delivered the opinion of the Court, except as to Part III-C.

In this case we consider whether the exhaustion rule in 28 U. S. C. §§ 2254(b), (c) requires a federal district court to dismiss a petition for a writ of habeas corpus containing any claims that have not been exhausted in the state courts. Because a rule requiring exhaustion of all claims furthers the purposes underlying the habeas statute, we hold that a district court must dismiss such “mixed petitions,” leaving the prisoner with the choice of returning to state court to exhaust his claims or of amending or resubmitting the habeas petition to present only exhausted claims to the district court.

HH

Following a jury trial, respondent Noah Lundy was convicted on charges of rape and crime against nature, and sentenced to the Tennessee State Penitentiary.1 After the Tennessee Court of Criminal Appeals affirmed the convictions and the Tennessee Supreme Court denied review, the respondent filed an unsuccessful petition for postconviction relief in the Knox County Criminal Court.

*511The respondent subsequently filed a petition in Federal District Court for a writ of habeas corpus under 28 U. S. C. § 2254, alleging four grounds for relief: (1) that he had been denied the right to confrontation because the trial court limited the defense counsel’s questioning of the victim; (2) that he had been denied the right to a fair trial because the prosecuting attorney stated that the respondent had a violent character; (3) that he had been denied the right to a fair trial because the prosecutor improperly remarked in his closing argument that the State’s evidence was uncontradicted; and (4) that the trial judge improperly instructed the jury that every witness is presumed to swear the truth. After reviewing the state-court records, however, the District Court concluded that it could not consider claims three and four “in the constitutional framework” because the respondent had not exhausted his state remedies for those grounds. The court nevertheless stated that “in assessing the atmosphere of the cause taken as a whole these items may be referred to collaterally.”2

Apparently in an effort to assess the “atmosphere” of the trial, the District Court reviewed the state trial transcript and identified 10 instances of prosecutorial misconduct, only 5 of which the respondent had raised before the state courts.3

*512In addition, although purportedly not ruling on the respondent’s fourth ground for relief — that the state trial judge improperly charged that “every witness is presumed to swear the truth” — the court nonetheless held that the jury instruction, coupled with both the restriction of counsel’s cross-examination of the victim and the prosecutor’s “personal testimony” on the weight of the State’s evidence, see n. 3, supra, violated the respondent’s right to a fair trial. In conclusion, the District Court stated:

“Also, subject to the question of exhaustion of state remedies, where there is added to the trial atmosphere the comment of the Attorney General that the only story presented to the jury was by the state’s witnesses there is such mixture of violations that one cannot be separated from and considered independently of the others.

“. . . Under the charge as given, the limitation of cross examination of the victim, and the flagrant prosecutorial misconduct this court is compelled to find that petitioner did not receive a fair trial, his Sixth Amendment rights *513were violated and the jury poisoned by the prosecutorial misconduct.”4

In short, the District Court considered several instances of prosecutorial misconduct never challenged in the state trial or appellate courts, or even raised in the respondent’s habeas petition.

The Sixth Circuit affirmed the judgment of the District Court, 624 F. 2d 1100 (1980), concluding in an unreported order that the court properly found that the respondent’s constitutional rights had been “seriously impaired by the improper limitation of his counsel’s cross-examination of the prosecutrix- and by the prosecutorial misconduct.” The court specifically rejected the State’s argument that the District Court should have dismissed the petition because it included both exhausted and unexhausted claims.

HH f — t

The petitioner urges this Court to apply a “total exhaustion” rule requiring district courts to dismiss every habeas corpus petition that contains both exhausted and unex-hausted claims.5 The petitioner argues at length that such a *514rule furthers the policy of comity underlying the exhaustion doctrine because it gives the state courts the first opportunity to correct federal constitutional errors and minimizes federal interference and disruption of state judicial proceedings. The petitioner also believes that uniform adherence to a total exhaustion rule reduces the amount of piecemeal ha-beas litigation.

Under the petitioner’s approach, a district court would dismiss a petition containing both exhausted and unexhausted claims, giving the prisoner the choice of returning to state court to litigate his unexhausted claims, or of proceeding with only his exhausted claims in federal court. The petitioner believes that a prisoner would be reluctant to choose the latter route since a district court could, in appropriate circumstances under Habeas Corpus Rule 9 (b), dismiss subsequent federal habeas petitions as an abuse of the writ.6 In other words, if the prisoner amended the petition to delete the unexhausted claims or immediately refiled in federal court a petition alleging only his exhausted claims, he could lose the opportunity to litigate his presently unexhausted claims in federal court. This argument is addressed in Part III-C of this opinion.

*515In order to evaluate the merits of the petitioner’s arguments, we turn to the habeas statute, its legislative history, and the policies underlying the exhaustion doctrine.

Ill

A

The exhaustion doctrine existed long before its codification by Congress in 1948. In Ex parte Royall, 117 U. S. 241, 251 (1886), this Court wrote that as a matter of comity, federal courts should not consider a claim in a habeas corpus petition until after the state courts have had an opportunity to act:

“The injunction to hear the case summarily, and thereupon ‘to dispose of the party as law and justice require’ does not deprive the court of discretion as to the time and mode in which it will exert the powers conferred upon it. That discretion should be exercised in the light of the relations existing, under our system of government, between the judicial tribunals of the Union and of the States, and in recognition of the fact that the public good requires that those relations be not disturbed by unnecessary conflict between courts equally bound to guard and protect rights secured by the Constitution.”

