Opinion

JUSTICE O'CONNOR, with whom THE CHIEF JUSTICE joins, concurring in part and dissenting in part.

Today the Court permits the federal courts to overturn on habeas the conviction of a double murderer, not on the basis of an inexorable constitutional or statutory command, but because it believes the result desirable from the standpoint of equity and judicial administration. Because the principles that inform our habeas jurisprudence-finality, federalism, and fairness-counsel decisively against the result the Court reaches, I respectfully dissent from this holding.

I

The Court does not sit today in direct review of a statecourt judgment of conviction. Rather, respondent seeks relief by collaterally attacking his conviction through the writ of habeas corpus. While petitions for the writ of habeas corpus are now commonplace-over 12,000 were filed in 1990, compared to 127 in 1941-their current ubiquity ought not detract from the writ's historic importance. See L. Mecham, Annual Report of the Director of the Administrative Office of the United States Courts 191 (1991) (1990 figures); Fay v. Noia, 372 U. S. 391, 446, n. 2 (1963) (Clark, J., dissenting) (1941 figures). "The Great Writ" can be traced through the common law to well before the founding of this Nation; its role as a "prompt and efficacious remedy for whatever society deems to be intolerable restraints" is beyond question. Fay, 372 U. S., at 401-402. As Justice Harlan explained:"Habeas corpus ad subjiciendum is today, as it has always been, a fundamental safeguard against unlawful custody.... Although the wording of earlier statutory provisions has been changed, the basic question before the court to which the writ is addressed has always been the same: in the language of the present statute, on the books since 1867, is the detention complained of 'in violation of the Constitution or laws or treaties of the United States'?" Id., at 449 (dissenting opinion).

Nonetheless, we repeatedly have recognized that collateral attacks raise numerous concerns not present on direct review. Most profound is the effect on finality. It goes without saying that, at some point, judicial proceedings must draw to a close and the matter deemed conclusively resolved; no society can afford forever to question the correctness of its every judgment. "[T]he writ," however, "strikes at finality," McCleskey v. Zant, 499 U. S. 467, 491 (1991), depriving the criminal law "of much of its deterrent effect," Teague v. Lane, 489 U. S. 288, 309 (1989) (plurality opinion), and sometimes preventing the law's just application altogether, see McCleskey, supra, at 491. "No one, not criminal defendants, not the judicial system, not society as a whole is benefited by a judgment providing a man shall tentatively go to jail today, but tomorrow and every day thereafter his continued incarceration shall be subject to fresh litigation." Mackey v. United States, 401 U. S. 667, 691 (1971) (Harlan, J., concurring in part and dissenting in part); see also Mc Cleskey, supra, at 492.

In our federal system, state courts have primary responsibility for enforcing constitutional rules in their own criminal trials. When a case comes before the federal courts on habeas rather than on direct review, the judicial role is "significantly different." Mackey, supra, at 682 (Harlan, J., concurring in part and dissenting in part). Accord, Teague, supra, at 306-308. Most important here, federal courts on direct review adjudicate every issue of federal law properly presented; in contrast, "federal courts have never had a similar obligation on habeas corpus." Mackey, supra, at 682 (Harlan, J., concurring in part and dissenting in part). As the Court explains today, federal courts exercising their habeas powers may refuse to grant relief on certain claims because of "prudential concerns" such as equity and federalism. Ante, at 686. This follows not only from the express language of the habeas statute, which directs the federal courts to "dispose of [habeas petitions] as law and justice require," 28 U. s. C. § 2243, but from our precedents as well. In Francis v. Henderson, 425 U. S. 536 (1976), we stated that "[t]his Court has long recognized that in some circumstances considerations of comity and concerns for the orderly administration of criminal justice require a federal court to forgo the exercise of its habeas corpus power." Id., at 539. Accord, Gomez v. United States Dist. Court for Northern Dist. of Cal., 503 U. S. 653, 653-654 (1992) ("Whether [a] claim is framed as a habeas petition or as a [42 U. S. C.] § 1983 action, [what is sought is] an equitable remedy"; as a result, equity must be "take[n] into consideration"); Fay v. Noia, supra, at 438 ("[H]abeas corpus has traditionally been regarded as governed by equitable principles"); Duckworth v. Eagan, 492 U. S. 195, 213 (1989) (O'CONNOR, J., concurring) ("[T]he Court has long recognized that habeas corpus [is] governed by equitable principles" (internal quotation marks omitted)).

