delivered the opinion of the Court.

Federal Rule of Appellate Procedure 23(c) provides that, when the Government appeals a decision granting a writ of habeas corpus, the habeas petitioner shall be released from custody “unless the court or justice or judge rendering the decision, or the court of appeals or the Supreme Court, or a judge or justice of either court shall otherwise order.” Federal Rule of Appellate Procedure 23(d) states that initial orders issued pursuant to Rule 23(c) shall “govern review in the court of appeals and in the Supreme Court unless for special reasons shown . . . the order shall be modified, or an independent order respecting custody, enlargement or surety shall be made.”1 In this case, we are asked to decide what factors these provisions allow a court to consider in determining whether to release a state prisoner pending appeal of a district court order granting habeas relief.

In January 1981, respondent Dana Braunskill was convicted in the Superior Court of New Jersey, Law Division, of sexual assault and unlawful possession of a weapon, in violation of N. J. Stat. Ann. §§2C:14-2, 2C:39-5(d) (West 1982 and Supp. 1986-1987), and was sentenced to eight years’ *773imprisonment. The Appellate Division of the Superior Court affirmed the convictions, and the New Jersey Supreme Court denied review.

Respondent then, in 1985, filed a petition for a writ of habeas corpus in the United States District Court for the District of New Jersey. Finding that respondent’s Sixth Amendment rights had been violated at his trial, the District Court granted respondent’s petition and ordered that “a writ of habeas corpus shall issue unless within 30 days the State of New Jersey shall afford [respondent] a new trial.” 629 F. Supp. 511, 526 (1986). Petitioners subsequently moved the District Court to stay its order pending appeal. Relying on Carter v. Rafferty, 781 F. 2d 993, 997 (CA3 1986), the District Court determined that it could grant petitioners’ request only if they demonstrated that there was risk that respondent would not appear for subsequent proceedings. The court found that petitioners had failed to make such a showing and denied the motion.

Petitioners then filed a motion in the United States Court of Appeals for the Third Circuit, seeking a stay of the District Court’s order releasing respondent. The Court of Appeals denied the motion by order dated May 27,1986. We granted certiorari to review the Court of Appeals’ denial of the stay, 479 U. S. 881 (1986), and now vacate and remand the case to the Court of Appeals.2

In Carter v. Rafferty, supra, the authority governing the Court of Appeals decision in this case,3 the court held that *774federal courts deciding whether to release a successful ha-beas petitioner pending appeal may consider the petitioner’s risk of flight, but not his danger to the community. The court observed that Rule 28(c) creates a presumption that a prisoner who has received habeas relief is entitled to release from custody. Moreover, the Carter court reasoned, the principal interests that a federal court may consider under Rules 23(c) and (d) are those of ensuring the appearance of the prisoner in subsequent federal proceedings and returning the prisoner to state custody if the State prevails on appeal of the award of habeas relief. To conclude otherwise, the court determined, would result in federal-court intrusion into matters of traditional state concern.

We do not believe that federal courts, in deciding whether to stay pending appeal a district court order granting relief to a habeas petitioner, are as restricted as the Carter court thought. Rule 23(c) undoubtedly creates a presumption of release from custody in such cases,4 but that presumption may be overcome if the judge rendering the decision, or an appellate court or judge, “otherwise orders.” Rule 23(d) creates a presumption of correctness for the order of a district court entered pursuant to Rule 23(c), whether that order enlarges the petitioner or refuses to enlarge him, but this presumption may be overcome in the appellate court “for special reasons shown.” We think a resort to the history of habeas practice in the federal courts and the traditional standards governing stays of civil judgments in those courts is helpful in illuminating the generality of these terms of Rules 23(c) and (d).

