515 U.S. 304 (1995)
JOHNSON et al.
United States Supreme Court.
Argued April 18, 1995.
Decided June 12, 1995.
CERTIORARI TO THE UNITED STATES COURT OF APPEALS FOR THE SEVENTH CIRCUIT
Breyer, J., delivered the opinion for a unanimous Court. Charles A. Rothfeld argued the cause for petitioners. With him on the briefs was Mark F. Smolens.
Cornelia T. L. Pillard argued the cause for the United States as amicus curiae urging reversal. With her on the brief were Solicitor General Days, Assistant Attorney Gen- eral Hunger, Deputy Solicitor General Bender, Barbara L. Herwig, and Richard A. Olderman.
Edward G. Proctor, Jr., argued the cause for respondent. With him on the brief was Anthony Pinelli.[*]
Justice Breyer, delivered the opinion of the Court.
This case concerns government officialsentitled to assert a qualified immunity defense in a "constitutional tort" actionwho seek an immediate appeal of a district court order denying their motions for summary judgment. The order in question resolved a fact -related dispute about the pretrial record, namely, whether or not the evidence in the pretrial record was sufficient to show a genuine issue of fact for trial. We hold that the defendants cannot immediately appeal this kind of fact-related district court determination. And, we affirm the similar holding of the Court of Appeals for the Seventh Circuit.
The plaintiff in this case, Houston Jones, is a diabetic. Police officers found him on the street while he was having an insulin seizure. The officers thought he was drunk, they arrested him, and they took him to the police station. Jones later found himself in a hospital, with several broken ribs. Subsequently, Jones brought this "constitutional tort" action against five named policemen. Rev. Stat. § 1979, as amended, 42 U. S. C. § 1983. Jones claimed that these policemen used excessive force when they arrested him and that they beat him at the station.
Three of the officers (the petitioners here) moved for summary judgment arguing that, whatever evidence Jones might have about the other two officers, he could point to no evidence that these three had beaten him or had been present while others did so. Jones responded by pointing to his deposition, in which he swore that officers (though he did not name them) had used excessive force when arresting him and, later, in the booking room at the station house. He also pointed to the three officers' own depositions, in which they admitted they were present at the arrest and in or near the booking room when Jones was there.
The District Court denied the officers' summary judgment motion. The court wrote that Seventh Circuit precedent indicated potential liability if the three officers "stood by and allowed others to beat the plaintiff." App. to Pet. for Cert. 7a. And, the court held that there was "sufficient circumstantial evidence supporting [Jones'] theory of the case." Id., at 8a.
The three officers immediately appealed the District Court's denial of their summary judgment motion. They argued, in relevant part, that the denial was wrong because the record contained "not a scintilla of evidence . . . that one or more" of them had "ever struck, punched or kicked the plaintiff, or ever observed anyone doing so." Brief for Appellants in No. 93-3777 (CA7), p. 10. But, the Seventh Circuit refused to consider this argumentnamely, that the District Court had improperly rejected their contention that the record lacked sufficient evidence even to raise a "genuine" (i. e., triable) issue of fact. The Seventh Circuit held that it "lack[ed] appellate jurisdiction over th[is] contention," i. e., of the "evidence insufficiency" contention that "we didn't do it." 26 F. 3d 727, 728 (1994). It consequently dismissed their appeal.
Courts of Appeals hold different views about the immediate appealability of such pretrial "evidence insufficiency" claims made by public official defendants who assert qualified immunity defenses. Compare, e. g., Kaminsky v. Rosenblum, 929 F. 2d 922, 926 (CA2 1991) (saying that no appellate jurisdiction exists); Giuffre v. Bissell, 31 F. 3d 1241, 1247 (CA3 1994) (same); Boulos v. Wilson, 834 F. 2d 504, 509 (CA5 1987) (same); Elliott v. Thomas, 937 F. 2d 338, 341-342 (CA7 1991) (same), cert. denied, 502 U. S. 1074, 1121 (1992); Crawford-El v. Britton, 951 F. 2d 1314, 1317 (CADC 1991) (same), with Unwin v. Campbell, 863 F. 2d 124, 128 (CA1 1988) (saying that appellate jurisdiction does exist); Turner v. Dammon, 848 F. 2d 440, 444 (CA4 1988) (same); Kelly v. Bender, 23 F. 3d 1328, 1330 (CA8 1994) (same); Burgess v. Pierce County, 918 F. 2d 104, 106, and n. 3 (CA9 1990) (per curiam) (same); Austin v. Hamilton, 945 F. 2d 1155, 1157, 1162-1163 (CA10 1991) (same). We therefore granted certiorari. 513 U. S. 1071 (1995).
