delivered the opinion of the Court.

In Edwards v. Arizona, 451 U. S. 477 (1981), this Court ruled that a criminal defendant’s rights under the Fifth and Fourteenth Amendments were violated by the use of his confession obtained by police-instigated interrogation — without counsel present — after he requested an attorney. This case presents the issue whether that ruling is applicable to a case pending on direct appeal in a state court at the time Edwards was decided.

I

There is no dispute as to the facts. Petitioner Kevin Michael Shea was charged in Louisiana with two counts of armed robbery.. He was arrested on July 2, 1979, and was taken to the police station at Shreveport. There he was turned over to Detectives Smith and Snell for questioning. His so-called Miranda rights, see Miranda v. Arizona, 384 U. S. 436 (1966), were read to him, and he signed a standard Miranda card. He said, however, that he did not wish to make any statement until he saw a lawyer. The interview thereupon was terminated.

The following afternoon, July 3, before petitioner had been in communication with any lawyer, Detective Snell returned. He informed petitioner that he was to be transferred from the city jail to the parish jail. Without inquiring of petitioner whether he had spoken with an attorney or whether he was indigent, and without any indication from petitioner that he now was willing to be interrogated, Snell asked if he wanted to talk about the case. Again, Miranda rights were read to petitioner and again he signed a Miranda card. He then orally confessed that he had committed the two robberies.

The charges against petitioner came on for trial in due course in the State District Court for Caddo Parish. At this point, the two counts were severed. Prior to his trial before a jury on the first count, petitioner formally moved to suppress the confession of July 3. App. 2. At the trial, which *53took place in 1980, the prosecution offered the confession in evidence. The defense objected, but the objection was overruled and the confession was admitted. Petitioner was convicted. He filed a like suppression motion with respect to the second charge. Id., at 6. When this was denied, he withdrew his original plea and entered a plea of guilty, with a reservation under state law, see State v. Crosby, 338 So. 2d 584, 588 (La. 1976), of his right to appeal the denial of the motion to suppress. App. 7-8.

Ón his appeal to the Supreme Court of Louisiana, petitioner raised the issue of the trial court’s error, in violation of Miranda, in admitting the confession. In its opinion, the Louisiana tribunal cited this Court’s decision in Edwards, which had come down in the meantime but subsequent to petitioner’s trial and convictions. The Louisiana court acknowledged the presence of an Edwards violation.1 It stated:

“In the present case it is undisputed that the police did initiate such an inquiry on July 3, after having been clearly informed by the defendant on the previous evening that he would not make any statements without counsel. Consequently, there was a violation of the additional standard governing police interrogation of a suspect imposed by Edwards v. Arizona . . . .” 421 So. 2d 200, 203 (1982).

The court, however, went on to hold that Edwards was not to be applied in petitioner’s case:

“As this [error] occurred before the decision in Edwards was rendered and we are convinced the United States *54Supreme Court will pronounce that decision is not retroactive, we so hold in this case.” 421 So. 2d, at 204.

Petitioner successfully obtained a rehearing on the ret-roactivity issue. On rehearing, although the Louisiana Supreme Court again acknowledged, id., at 210, that petitioner’s confession, under Edwards, was not admissible, that court adhered, over two dissents, to its position that Edwards was not to be given retroactive effect. It stated that that decision was a “clear break with the past,” was a new ruling, and was not retroactive. 421 So. 2d, at 210.

Because of the importance of the issue and because of conflicting decisions elsewhere,2 we granted certiorari. 466 U. S. 957 (1984).