*775Federal habeas corpus practice, as reflected by the decisions of this Court, indicates that a court has broad discretion in conditioning a judgment granting habeas relief. Federal courts are authorized, under 28 U. S. C. § 2243, to dispose of habeas corpus matters “as law and justice require.” In construing § 2243 and its predecessors, this Court has repeatedly stated that federal courts may delay the release of a successful habeas petitioner in order to provide the State an opportunity to correct the constitutional violation found by the court. See, e. g., Rogers v. Richmond, 365 U. S. 534, 549 (1961); Dowd v. United States ex rel. Cook, 340 U. S. 206, 210 (1951); In re Bonner, 151 U. S. 242, 261-262 (1894). Even in 1894, when this Court’s Rule 34 indicated that enlargement of successful habeas petitioners pending the State’s appeal was mandatory, see n. 4, swpra, the Court interpreted the predecessor of §2243 as vesting a federal court “with the largest power to control and direct the form of judgment to be entered in cases brought up before it on habeas corpus.” Id., at 261. We think it would make little sense if this broad discretion allowed in fashioning the judgment granting relief to a habeas petitioner were to evaporate suddenly when either the district court or the court of appeals turns to consideration of whether the judgment granting habeas relief should be stayed pending appeal. Although the predecessor of Rule 23 apparently required this strange result, see n. 4, supra, the language of the current Rule undoubtedly permits a more sensible interpretation.

In those instances where a Member of this Court has been confronted with the question whether a prevailing habeas petitioner should be released pending the Court’s disposition of the State’s petition for certiorari, our approach has been to follow the general standards for staying a civil judgment. See Tate v. Rose, 466 U. S. 1301 (1984) (O’Connor, J., in chambers); cf. Sumner v. Mata, 446 U. S. 1302 (19801 (Rehnquist, J., in chambers). This practice reflects the common-sense notion that a court’s denial of enlargement to a *776successful habeas petitioner pending review of the order granting habeas relief has the same effect as the court’s issuance of a stay of that order. Our decisions have consistently recognized that habeas corpus proceedings are civil in nature. See, e. g., Browder v. Director, Illinois Dept. of Corrections, 434 U. S. 257, 269 (1978).5 It is therefore logical to conclude that the general standards governing stays of civil judgments should also guide courts when they must decide whether to release a habeas petitioner pending the State’s appeal; and such a conclusion is quite consistent with the general language contained in Rules 23(c) and (d).

Different Rules of Procedure govern the power of district courts and courts of appeals to stay an order pending appeal. See Fed. Rule Civ. Proc. 62(c); Fed. Rule App. Proc. 8(a). Under both Rules, however, the factors regulating the issuance of a stay are generally the same: (1) whether the stay applicant has made a strong showing that he is likely to succeed on the merits; (2) whether the applicant will be irreparably injured absent a stay; (3) whether issuance of the stay will substantially injure the other parties interested in the proceeding; and (4) where the public interest lies. See, e. g., Virginia Petroleum Jobbers Assn. v. FPC, 104 U. S. App. D. C. 106, 110, 259 F. 2d 921, 925 (1958); Washington Metropolitan Area Comm’n v. Holiday Tours, Inc., 182 U. S. App. D. C. 220, 221-222, 559 F. 2d 841, 842-844 (1977); Garcia-Mir v. Meese, 781 F. 2d 1450, 1453 (CA11 1986); Accident Fund v. Baerwaldt, 579 F. Supp. 724, 725 (WD Mich. *7771984); see generally 11 C. Wright & A. Miller, Federal Practice and Procedure § 2904 (1973).

For the reasons stated, we think that a court making an initial custody determination under Rule 23(c) should be guided not only by the language of the Rule itself but also by the factors traditionally considered in deciding whether to stay a judgment in a civil case. There is presumption in favor of enlargement of the petitioner with or without surety, but it may be overcome if the traditional stay factors tip the balance against it. A court reviewing an initial custody determination pursuant to Rule 23(d) must accord a presumption of correctness to the initial custody determination made pursuant to Rule 23(c), whether that order directs release or continues custody, but that presumption, too, may be overcome if the traditional stay factors so indicate. The construction of Rule 23 we here adopt accords both the court making the initial custody determination and the court reviewing that determination considerably more latitude than that apparently thought appropriate by the Court of Appeals for the Third Circuit in this case.

Since the traditional stay factors contemplate individualized judgments in each case, the formula cannot be reduced to a set of rigid rules. The Court of Appeals in Carter v. Rafferty, 781 F. 2d 993 (CA3 1986), agreed that the possibility of flight should be taken into consideration, and we concur in that determination. We also think that, if the State establishes that there is a risk that the prisoner will pose a danger to the public if released, the court may take that factor into consideration in determining whether or not to enlarge him. The State’s interest in continuing custody and rehabilitation pending a final determination of the case on appeal is also a factor to be considered; it will be strongest where the remaining portion of the sentence to be served is long, and weakest where there is little of the sentence remaining to be served.