Three background principles guide our effort to decide this issue. First, the relevant statute grants appellate courts jurisdiction to hear appeals only from "final decisions" of district courts. 28 U. S. C. § 1291. Given this statute, interlocutory appealsappeals before the end of district court proceedingsare the exception, not the rule. The statute recognizes that rules that permit too many interlocutory appeals can cause harm. An interlocutory appeal can make it more difficult for trial judges to do their basic jobsupervising trial proceedings. It can threaten those proceedings with delay, adding costs and diminishing coherence. It also risks additional, and unnecessary, appellate court work either when it presents appellate courts with less developed records or when it brings them appeals that, had the trial simply proceeded, would have turned out to be unnecessary. See Richardson-Merrell Inc. v. Koller, 472 U. S. 424, 430 (1985); Flanagan v. United States, 465 U. S. 259, 263-264 (1984); Firestone Tire & Rubber Co. v. Risjord, 449 U. S. 368, 374 (1981).
Of course, sometimes interlocutory appellate review has important countervailing benefits. In certain cases, it may avoid injustice by quickly correcting a trial court's error. It can simplify, or more appropriately direct, the future course of litigation. And, it can thereby reduce the burdens of future proceedings, perhaps freeing a party from those burdens entirely. Congress consequently has authorized, through other statutory provisions, immediate appeals (or has empowered courts to authorize immediate appeals) in certain classes of casesclasses in which these countervailing benefits may well predominate. None of these special "immediate appeal" statutes, however, is applicable here. See 28 U. S. C. § 1292 (immediate appeal of, e. g., orders granting or denying injunctions; authority to "certify" certain important legal questions); Fed. Rule Civ. Proc. 54(b) (authorizing district courts to "direct the entry of a final judgment as to one or more but fewer than all of the claims or parties"); 28 U. S. C. §§ 1292(e), 2072(c) (1988 ed., Supp. V) (authorizing this Court to promulgate rules designating certain kinds of orders as immediately appealable); cf. 28 U. S. C. § 1651 (authorizing federal courts to "issue all writs necessary or appropriate," including writs of mandamus).
Second, in Cohen v. Beneficial Industrial Loan Corp., 337 U. S. 541 (1949), this Court held that certain so-called collateral orders amount to "final decisions," immediately appealable under the here-relevant statute, 28 U. S. C. § 1291, even though the district court may have entered those orders before (perhaps long before) the case has ended. These special "collateral orders" were those that fell within
"that small class which finally determine claims of right separable from, and collateral to, rights asserted in the action, 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." Cohen, supra, at 546.
More recently, this Court has restated Cohen as requiring that the order "` 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.' " Puerto Rico Aqueduct and Sewer Authority v. Metcalf & Eddy, Inc., 506 U. S. 139, 144 (1993) (brackets in original) (quoting Coopers & Lybrand v. Livesay, 437 U. S. 463, 468 (1978)).
In determining which "collateral orders" amount to "final decisions," these requirements help qualify for immediate appeal classes of orders in which the considerations that favor immediate appeals seem comparatively strong and those that disfavor such appeals seem comparatively weak. The requirement that the issue underlying the order be "`effectively unreviewable' " later on, for example, means that failure to review immediately may well cause significant harm. See 15A C. Wright, A. Miller, & E. Cooper, Federal Practice and Procedure § 3911, pp. 334-335 (1992) (hereinafter Wright & Miller). The requirement that the district court's order "conclusively determine" the question means that appellate review is likely needed to avoid that harm. Id., at 333. The requirement that the matter be separate from the merits of the action itself means that review now is less likely to force the appellate court to consider approximately the same (or a very similar) matter more than once, and also seems less likely to delay trial court proceedings (for, if the matter is truly collateral, those proceedings might continue while the appeal is pending). Id., at 333-334.
Third, in Mitchell v. Forsyth, 472 U. S. 511 (1985), this Court held that a district court's order denying a defendant's motion for summary judgment was an immediately appealable "collateral order" (i. e., a "final decision") under Cohen, where (1) the defendant was a public official asserting a defense of "qualified immunity," and (2) the issue appealed concerned, not which facts the parties might be able to prove, but, rather, whether or not certain given facts showed a violation of "clearly established" law. 472 U. S., at 528; see Harlow v. Fitzgerald, 457 U. S. 800, 818 (1982) (holding that public officials are entitled to a "qualified immunity" from "liability for civil damages insofar as their conduct does not violate clearly established . . . rights of which a reasonable person would have known"). Applying Cohen `s criteria, the Mitchell Court held that this kind of summary judgment order was, in a sense, "effectively unreviewable," for review after trial would come too late to vindicate one important purpose of "qualified immunity"namely, protecting public officials, not simply from liability,but also from standing trial. Mitchell, supra, at 525-527. For related reasons, the Court found that the order was conclusive, i. e., it "conclusively" settled the question of the defendant's immunity from suit. 472 U. S., at 527.
