dissenting.
Petitioner Kaufman was convicted of robbing a federally insured savings and loan association while armed with a pistol. Part of the evidence used against him was a revolver, some of the stolen traveler’s checks, a money-order receipt, a traffic summons, and gasoline receipts. During the trial petitioner’s counsel conceded that petitioner had committed the robbery but contended he was not responsible for the crime because he was mentally ill at the time. An appeal from his conviction was rejected by the Court of Appeals, 350 F. 2d 408 (C. A. 8th Cir. 1965), and we denied certiorari, 383 U. S. 951 (1966). Three months later — after the *232decision had become what is generally considered “final” — he filed in the Federal District Court the present motion under 28 U. S. C. § 2255, asking that his sentence be vacated on the ground, among others, that the trial court had committed error in not suppressing the evidence against him because the articles had been obtained by an unlawful search and seizure. Despite the fact that he has never, either in his trial or in this proceeding, asserted that he had not actually physically committed the robbery with a pistol, and despite the fact that this plainly reliable evidence clearly shows, along with the other evidence at trial, that he was not insane, the Court is reversing his case, holding that he can collaterally attack the judgment after it had become final. I dissent.
My dissent rests on my belief that not every conviction based in part on a denial of a constitutional right is subject to attack by habeas corpus or § 2255 proceedings after a conviction has become final. This conclusion is supported by the language of § 2255 which clearly suggests that not every constitutional claim is intended to be a basis for collateral relief.1 And, as this Court has said in Fay v. Noia, with reference to habeas corpus,
“Discretion is implicit in the statutory command that the judge . . . 'dispose of the matter as law and justice require,' 28 U. S. C. § 2243 . . . .” 372 U. S. 391, 438.
Of course one important factor that would relate to whether the conviction should be vulnerable to collateral *233attack is the possibility of the applicant’s innocence. For illustration, few would think that justice requires release of a person whose allegations clearly show that he was guilty of the crime of which he had been convicted.
I agree with the Court’s conclusion that the scope of collateral attack is substantially the same in federal habeas corpus cases which involve challenges to state convictions, as it is in § 2255 cases which involve challenges to federal convictions. The crucial question, however, is whether certain types of claims, such as a claim to keep out relevant and trustworthy evidence because the result of an unconstitutional search and seizure, should normally be open in these collateral proceedings. This question was fully and carefully considered by the Court of Appeals for the District of Columbia Circuit in Thornton v. United States, 125 U. S. App. D. C. 114, 368 F. 2d 822 (1966), and I agree substantially with the opinion of Judge Leventhal for the majority of that court, which states:2
“[Generally a claim by a federal prisoner that evidence admitted at his trial was the fruit of an unconstitutional search or seizure is not properly the ground of a collateral attack on his conviction. As further noted below, this rule is subject to an exception for special circumstances ....
“Many opinions declare that collateral attack, as by habeas corpus, is available to correct the denial of a constitutional right. This is the general rule but it is not an absolute. . . .
“The courts are called on to evolve and provide procedures and remedies that are effective to vindicate constitutional rights. However, where effective procedures are available in the direct proceeding, there *234is no imperative to provide an additional, collateral review, leaving no stone unturned, when exploration of all avenues of justice at the behest of individual petitioners may impair judicial administration of the federal courts, as by making criminal litigation interminable, and diverting resources of the federal judiciary.”