Concerns for equity and federalism resonate throughout our habeas jurisprudence. In 1886, only eight years after Congress gave the federal courts power to issue writs ordering the release of state prisoners, this Court explained that courts could accommodate federalism and comity concerns by withholding relief until after state proceedings had terminated. Ex parte Royall, 117 U. S. 241, 251-253. Accord, Fay, supra, at 418-419. More recently, we relied on those same concerns in holding that new constitutional rules of criminal procedure do not apply retroactively on habeas. Teague, supra, at 306. Our treatment of successive petitions and procedurally defaulted claims similarly is governed by equitable principles. McCleskey, 499 U. S., at 489-491 (successive petitions); id., at 490 (procedurally defaulted claims); Fay, supra, at 438 (procedurally defaulted claims). Most telling of all, this Court continuously has recognized that the ultimate equity on the prisoner's side-a sufficient showing of actual innocence-is normally sufficient, standing alone, to outweigh other concerns and justify adjudication of the prisoner's constitutional claim. See Sawyer v. Whitley, 505 U. S. 333, 340-347 (1992) (actual innocence of penalty); Murray v. Carrier, 477 U. S. 478, 496 (1986) (federal courts may reach procedurally defaulted claims on a showing that a constitutional violation probably resulted in the conviction of an actually innocent person); Kuhlmann v. Wilson, 477 U. S. 436, 454 (1986) (colorable showing of actual innocence suffices to excuse successive claim); see also Teague v. Lane, supra, at 313 (where absence of procedure seriously diminishes the likelihood of an accurate conviction, a new rule requiring the procedure may be applied retroactively on habeas).

Nonetheless, decisions concerning the availability of habeas relief warrant restraint. Nowhere is the Court's restraint more evident than when it is asked to exclude a substantive category of issues from relitigation on habeas. Although we recognized the possibility of excluding certain types of claims long ago, see Mackey, supra, at 683 (Harlan, J., concurring in part and dissenting in part), only once has this Court found that the concerns of finality, federalism, and fairness supported such a result; that was in Stone v. Powell, 428 U. S. 465 (1976). Ante, at 686. Since then, the Court has refused to bar additional categories of claims on three different occasions. Ante, at 687-688.

Today we face the question whether Stone v. Powell should extend to bar claims on habeas that the prophylactic rule of Miranda v. Arizona, 384 U. S. 436 (1966), had been violated. Continuing the tradition of caution in this area, the Court answers that question in the negative. This time I must disagree. In my view, the "prudential concerns," ante, at 686, that inform our habeas jurisprudence counsel the exclusion of Miranda claims just as strongly as they did the exclusionary rule claims at issue in Stone itself.

II

In Stone, the Court explained that the exclusionary rule of Mapp v. Ohio, 367 U. S. 643 (1961), was not an inevitable product of the Constitution but instead "'a judicially created remedy.''' Stone, supra, at 486 (quoting United States v. Calandra, 414 U. S. 338, 349 (1974)). By threatening to exclude highly probative and sometimes critical evidence, the exclusionary rule "is thought to encourage those who formulate law enforcement policies, and the officers who implement them, to incorporate Fourth Amendment ideals into their value system." Stone, 428 U. S., at 492. The deterrent effect is strong: Any transgression of the Fourth Amendment carries the risk that evidence will be excluded at trial. Nonetheless, this increased sensitivity to Fourth Amendment values carries a high cost. Exclusion not only deprives the jury of probative and sometimes dispositive evidence, but also "deflects the truthfinding process and often frees the guilty." Id., at 490. When that happens, it is not just the executive or the judiciary but all of society that suffers: The executive suffers because the police lose their suspect and the prosecutor the case; the judiciary suffers because its processes are diverted from the central mission of ascertaining the truth; and society suffers because the populace again finds a guilty and potentially dangerous person in its midst, solely because a police officer bungled.