The Court in Mitchell found more difficult the "separability" question, i. e., whether or not the "qualified immunity" issue was "completely separate from the merits of the action," supra, at 310. The Court concluded that:
"it follows from the recognition that qualified immunity is in part an entitlement not to be forced to litigate the consequences of official conduct that a claim of immunity is conceptually distinct from the merits of the plaintiff's claim that his rights have been violated." Mitchell, supra, at 527-528 (emphasis added).
And, the Court said that this "conceptual distinctness" made the immediately appealable issue "separate" from the merits of the plaintiff's claim, in part because an
"appellate court reviewing the denial of the defendant's claim of immunity need not consider the correctness of the plaintiff's version of the facts, nor even determine whether the plaintiff's allegations actually state a claim. All it need determine is a question of law: whether the legal norms allegedly violated by the defendant were clearly established at the time of the challenged actions or, in cases where the district court has denied summary judgment for the defendant on the ground that even under the defendant's version of the facts the defendant's conduct violated clearly established law, whether the law clearly proscribed the actions the defendant claims he took." Id., at 528 (footnote omitted).
We now consider the appealability of a portion of a district court's summary judgment order that, though entered in a "qualified immunity" case, determines only a question of "evidence sufficiency," i. e., which facts a party may, or may not, be able to prove at trial. This kind of order, we conclude, is not appealable. That is, the District Court's determination that the summary judgment record in this case raised a genuine issue of fact concerning petitioners' involvement in the alleged beating of respondent was not a "final decision" within the meaning of the relevant statute. We so decide essentially for three reasons.
First, consider Mitchell itself, purely as precedent. The dispute underlying the Mitchell appeal involved the application of "clearly established" law to a given (for appellate purposes undisputed) set of facts. And, the Court, in its opinion, explicitly limited its holding to appeals challenging, not a district court's determination about what factual issues are "genuine," Fed. Rule Civ. Proc. 56(c), but the purely legal issue what law was "clearly established." The opinion, for example, referred specifically to a district court's "denial of a claim of qualified immunity, to the extent that it turns on an issue of law. " 472 U. S., at 530 (emphasis added). It "emphasize[d] . . . that the appealable issue is a purely legal one: whether the facts alleged (by the plaintiff, or, in some cases, the defendant) support a claim of violation of clearly established law." Id., at 528, n. 9. It distinguished precedent not permitting interlocutory appeals on the ground that "a qualified immunity ruling . . . is . . . a legal issue that can be decided with reference only to undisputed facts and in isolation from the remaining issues of the case." Id., at 530, n. 10. And, it explained its separability holding by saying that "[a]n appellate court reviewing the denial of the defendant's claim of immunity need not consider the correctness of the plaintiff's version of the facts." Id., at 528. Although there is some language in the opinion that sounds as if it might imply the contrary, it does not do so when read in context. See, e. g., id., at 526 (referring to defendant's entitlement to summary judgment, not to appealability, by saying that "defendant is entitled to summary judgment if discovery fails to uncover evidence sufficient to create a genuine issue").
Second, consider, in the context of an "evidence sufficiency" claim, Cohen `s conceptual theory of appealability the theory that brings immediate appealability within the scope of the jurisdictional statute's "final decision" requirement. That theory finds a "final" district court decision in part because the immediately appealable decision involves issues significantly different from those that underlie the plaintiff's basic case. As we have just pointed out, Mitchell rested upon the view that "a claim of immunity is conceptually distinct from the merits of the plaintiff's claim." 472 U. S., at 527. It held that this was so because, although sometimes practically intertwined with the merits, a claim of immunity nonetheless raises a question that is significantly different from the questions underlying plaintiff's claim on the merits (i. e., in the absence of qualified immunity). Id., at 528.