It was formerly the rule in this country that judgments were so impervious to collateral attack that a defendant could not collaterally attack his conviction even after the Government had admitted his innocence. That rule, obviously a harsh and what might seem to most people an indefensible one, has of course now been abandoned. It was finally put to rest in Fay v. Noia, 372 U. S. 391 (1963). It is this element of probable or possible innocence that I think should be given weight in determining whether a judgment after conviction and appeal and affirmance should be open to collateral attack, for the great historic role of the writ of habeas corpus has been to insure the reliability of the guilt-determining process.3 In Fay v. Noia, Noia and his two codefendants had been convicted of felony murder in New York state court and each had been sentenced to life imprisonment. The sole evidence against each defendant was his confession. While his codefendants appealed, Noia did not, for fear that if he secured a reversal and was reconvicted at a second trial, he would be sentenced to death. The confessions of one codefendant were subsequently found by the Court of Appeals for the Second Circuit to have been coerced by practices which that court described as “satanic.” United States ex rel. Caminito v. Murphy, 222 F. 2d 698, *235701 (1955). The conviction of Noia’s other codefendant was also reversed, People v. Bonino, 1 N. Y. 2d 752, 135 N. E. 2d 51, and since there was no evidence other than the coerced confessions that they were guilty, the State apparently declined to reprosecute them and they were set free. See United States ex rel. Noia v. Fay, 183 F. Supp. 222, 227, n. 6 (1960). Noia, however, languished in prison, even though his alleged confederates had been released. Both the New York courts and the Federal District Court declined to review his case on the ground that his 1942 failure to appeal made his conviction “final.” He remained in jail despite the fact that he “continuously asserted his innocence of the crime for which he [had] been convicted” in his petition for habeas corpus and elsewhere. See Transcript of Record in No. 84, October Term, 1962, p. 8.
It was under these circumstances, strongly appealing to the Court’s sense of what justice required, that this Court held that Noia was entitled to challenge his conviction even though it had previously become “final.” My Brother Harlan, dissenting, concluded that no matter how appealing the circumstances, one wrongly convicted must be consigned to the slow, tedious, and uncertain road to whatever relief he might possibly get from the Chief Executive. On the contrary, I agreed with Fay v. Noia as one of the bright landmarks in the administration of criminal justice. But I did not think then and do not think now that it laid down an inflexible rule compelling the courts to release every prisoner who alleges in collateral proceedings some constitutional flaw, regardless of its nature, regardless of his guilt or innocence, and regardless of the circumstances of the case. The Court’s opinion in Noia shows, from beginning to end, that the defendant’s guilt or innocence is at least one of the vital considerations in determining whether collateral relief should be available to a convicted de*236fendant. The Court repeatedly emphasized that the only evidence against Noia was a coerced confession and that he remained in jail while the State permitted his alleged confederates to go free. The Court made it clear that equitable considerations such as these should play a part in determining the availability of federal habeas corpus:
“Although we hold that the jurisdiction of the federal courts on habeas corpus is not affected by procedural defaults incurred by the applicant during the state court proceedings, we recognize a limited discretion in the federal judge to deny relief to an applicant under certain circumstances. Discretion is implicit in the statutory command that the judge, after granting the writ and holding a hearing of appropriate scope, 'dispose of the matter as law and justice require/ 28 U. S. C. §2243; and discretion was the flexible concept employed by the federal courts in developing the exhaustion rule. Furthermore, habeas corpus has traditionally been regarded as governed by equitable principles.” 4
And in its closing paragraph, the Court stressed:
“Today as always few indeed is the number of state prisoners who eventually win their freedom by means of federal habeas corpus. Those few who are ultimately successful are persons whom society has grievously wronged and for whom belated liberation is little enough compensation.” 5
Surely, it cannot be said of Kaufman, an admitted armed robber, that he is a person whom “society has grievously wronged and for whom belated liberation is little enough compensation.”
*237Although, as the Court of Appeals indicated in the Thornton case, habeas corpus has been thought of broadly as a means of securing redress for the violation of any “constitutional right,” it was true until Mapp v. Ohio, 367 U. S. 643 (1961), that almost every “constitutional right” referred to in this sense played a central role in assuring that the trial would be a reliable means of testing guilt. It is true that the prohibition against coerced confessions has been vigorously enforced even in the absence of proof that the confession itself was unreliable, e. g., Rogers v. Richmond, 365 U. S. 534 (1961), but even this prohibition rests to a substantial extent on recognition that all such confessions “may be and have been, to an unascertained extent, found to be untrustworthy,” id., at 541.