The interest of the habeas petitioner in release pending appeal, always substantial, will be strongest where the factors *778mentioned in the preceding paragraph are weakest. The balance may depend to a large extent upon determination of the State’s prospects of success in its appeal. Where the State establishes that it has a strong likelihood of success on appeal, or where, failing that, it can nonetheless demonstrate a substantial case on the merits, continued custody is permissible if the second and fourth factors in the traditional stay analysis militate against release. Cf. McSurely v. McClellan, 225 U. S. App. D. C. 67, 75, 697 F. 2d 309, 317 (1982); O’Bryan v. Estelle, 691 F. 2d 706, 708 (CA5 1982), cert. denied, 465 U. S. 1013 (1984); Ruiz v. Estelle, 650 F. 2d 555, 565-566 (CA5 1981). Where the State’s showing on the merits falls below this level, the preference for release should control.

Respondent contends, and the Court of Appeals apparently agreed, that matters of “traditional state concern” such as the petitioner’s danger to the community ought not to be considered in determining whether a successful habeas petitioner should be enlarged pending appeal. Respondent supports his argument by stating that this Court’s decisions embody the view that state governments should have the opportunity to vindicate state interests in their own court systems. We do not at all dispute this observation, but note that here we have the Attorney General of New Jersey speaking for that State and seeking a stay of the District Court order enlarging a habeas petitioner pending appeal. Whatever strain on federal-state relations arising as a result of federal habeas jurisdiction comes because of the granting of habeas relief itself, and not the existence of any discretion in habeas courts to refuse enlargement of a successful habeas petitioner pending appeal. Until the final determination of the petitioner’s habeas claim, federal courts must decide applications for stay of release using factors similar to those used in deciding whether to stay other federal-court judgments.

Respondent finally contends that staying the release of a successful habeas petitioner pending appeal because of dan*779gerousness, even when guided by the standards we have enunciated, is “repugnant to the concept of substantive due process, which . . . prohibits the total deprivation of liberty simply as a means of preventing future crimes.” United States v. Salerno, 794 F. 2d 64, 71-72 (CA2 1986). We have just held in reversing the judgment of the Court of Appeals for the Second Circuit in Salerno, however, that the quoted language is an incorrect statement of constitutional law. Ante, p. 739. But we also think that a successful habeas petitioner is in a considerably less favorable position than a pretrial arrestee, such as the respondent in Salerno, to challenge his continued detention pending appeal. Unlike a pretrial arrestee, a state habeas petitioner has been adjudged guilty beyond a reasonable doubt by a judge or jury, and this adjudication of guilt has been upheld by the appellate courts of the State. Although the decision of a district court granting habeas relief will have held that the judgment of conviction is constitutionally infirm, that determination itself may be overturned on appeal before the State must retry the petitioner. This being the case, we do not agree that the Due Process Clause prohibits a court from considering, along with the other factors that we previously described, the dangerousness of a habeas petitioner as part of its decision whether to release the petitioner pending appeal.

We think that the District Court and the Court of Appeals, in relying on the latter’s decision in Carter v. Rafferty, supra, took too limited a view of the discretion allowed to federal courts under Rules 23(c) and (d) in staying pending appeal an order directing the release of a habeas petitioner. We therefore vacate the judgment of the Court of Appeals denying petitioner’s application for a stay in this case, and remand the case to that court for further proceedings consistent with this opinion.

It is so ordered.

with whom Justice Brennan and Justice Blackmun join, dissenting.

This Court construes Federal Rules of Appellate Procedure 23(c) and 23(d) to invest federal courts with broad discretion to keep a successful habeas petitioner in custody-pending appeal by the State. Because I believe that this novel approach allows federal courts to usurp the role of the state courts and undermine the purpose of habeas corpus proceedings, I dissent.

I

In our federal system, state courts are the appropriate forums for deciding questions of release for those charged with state offenses. The law that applies in these state proceedings is state bail law; in this case, state law grants respondent a right to be admitted to bail. Federal courts are not free to deprive respondent of that right, merely because the State’s representative asks them to. Federal Rules of Procedure cannot supplant either substantive rights guaranteed under the state law or the state processes developed to enforce those rights.