While that cost is considered acceptable when a case is on direct review, the balance shifts decisively once the case is on habeas. There is little marginal benefit to enforcing the exclusionary rule on habeas; the penalty of exclusion comes too late to produce a noticeable deterrent effect. Id., at 493. Moreover, the rule "divert[s attention] from the ultimate question of guilt," squanders scarce federal judicial re sources, intrudes on the interest in finality, creates friction between the state and federal systems of justice, and upsets the "'constitutional balance upon which the doctrine of federalism is founded.'" Id., at 490, 491, n. 31 (quoting Schneckloth v. Bustamonte, 412 U. S. 218, 259 (1973) (Powell, J., concurring)). Because application of the exclusionary rule on habeas "offend[s] important principles of federalism and finality in the criminal law which have long informed the federal courts' exercise of habeas jurisdiction," Duckworth, 492 U. S., at 208 (O'CONNOR, J., concurring), we held in Stone that such claims would no longer be cognizable on habeas so long as the State already had provided the defendant with a full and fair opportunity to litigate.

I continue to believe that these same considerations apply to Miranda claims with equal, if not greater, force. See Duckworth, supra, at 209 (O'CONNOR, J., concurring). Like the suppression of the fruits of an illegal search or seizure, the exclusion of statements obtained in violation of Miranda is not constitutionally required. This Court repeatedly has held that Miranda's warning requirement is not a dictate of the Fifth Amendment itself, but a prophylactic rule. See, e. g., McNeil v. Wisconsin, 501 U. S. 171, 176 (1991); Michigan v. Harvey, 494 U. S. 344, 350 (1990); Duckworth, supra, at 203; New York v. Quarles, 467 U. S. 649, 654 (1984); Michigan v. Tucker, 417 U. S. 433, 442-446 (1974). Because Miranda "sweeps more broadly than the Fifth Amendment itself," it excludes some confessions even though the Constitution would not. Oregon v. Elstad, 470 U. S. 298, 306 (1985). Indeed, "in the individual case, Miranda's preventive medicine [often] provides a remedy even to the defendant who has suffered no identifiable constitutional harm." Id., at 307.

Miranda's overbreadth, of course, is not without justification. The exclusion of unwarned statements provides a strong incentive for the police to adopt "procedural safeguards," Miranda, 384 U. S., at 444, against the exaction of compelled or involuntary statements. It also promotes institutional respect for constitutional values. But, like the exclusionary rule for illegally seized evidence, Miranda's prophylactic rule does so at a substantial cost. Unlike involuntary or compelled statements-which are of dubious reliability and are therefore inadmissible for any purpose-confessions obtained in violation of Miranda are not necessarily untrustworthy. In fact, because voluntary statements are "trustworthy" even when obtained without proper warnings, Johnson v. New Jersey, 384 U. S. 719, 731 (1966), their suppression actually impairs the pursuit of truth by concealing probative information from the trier of fact. See Harvey, supra, at 350 (Miranda "result[s] in the exclusion of some voluntary and reliable statements"); Elstad, supra, at 312 (loss of "highly probative evidence of a voluntary confession" is a "high cost [for] law enforcement"); McNeil, supra, at 181 (because "the ready ability to obtain uncoerced confessions is not an evil but an unmitigated good," the exclusion of such confessions renders society "the loser"); Tucker, supra, at 461 (WHITE, J., concurring in judgment) ("[H]aving relevant and probative testimony, not obtained by actual coercion... aides] in the pursuit of truth"); Miranda, supra, at 538 (WHITE, J., dissenting) ("Particularly when corroborated,... such [voluntary] confessions have the highest reliability and significantly contribute to the certitude with which we may believe the accused is guilty").