II

Edwards, the case at the center of the present controversy, involved facts startlingly similar to those of the present case. Police officers informed Edwards of his Miranda rights and questioned him until he said he wanted an attorney. At that point questioning ceased. The next day, however, other officers visited Edwards, stated they wanted to talk to him, informed him of his Miranda rights, and obtained an oral confession. This Court was positive and clear in its ruling:

“[Although we have held that after initially being advised of his Miranda rights, the accused may himself *55validly waive his rights and respond to interrogation, . . . the Court has strongly indicated that additional safeguards are necessary when the accused asks for counsel; and we now hold that when an accused has invoked his right to have counsel present during custodial interrogation, a valid waiver of that right cannot be established by showing only that he responded to further police-initiated custodial interrogation even if he has been advised of his rights. We further hold that an accused, such as Edwards, having expressed his desire to deal with the police only through counsel, is not subject to further interrogation by the authorities until counsel has been made available to him, unless the accused himself initiates further communication, exchanges, or conversations with the police” (footnote omitted). 451 U. S., at 484-485.

See also Rhode Island v. Innis, 446 U. S. 291, 298 (1980); Fare v. Michael C., 442 U. S. 707, 719 (1979); Michigan v. Mosley, 423 U. S. 96, 104, n. 10 (1975), and id., at 109-111 (opinion concurring in result); Miranda v. Arizona, 384 U. S., at 444-445, 474.

The legal principle, thus, is established and is uncontested here. The only question before us in this case is whether that ruling applies retroactively with respect to petitioner’s convictions when the issue was raised and his case was pending and undecided on direct appeal in the state system at the time Edwards was decided.3

*56HH h-H I — I

Two of this Court’s recent cases bear importantly upon the issue. The first is United States v. Johnson, 457 U. S. 537 (1982). In that case, we held that a decision of this Court concerning Fourth Amendment rights was to be applied retroactively to all convictions that were not yet final at the time the decision was rendered, except in those situations that would be clearly controlled by existing retroactivity precedents to the contrary. . Specifically, the Court held that Payton v. New York, 445 U. S. 573 (1980), was to be applied retroactively to Johnson’s case.

The Court in Johnson found persuasive Justice Harlan’s earlier reasoning that application of a new rule of law to cases pending on direct review is necessary in order for the Court to avoid being in the position of a super-legislature, selecting one of several cases before it to use to announce the new rule and then letting all other similarly situated persons be passed by unaffected and unprotected by the new rule. See Desist v. United States, 394 U. S. 244, 256 (1969) (dissenting opin*57ion); Mackey v. United States, 401 U. S. 667, 675 (1971) (separate opinion). The Court noted that, at a minimum, “‘all “new” rules of constitutional law must ... be applied to all those cases which are still subject to direct review by this Court at the time the “new” decision is handed down.’” United States v. Johnson, 457 U. S., at 548, quoting from the dissent in Desist v. United States, 394 U. S., at 258. In Johnson the Court, “[t]o the extent necessáry to decide today’s case, . . . embrace[d] Justice Harlan’s views in Desist and Mackey.” 457 U. S., at 562. It thus determined that unless the rule is so clearly a break with the past that prior precedents mandate nonretroactivity, a new Fourth Amendment rule is to be applied to cases pending on direct review when the rule was adopted.

In considering the retroactivity of Payton, the Court then concluded that the question was to be resolved fairly by applying the Payton ruling to all cases pending on direct review when Payton was decided. So to do (a) would provide a principle of decisionmaking consonant with the Court’s original understanding in Linkletter v. Walker, 381 U. S. 618 (1965), and Tehan v. United States ex rel. Shott, 382 U. S. 406 (1966), (b) would comport with this Court’s judicial responsibility to do justice to each litigant on the merits of his own case, and (c) would further the goal of treating similarly situated defendants similarly.

The second case is Solem v. Stumes, 465 U. S. 638 (1984). It, too, clearly involved an obvious Edwards violation that took place in 1973, more than seven years before Edwards. After Stumes’ state-court conviction had been finally affirmed by the Supreme Court of South Dakota, he sought federal habeas relief. His petition for a writ, however, was denied by the Federal District Court. While Stumes’ appeal was pending in the Court of Appeals, Edwards was decided here. The Court of Appeals then ruled that, under Edwards, the police had acted unconstitutionally. This *58Court, by a divided vote, reversed, holding that Edwards was not to be applied retroactively in the Stumes situation. Justice Powell concurred in the judgment, 465 U. S., at 651, for he would not impose upon the State the costs that accrue by retroactive application of a new rule of constitutional law on habeas corpus; those costs, in his view, “generally far outweigh the benefits of this application.” Id., at 654.

