’s argument may exert some pull on a narrow analysis, it does not hold up under the broad scrutiny to which all claims of immediate appealability under §1291 must be subjected. To be sure, Abney and Mitchell are fairly cited for the proposition that orders denying certain immunities are strong candidates for prompt appeal under § 1291. But Digital’s larger contention, that a party’s ability to characterize a district court’s decision as denying an irreparable “right not to stand trial” altogether is sufficient as well as necessary for a collateral order appeal, is neither an accurate distillation of our case law nor an appealing prospect for adding to it.

Even as they have recognized the need for immediate appeals under § 1291 to vindicate rights that would be “irretrievably lost,” Richardson-Merrell, 472 U. S., at 431, if re *872 view were confined to final judgments only, our cases have been at least as emphatic in recognizing that the jurisdiction of the courts of appeals should not, and cannot, depend on a party’s agility in so characterizing the right asserted. This must be so because the strong bias of § 1291 against piecemeal appeals almost never operates without some cost. A fully litigated case can no more be untried than the law’s proverbial bell can be unrung, and almost every pretrial or trial order might be called “effectively unreviewable” in the sense that relief from error can never extend to rewriting history. Thus, erroneous evidentiary rulings, grants or denials of attorney disqualification, see, e. g., RichardsonMerrell, supra, and restrictions on the rights of intervening parties, see Stringfellow v. Concerned Neighbors in Action, 480 U. S. 370 (1987), may burden litigants in ways that are only imperfectly reparable by appellate reversal of a final district court judgment, cf. Carroll, 354 U. S., at 406; Parr v. United States, 351 U. S. 513, 519-520 (1956); and other errors, real enough, will not seem serious enough to warrant reversal at all, when reviewed after a long trial on the merits, see Stringfellow, supra. In still other cases, see Coopers & Lybrand v. Livesay, 437 U. S. 463 (1978), an erroneous district court decision will, as a practical matter, sound the “death knell” for many plaintiffs’ claims that might have gone forward if prompt error correction had been an option. But if immediate appellate review were available every such time, Congress’s final decision rule would end up a pretty puny one, and so the mere identification of some interest that would be “irretrievably lost” has never sufficed to meet the third Cohen requirement. See generally Lauro Lines s.r.l. v. Chasser, 490 U. S. 495, 499 (1989) (“It is always true, however, that ‘there is value ... in triumphing before trial, rather than after it’ ”) (quoting United States v. MacDonald, 435 U. S. 850, 860, n. 7 (1978)); Richardson-Merrell, supra, at 436.

*873 Nor does limiting the focus to whether the interest asserted may be called a “right not to stand trial” offer much protection against the urge to push the § 1291 limits. We have, after all, acknowledged that virtually every right that could be enforced appropriately by pretrial dismissal might loosely be described as conferring a “right not to stand trial,” see, e. g., Midland Asphalt, 489 U. S., at 501; Van Cauwenberghe v. Biard, 486 U. S., at 524. Allowing immediate appeals to vindicate every such right would move § 1291 aside for claims that the district court lacks personal jurisdiction, see Van Cauwenberghe, supra, that the statute of limitations has run, see 15B C. Wright, A. Miller, & E. Cooper, Federal Practice and Procedure § 3918.5, and n. 65, p. 521 (1992), that the movant has been denied his Sixth Amendment right to a speedy trial, see MacDonald, supra, that an action is barred on claim preclusion principles, that no material fact is in dispute and the moving party is entitled to judgment as a matter of law, or merely that the complaint fails to state a claim. Such motions can be made in virtually every case, see generally id., at 862; United States v. Hollywood Motor Car Co., 458 U. S., at 270, and it would be no consolation that a party’s meritless summary judgment motion or res judicata claim was rejected on immediate appeal; the damage to the efficient and congressionally mandated allocation of judicial responsibility would be done, and any improper purpose the appellant might have had in saddling its opponent with cost and delay would be accomplished. Cf. Richardson-Merrell, supra, at 434 (appeals from “entirely proper” decisions impose the same costs as do appeals from “injudicious” ones). Thus, precisely because candor forces us to acknowledge that there is no single, “obviously correct way to characterize” an asserted right, Lauro Lines, supra, at 500, we have held that §1291 requires courts of appeals to view claims of a “right not to be tried” with skepticism, if not a jaundiced eye. Cf. Van Cauwenberghe, supra, at 524-525.