When the case is on direct review, that damage to the truth-seeking function is deemed an acceptable sacrifice for the deterrence and respect for constitutional values that the Miranda rule brings. But once a case is on collateral review, the balance between the costs and benefits shifts; the interests of federalism, finality, and fairness compel Mi randa's exclusion from habeas. The benefit of enforcing Miranda through habeas is marginal at best. To the extent Miranda ensures the exclusion of involuntary statements, that task can be performed more accurately by adjudicating the voluntariness question directly. See Johnson, supra, at 730-731. And, to the extent exclusion of voluntary but unwarned confessions serves a deterrent function, "[t]he awarding of habeas relief years after conviction will often strike like lightning, and it is absurd to think that this added possibility... will have any appreciable effect on police training or behavior." Duckworth, supra, at 211 (O'CONNOR, J., concurring). Judge Friendly made precisely the same point 18 years earlier: "[T]he deterrent value of permitting collateral attack," he explained, "goes beyond the point of diminishing returns." Friendly, Is Innocence Irrelevant? Collateral Attack on Criminal Judgments, 38 U. Chi. L. Rev. 142, 163 (1970).

Despite its meager benefits, the relitigation of Miranda claims on habeas imposes substantial costs. Just like the application of the exclusionary rule, application of Miranda's prophylactic rule on habeas consumes scarce judicial resources on an issue unrelated to guilt or innocence. No less than the exclusionary rule, it undercuts finality. It creates tension between the state and federal courts. And it upsets the division of responsibilities that underlies our federal system. But most troubling of all, Miranda's application on habeas sometimes precludes the just application of law altogether. The order excluding the statement will often be issued "years after trial, when a new trial may be a practical impossibility." Duckworth, 492 U. S., at 211 (O'CONNOR, J., concurring). Whether the Court admits it or not, the grim result of applying Miranda on habeas will be, time and time again, "the release of an admittedly guilty individual who may pose a continuing threat to society." Ibid.

Any rule that so demonstrably renders truth and society "the loser," McNeil v. Wisconsin, 501 U. S., at 181, "'bear[s] a heavy burden of justification, and must be carefully limited to the circumstances in which it will pay its way by deterring official lawlessness,'" United States v. Leon, 468 U. S. 897, 908, n. 6 (1984) (quoting Illinois v. Gates, 462 U. S. 213, 257 258 (1983) (WHITE, J., concurring in judgment)). That burden is heavier still on collateral review. In light of the meager deterrent benefit it brings and the tremendous costs it imposes, in my view application of Miranda's prophylactic rule on habeas "falls short" of justification. Ante, at 688.

III

The Court identifies a number of differences that, in its view, distinguish this case from Stone v. Powell. Ante, at 691-695. I am sympathetic to the Court's concerns but find them misplaced nonetheless.

The first difference the Court identifies concerns the nature of the right protected. Miranda, the Court correctly points out, fosters Fifth Amendment, rather than Fourth Amendment, values. Ante, at 691. The Court then offers a defense of the Fifth Amendment, reminding us that it is "'a fundamental trial right' " that reflects" 'principles of humanity and civil liberty'''; that it was secured" 'after years of struggle'''; and that it does not serve "some value necessarily divorced from the correct ascertainment of guilt." Ante, at 691-692 (quoting United States v. Verdugo-Urquidez, 494 U. S. 259, 364 (1990), and Bram v. United States, 168 U. S. 532, 544 (1897)). The Court's spirited defense of the Fifth Amendment is, of course, entirely beside the point. The question is not whether true Fifth Amendment claims-the extraction and use of compelled testimony-should be cognizable on habeas. It is whether violations of Miranda's prophylactic rule, which excludes from trial voluntary confessions obtained without the benefit of Miranda's now-familiar warnings, should be. The questions are not the same; nor are their answers.

To say that the Fifth Amendment is a "'fundamental trial right,'" ante, at 691 (quoting United States v. VerdugoUrquidez, supra, at 264), is thus both correct and irrelevant. Miranda's warning requirement may bear many labels, but "fundamental trial right" is not among them. Long before Miranda was decided, it was well established that the Fifth Amendment prohibited the introduction of compelled or involuntary confessions at trial. And long before Miranda, the courts enforced that prohibition by asking a simple and direct question: Was "the confession the product of an essentially free and unconstrained choice," or was the defendant's will "overborne"? Schneckloth v. Bustamonte, 412 U. S., at 225 (quoting Culombe v. Connecticut, 367 U. S. 568, 602 (1961)); see ante, at 688-689; see, e. g., Bram v. United States, supra. Miranda's innovation was its introduction of the warning requirement: It commanded the police to issue warnings (or establish other procedural safeguards) before obtaining a statement through custodial interrogation. And it backed that prophylactic rule with a similarly prophylactic remedy-the requirement that unwarned custodial statements, even if wholly voluntary, be excluded at trial. Mi randa, 384 U. S., at 444. Excluding violations of Miranda's prophylactic suppression requirement from habeas would not leave true Fifth Amendment violations unredressed. Prisoners still would be able to seek relief by "invok[ing] a substantive test of voluntariness" or demonstrating prohibited coercion directly. Johnson, 384 U. S., at 730; Elstad, 470 U. S., at 307-308 (statements falling outside Miranda's sweep analyzed under voluntariness standard). The Court concedes as much. Ante, at 693 ("[E]liminating habeas review of Miranda issues would not prevent a state prisoner from simply converting his barred Miranda claim into a due process claim that his conviction rested on an involuntary confession").