Where, however, a defendant simply wants to appeal a district court's determination that the evidence is sufficient to permit a particular finding of fact after trial, it will often prove difficult to find any such "separate" questionone that is significantly different from the fact-related legal issues that likely underlie the plaintiff's claim on the merits. See Anderson v. Liberty Lobby, Inc., 477 U. S. 242, 248 (1986) (district court's task, in deciding whether there is a "genuine" issue of fact, is to determine "if the evidence is such that a reasonable jury could return a verdict for the nonmoving party"); see also Elliott v. Thomas, 937 F. 2d, at 341 ("[W]hether the defendants did the deeds alleged . . . is precisely the question for trial") (emphasis in original), cert. denied, 502 U. S. 1074, 1121 (1992); Wright v. South Arkansas Regional Health Center, Inc., 800 F. 2d 199, 203 (CA8 1986) (saying that this question "is . . . less clearly separable from the merits" than the question in Mitchell ); see also Brief for United States 18 ("In one sense, a ruling regarding the sufficiency of the evidence is closely intertwined with the merits").
It has been suggested that Mitchell implicitly recognized that "the need to protect officials against the burdens of further pretrial proceedings and trial" justifies a relaxation of the separability requirement. 15A Wright & Miller § 3914.10, at 656; see id., § 3911, at 344-345; id., § 3911.2, at 387; see also Tr. of Oral Arg. 20 ("[W]here the right not to be tried is at stake, [closer] association with the merits is tolerated") (argument of the United States). Assuming that to be so, and despite a similar interest in avoiding trial in the kind of case here at issue, we can find no separability. To take what petitioners call a small step beyond Mitchell, Brief for Petitioners 18, would more than relax the separability requirementit would in many cases simply abandon it.
Finally, consider the competing considerations that underlie questions of finality. See supra, at 309-310. We of course decide appealability for categories of orders rather than individual orders. See Digital Equipment Corp. v. Desktop Direct, Inc., 511 U. S. 863, 868 (1994). Thus, we do not now in each individual case engage in ad hoc balancing to decide issues of appealability. See generally P. Bator, D. Meltzer, P. Mishkin, & D. Shapiro, Hart and Wechsler's The Federal Courts and The Federal System 1810 (3d ed. 1988). But, that does not mean that, in delineating appealable categories, we should not look to "the competing considerations underlying all questions of finality'the inconvenience and costs of piecemeal review on the one hand and the danger of denying justice by delay on the other.' " Eisen v. Carlisle & Jacquelin, 417 U. S. 156, 171 (1974) (quoting Dickinson v. Petroleum Conversion Corp., 338 U. S. 507, 511 (1950)). And, those considerations, which we discussed above in Part IIA, argue against extending Mitchell to encompass orders of the kind before us.
For one thing, the issue here at stakethe existence, or nonexistence, of a triable issue of factis the kind of issue that trial judges, not appellate judges, confront almost daily. Institutionally speaking, appellate judges enjoy no comparative expertise in such matters. Cf. Pierce v. Underwood, 487 U. S. 552, 560-561 (1988); id., at 584 (White, J., concurring in part and dissenting in part) (noting that the "special expertise and experience of appellate courts" lies in "assessing the relative force of . . . applications of legal norms") (internal quotation marks omitted). And, to that extent, interlocutory appeals are less likely to bring important errorcorrecting benefits here than where purely legal matters are at issue, as in Mitchell. Cf. Richardson-Merrell, 472 U. S., at 434 (stating that the fact that "[m]ost pretrial orders [of the kind there at issue] are ultimately affirmed by appellate courts" militated against immediate appealability).
For another thing, questions about whether or not a record demonstrates a "genuine" issue of fact for trial, if appealable, can consume inordinate amounts of appellate time. Many constitutional tort cases, unlike the simple "we didn't do it" case before us, involve factual controversies about, for example, intentcontroversies that, before trial, may seem nebulous. To resolve those controversiesto determine whether there is or is not a triable issue of fact about such a matter may require reading a vast pretrial record, with numerous conflicting affidavits, depositions, and other discovery materials. This fact means, compared with Mitchell, greater delay.
For a third thing, the close connection between this kind of issue and the factual matter that will likely surface at trial means that the appellate court, in the many instances in which it upholds a district court's decision denying summary judgment, may well be faced with approximately the same factual issue again, after trial, with just enough change (brought about by the trial testimony) to require it, once again, to canvass the record. That is to say, an interlocutory appeal concerning this kind of issue in a sense makes unwise use of appellate courts' time, by forcing them to decide in the context of a less developed record, an issue very similar to one they may well decide anyway later, on a record that will permit a better decision. See 15A Wright & Miller § 3914.10, at 664 ("[I]f [immunity appeals] could be limited to. . . issues of law . . . there would be less risk that the court of appeals would need to waste time in duplicating investigations of the same facts on successive appeals").