In holding that the federal courts can consider a prevailing habeas petitioner’s danger to the community, the majority rejects the Third Circuit’s well-reasoned decision to the contrary in Carter v. Rafferty, 781 F. 2d 993 (1986).1 In that case, a Federal District Court had granted a writ of habeas corpus to Rubin “Hurricane” Carter, who had previously been convicted of murder in a New Jersey state court, and ordered him released from state custody. The State maintained that Carter was a danger to the community and sought an order from the Court of Appeals, pursuant to Rule 23(d), to keep him in custody pending appeal. The court’s analysis of Rules 23(c) and 23(d) started with several general principles: first, there is a presumption that a successful habeas *781petitioner is entitled to release “immediately or, more commonly, after an appropriately circumscribed period to allow the state time to retry the accused.” 781 F. 2d, at 994. Second, a federal court has a strong interest in ensuring the appearance of the petitioner in subsequent federal proceedings or, if the decision is overturned on appeal, in returning the petitioner to state custody. Id., at 995. Third, neither federal nor state bail standards govern the release of state prisoners in federal habeas proceedings. Ibid.

Based on these principles and on the limited role of the federal courts in habeas corpus proceedings, namely, “to determine whether or not a constitutional infirmity infected the defendant’s trial,” id., at 996, the court concluded that release of a state prisoner who prevailed in the district court can “only be challenged ... if matters are put in issue relating to a petitioner’s ability to respond to federal process, or which in some other respect relate to the federal interest.” Id., at 996-997. Because the sole reason advanced for Carter’s incarceration was his alleged dangerousness, “a matter traditionally reserved to the state authorities to decide,” id., at 996, the court denied the State’s motion to revoke Carter’s release. It emphasized that its holding did not leave the State without recourse:

“In those instances where the state is of the view that a petitioner should not, for other compelling reasons, remain at large, it may proceed before the state courts. Since questions of dangerousness per se and related issues are traditionally state concerns and since the victorious habeas petitioner generally still faces trial on a state indictment, the appropriate forum before which state authorities may seek relief is the state court with responsibility for pending or future proceedings concerning the underlying indictment.” Id., at 997-998.

The decision in Carter was based on traditional notions of federalism and comity. The majority rejects this approach, deferring instead to the State’s interest as an adversary *782party in litigation. This peculiar brand of federalism finds no support in our prior cases, which reflect deference to state courts and state-court decisions, not litigants representing the State.2

Even more disturbing is the fact that the majority’s result has no apparent basis in state law. The Attorney General for the State of New Jersey has failed to cite a single state statute, state rule of court, or state decision that permits preventive detention pending trial or, for that matter, pending appeal of an order granting state postconviction relief. This is hardly surprising, since New Jersey law does not permit a state court to consider a defendant’s future dangerousness in determining whether to order pretrial confinement. State v. Johnson, 61 N. J. 351, 294 A. 2d 245 (1972). Except in capital cases, the State Constitution provides a right to bail. See N. J. Const., Art. I, ¶ 11; see also N. J. Rule Crim. Prac. 3:26 — 1(a); N. J. Rule App. Prac. 2:9-3(d). The State Attorney General has asked the federal courts to confine respondent on a basis that New Jersey courts hold invalid. Such a request is clearly not proper, much less deserving of deference.

The majority suggests that refusal to allow federal courts to consider danger to the community is somehow inconsistent with the practice of granting “conditional writs” of habeas corpus,3 in which a federal court orders that the State re*783lease the habeas petitioner within a specified period unless it retries him within that time. I do not believe that the traditional practice of issuing conditional writs is implicated by the decision in this case, which turns on fundamental principles of federal noninterference with the procedures for vindication of state-law rights in state courts. I note, however, that the practice is entirely consistent with the traditional concept of deference to state courts. By delaying issuance of the writ for a reasonable period, the federal court gives the State an opportunity to correct the constitutional defect itself through retrial in its own courts.