A claim of illegal search and seizure under the Fourth Amendment is crucially different from many other constitutional rights; ordinarily the evidence seized can in no way have been rendered untrustworthy by the means of its seizure and indeed often this evidence alone establishes beyond virtually any shadow of a doubt that the defendant is guilty. A good example of such a case is one in which I filed a dissent today, Harris v. Nelson, post, p. 286. The prisoner in Harris was convicted on a charge that he had been in possession of marihuana, possession alone being a crime under state law. He later collaterally attacked that conviction, alleging that the marihuana had been unlawfully seized from his home, where he had been in illegal possession of it. He did not and evidently could not allege a single fact that would indicate the slightest possibility that he actually was innocent of the crime charged. Under these circumstances it implies no disrespect for the importance of the Fourth Amendment to recognize the simple proposition that treatment of search-and-seizure claims should *238correspond to the purpose of the Fourth Amendment exclusionary rule. As the Court of Appeals said in Thornton:6
“Our rejection of the availability of collateral review for claims of unreasonable search and seizure (in the absence of exceptional circumstances) is not attributable to a low regard for the significance of the Fourth Amendment in our times and civilization. On the contrary, the magnitude of the Fourth Amendment in our constitutional constellation has prompted unusual remedies by Congress, as well as the courts. . . .
“. . . The corollary, however, is a contraction of the need for enlarging collateral review in order to assure effective vindication of the constitutional interests involved.”
The purpose of the exclusionary rule, unlike most provisions of the Bill of Rights, does not include, even to the slightest degree, the goal of insuring that the guilt-determining process be reliable. Rather, as this Court has said time and again, the rule has one primary and overriding purpose, the deterrence of unconstitutional searches and seizures by the police. As the Court said in Linkletter v. Walker, 381 U. S. 618, 636-637 (1965):
“Mapp had as its prime purpose the enforcement of the Fourth Amendment through the inclusion of the exclusionary rule within its rights. This, it was found, was the only effective deterrent to lawless police action. Indeed, all of the cases since Wolf [v. Colorado, 338 U. S. 25 (1949)] requiring the exclusion of illegal evidence have been based on the necessity for an effective deterrent to illegal police action.”
How this purpose can be served by the broad and unqualified rule adopted by the Court today is something *239of a mystery. Of course, the shortcomings inherent in any human system make it impossible to eliminate entirely all the incentives to conduct an illegal search. It would seem rather fanciful, however, to suggest that these inevitable incentives would be decreased to any significant extent by the fact that if a conviction is obtained, after adequate opportunities have been provided to litigate constitutional claims, and if this conviction is upheld by all the reviewing courts, the validity of the search and seizure may later be questioned in a collateral proceeding. Understandably, the Court does not make any such suggestion and indeed makes no effort to justify its result in terms of the long-recognized deterrent purpose of the exclusionary rule. The Court instead simply provides us with a string of citations that supposedly settle the question, at least as to state convictions, ante, at 225, but the Court neglects to mention that not one of the cases it cites contains a single intimation that the issue before us now was even considered.7
The only other justification for the Court’s result that can be gleaned from its opinion is the statement that the reasoning of the Court of Appeals in Thornton “exalts the value of finality in criminal judgments at the expense of the interest of each prisoner in the vindication of his constitutional rights.” Ante, at 228. This astonishing statement is directly contrary to the principles this Court has consistently applied on this subject, as for example in Elkins v. United States, 364 U. S. 206, 217 (1960), where we said: “The [exclusionary] rule is calculated to prevent, not to repair. Its purpose is to deter— *240to compel respect for the constitutional guaranty in the only effectively available way — by removing the incentive to disregard it.” This same recognition that no personal right of the prisoner can be vindicated in these Fourth Amendment cases was stressed in this Court’s opinion in Linkletter, supra: “We cannot say that this [deterrent] purpose would be advanced by making the rule retrospective. The misconduct of the police prior to Mapp has already occurred and will not be corrected by releasing the prisoners involved. . . . Finally, the ruptured privacy of the victims’ homes and effects cannot be restored. Reparation comes too late.” 381 U. S., at 637.
The Court’s consistent adherence to this approach has continued through all of the various “retroactivity” cases, including today’s decision in Desist v. United States, post, p. 244, in which the Court emphasizes, quoting from Linkletter, that “ ‘[t]he misconduct of the police . . . has already occurred and will not be corrected by releasing the prisoners involved,’ ” and that “the exclusionary rule is but a ‘procedural weapon that has no bearing on guilt.’ ” It would be hard to find a more apt summary of this Court’s holdings in these “retro-activity” cases than the statement that they “exal[t] the value of finality in criminal judgments at the expense of the interest of each prisoner in the vindication of his constitutional rights.” But since this is the course the Court has chosen to steer, I think it would not be amiss to suggest that the Court at least decide this case on the same principles and seek to achieve a modest semblance of consistency. Instead the Court adopts a rule that offers no discernible benefits in enforcing the Fourth Amendment and insures that prisoners who are undoubtedly guilty will be set free.