Excluding Miranda claims from habeas, then, denies collateral relief only in those cases in which the prisoner's statement was neither compelled nor involuntary but merely obtained without the benefit of Miranda's prophylactic warnings. The availability of a suppression remedy in such cases cannot be labeled a "fundamental trial right," for there is no constitutional right to the suppression of voluntary state ments. Quite the opposite: The Fifth Amendment, by its terms, prohibits only compelled self-incrimination; it makes no mention of "unwarned" statements. U. S. Const., Arndt. 5 ("No person... shall be compelled in any criminal case to be a witness against himself" (emphasis added)). On that much, our cases could not be clearer. See, e. g., Michigan v. Tucker, 417 U. S., at 448 ("Cases which involve the SelfIncrimination Clause must, by definition, involve an element of coercion, since the Clause provides only that a person shall not be compelled to give evidence against himself"); see Elstad, supra, at 306-307; New York v. Quarles, 467 U. S., at 654-655, and n. 5. As a result, the failure to issue warnings does "not abridge [the] constitutional privilege against compulsory self-incrimination, but depart[s] only from the prophylactic standards later laid down by this Court in Mi randa." Tucker, supra, at 446. If the principles of federalism, finality, and fairness ever counsel in favor of withholding relief on habeas, surely they do so where there is no constitutional harm to remedy.

Similarly unpersuasive is the Court's related argument, ante, at 692, that the Fifth Amendment trial right is not "necessarily divorced" from the interest of reliability. Whatever the Fifth Amendment's relationship to reliability, Mi randa's prophylactic rule is not merely "divorced" from the quest for truth but at war with it as well. The absence of Miranda warnings does not by some mysterious alchemy convert a voluntary and trustworthy statement into an involuntary and unreliable one. To suggest otherwise is both unrealistic and contrary to precedent. As I explained above, we have held over and over again that the exclusion of unwarned but voluntary statements not only fails to advance the cause of accuracy but impedes it by depriving the jury of trustworthy evidence. Supra, at 703. In fact, we have determined that the damage Miranda does to the truthseeking mission of the criminal trial can become intolerable. We therefore have limited the extent of the suppression rem edy, see Harris v. New York, 401 U. S. 222, 224-226 (1971) (unwarned but voluntary statement may be used for impeachment), and dispensed with it entirely elsewhere, see Quarles, supra (unwarned statement may be used for any purpose where statement was obtained under exigent circumstances bearing on public safety). And at least one Member of this Court dissented from Miranda itself because it "establish[ed] a new... barrier to the ascertainment of truth by the judicial process." Miranda, supra, at 542 (opinion of WHITE, J.). Consequently, I agree with the Court that Miranda's relationship to accurate verdicts is an important consideration when deciding whether to permit Miranda claims on habeas. But it is a consideration that weighs decisively against the Court's decision today.