The upshot is that, compared with Mitchell, considerations of delay, comparative expertise of trial and appellate courts, and wise use of appellate resources argue in favor of limiting interlocutory appeals of "qualified immunity" matters to cases presenting more abstract issues of law. Considering these "competing considerations," we are persuaded that "[i]mmunity appeals . . . interfere less with the final judgment rule if they [are] limited to cases presenting neat abstract issues of law." 15A Wright & Miller § 3914.10, at 664; cf. Puerto Rico Aqueduct, 506 U. S., at 147 (noting the argument for a distinction between fact-based and law-based appeals, but seeing no "basis for drawing" it with respect to the particular kind of order at hand); 15A Wright & Miller § 3914.10, at 85 (1995 Supp.).
We recognize that, whether a district court's denial of summary judgment amounts to (a) a determination about preexisting "clearly established" law, or (b) a determination about "genuine" issues of fact for trial, it still forces public officials to trial. See Brief for Petitioners 11-16. And, to that extent, it threatens to undercut the very policy (protecting public officials from lawsuits) that (the Mitchell Court held) militates in favor of immediate appeals. Nonetheless, the countervailing considerations that we have mentioned (precedent, fidelity to statute, and underlying policies) are too strong to permit the extension of Mitchell to encompass appeals from orders of the sort before us.
We mention one final point. Petitioners argue that our effort to separate reviewable from unreviewable summary judgment determinations will prove unworkable. First, they say that the parties can easily manipulate our holding. A defendant seeking to create a reviewable summary judgment order might do so simply by adding a reviewable claim to a motion that otherwise would create an unreviewable order. "[H]ere, for example," they say, "petitioners could have contended that the law was unclear on how much force may be exerted against suspects who resist arrest." Brief for Petitioners 29, n. 11.
We do not think this is a serious problem. We concede that, if the District Court in this case had determined that beating respondent violated clearly established law, petitioners could have sought review of that determination. But, it does not automatically follow that the Court of Appeals would also have reviewed the here more important determination that there was a genuine issue of fact as to whether petitioners participated in (or were present at) a beating. Even assuming, for the sake of argument, that it may sometimes be appropriate to exercise "pendent appellate jurisdiction" over such a matter, but cf. Swint v. Chambers County Comm'n, 514 U. S. 35, 50-51 (1995), it seems unlikely that courts of appeals would do so in a case where the appealable issue appears simply a means to lead the court to review the underlying factual matter, see, e. g., Natale v. Ridgefield, 927 F. 2d 101, 104 (CA2 1991) (saying exercise of pendent appellate jurisdiction is proper only in "exceptional circumstances"); United States ex rel. Valders Stone & Marble, Inc. v. C-Way Constr. Co., 909 F. 2d 259, 262 (CA7 1990) (saying exercise of such jurisdiction is proper only where there are "`compelling reasons' ").
Second, petitioners add, if appellate courts try to separate an appealed order's reviewable determination (that a given set of facts violates clearly established law) from its unreviewable determination (that an issue of fact is "genuine"), they will have great difficulty doing so. District judges may simply deny summary judgment motions without indicating their reasons for doing so. How, in such a case, will the court of appeals know what set of facts to assume when it answers the purely legal question about "clearly established" law?
This problem is more serious, but not serious enough to lead us to a different conclusion. When faced with an argument that the district court mistakenly identified clearly established law, the court of appeals can simply take, as given, the facts that the district court assumed when it denied summary judgment for that (purely legal) reason. Knowing that this is "extremely helpful to a reviewing court," Anderson, 477 U. S., at 250, n. 6, district courts presumably will often state those facts. But, if they do not, we concede that a court of appeals may have to undertake a cumbersome review of the record to determine what facts the district court, in the light most favorable to the nonmoving party, likely assumed. Regardless, this circumstance does not make a critical difference to our result, for a rule that occasionally requires a detailed evidence-based review of the record is still, from a practical point of view, more manageable than the rule that petitioners urge us to adopt. Petitioners' approach would make that task, not the exception, but the rule. We note, too, that our holding here has been the law in several Circuits for some time. See supra, at 308-309. Yet, petitioners have not pointed to concrete examples of the unmanageability they fear.
For these reasons, we hold that a defendant, entitled to invoke a qualified immunity defense, may not appeal a district court's summary judgment order insofar as that order determines whether or not the pretrial record sets forth a "genuine" issue of fact for trial. The judgment of the Court of Appeals for the Seventh Circuit is therefore