Subsequent cases refined the principle that state remedies must be exhausted except in unusual circumstances. See, e. g., United States ex rel. Kennedy v. Tyler, 269 U. S. 13, 17-19 (1925) (holding that the lower court should have dismissed the petition because none of the questions had been raised in the state courts. “In the regular and ordinary course of procedure, the power of the highest state court in respect of such questions should first be exhausted”). In Ex parte Hawk, 321 U. S. 114, 117 (1944), this Court reiterated that comity was the basis for the exhaustion doctrine: “it is a principle controlling all habeas corpus petitions to the federal courts, that those courts will interfere with the administration of justice in the state courts only ‘in rare cases where exceptional circumstances of peculiar urgency are shown to *516exist.’”7 None of these cases, however, specifically applied the exhaustion doctrine to habeas petitions containing both exhausted and unexhausted claims.

In 1948, Congress codified the exhaustion doctrine in 28 U. S. C. §2254, citing Ex parte Hawk as correctly stating the principle of exhaustion.8 Section 2254,9 however, does not directly address the problem of mixed petitions. To be sure, the provision states that a remedy is not exhausted if there exists a state procedure to raise “the question presented,” but we believe this phrase to be too ambiguous to sustain the conclusion that Congress intended to either permit or prohibit review of mixed petitions. Because the legislative history of § 2254, as well as the pre-1948 cases, contains *517no reference to the problem of mixed petitions,10 in all likelihood Congress never thought of the problem.11 Consequently, we must analyze the policies underlying the statutory provision to determine its proper scope. Philbrook v. Glodgett, 421 U. S. 707, 713 (1975) (“‘In expounding a statute, we must. . . look to the provisions of the whole law, and to its object and policy’ ” (citations omitted)); United States v. Bacto-Unidisk, 394 U. S. 784, 799 (1969) (“where the statute’s language seem[s] insufficiently precise, the ‘natural way’ to draw the line ‘is in light of the statutory purpose’ ” (citation omitted)); United States v. Sisson, 399 U. S. 267, 297-298 (1970) (“The axiom that courts should endeavor to give statutory language that meaning that nurtures the poli*518cies underlying legislation is one that guides us when circumstances not plainly covered by the terms of a statute are subsumed by the underlying policies to which Congress was committed”); Unexcelled Chemical Corp. v. United States, 345 U. S. 59, 64 (1953) (“Arguments of policy are relevant when for example a statute has an hiatus that must be filled or there are ambiguities in the legislative language that must be resolved”).

B

The exhaustion doctrine is principally designed to protect the state courts’ role in the enforcement of federal law and prevent disruption of state judicial proceedings. See Braden v. 30th Judicial Circuit Court of Kentucky, 410 U. S. 484, 490-491 (1973).12 Under our federal system, the federal and state “courts [are] equally bound to guard and protect rights secured by the Constitution.” Ex parte Royall, 117 U. S., at 251. Because “it would be unseemly in our dual system of government for a federal district court to upset a state court conviction without an opportunity to the state courts to correct a constitutional violation,” federal courts apply the doctrine of comity, which “teaches that one court should defer action on causes properly within its jurisdiction until the courts of another sovereignty with concurrent powers, and already cognizant of the litigation, have had' an opportunity to pass upon the matter.” Darr v. Burford, 339 U. S. 200, 204 (1950). See Duckworth v. Serrano, 454 U. S. 1, 2 (1981) (per curiam) (noting that the exhaustion requirement “serves to minimize friction between our federal and state systems of justice by allowing the State an initial opportunity to pass upon and correct alleged violations of prisoners’ federal rights”).

A rigorously enforced total exhaustion rule will encourage state prisoners to seek full relief first from the state courts, thus giving those courts the first opportunity to review all *519claims of constitutional error. As the number of prisoners who exhaust all of their federal claims increases, state courts may become increasingly familiar with and hospitable toward federal constitutional issues. See Braden v. 30th Judicial Circuit Court of Kentucky, supra, at 490. Equally as important, federal claims that have been fully exhausted in state courts will more often be accompanied by a complete factual record to aid the federal courts in their review. Cf. 28 U. S. C. § 2254(d) (requiring a federal court reviewing a ha-beas petition to presume as correct factual findings made by a state court).

The facts of the present case underscore the need for a rule encouraging exhaustion of all federal claims. In his opinion, the District Court Judge wrote that “there is such mixture of violations that one cannot be separated from and considered independently of the others.” Because the two unexhausted claims for relief were intertwined with the exhausted ones, the judge apparently considered all of the claims in ruling on the petition. Requiring dismissal of petitions containing both exhausted and unexhausted claims will relieve the district courts of the difficult if not impossible task of deciding when claims are related, and will reduce the temptation to consider unexhausted claims.

In his dissent, Justice Stevens suggests that the District Court properly evaluated the respondent’s two exhausted claims “in the context of the entire trial.” Post, at 541. Unquestionably, however, the District Court erred in considering unexhausted claims, for § 2254(b) expressly requires the prisoner to exhaust “the remedies available in the courts of the State.” See n. 9, supra. Moreover, to the extent that exhausted and unexhausted claims are interrelated, the general rule among the Courts of Appeals is to dismiss mixed ha-beas petitions for exhaustion of all such claims. See, e. g., Triplett v. Wyrick, 549 F. 2d 57 (CA8 1977); Miller v. Hall, 536 F. 2d 967 (CA1 1976); Hewett v. North Carolina, 415 F. 2d 1316 (CA4 1969).

Rather than an “adventure in unnecessary lawmaking” (Stevens, J., post, at 539), our holdings today reflect our in*520terpretation of a federal statute on the basis of its language and legislative history, and consistent with its underlying policies. There is no basis to believe that today’s holdings will “complicate and delay” the resolution of habeas petitions (Stevens, J., post, at 550), or will serve to “trap the unwary pro se prisoner.” (Blackmun, J., post, at 530.) On the contrary, our interpretation of §§ 2254(b), (c) provides a simple and clear instruction to potential litigants: before you bring any claims to federal court, be sure that you first have taken each one to state court. Just as pro se petitioners have managed to use the federal habeas machinery, so too should they be able to master this straightforward exhaustion requirement. Those prisoners who misunderstand this requirement and submit mixed petitions nevertheless are entitled to resubmit a petition with only exhausted claims or to exhaust the remainder of their claims.