The primary difference between Johnson, on the one hand, and Stumes, on the other, is the difference between a pending and undecided direct review of a judgment of conviction and a federal collateral attack upon a state conviction which has become final.4 We must acknowledge, of course, that Johnson does not directly control the disposition of the present case. In Johnson, the Court specifically declined to address the implications of its holding for a case in a constitutional area other than the Fourth Amendment, or for a case in which a Fourth Amendment issue is raised on collateral *59attack.5 457 U. S., at 562. We now conclude, however, that there is no reason to reach in this case a result that is different from the one reached in Johnson. See Mack v. Oklahoma, 459 U. S. 900 (1982). There is nothing about a Fourth Amendment rule that suggests that in this context it should be given greater retroactive effect than a Fifth Amendment rule. Indeed, a Fifth Amendment violation may be more likely to affect the truth-finding process than a Fourth Amendment violation. And Justice Harlan’s reasoning — that principled decisionmaking and fairness to similarly situated petitioners require application of a new rule to all cases pending on direct review — is applicable with equal force to the situation presently before us. We hold that our analysis in Johnson is relevant for petitioner’s direct-review Fifth Amendment claim under Edwards. He is entitled to the benefit of the ruling in that case.

> I — I

Other arguments that have been made m support of the judgment below are not persuasive. First, it is said that drawing a distinction between a case pending on direct review and a case on collateral attack produces inequities and injustices that are not any different from those that Johnson purported to cure. The argument is that the litigant whose Edwards claim will not be considered because it is presented on collateral review will be just as unfairly treated as the direct-review litigant whose claim would be bypassed were Edwards not the law. The distinction, however, properly *60rests on considerations of finality in the judicial process. The one litigant already has taken his case through the primary system. The other has not. For the latter, the curtain of finality has not been drawn. Somewhere, the closing must come. Justice Powell stressed this in his opinion concurring in the judgment in Solem v. Stumes, 465 U. S., at 653-654. He said specifically: “[I]t is particularly difficult in such cases to justify imposing upon the State the costs of collateral review. These are not insubstantial.” Id., at 654.

Next, it is said that the application of Edwards to cases pending on direct review will result in the nullification of many convictions and will relegate prosecutors to the difficult position of having to retry cases concerning events that took place years ago. We think this concern is overstated. We are given no empirical evidence in its support, and Louisiana states that any such evidence is unavailable. Brief for Respondent 11. We note, furthermore, that several courts have applied Edwards to cases pending on direct review without expressing concern about lapse of time or retro-activity and without creating any apparent administrative difficulty. See n. 2, supra. And if a case is unduly slow in winding its way through a State’s judicial system, that could be as much the State’s fault as the defendant’s, and should not serve to penalize the defendant.

In addition, it is said that in every case, Edwards alone excepted, reliance on existing law justifies the nonapplication of Edwards. But, as we have pointed out, there is no difference between the petitioner in Edwards and the petitioner in the present case. If the Edwards principle is not to be applied retroactively, the only way to dispense equal justice to Edwards and to Shea would be a rule that confined the Edwards principle to prospective application unavailable even to Edwards himself.

Finally, it is said that the Edwards rule is only prophylactic in- character, and is not one designed to enhance accuracy in criminal jurisprudence. This argument, of course, is *61taken from Michigan v. Payne, 412 U. S. 47 (1973), where the retroactivity of North Carolina v. Pearce, 395 U. S. 711 (1969), was under consideration. The argument, we feel, is fully answered by the decision in United States v. Johnson, and by what we have said above in this opinion.

The judgment of the Supreme Court of Louisiana is reversed, and the case is remanded to that court for further proceedings not inconsistent with this opinion.

It is so ordered.

dissenting.