The consideration the Court identifies as being "most importan[t]" of all, ante, at 693, is an entirely pragmatic one. Specifically, the Court "project[s]" that excluding Miranda questions from habeas will not significantly promote efficiency or federalism because some Miranda issues are relevant to a statement's voluntariness. Ante, at 693-695. It is true that barring Miranda claims from habeas poses no barrier to the adjudication of voluntariness questions. But that does not make it "reasonable to suppose that virtually all Miranda claims [will] simply be recast" and litigated as voluntariness claims. Ante, at 693. Involuntariness requires coercive state action, such as trickery, psychological pressure, or mistreatment. Colorado v. Connelly, 479 U. S. 157, 167 (1986) ("[C]oercive police activity is a necessary predicate to the finding that a confession is not 'voluntary' "); ante, at 693 (referring to "the crucial element of police coercion"). A Miranda claim, by contrast, requires no evidence of police overreaching whatsoever; it is enough that law enforcement officers commit a technical error. Even the forgetful failure to issue warnings to the most wary, knowledgeable, and sea soned of criminals will do. Miranda, 384 U. S., at 468 ("[WJe will not pause to inquire in individual cases whether the defendant was aware of his rights without a warning being given"). Given the Court's unqualified trust in the willingness of police officers to satisfy Miranda's requirements, ante, at 695, its suggestion that their every failure to do so involves coercion seems to me ironic. If the police have truly grown in "constitutional... sophistication," ibid., then certainly it is reasonable to suppose that most technical errors in the administration of Miranda's warnings are just that.

In any event, I see no need to resort to supposition. The published decisions of the lower federal courts show that what the Court assumes to be true demonstrably is not. In case after case, the courts are asked on habeas to decide purely technical Miranda questions that contain not even a hint of police overreaching. And in case after case, no voluntariness issue is raised, primarily because none exists. Whether the suspect was in "custody," 1 whether or not there was "interrogation," 2 whether warnings were given or were adequate,3 whether the defendant's equivocal statement constituted an invocation of rights,4 whether waiver was knowing and intelligent 5-this is the stuff that Miranda claims are made of. While these questions create litigable issues under Miranda, they generally do not indicate the existence of coercion-pressure tactics, deprivations, or exploitations of the defendant's weaknesses-sufficient to establish involuntariness. Even assuming that many Miranda claims could "simply be recast" as voluntariness claims, it does not follow that barring Miranda's prophylactic rule from habeas would unduly complicate their resolution. The Court labels Miranda a "bright-line (or, at least, brighter-line) rul[eJ" and involuntariness an "exhaustive totality-of-circumstances approach," ante, at 694, but surely those labels overstate the differences. Miranda, for all its alleged brightness, is not without its difficulties; and voluntariness is not without its strengths. JusTICE WHITE so observed in his Miranda dissent, noting that the Court could not claim that"judicial time and effort... will be conserved because of the ease of application of the [Miranda] rule. [Miranda] leaves open such questions as whether the accused was in custody, whether his statements were spontaneous or the product of interrogation, whether the accused has effectively waived his rights,... all of which are certain to prove productive of uncertainty during investigation and litigation during prosecution." Miranda, supra, at 544-545.

Experience has proved JUSTICE WHITE'S prediction correct. Miranda creates as many close questions as it resolves. The task of determining whether a defendant is in "custody" has proved to be "a slippery one." Elstad, 470 U. S., at 309; see, e. g., n. 1, supra (custody cases). And the supposedly "bright" lines that separate interrogation from spontaneous declaration, the exercise of a right from waiver, and the adequate warning from the inadequate, likewise have turned out to be rather dim and ill defined. See Rhode Island v. Innis, 446 U. S. 291 (1980) (interrogation); n. 2, supra (interrogation); nn. 4 and 5, supra (waiver and invocation); n. 3, supra (adequacy of warnings). Yet Miranda requires those lines to be drawn with precision in each case.

The totality-of-the-circumstances approach, on the other hand, permits each fact to be taken into account without re sort to formal and dispositive labels. By dispensing with the difficulty of producing a yes-or-no answer to questions that are often better answered in shades and degrees, the voluntariness inquiry often can make judicial decisionmaking easier rather than more onerous. Thus, it is true that the existence of warnings is still a consideration under the totality-of-the-circumstances approach, ante, at 693-694, but it is unnecessary to determine conclusively whether "custody" existed and triggered the warning requirement, or whether the warnings given were sufficient. It is enough that the habeas court look to the warnings or their absence, along with all other factors, and consider them in deciding what is, after all, the ultimate question: whether the confession was compelled and involuntary or the product of a free and unimpaired will. See Schneckloth v. Bustamonte, 412 U. S., at 225-226.