Rather than increasing the burden on federal courts, strict enforcement of the exhaustion requirement will encourage habeas petitioners to exhaust all of their claims in state court and to present the federal court with a single habeas petition. To the extent that the exhaustion requirement reduces piecemeal litigation, both the courts and the prisoners should benefit, for as a result the district court will be more likely to review all of the prisoner’s claims in a single proceeding, thus providing for a more focused ánd thorough review.

C

The prisoner’s principal interest, of course, is in obtaining speedy federal relief on his claims. See Braden v. 30th Judicial Circuit Court of Kentucky, supra, at 490. A total exhaustion rule will not impair that interest since he can always amend the petition to delete the unexhausted claims, rather than returning to state court to exhaust all of his claims. By invoking this procedure, however, the prisoner would risk forfeiting consideration of his unexhausted claims in federal court. Under 28 U. S. C. §2254 Rule 9(b), a district court *521may dismiss subsequent petitions if it finds that “the failure of the petitioner to assert those [new] grounds in a prior petition constituted an abuse of the writ.” See n. 6, supra. The Advisory Committee to the Rules notes that Rule 9(b) incorporates the judge-made principle governing the abuse of the writ set forth in Sanders v. United States, 373 U. S. 1, 18 (1963), where this Court stated:

“[I]f a prisoner deliberately withholds one of two grounds for federal collateral relief at the time of filing his first application, in the hope of being granted two hearings rather than one or for some other such reason, he may be deemed to have waived his right to a hearing on a second application presenting the withheld ground. The same may be true if, as in Wong Doo, the prisoner deliberately abandons one of his grounds at the first hearing. Nothing in the traditions of habeas corpus requires the federal courts to tolerate needless piecemeal litigation, or to entertain collateral proceedings whose only purpose is to vex, harass, or delay.”13

See Advisory Committee Note to Habeas Corpus Rule 9(b), 28 U. S. C., p. 273. Thus a prisoner who decides to proceed only with his exhausted claims and deliberately sets aside his unexhausted claims risks dismissal of subsequent federal petitions.

*522IV

In sum, because a total exhaustion rule promotes comity and does not unreasonably impair the prisoner’s right to relief, we hold that a district court must dismiss habeas petitions containing both unexhausted and exhausted claims.14 Accordingly, the judgment of the Court of Appeals is reversed, and the case is remanded for proceedings consistent with this opinion.

It is so ordered.

concurring in the judgment.

The important issue before the Court in this case is whether the conservative “total exhaustion” rule espoused now by two Courts of Appeals, the Fifth and the Ninth Circuits, see ante, at 513, n. 5, is required by 28 U. S. C. §§ 2254(b) and (c), or whether the approach adopted by eight other Courts of Appeals — that a district court may review the exhausted claims of a mixed petition — is the proper interpretation of the statute. On this basic issue, I firmly agree with the majority of the Courts of Appeals.

I do not dispute the value of comity when it is applicable and productive of harmony between state and federal courts, nor do I deny the principle of exhaustion that §12254(b) and (c) so clearly embrace. What troubles me is that the “total. exhaustion” rule, now adopted by this Court, can be read ipto the statute, as the Court concedes, ante, at 516-517, only by sheer force; that it operates as a trap for the uneducated and indigent pro se prisoner-applicant; that it delays the resolution of claims that are not frivolous; and that it tends to increase, rather than to alleviate, the caseload burdens on both state and federal courts. To use the old expression, the Court’s ruling seems to me to “throw the baby out with the bath water.”

*523Although purporting to rely on the policies upon which the exhaustion requirement is based, the Court uses that doctrine as “a blunderbuss to shatter the attempt at litigation of constitutional claims without regard to the purposes that underlie the doctrine and that called it into existence.” Braden v. 30th Judicial Circuit Court of Ky., 410 U. S. 484, 490 (1973). Those purposes do not require the result the Court reaches; in fact, they support the approach taken by the Court of Appeals in this case and call for dismissal of only the unexhausted claims of a mixed habeas petition. Moreover, to the extent that the Court’s ruling today has any impact whatsoever on the workings of federal habeas, it will alter, I fear, the litigation techniques of very few habeas petitioners.

I

A

The Court correctly observes, ante, at 516-517, that neither the language nor the legislative history of the exhaustion provisions of §§ 2254(b) and (e) mandates dismissal of a habeas petition containing both exhausted and unexhausted claims. Nor does precedent dictate the result reached here. In Picard v. Connor, 404 U. S. 270 (1971), for example, the Court ruled that “once the federal claim has been fairly presented to the state courts, the exhaustion requirement is satisfied.” Id., at 275 (emphasis supplied). .Respondent complied with the direction in Picard with respect to his challenges to the trial court’s limitation of cross-examination of the victim and to at least some of the prosecutor’s allegedly improper comments.

The Court fails to note, moreover, that prisoners are not compelled to utilize every available state procedure in order to satisfy the exhaustion requirement. Although this Court’s precedents do not address specifically the appropriate treatment of mixed habeas petitions, they plainly suggest that state courts need not inevitably be given every opportunity to safeguard a prisoner’s constitutional rights and to pro*524vide him relief before a federal court may entertain his ha-beas petition.1

B

In reversing the judgment of the Sixth Circuit, the Court focuses, as it must, on the purposes the exhaustion doctrine is intended to serve. I do not dispute the importance of the exhaustion requirement or the validity of the policies on which it is based. But I cannot agree that those concerns will be sacrificed by permitting district courts to consider exhausted habeas claims.