*874 In Midland Asphalt, for example, we had no trouble in dispatching a defendant’s claim of entitlement to an immediate appeal from an order denying dismissal for alleged violation of Federal Rule of Criminal Procedure 6(e), forbidding disclosure of secret grand jury information. Noting “ ‘a crucial distinction between a right not to be tried and a right whose remedy requires the dismissal of charges,’ ” 489 U. S., at 801, quoting Hollywood Motor Car, supra, at 269, we observed that Rule 6(e) “contains no hint,” 489 U. S., at 802, of an immunity from trial, and we contrasted that Rule with the Fifth Amendment’s express provision that “[n]o person shall be held to answer” for a serious crime absent grand jury indictment. Only such an “explicit statutory or constitutional guarantee that trial will not occur,” we suggested, id., at 801, could be grounds for an immediate appeal of right under § 1291. 4

The characterization issue surfaced again (and more ominously for Digital, see infra, at 880) in Lauro Lines, supra, where a defendant sought to appeal under §1291 from an order denying effect to a contractual provision that a Neapolitan court would be the forum for trying all disputes arising from the parties’ cruise-ship agreement. While we realized of course that the value of the forum-selection clause would be diminished if the defendant could be tried before appealing, we saw the contractual right to limit trial to an Italian forum as “different in kind” from the entitlement to “avoid *875 suit altogether” that Abney and Mitchell held could be “adequately vindicated]” only on immediate appeal. 490 U. S., at 501.

E

As Digital reads the cases, the only things standing in the way of an appeal to perfect its claimed rights under the settlement agreement are the lone statement in Midland Asphalt, to the effect that only explicit statutory and constitutional immunities may be appealed immediately under § 1291, and language (said to be stray) repeated in many of our collateral order decisions, suggesting that the “importance” of the right asserted is an independent condition of appeal-ability. See Brief for Petitioner 28-34. The first, Digital explains, cannot be reconciled with Mitchell’s holding, that denial of qualified immunity (which we would be hard pressed to call “explicitly ... guarantee^] ” by a particular constitutional or statutory provision) is a collateral order under § 1291; as between Mitchell and the Midland Asphalt dictum, Digital says, the dictum must give way.. As for the second obstacle, Digital adamantly maintains that “importance” has no place in a doctrine justified as supplying a gloss on Congress’s “final decision” language.

1

These arguments miss the mark. First, even if Mitchell could not be squared fully with the literal words of the Midland Asphalt sentence (but cf. Lauro Lines, 490 U. S., at 499, noting that Midland Asphalt was a criminal case and Mitchell was not), that would be only because the qualified immunity right is inexplicit, not because it lacks a good pedigree in public law. Indeed, the insight that explicitness may not be needed for jurisdiction consistent with § 1291 only leaves Digital with the unenviable task of explaining why other rights that might fairly be said to include an (implicit) “right to avoid trial” aspect are less in need of protection by immediate review, or more readily vindicated on appeal from final *876 judgment, than the (claimed) privately negotiated right to be free from suit. It is far from clear, for example, why § 1291 should bless a party who bargained for the right to avoid trial, but not a party who “purchased” the right by having once prevailed at trial and now pleads res judicata, see In re Corrugated Container Antitrust Litigation v. Willamette Industries, Inc., 694 F. 2d 1041 (CA5 1983); or a party who seeks shelter under the statute of limitations, see, e.g., United States v. Weiss, 7 F. 3d 1088 (CA2 1993), which is usually understood to secure the same sort of “repose” that Digital seeks to vindicate here, see Brief for Petitioner 25; or a party not even subject to a claim on which relief could be granted. See also Cobbledick, 309 U. S., at 325 (“Bearing the discomfiture and cost of a prosecution for crime even by an innocent person is one of the painful obligations of citizenship”); Firestone Tire & Rubber Co. v. Risjord, 449 U. S. 368, 378 (1981) (“[Potential harm” should be compared to “the harm resulting from other interlocutory orders that may be erroneous”) (internal quotation marks omitted).

Digital answers that the status under § 1291 of these other (seemingly analogous) rights should not give us pause, because the text and structure of this particular settlement with Desktop confer what no res judicata claimant could ever have, an express right not to stand trial. 5 But we cannot attach much significance one way or another to the supposed clarity of the agreement’s terms in this case. To ground a ruling here on whether this settlement agreement in terms confers the prized “right not to stand trial” (a point Desktop by no means concedes) would flout our own frequent admonitions, see, e. g., Van Cauwenberghe, 486 U. S., at 529, that availability of collateral order appeal must be determined at *877 a higher level of generality. Indeed, just because it would be the rare settlement agreement that could not be construed to include (at least an implicit) freedom-from-trial “aspect,” we decide this case on the assumption that if Digital prevailed here, any district court order denying effect to a settlement agreement could be appealed immediately. (And even if form were held to matter, settlement agreements would all include “immunity from suit” language a good deal plainer than what Digital relies on here, see Tr. of Oral Arg. 44.) See also Van Cauwenberghe, swpra, at 524 (“For purposes of determining appealability, ... we will assume, but do not decide, that petitioner has presented a substantial claim” on the merits). 6