Nor does continued application of Miranda's prophylactic rule on habeas dispense with the necessity of testing confessions for voluntariness. While Miranda's conclusive presumption of coercion may sound like an impenetrable barrier to the introduction of compelled testimony, in practice it leaks like a sieve. Miranda, for example, does not preclude the use of an unwarned confession outside the prosecution's case in chief, Harris v. New York, 401 U. S. 222 (1971); Oregon v. Hass, 420 U. S. 714 (1975); involuntary statements, by contrast, must be excluded from trial for all purposes, Mincey v. Arizona, 437 U. S. 385, 398 (1978). Miranda does not preclude admission of the fruits of an unwarned statement, see Oregon v. Elstad, supra; but under the Fifth and Fourteenth Amendments, we require the suppression of not only compelled confessions but tainted subsequent confessions as well, Clewis v. Texas, 386 U. S. 707, 710 (1967). Finally, Miranda can fail to exclude some truly involuntary statements: It is entirely possible to extract a compelled statement despite the most precise and accurate of warnings. See Johnson, 384 U. S., at 730 (warnings are only one factor in determining voluntariness).

The Court's final rationale is that, because the federal courts rarely issue writs for Miranda violations, eliminating Miranda claims from habeas will not decrease state-federal tensions to an appreciable degree. Ante, at 694-695. The relative infrequency of relief, however, does not diminish the intrusion on state sovereignty; it diminishes only our justification for intruding in the first place. After all, even if relief is denied at the end of the day, the State still must divert its scarce prosecutorial resources to defend an otherwise final conviction. If relief is truly rare, efficiency counsels in favor of dispensing with the search for the prophylactic rule violation in a haystack; instead, the federal courts should concentrate on the search for true Fifth Amendment violations by adjudicating the questions of voluntariness and compulsion directly. I therefore find it of little moment that the Police Foundation et al. support respondent. Ante, at 695, n. 6. Those who bear the primary burden of defending state convictions in federal courts-including 36 States and the National District Attorneys Associationresoundingly support the opposite side. See Brief for California et al. as Amici Curiae; Brief for Americans for Effective Law Enforcement, Inc., and the National District Attorneys Association, Inc., as Amici Curiae; see also Brief for United States as Amicus Curiae (United States must defend against claims raised by federal prisoners under 28 U. S. C. § 2255).

The Court's response, that perhaps the police respect the Miranda rule as a result of "the existence of [habeas] review," ante, at 695, is contrary to both case law and common sense. As explained above, there is simply no reason to think that habeas relief, which often "'strike[s] like lightning'" years after conviction, contributes much additional deterrence beyond the threat of exclusion during state proceedings. See supra, at 704 (quoting Duckworth, 492 U. S., at 211 (O'CONNOR, J., concurring)). Accord, Friendly, 38 U. Chi. L. Rev., at 163. And our decision in Stone expressly so held: "The view that the deterrence... would be furthered rests on the dubious assumption that law enforcement authorities would fear that federal habeas review might reveal flaws... that went undetected at trial and on appeal." 428 U. S., at 493 (footnote omitted). The majority offers no justification for disregarding our decision in Stone; nor does it provide any reason to question the truth of Stone's observation.

IV

As the Court emphasizes today, Miranda's prophylactic rule is now 27 years old; the police and the state courts have indeed grown accustomed to it. Ante, at 695. But it is precisely because the rule is well accepted that there is little further benefit to enforcing it on habeas. We can depend on law enforcement officials to administer warnings in the first instance and the state courts to provide a remedy when law enforcement officers err. None of the Court's asserted justifications for enforcing Miranda's prophylactic rule through habeas-neither reverence for the Fifth Amendment nor the concerns of reliability, efficiency, and federalism-counsel in favor of the Court's chosen course. Indeed, in my view they cut in precisely the opposite direction. The Court may reconsider its decision when presented with empirical data. See ante, at 693 (noting absence of empirical data); ante, at 688 (holding only that today's argument in favor of extending Stone "falls short"). But I see little reason for such a costly delay. Logic and experience are at our disposal now. And they amply demonstrate that applying Miranda's prophylactic rule on habeas does not increase the amount of justice dispensed; it only increases the frequency with which the admittedly guilty go free. In my view, Miranda imposes such grave costs and produces so little benefit on habeas that its continued application is neither tolerable nor justified. Accordingly, I join Part III of the Court's opinion but respectfully dissent from the remainder.