The first interest relied on by the Court involves an offshoot of the doctrine of federal-state comity. ■ The Court hopes to preserve the state courts’ role in protecting constitutional rights, as well as to afford those courts an opportunity to correct constitutional errors and — somewhat patronizingly — to “become increasingly familiar with and hospitable toward federal constitutional issues.” Ante, at 519. My proposal, however, is not inconsistent with the Court’s concern for comity: indeed, the state courts have occasion to rule first on every constitutional challenge, and *525have ample opportunity to correct any such error, before it is considered by a federal court on habeas.

In some respects, the Court’s ruling appears more destructive than solicitous of federal-state comity. Remitting a ha-beas petitioner to state court to exhaust a patently frivolous claim before the federal court may consider a serious, exhausted ground for relief hardly demonstrates respect for the state courts. The state judiciary’s time and resources are then spent rejecting the obviously meritless unexhausted claim, which doubtless will receive little or no attention in the subsequent federal proceeding that focuses on the substantial exhausted claim. I can “conceive of no reason why the State would wish to burden its judicial calendar with a narrow issue the resolution of which is predetermined by established federal principles.” Roberts v. LaVallee, 389 U. S. 40, 43 (1967).2

The second set of interests relied upon by the Court involves those of federal judicial administration — ensuring that a § 2254 petition is accompanied by a complete factual record to facilitate review and relieving the district courts of the responsibility for determining when exhausted and unex-*526hausted claims are interrelated. If a prisoner has presented a particular challenge in the state courts, however, the ha-beas court will have before it the complete factual record relating to that claim.3 And the Court’s Draconian approach is hardly necessary to relieve district courts of the obligation to consider exhausted grounds for relief when the prisoner also has advanced interrelated claims not yet reviewed by the state courts. When the district court believes, on the facts of the case before it, that the record is inadequate or that full consideration of the exhausted claims is impossible, it has always been free to dismiss the entire habeas petition pending resolution of unexhausted claims in the state courts. Certainly, it makes sense to commit these decisions to the discretion of the lower federal courts, which will be familiar with the specific factual context of each case.

The federal courts that have addressed the issue of interrelatedness have had no difficulty distinguishing related from unrelated habeas claims. Mixed habeas petitions have been dismissed in toto when “the issues before the federal court logically depend for their relevance upon resolution of an un-exhausted issue,” Miller v. Hall, 536 F. 2d 967, 969 (CA1 1976), or when consideration of the exhausted claim “would necessarily be affected ...” by the unexhausted claim, United States ex rel. McBride v. Fay, 370 F. 2d 547, 548 (CA2 1966). Thus, some of the factors to be considered in determining whether a prisoner’s grounds for collateral relief are interrelated are whether the claims are based on the same constitutional right or factual issue, and whether they require an understanding of the totality of the circumstances and therefore necessitate examination of the entire record. Compare Johnson v. United States District Court, 519 F. 2d 738, 740 (CA8 1975) (prisoner’s challenge to the voluntariness of his guilty plea intertwined with his claims that at the time *527of the plea he was mentally incompetent and without effective assistance of counsel); United States ex rel. DeFlumer v. Mancusi, 380 F. 2d 1018, 1019 (CA2 1967) (dispute regarding the voluntariness of the prisoner’s guilty plea “would necessarily affect the consideration of the coerced confession claim, because a voluntary guilty plea entered on advice of counsel is a waiver of all non-jurisdictional defects in any prior stage of the proceedings”); United States ex rel. McBride v. Fay, 370 F. 2d, at 548; and United States ex rel. Martin v. McMann, 348 F. 2d 896, 898 (CA2 1965) (defendant’s challenge to the voluntariness of his confession related to his claim that the confession was obtained in violation of his right to the assistance of counsel and without adequate warnings), with Miller v. Hall, 536 F. 2d, at 969 (no problem of interrelationship when exhausted claims involved allegations that the police lacked probable cause to search defendant’s van and had no justification for failing to secure a search warrant, and unexhausted claim maintained that the arresting officer had committed perjury at the suppression hearing); and United States ex rel. Levy v. McMann, 394 F. 2d 402, 404 (CA2 1968).

The Court’s interest in efficient administration of the federal courts therefore does not require dismissal of mixed ha-beas petitions. In fact, that concern militates against the approach taken by the Court today. In order to comply with the Court’s ruling, a federal court now will have to review the record in a § 2254 proceeding at least summarily in order to determine whether all claims have been exhausted. In many cases a decision on the merits will involve only negligible additional effort. And in other cases the court may not realize that one of a number of claims is unexhausted until after substantial work has been done. If the district court must nevertheless dismiss the entire petition until all grounds for relief have been exhausted, the prisoner will likely return to federal court eventually, thereby necessitating duplicative examination of the record and consideration of the exhausted *528claims — perhaps by another district judge. See Justice Stevens’ dissenting opinion, post, at 545. Moreover, when the § 2254 petition does find its way back to federal court, the record on the exhausted grounds for relief may well be stale and resolution of the merits more difficult.4

The interest of the prisoner and of society in “preserving] the writ of habeas corpus as a ‘swift and imperative remedy in all cases of illegal restraint or confinement,’” Braden v. 30th Judicial Circuit Court of Ky., 410 U. S., at 490, is the final policy consideration to be weighed in the balance. Compelling the habeas petitioner to repeat his journey through the entire state and federal legal process before receiving a ruling on his exhausted claims obviously entails substantial delay.5 And if the prisoner must choose between undergoing that delay and forfeiting unexhausted claims, see ante, at 520-521, society is likewise forced to sacrifice either the swiftness of habeas or its availability to remedy all unconstitutional imprisonments.6 Dismissing only unexhausted *529grounds for habeas relief, while ruling on the merits of all unrelated exhausted claims, will diminish neither the promptness nor the efficacy of the remedy and, at the same time, will serve the state and federal interests described by the Court.7

II

The Court’s misguided approach appears to be premised on the specter of “the sophisticated litigious prisoner intent upon a strategy of piecemeal litigation . . . ,” whose aim is to have more than one day in court. Galtieri v. Wainwright, 582 F. 2d 348, 369 (CA5 1978) (en banc) (dissenting opinion). Even if it could be said that the Court’s view accurately reflects reality, its ruling today will not frustrate the Perry Masons of the prison populations. To avoid dismissal, they will simply include only exhausted claims in each of many successive habeas petitions. Those subsequent petitions may be dismissed, as Justice Brennan observes, only if the prisoner has “abused the writ” by deliberately choosing, for purposes of delay, not to include all his claims in one petition. See post, at 535-536 (opinion concurring in part and dissenting in part). And successive habeas petitions that meet the *530“abuse of the writ” standard have always been subject to dismissal, irrespective of the Court’s treatment of mixed petitions today. The Court’s ruling in this case therefore provides no additional incentive whatsoever to consolidate all grounds for relief in one § 2254 petition.