Last Term, in Solem v. Stumes, 465 U. S. 638 (1984), we held that the rule announced by the Court in Edwards v. Arizona, 451 U. S. 477 (1981), should not be applied retroactively in collateral attacks on criminal convictions. We concluded that the prophylactic purpose of the Edwards rule, the justifiable failure of police and prosecutors to foresee the Court’s decision in Edwards, and the substantial disruption of the criminal justice system that retroactive application of Edwards would entail all indicated the wisdom of holding Edwards nonretroactive. Today, however, the majority concludes that notwithstanding the substantial reasons for restricting the application of Edwards to cases involving interrogations that postdate the Court’s opinion in Edwards, the Edwards rule must be applied retroactively to all cases in which the process of direct appeal had not yet been completed when Edwards was decided. In so holding, the majority apparently adopts a rule long advocated by a shifting minority of Justices and endorsed in limited circumstances by the majority in United States v. Johnson, 457 U. S. 537 (1982): namely, the rule that any new constitutional decision — except, perhaps, one that constitutes a “clear break with the past” — must be applied to all cases pending on direct appeal at the time it is handed down.

*62Two concerns purportedly underlie the majority’s decision. The first is that retroactivity is somehow an essential attribute of judicial decisionmaking, and that when the Court announces a new rule and declines to give it retroactive effect, it has abandoned the judicial role and assumed the function of a legislature — or, to use the term Justice Harlan employed in describing the problem, a “super-legislature.” Desist v. United States, 394 U. S. 244, 259 (1969) (Harlan, J., dissenting). The second (and not completely unrelated) concern is fairness. It is the business of a court, the majority reasons, to treat like cases alike; accordingly, it is unfair for one litigant to receive the benefit of a new decision when another, identically situated, is denied the same benefit. The majority’s concerns are no doubt laudable, but I cannot escape the conclusion that the rule they have spawned makes no sense.

As a means of avoiding what has come to be known as the super-legislature problem, the rule announced by the majority is wholly inadequate. True, the Court is not and cannot be a legislature, super or otherwise. But I should think that concerns about the supposed usurpation of legislative authority by this Court generally go more to the substance of the Court’s decisions than to whether or not they are retroactive. Surely those who believe that the Court has overstepped the bounds of its legitimate authority in announcing a new rule of constitutional law will find little solace in a decision holding the new rule retroactive. If a decision is in some sense illegitimate, making it retroactive is a useless gesture that will fool no one. If, on the other hand, the decision is a salutary one, but one whose purposes are ill-served by retroactive application, retroactivity may be worse than useless, imposing costs on the criminal justice system that will likely be uncompensated for by any perceptible gains in “judicial legitimacy.”

The futility of this latest attempt to use retroactivity doctrine to avoid the super-legislature difficulty is highlighted by *63the majority’s unwillingness to commit itself to the logic of its position. For even as it maintains that retroactivity is essential to the judicial function, today’s majority, like the majority in Johnson, supra, continues to hold out the possibility that a “really” new rule — one that marks a clear break with the past — may not have to be applied retroactively even to cases pending on direct review at the time the new decision is handed down. See ante, at 57 and 59, n. 5; Johnson, supra, at 549-550, 551-554. Of course, if the majority were truly concerned with the super-legislature problem, it would be “clear break” decisions that would trouble it the most. Indeed, one might expect that a Court as disturbed about the problem as the majority purports to be would swear off such decisions altogether, not reserve the power both to issue them and to decline to apply them retroactively. In leaving open the possibility of an exception for “clear break” decisions, the majority demonstrates the emptiness of its proposed solution to the super-legislature problem.

The claim that the majority’s rule serves the interest of fairness is equally hollow. Although the majority finds it intolerable to apply a new rule to one case on direct appeal but not to another, it is perfectly willing to tolerate disparate treatment of defendants seeking direct review of their convictions and prisoners attacking their convictions in collateral proceedings. As I have stated before, see Johnson, supra, at 566-568 (White, J., dissenting); Williams v. United States, 401 U. S. 646, 656-659 (1971) (plurality opinion), it seems to me that the attempt to distinguish between direct and collateral challenges for purposes of retroactivity is misguided. Under the majority’s rule, otherwise identically situated defendants may be subject to different constitutional rules, depending on just how long ago now-unconstitutional conduct occurred and how quickly cases proceed through the criminal justice system. The disparity is no different in kind from that which occurs when the benefit of a new constitutional rule is retroactively afforded to the defendant in whose *64case it is announced but to no others; the Court’s new approach equalizes nothing except the numbers of defendants within the disparately treated classes.