Notes

1 See, e. g., Schiro v. Clark, 963 F.2d 962, 974-975 (CA7 1992) (defendant approached officer in halfway house and asked to speak to him; not in custody); Tart v. Massachusetts, 949 F.2d 490,504 (CA11991) (fisherman asked to produce document on board his own, docked boat; no custody); Williams v. Chrans, 945 F.2d 926, 950-952 (CA7 1991) (voluntary appearance for presentence report interview; not in custody), cert. denied, 505 U. S. 1208 (1992); Carlson v. State, 945 F.2d 1026, 1028-1029 (CA8 1991) (suspect questioned at his home; no custody); Davis v. Kemp, 829 F.2d 1522, 1535 (CAll 1987) (defendant voluntarily went to police station in absence of evidence that there was probable cause for arrest; not in custody), cert. denied, 485 U. S. 929 (1988); Cobb v. Perini, 832 F.2d 342, 345-347 (CA6 1987) (investigatory Terry-stop; not in custody), cert. denied, 486 U. S. 1024 (1988); Leviston v. Black, 843 F.2d 302, 304 (CA8) (injail interview initiated by incarcerated defendant; no custody), cert. denied, 488 U. S. 865 (1988); Cordoba v. Hanrahan, 910 F.2d 691, 693-694 (CAW) (drunk driver questioned at accident scene before arrest; not in custody), cert. denied, 498 U. S. 1014 (1990).

2 See, e. g., Endress v. Dugger, 880 F.2d 1244, 1246-1250 (CAll 1989) (defendant volunteered information without questioning), cert. denied, 495 U. S. 904 (1990); United States ex rel. Church v. De Robertis, 771 F.2d 1015, 1018-1020 (CA7 1985) (placing defendant's brother in cell with him not interrogation); Harryman v. Estelle, 616 F.2d 870, 873-875 (CA5) (en banc) (officer's surprised exclamation, "What is this?" upon finding condom filled with white powder, constituted interrogation), cert. denied, 449 U. S. 860 (1980); Phillips v. Attorney General of California, 594 F.2d 1288, 1290-1291 (CA9 1979) (defendant volunteered information after officer stated that he wished to see interior of defendant's plane).

3 See, e. g., Chambers v. Lockhart, 872 F.2d 274, 275-276 (CA8) (omission of right to free appointed counsel), cert. denied, 493 U. S. 938 (1989); Gates v. Zant, 863 F.2d 1492, 1500-1501 (CAll) (no warning that videotape of confession could be used), cert. denied, 493 U. S. 945 (1989); Crespo v. Armontrout, 818 F.2d 684, 685-686 (CA8) (when and whether warnings were given), cert. denied, 484 U. S. 978 (1987); De La Rosa v. Texas, 743 F.2d 299, 301-302 (CA5 1984) (officer's explanation of the warnings alleged to be misleading), cert. denied, 470 U. S. 1065 (1985); Stanley v. Zant, 697 F. 2d 955, 972 (CA111983) (allegedly misleading waiver form), cert. denied, 467 U. S. 1219 (1984).

4 See, e. g., Bobo v. Kolb, 969 F.2d 391, 395-398 (CA7 1992) (standing mute); Christopher v. Florida, 824 F.2d 836, 841-843 (CA111987) (equivocal invocation of right to silence), cert. denied, 484 U. S. 1077-1078 (1988); Lightbourne v. Dugger, 829 F.2d 1012, 1017-1019 (CA111987) (spontaneous resumption of discussion after cutting off questioning), cert. denied, 488 U. S. 934 (1988).

5 See, e. g., Terrovona v. Kincheloe, 912 F.2d 1176, 1179-1180 (CA9 1990) (validity of implied waiver in light of defendant's "background, experience, and conduct"), cert. denied, 499 U. S. 979 (1991); Fike v. James, 833 F.2d 1503, 1506-1507 (CA111987) (defendant's initiation of contact waived previous invocation of rights).