Instead of deterring the sophisticated habeas petitioner who understands, and wishes to circumvent, the rules of exhaustion, the Court’s ruling will serve to trap the unwary pro se prisoner who is not knowledgeable about the intricacies of the exhaustion doctrine and whose only aim is to secure a new trial or release from prison. He will consolidate all conceivable grounds for relief in an attempt to accelerate review and minimize costs. But, under the Court’s approach, if he unwittingly includes in a § 2254 motion a claim not yet presented to the state courts, he risks dismissal of the entire petition and substantial delay before a ruling on the merits of his exhausted claims.

The Court suggests that a prisoner who files a mixed ha-beas petition will have the option of amending or resubmitting his complaint after deleting the unexhausted claims. See ante, at 510, 520. To the extent that prisoners are permitted simply to strike unexhausted claims from a § 2254 petition and then proceed as if those claims had never been presented, I fail to understand what all the fuss is about. In that event, the Court’s approach is virtually indistinguishable from that of the Court of Appeals, which directs thé district court itself to dismiss unexhausted grounds for relief.

I fear, however, that prisoners who mistakenly submit mixed petitions may not be treated uniformly. A prisoner’s opportunity to amend a §2254 petition may depend on his awareness of the existence of that alternative or on a sympathetic district judge who informs him of the option and permits the amendment. See Fed. Rule Civ. Proc. 15(a). If the prisoner is required to refile the petition after striking the unexhausted claims, he may have to begin the process anew and thus encounter substantial delay before his com*531plaint again comes to the district court’s attention. See Stevens, J., post, at 546, n. 15.

Adopting a rule that will afford knowledgeable prisoners more favorable treatment is, I believe, antithetical to the purposes of the habeas writ. Instead of requiring a habeas petitioner to be familiar with the nuances of the exhaustion doctrine and the process of amending a complaint, I would simply permit the district court to dismiss unexhausted grounds for relief and consider exhausted claims on the merits.

Ill

Although I would affirm the Court of Appeals’ ruling that the exhaustion doctrine requires dismissal of only the unex-hausted claims in a mixed habeas petition, I would remand the case for reconsideration of the merits of respondent’s constitutional arguments. As the Court notes, the District Court erred in considering both exhausted and unexhausted claims when ruling on Lundy’s § 2254 petition. See ante, at 511-513. The Court of Appeals attempted to recharacterize the District Court’s grant of relief as premised on only the exhausted claims and ignored the District Court’s conclusion that the exhausted and unexhausted claims were interrelated. See App. 95-96.8

Even were the Court of Appeals’ recharacterization accurate, that court affirmed the District Court on the ground that respondent’s constitutional rights had been “seriously impaired by the improper limitation of his counsel’s cross-examination of the prosecutrix and by the prosecutorial mis*532conduct.” Id., at 96. The court does not appear to have specified which allegations of prosecutorial misconduct it considered in reaching this conclusion, and the record does not reflect whether the court improperly took into account instances of purported misconduct that respondent has never challenged in state court. See ante, at 511-512, n. 3. This ambiguity is of some importance because the court’s general statement does not indicate whether the court would have granted habeas relief on the confrontation claim alone, or whether its judgment is based on the combined effect of the limitation of cross-examination and the asserted prose-cutorial misconduct.

I therefore would remand the case, directing that the courts below dismiss respondent’s unexhausted claims and examine those that have been properly presented to the state courts in order to determine whether they are interrelated with the unexhausted grounds and, if not, whether they warrant collateral relief.

concurring in part and dissenting in part.

I join the opinion of the. Court (Parts I, II, III-A, III — B, and IV, ante), but I do not join in the opinion of the plurality (Part III — C, ante). I agree with the Court’s holding that the exhaustion requirement of 28 U. S. C. §§ 2254(b), (c) obliges a federal district court to dismiss, without consideration on the merits, a habeas corpus petition from a state prisoner when that petition contains claims that have not been exhausted in the state courts, “leaving the prisoner with the choice of returning to state court to exhaust his claims or of amending or resubmitting the habeas petition to present only exhausted claims to the district court.” Ante, at 510. But I disagree with the plurality’s view, in Part III — C, that a ha-beas petitioner must “risk forfeiting consideration of his un-exhausted claims in federal court” if he “decides to proceed only with his exhausted claims and deliberately sets aside his *533unexhausted claims” in the face of the district court’s refusal to consider his “mixed” petition. Ante, at 520, 521. The issue of Rule 9(b)’s proper application to successive petitions brought as the result of our decision today is not before us — it was not among the questions presented by petitioner, nor was it briefed and argued by the parties. Therefore, the issue should not be addressed until we have a case presenting it. In any event, I disagree with the plurality’s proposed disposition of the issue. In my view, Rule 9(b) cannot be read to permit dismissal of a subsequent petition under the circumstances described in the plurality’s opinion.