It is seemingly becoming more and more difficult to gain acceptance for the proposition that punishment, of *241the guilty is desirable, other things being equal. One commentator, who attempted in vain to dissuade this Court from today’s holding, thought it necessary to point out that there is “a strong public interest in convicting the guilty.”8 Indeed the day may soon come when the ever-cautious law reviews will actually be forced to offer the timid and uncertain contention, recently suggested satirically, that “crime may be thought socially undesirable, and its control a 'valid governmental objective’ to which the criminal law is 'rationally related.’ ”9
I cannot agree to a rule, however technical it may seem, that leads to these results. I would not let any criminal conviction become invulnerable to collateral attack where there is left remaining the probability or possibility that constitutional commands related to the integrity of the fact-finding process have been violated. In such situations society has failed to perform its obligation to prove beyond a reasonable doubt that the defendant committed the crime. But it is quite a different thing to permit collateral attack on a conviction after a trial according to due process when the defendant clearly is, by the proof and by his own admission, guilty of the crime charged. There may, of course, as the Court of Appeals held in the Thornton case, be some special circumstances in which allowance of a Fourth Amendment claim in a collateral proceeding would be justified in terms of the relevant and applicable constitutional principles. Some of the situations possibly falling in this category have been enumerated and examined by others,10 and there are circumstances alleged here that might lead *242to such a disposition of this case.11 But the Court does not rest its judgment on this narrow ground, and I therefore do not attempt to pass on it. I do contend, however, that the court below was right in refusing to follow the broad rule that this Court is announcing today. In collateral attacks whether by habeas corpus or by § 2255 proceedings, I would always require that the convicted defendant raise the kind of constitutional claim that casts some shadow of a doubt on his guilt. This defendant is permitted to attack his conviction collaterally although he conceded at the trial and does not now deny that he had robbed the savings and loan association and although the evidence makes absolutely clear that he knew what he was doing. Thus, his guilt being certain, surely he does not have a constitutional right to get a new trial. I cannot possibly agree with the Court.
“If the court finds that the judgment was rendered without jurisdiction, or that the sentence imposed was not authorized by law or otherwise open to collateral attack, or that there has been such a denial or infringement of the constitutional rights of the prisoner as to render the judgment vulnerable to collateral attack, the court shall vacate and set the judgment aside and shall discharge the prisoner or resentence him or grant a new trial or correct the sentence as may appear appropriate.” 28 U. S. C. §2255. (Emphasis supplied.)
125 U. S. App. D. C., at 116-118, 368 F. 2d, at 824-826.
See Mishkin, The Supreme Court, 1964 Term — Foreword: The High Court, The Great Writ, and the Due Process of Time and Law, 79 Harv. L. Rev. 56, 79-86 (1965).
372 U. S. 391, 438.
Id., at 440-441.
125 U. S. App. D. C., at 118, 368 F. 2d, at 826.
Only one of these decisions, Mancusi v. DeForte, 392 U. S. 364 (1968), actually ordered the granting of habeas relief on the basis of a search-and-seizure claim, and in Mancusi (as in Warden v. Hayden, 387 U. S. 294 (1967)) the issue was not even theoretically before us since only in the most exceptional case would we have considered a question not mentioned in the State Warden’s petition for a writ of certiorari.
Amsterdam, Search, Seizure, and Section 2255: A Comment, 112 U. Pa. L. Rev. 378, 389 (1964).
79 Harv. L. Rev. (parody ed.) 10, 12 (March 1966).
Thornton v. United States, supra; Amsterdam, supra, n. 8, at 391-392, n. 60.
Petitioner’s allegations suggest that he may have been unjustifiably frustrated in his efforts to raise the search-and-seizure issue on direct appeal from his conviction. See the Court’s opinion, ante, at 220, n. 3.