Under today’s decision a federal court can disregard both state law and state processes and authorize the indefinite detention of a successful habeas petitioner, without a full-blown adversary hearing, without appointing counsel, without providing immediate appellate review of its decision, and without satisfying any elevated burden of proof. Compare United States v. Salerno, ante, at 747, 751-752. The Court’s analysis in this area strikes me as result oriented, to say the least. Writing for the Court in Salerno, The Chief Justice chose to rely on the “numerous procedural safeguards” contained in the Bail Reform Act of 1984 to sustain the statute’s constitutionality. Ante, at 755. Recognizing the “individual’s strong interest in liberty,” Chief Justice Rehnquist stated:

*784“We do not minimize the importance and fundamental nature of this right. But, as our cases hold, this right may, in circumstances where the government’s interest is sufficiently weighty, be subordinated to the greater needs of society. We think that Congress’ careful delineation of the circumstances under which detention will be permitted satisfies this standard.” Ante, at 750-751 (emphasis added).

Yet in this case, where the same important and fundamental right is at stake, The Chief Justice, again writing for the Court, disregards the total absence of safeguards against erroneous or unnecessary deprivations of liberty.

The majority attempts to distinguish the successful habeas petitioner from the pretrial detainee in Salerno, observing that “a state habeas petitioner has been adjudged guilty beyond a reasonable doubt by a judge or jury, and this adjudication of guilt has been upheld by the appellate courts of the State.” Ante, at 779. The Court concedes, as it must, that this conviction has been found constitutionally infirm by a Federal District Court, but it notes that this “determination itself may be overturned on appeal before the State must retry the petitioner.” Ibid. This observation trivializes the District Court’s ruling that the State obtained its conviction in violation of respondent’s constitutional rights. Respondent’s conviction has been rendered null and void by a federal court of competent jurisdiction; it provides no basis for continuation of punishment or, as the majority so delicately puts it, “continuing custody and rehabilitation.” Ante, at 777. See Bell v. Wolfish, 441 U. S. 520, 535 (1979) (“[A] detainee may not be punished prior to an adjudication of guilt in accordance with due process of law”). The fact that the ruling might later be reversed does not diminish its current validity. We do not discount federal-court rulings simply because they “may be overturned on appeal.”

Granting broad discretion to deny release pending appeal undermines the central purpose of habeas corpus proceed*785ings: to provide “protection against illegal custody.” Brown v. Allen, 344 U. S. 443, 465 (1953); see also Preiser v. Rodriguez, 411 U. S. 475, 485-486 (1973). In this case, at the time the writ issued, respondent had spent five years in prison. He would have been eligible for parole in approximately eight months.4 Had the State obtained a stay of release, he undoubtedly would have to serve the entire sentence imposed pursuant to a conviction now determined to be unconstitutional. The writ of habeas corpus would have provided him no protection against illegal custody. If a prisoner’s confinement is to continue pending appeal, it should only be for reasons consistent with, or at least not in conflict with, the primary purpose of habeas corpus. These reasons cannot include continuation of punishment, as the majority implies. See ante, at 777.

Finally, continued reliance on the state conviction in this case ignores the nature of the constitutional defect identified by the District Court: the error in this case directly implicates the truth-finding process. Respondent has consistently maintained that this is a case of mistaken identity and that he was elsewhere on the night of the crime. As part of his defense, he sought to introduce the testimony of an alibi witness. Because his counsel failed to file a timely notice of alibi testimony, the trial court refused to allow him to do so. Even without the benefit of the witness’ testimony, the jury deliberated for 214 days before returning a guilty verdict. The District Court noted that an alibi witness would have strengthened respondent’s case and created reasonable doubt, 629 F. Supp. 511, 523 (NJ 1986); it concluded that the trial court’s refusal to allow respondent to introduce this testimony violated his Sixth Amendment right to present witnesses to establish a defense. Ibid.

*786I — I hH HH

Title 28 U. S. C. § 2243, cited by the majority, ante, at 775, authorizes federal courts to dispose of habeas corpus matters “as law and justice require.” The majority’s construction of Rules 23(c) and 23(d) is contrary to both law and justice. It is inconsistent with this Court’s longstanding notions of federalism and comity. It allows federal courts to substitute their own ad hoc standards for the rules and procedures the States have established for regulating the pretrial release of those accused of state-law offenses.

I therefore dissent.