I

The plurality recognizes, as it must, that in enacting Rule 9(b) Congress explicitly adopted the “abuse of the writ” standard announced in Sanders v. United States, 373 U. S. 1 (1963). Ante, at 521. The legislative history of Rule 9(b) illustrates the meaning of that standard. As transmitted by this Court to Congress, Rule 9(b) read as follows:

“Successive Petitions. A second or successive petition may be dismissed if the judge finds that it fails to allege new or different grounds for relief and the prior determination was on the merits or, if new and different grounds are alleged, the judge finds that the failure of the petitioner to assert those grounds in a prior petition is not excusable.” H. R. Rep. No. 94-1471, p. 8 (1976) (emphasis added).

The interpretive gloss placed upon proposed Rule 9(b) by this Court’s Advisory Committee on the- Rules Governing § 2254 Cases in the United States District Courts was that:

“With reference to a successive application asserting a new ground or one not previously decided on the merits, the court in Sanders noted:

[‘]In either case, full consideration of the merits of the new application can be avoided only if there has *534been an abuse of the writ * * * and this the Government has the burden of pleading. * * *

□Thus, for example, if a prisoner deliberately withholds one of two grounds for federal collateral relief at the time of filing his first application, * * * he may be deemed to have waived his right to a hearing on a second application presenting the withheld ground. [’]

“373 U. S., at 17-18. .

“Subdivision (b) [of Rule 9] has incorporated this principle and requires that the judge find petitioner’s failure to have asserted the new grounds in the prior petition to be inexcusable.” Advisory Committee Note to Rule 9(b), 28 U. S. C., p. 273 (emphasis added).

But Congress did not believe that this Court’s transmitted language, and the Advisory Committee Note explaining it, went far enough in protecting a state prisoner’s right to gain habeas relief. In its Report on proposed Rule 9(b), the House Judiciary Committee stated that, in its view, “the ‘not excusable’ language [of the proposed Rule] created a new and undefined standard that gave a judge too broad a discretion to dismiss a second or successive petition.” H. R. Rep. No. 94-1471, supra, at 5 (emphasis added). The Judiciary Committee thus recommended that the words, “is not excusable,” be replaced by the words, “constituted an abuse of the writ.” Id., at 5, 8. This change, the Committee believed, would bring Rule 9(b) “into conformity with existing law.” Id., at 5. It was in the Judiciary Committee’s revised form — employing the “abusive” standard for dismissal — that Rule 9(b) became law.

II

It is plain that a proper construction of Rule 9(b) must be consistent with its legislative history. This necessarily entails an accurate interpretation of the Sanders standard, on which the Rule is based. It also requires consideration of *535the explanatory language of the Advisory Committee, and Congress’ subsequent strengthening amendment to the text of the Rule. But the plurality, entirely misreading Sanders, embraces an interpretation of the Rule 9(b) standard that is manifestly incorrect, and patently inconsistent with the Advisory Committee’s exposition and Congress’ expressed expectations.

The relevant language from Sanders, quoted by the plurality, ante, at 521, is as follows:

“[I]f a prisoner deliberately withholds one of two grounds for federal collateral relief at the time of filing his first application, in the hope of being granted two hearings rather than one or for some other such reason, he may be deemed to have waived his right to a hearing on a second application presenting the withheld ground. The same may be true if, as in Wong Doo, the prisoner deliberately abandons one of his grounds at the first hearing. Nothing in the traditions of habeas corpus requires the federal courts to tolerate needless piecemeal litigation, or to entertain collateral proceedings whose only purpose is to vex, harass, or delay.” 373 U. S., at 18.

From this language the plurality concludes: “Thus a prisoner who decides to proceed only with his exhausted claims and deliberately sets aside his unexhausted claims risks dismissal of subsequent federal petitions.” Ante, at 521.

The plurality’s conclusion simply distorts the meaning of the quoted language. Sanders was plainly concerned with “a prisoner deliberately withholding] one of two grounds” for relief “in the hope of being granted two hearings rather than one or for some other such reason.” Sanders also notes that waiver might be inferred where “the prisoner deliberately abandons one of his grounds at the first hearing.” Finally, Sanders states that dismissal is appropriate either when the court is faced with “needless piecemeal litigation” or with *536“collateral proceedings whose only 'purpose is to vex, harass, or delay.” Thus Sanders made it crystal clear that dismissal for “abuse of the writ” is only appropriate when a prisoner was free to include all of his claims in his first petition, but knowingly and deliberately chose not to do so in order to get more than “one bite at the apple.” The plurality’s interpretation obviously would allow dismissal in a much broader class of cases than Sanders permits.

This Court is free, of course, to overrule Sanders. But even that course would not support the plurality’s conclusion. For Congress incorporated the “judge-made” Sanders principle into positive law when it enacted Rule 9(b). That principle, as explained by the Advisory Committee’s Note, at least “requires that the [habeas] judge find petitioner’s failure to have asserted the new grounds in the prior petition to be inexcusable.” Indeed, Congress went beyond the Advisory Committee’s language, believing that the “inexcusable” standard made the dismissal of successive petitions too easy. Congress instead required the habeas court to find a successive petitioner’s behavior “abusive” before the drastic remedy of dismissal could be employed. That is how Congress understood the Sanders principle, and the plurality is simply not free to ignore that understanding, because it is now embedded in the statutory language of Rule 9(b).

HH ► — I

The plurality’s attempt to apply its interpretation of Sanders only reinforces my conclusion that the plurality has misread that case. The plurality hypothesizes a prisoner who presents a “mixed” habeas petition that is dismissed without any examination of its claims on the merits, and who, after his exhausted claims are rejected, presents a second petition containing the previously unexhausted claims. The plurality then equates the position of such a prisoner with that of the “abusive” habeas petitioner discussed in the Sanders passage. But in my view, the position of the plurality’s hypo*537thetical prisoner is obviously very different. If the habeas court refuses to entertain a “mixed” petition — as it must under the plurality’s view — then the prisoner’s “abandonment” of his unexhausted claims cannot in any meaningful sense be termed “deliberate,” as that term was used in Sanders. There can be no “abandonment” when the prisoner is not permitted to proceed with his unexhausted claims. If he is to gain “speedy federal relief on his claims” — to which he is entitled, as the Court recognizes with its citation to Braden, ante, at 520 — then the prisoner must proceed only with his exhausted claims. Thus the prisoner in such a case cannot be said to possess a “purpose to vex, harass, or delay,” nor any “hope of being granted two hearings rather than one.”