2

The more fundamental response, however, to the claim that an agreement's provision for immunity from trial can *878 distinguish it from other arguable rights to be trial free is simply that such a right by agreement does not rise to the level of importance needed for recognition under §1291. This, indeed, is the bone of the fiercest contention in the case. In disparaging any distinction between an order denying a claim grounded on an explicit constitutional guarantee of immunity from trial and an order at odds with an equally explicit right by private agreement of the parties, Digital stresses that the relative “importance” of these rights, heavily relied upon by the Court of Appeals, is a rogue factor. No decision of this Court, Digital maintains, has held an order unappealable as “unimportant” when it has otherwise met the three Cohen requirements, and whether a decided issue is thought “important,” it says, should have no bearing on whether it is “final” under § 1291.

If “finality” were as narrow a concept as Digital maintains, however, the Court would have had little reason to go beyond the first factor in Cohen, see also United States v. 243.22 Acres of Land in Babylon, Suffolk Cty., 129 F. 2d 678, 680 (CA2 1942) (Frank, J.) (“ ‘Final’ is not a clear one-purpose word”). And if “importance” were truly aberrational, we would not find it featured so prominently in the Cohen opinion itself, which describes the “small class” of immediately appealable prejudgment decisions in terms of rights that are “too important to be denied review” right away, see 337 U. S., at 546. To be sure, Digital may validly question whether “importance” is a factor “beyond” the three Cohen conditions or whether it is best considered, as we have sometimes suggested it should be, in connection with the second, “separability,” requirement, see, e.g., Coopers & Lybrand, 437 U. S., at 468; Lauro Lines, 490 U. S., at 498, but neither enquiry could lead to the conclusion that “importance” is itself unimportant. To the contrary, the third Cohen question, whether a right is “adequately vindicable” or “effectively reviewable,” simply cannot be answered without a judgment about the value of the interests that would be lost *879 through rigorous application of a final judgment requirement. See generally Van Cauwenberghe, supra, at 524 0“[T]he substance of the rights entailed, rather than the advantage to a litigant in winning his claim sooner,’ ” is dis-positive) (quoting MacDonald, 435 U. S., at 860, n. 7); Lauro Lines, supra, at 502-503 (Scalia, J., concurring).

While there is no need to decide here that a privately conferred right could never supply the basis of a collateral order appeal, but cf. n. 7, infra (discussing 9 U. S. C. § 16), there are surely sound reasons for treating such rights differently from those originating in the Constitution or statutes. When a policy is embodied in a constitutional or statutory provision entitling a party to immunity from suit (a rare form of protection), there is little room for the judiciary to gainsay its “importance.” Including a provision in a private contract, by contrast, is barely a prima facie indication that the right secured is “important” to the benefited party (contracts being replete with boilerplate), let alone that its value exceeds that of other rights not embodied in agreements (e. g., the right to be free from a second suit based on a claim that has already been litigated), or that it qualifies as “important” in Cohen’s sense, as being weightier than the societal interests advanced by the ordinary operation of final judgment principles. Where statutory and constitutional rights are concerned, “irretrievably] los[s]” can hardly be trivial, and the collateral order doctrine might therefore be understood as reflecting the familiar principle of statutory construction that, when possible, courts should construe statutes (here § 1291) to foster harmony with other statutory and constitutional law, see, e. g., Ruckelshaus v. Monsanto Co., 467 U. S. 986, 1018 (1984); United States ex rel. Milwaukee Social Democratic Publishing Co. v. Burleson, 255 U. S. 407, 437-438 (1921) (Holmes, J., dissenting). But it is one thing to say that the policy of § 1291 to avoid piecemeal litigation should be reconciled with policies embodied in other statutes or the Constitution, and quite another to suggest that this *880 public policy may be trumped routinely by the expectations or clever drafting of private parties. 7

Indeed, we do not take issue with the Tenth Circuit’s observation that this case shares more in common with Lauro Lines than with Mitchell. It is hard to see how, for purposes of § 1291, the supposedly explicit “right not to be tried” element of the settlement agreement in this case differs from the unarguably explicit, privately negotiated “right not to be tried in any forum other than Naples, Italy,” in that one. There, no less than here (if Digital reads the settlement agreement correctly), one private party secured from another a promise not to bring suit for reasons that presumably included avoiding the burden, expense, and perhaps embarrassment of a certain class of trials (all but Neapolitan ones or, here, all prompted by Desktop). Cf. Lauro Lines, supra, at 501 (asserted right was “surely as effectively vindicable” on final judgment appeal as was the right in Van Cauwenberghe). 8 The losing argument in Lauro Lines should be a losing argument here.

*881 Nor are we swayed by Digital’s last-ditch effort to come within Cohen’s sense of “importance” by trying to show that settlement-agreement “immunities” merit first-class tre