The majority recognizes that the distinction between direct review and habeas is problematic, but justifies its differential treatment by appealing to the need to draw “the curtain of finality,” ante, at 60, on those who were unfortunate enough to have exhausted their last direct appeal at the time Edwards was decided. Yet the majority offers no reasons for its conclusion that finality should be the decisive factor. When a conviction is overturned on direct appeal on the basis of an Edwards violation, the remedy offered the defendant is a new trial at which any inculpatory statements obtained in violation of Edwards will be excluded. It is not clear to me why the majority finds such a burdensome remedy more acceptable when it is imposed on the State on direct review than when it is the result of a collateral attack. The disruption attendant upon the remedy does not vary depending on whether it is imposed on direct review or habeas;1 accord*65ingly, if the remedy must be granted to defendants on direct appeal, there is no strong reason to deny it to prisoners attacking their convictions collaterally. Conversely, if it serves no worthwhile purpose to grant the remedy to a defendant whose conviction was final before Edwards, it is hard to see why the remedy should be available on direct review.

The underlying flaw of the majority’s opinion is its failure to articulate the premises on which the retroactivity doctrine it announces actually rests. In recognizing that a decision marking a clear break from the past may not be retroactive and in holding that the concern of finality trumps considerations of fairness that might otherwise dictate retroactivity in collateral proceedings, the majority implicitly recognizes that there is in fact more at issue in decisions involving retro-activity than treating like cases alike. In short, the majority recognizes that there are reasons why certain decisions ought not be retroactive. But the rules the majority announces fail to reflect any thoughtful inquiry into what those reasons might be. By contrast, the principles of retroactivity set forth in Linkletter v. Walker, 381 U. S. 618 (1965), and most recently applied in Solem v. Stumes, 465 U. S. 638 (1984), provide a rational framework for thinking about the question whether retroactive application of any particular decision makes sense — that is, whether the benefits of retroactivity outweigh its costs. Because the Court has already deter*66mined that the relevant considerations set forth in Linkletter (the purpose of the new rule, the extent of law enforcement officials’ justifiable reliance on the prior rule, and the effects on the criminal justice system of retroactivity) dictate non-retroactive application of the rule in Edwards, I cannot join in the majority’s conclusion that that rule should be applied retroactively to cases pending on direct review at the time of our decision in Edwards. More importantly, I cannot concur in the approach to retroactivity adopted by today’s majority — an approach that, if our precedents regarding the non-retroactivity of decisions marking a clear break with the past remain undisturbed, merely adds a confusing and unjustifiable addendum to our retroactivity jurisprudence.2

I respectfully dissent.

dissenting.

I would be willing to join the result reached by the Court in this case if the majority were willing to adopt both aspects of the approach to retroactivity propounded by Justice Harlan in his concurrence in Mackey v. United States, 401 U. S. 667, 675 (1971). Under his approach, new constitutional rules prescribed by this Court for the conduct of criminal prosecutions would apply retroactively to all cases on direct appeal at the time the new rule was announced and, with narrow exceptions, would not apply in collateral proceedings challenging convictions that had become final before the new rule was announced. I will not attempt to summarize the justifica*67tions for this approach so thoughtfully articulated by Justice Harlan.

Because the Court apparently is not willing to adopt in entirety Justice Harlan’s bright-line distinction between direct appeals and collateral attacks, I join Justice White’s dissent, agreeing with him that there is little logic to the Court’s analysis and its rejection of the sound reasons given in Solem v. Stumes, 465 U. S. 638 (1984), for making Edwards v. Arizona, 451 U. S. 477 (1981), nonretroactive.*