Moreover, the plurality’s suggested treatment of its hypothetical prisoner flatly contradicts the Rule 9(b) standard as explained by the Advisory Committee, and a fortiori contradicts that standard as strengthened and extended by Congress. After the prisoner’s first, “mixed” petition has been mandatorily dismissed without any scrutiny, after his exhausted claims have been rejected, and after he has then presented his previously unexhausted claims in a second petition, there is simply no way in which a habeas court could “find petitioner’s failure to have asserted the new grounds in the prior petition to be inexcusable.” On the contrary, petitioner’s failure to have asserted the “new,” previously unex-hausted claims in the prior petition could only be found to have been required by the habeas court itself, as a condition for its consideration of the exhausted claims. If the plurality’s interpretation of Rule 9(b) cannot satisfy the Advisory Committee’s “inexcusable” standard, then it falls even further short of the higher, “abusive” standard eventually adopted by Congress.

IV

I conclude that when a prisoner’s original, “mixed” habeas petition is dismissed without any examination of its claims on the merits, and when the prisoner later brings a second peti*538tion based on the previously unexhausted claims that had earlier been refused a hearing, then the remedy of dismissal for “abuse of the writ” cannot be employed against that second petition, absent unusual factual circumstances truly suggesting abuse. This conclusion is to my mind inescapably compelled not only by Sanders, but also by the Advisory Committee explanation of the Rule, and by Congress’ subsequent incorporation of the higher, “abusive” standard into the Rule. The plurality’s conclusion, in contrast, has no support whatever from any of these sources. Nor, of course, does it have the support of a majority of the Court.*

concurring in part and dissenting in part.

I agree with most of Justice Brennan’s opinion; but like Justice Blackmun, I. would not require a “mixed” petition to be dismissed in its entirety, with leave'to resubmit the exhausted claims. The trial judge cannot rule on the unex-hausted issues and should dismiss them. . But. he should rule on the exhausted claims unless they are intertwined with those he must dismiss or unless the habeas petitioner prefers to have his entire petition dismissed. In any event, if the judge' rules on those issues that are ripe and dismisses those that are not, .1 would not tax the petitioner with abuse of the writ if he returns with the latter claims after seeking state relief.

dissenting.

This case raises important questions about the authority of federal judges. In my opinion the District Judge properly exercised his statutory duty to consider the merits of the claims advanced by respondent that previously had been rejected by the Tennessee courts. The District Judge ex*539ceeded, however, what I regard as proper restraints on the scope of collateral review of state-court judgments. Ironically, instead of correcting his error, the Court today fashions a new rule of law that will merely delay the final disposition of this case and, as Justice Blackmun demonstrates, impose unnecessary burdens on both state and federal judges.

An adequate explanation of my disapproval of the Court’s adventure in unnecessary lawmaking requires some reference to the facts of this case and to my conception of the proper role of the writ of habeas corpus in the administration of justice in the United States.

I

Respondent was convicted in state court of rape and a crime against nature. The testimony of the victim was corroborated by another eyewitness who was present during the entire sadistic episode. The evidence of guilt is not merely sufficient; it is convincing. As is often the case in emotional, controverted, adversary proceedings,.trial error occurred. Two of those errors—a remark by the prosecutor1 and a limitation on defense counsel’s cross-examination *540of the victim2 — were recognized by the Tennessee Court of Criminal Appeals, but held to be harmless in the context of the entire case. Because the state appellate court considered and rejected these two errors as a basis for setting aside his conviction, respondent has exhausted his state remedies with respect to these two claims.

In his application in federal court for a writ of habeas corpus, respondent alleged that these trial errors violated his constitutional rights to confront the witnesses against him and to obtain a fair trial. In his petition, respondent also al*541leged that the prosecutor had impermissibly commented on his failure to testify3 and that the trial judge had improperly instructed the jury that “every witness is presumed to swear the truth.”4 Because these two additional claims had not been presented to the Tennessee Court of Criminal Appeals, the Federal District Judge concluded that he could “not consider them in the constitutional framework.” App. 88. He added, however, that “in assessing the atmosphere of the cause taken as a whole these items may be referred to collaterally.”5

In considering the significance of respondent’s two exhausted claims, the District Court thus evaluated them in the context of the entire trial record. That is precisely what the Tennessee Court of Criminal Appeals did in arriving at its conclusion that these claims, identified as error, were not sufficiently prejudicial to justify reversing the conviction and ordering a retrial.6 In considering whether the error in these two exhausted claims was sufficient to justify a grant of ha-*542beas corpus relief, the federal court — like the state court— had a duty to look at the context in which the error occurred to determine whether it was either aggravated or mitigated by other aspects of the proceeding.7 The state court and the federal court formed differing judgments based on that broad review. I happen to share the appraisal of the state court on the merits, but I believe that the procedure followed by the federal court was entirely correct.

The Court holds, however, that the District Court committed two procedural errors. “Unquestionably,” according to the Court, it was wrong for the District Court to consider the portions of the trial record described in the unexhausted claims in evaluating those claims that had been exhausted. Ante, at 519. More fundamentally, according to the Court, it was wrong for the District Court even to consider the merits of the exhausted claims because the prisoner had included unexhausted claims in his pleadings. Both of the Court’s holdings are unsatisfactory for the same basic reason: the Court assumes that the character of all claims alleged in ha-beas corpus petitions is the same. Under the Court’s analysis, any unexhausted claim asserted in a habeas corpus petition — no matter how frivolous — is sufficient to command the district judge to postpone relief on a meritorious exhausted claim, no matter how obvious and outrageous the constitutional violation may be.

*543In my opinion claims of constitutional error are not fungible. There are at least four types. The one most frequently encountered is a claim that attaches a constitutional label to a set of facts that does not disclose a violation of any constitutional right. In my opinion, each of the four claims asserted in this case falls in that category. The second class includes constitutional violations that are not of sufficient import in a particular case to justify reversal even on direct appeal, when the evidence is still fresh and a fair retrial could be promptly conducted. Chapman v. California, 386 U. S. 18, 22; Harrington v. California, 395 U. S. 250, 254. A third category includes errors that are important enough to require reversal on direct appeal but do not reveal the kind of fundamental unfairness to the accused that will support a collateral attack on a final judgment. See, e. g., Stone v. Powell, 428 U. S. 465.8 The fourth category includes those er*544rors that are so fundamental that they infect the validity of the underlying judgment itself, or the integrity of the process by which that judgment was obtained. This category cannot be defined precisely; concepts of “fundamental fairness” are not frozen in time. But the kind of error that falls in this category is best illustrated by recalling the classic grounds for the issuance of a writ of habeas corpus — that the proceeding was dominated by mob violence;9 that the prosecutor knowingly made use of perjured testimony;10 or that the conviction was based on a confession extorted from the defendant by brutal methods.11 Errors of this kind justify collateral relief no matter how long a judgment may have been final12 and even though they may not have been preserved properly in the original trial.13

*545In this case, I think it is clear that neither the exhausted claims nor the unexhausted claims describe any error demonstrating that respondent’s trial was fundamentally unfair. Since his lawyer found insufficient merit in the two unex-hausted claims to object to the error at trial or to raise the claims on direct appeal,141 would expect that the Tennessee courts will consider them to have been waived as a matter of state law; thereafter, under the teaching of cases such as Wainwright v. Sykes, 433 U. S. 72, they undoubtedly will not support federal relief. This case is thus destined to return to the Federal District Court and the Court of Appeals where, it is safe to predict, those courts will once again come to the conclusion that the writ should issue. The additional procedure that the Court requires before considering the merits will be totally unproductive.

If my appraisal of respondent’s exhausted claims is incorrect — if the trial actually was fundamentally unfair to the respondent — postponing relief until another round of review in the state and federal judicial systems has been completed is truly outrageous. The unnecessary delay will make it more difficult for the prosecutor to obtain a conviction on retrial if respondent is in fact guilty; if he is innocent, requiring him to languish in jail because he made a pleading error is callous indeed.

There are some situations in which a district judge should refuse to entertain a mixed petition until all of the prisoner’s claims have been exhausted. If the unexhausted claim appears to involve error of the most serious kind and if it is reasonably clear that the exhausted claims do not, addressing the merits of the exhausted claims will merely delay the ulti*546mate disposition of the case. Or if an evidentiary hearing is necessary to decide the merits of both the exhausted and un-exhausted claims, a procedure that enables all fact questions to be resolved in the same hearing should be followed. I therefore would allow district judges to exercise discretion to determine whether the presence of an unexhausted claim in a habeas corpus application makes it inappropriate to consider the merits of a properly pleaded exhausted claim. The inflexible, mechanical rule the Court adopts today arbitrarily denies district judges the kind of authority they need to administer their calendars effectively.15

W h — I

In recent years federal judges at times have lost sight of the true office of the great writ of habeas corpus. It is quite unlike the common-law writ of error that enabled a higher court to correct errors committed by a nisi prius tribunal in the trial of civil or criminal cases by ordering further proceedings whenever trial error was detected. The writ of ha-beas corpus is a fundamental guarantee of liberty.16

*547The fact that federal judges have at times construed their power to issue writs of habeas corpus as though it were tantamount to the authority of an appellate court considering a direct appeal from a trial court judgment has had two unfortunate consequences. First, it has encouraged prisoners to file an ever-increasing volume of federal applications that often amount to little more than a request for further review of asserted grounds for reversal that already have been adequately considered and rejected on direct review. Second, it has led this Court into the business of creating special procedural rules for dealing with this flood of litigation. The doctrine of nonretroactivity, the emerging “cause and prejudice” doctrine, and today’s “total exhaustion” rule are examples of judicial lawmaking that might well have been avoided by confining the availhbility of habeas corpus relief to cases that truly involve fundamental unfairness.

When that high standard is met, there should be no question about the retroactivity of the constitutional rule being enforced. Nor do I believe there is any need to fashion definitions of “cause” and “prejudice” to determine whether an error that was not preserved at trial or on direct appeal is subject to review in a collateral federal proceeding.17 The *548availability of habeas corpus relief should depend primarily on the character of the alleged constitutional violation and not on the procedural history underlying the claim.18

*549The “total exhaustion” rule the Court crafts today demeans the high office of the great writ. Perhaps a rule of this kind would be an appropriate response to a flood of litigation requesting review of minor disputes. An assumption that most of these petitions are groundless might be thought to justify technical pleading requirements that would provide a mechanism for reducing the sheer number of cases in which the merits must be considered. But the Court’s experience has taught us not only that most of these petitions lack merit, but also that there are cases in which serious injustice must *550be corrected by the issuance of the writ.19 In such cases, the statutory requirement that adequate state remedies be exhausted must, of course, be honored. When a person’s liberty is at stake, however, there surely is no justification for the creation of needless procedural hurdles.20

Procedural regularity is a matter of fundamental importance in the administration of justice. But procedural niceties that merely complicate and delay the resolution of disputes are another matter. In my opinion the federal habeas corpus statute should be construed to protect the former and, whenever possible, to avoid the latter.

I respectfully dissent.