delivered the opinion of the Court.

Petitioner Horace Butler was convicted and sentenced to death for the murder of Pamela Lane. After his conviction became final on direct appeal, Butler collaterally attacked his conviction by way of a petition for federal habeas corpus. Butler relied on our decision in Arizona v. Roberson, 486 *409U. S. 675 (1988), decided after his conviction became final on direct appeal. We have held, however, that a new decision generally is not applicable in cases on collateral review unless the decision was dictated by precedent existing at the time the petitioner’s conviction became final. Penry v. Lynaugh, 492 U. S. 302 (1989); Teague v. Lane, 489 U. S. 288 (1989). We hold that our ruling in Roberson was not so dictated and that Butler’s claim is not within either of two narrow exceptions to the general rule.

Pamela Lane, a clerk at a convenience store near Charleston, South Carolina, was last seen alive when she left work riding a moped late in the evening of July 17, 1980. The next day several fishermen discovered Lane’s body near a bridge, and the following day a local minister found Lane’s moped submerged in a pond behind his church.

Petitioner Butler was arrested six weeks later on an unrelated assault and battery charge and placed in the Charleston County Jail. After invoking his Fifth Amendment right to counsel, Butler retained counsel who appeared with him at a bond hearing on August 31, 1980. He was unable to make bond, however, and was returned to the county jail. Butler’s attorney would later contend in state collateral relief proceedings that after the bond hearing, he had told the police officers not to question Butler further. The officers testified that they remembered no such instruction.

Early in the morning of September 1, 1980, Butler was taken from the jail to the Charleston County Police station. He was then informed for the first time that he was a suspect in Lane’s murder. After receiving Miranda warnings, see Miranda v. Arizona, 384 U. S. 436 (1966), Butler indicated that he understood his rights and signed two “waiver of rights” forms. The police then interrogated Butler about the murder. Butler did not request his attorney’s presence at any time during the interrogation.

Butler offered two explanations for Lane’s death. First, he claimed that a friend, one White, killed Lane and then sought Butler’s help in disposing of the moped. When his in*410terrogators evidenced skepticism over this statement, Butler tried again. He said that he had come upon Lane in his car and had motioned her over to the side of the road. She then voluntarily accompanied him in a drive to a nearby wooded area where the two engaged in consensual sex. Afterwards Lane threatened to accuse Butler of rape when she realized she would be late getting home. Butler maintained that he panicked, shot Lane with a handgun, and dumped her body off a bridge. In this version of the story, Butler asserted that White helped him dispose of the moped. Butler later took the police to the locations of the various events culminating in Lane’s death.

The State indicted Butler and brought him to trial on a charge of first-degree murder. The trial court denied Butler’s motion to suppress the statements given to police, and the statements were introduced into evidence. The jury found Butler guilty and, in a separate proceeding, sentenced him to death concluding that hé committed the murder during the commission of a rape. The Supreme Court of South Carolina upheld Butler’s conviction on direct appeal, State v. Butler, 277 S. C. 452, 290 S. E. 2d 1, and we denied certiorari. Butler v. South Carolina, 459 U. S. 932 (1982). Subsequently, Butler unsuccessfully petitioned for collateral relief in the State’s courts, see Butler v. State, 286 S. C. 441, 334 S. E. 2d 813 (1985), and we again denied certiorari. Butler v. South Carolina, 474 U. S. 1094 (1986).

In May 1986, Butler filed this petition for federal habeas relief pursuant to 28 U. S. C. §2254. As characterized by the District Court, one question raised in the petition was “whether police had the right to initiate questioning about the murder knowing petitioner had retained an attorney for the assault charge.” App. 119. The District Court dismissed the petition on respondents’ motion for summary judgment.

On appeal to the United States Court of Appeals for the Fourth Circuit, see Butler v. Aiken, 846 F. 2d 255 (1988), *411Butler argued that Edwards v. Arizona, 451 U. S. 477 (1981), requires the police, during continuous custody, to refrain from all further questioning once an accused invokes his right to counsel on any offense. In support of his argument, Butler relied principally on United States ex rel. Espinoza v. Fairman, 813 F. 2d 117 (CA7 1987). The Court of Appeals rejected Butler’s Espinoza-based contention, finding the Seventh Circuit’s ruling an unpersuasive and “dramatic” extension of Edivards. Butler, 846 F. 2d, at 258.

The court concluded that Butler’s statements were preceded by appropriate warnings and a voluntary waiver of Fifth Amendment protections. The statements, therefore, were not obtained in violation of his constitutional rights or Edwards’ prophylactic rule. According to the court, a properly initiated interrogation on an entirely different charge does not intrude into an accused’s previously invoked rights but instead offers the accused an opportunity to weigh his rights intelligently in light of changed circumstances. When, as occurred in this case, the accused then freely waives any constitutional right to counsel and provides voluntary statements of an incriminating nature, there is no justification for undermining the search for the truth by suppressing those statements. Butler, 846 F. 2d, at 259. The Court of Appeals affirmed the dismissal of Butler’s petition, and approximately one month later, denied Butler’s request for rehearing and suggestion for rehearing en banc.

On the same day the court denied Butler’s rehearing petitions, we handed down our decision in Roberson. We held in Roberson that the Fifth Amendment bars police-initiated interrogation following a suspect’s request for counsel in the context of a separate investigation. 486 U. S., at 682. On Butler’s motion for reconsideration, the original Fourth Circuit panel considered Butler’s new contention that Roberson requires suppression of his statements taken in the separate investigation of Lane’s murder. Although the panel conceded that the substance of its prior conclusion “was cast into *412immediate and serious doubt” by our subsequent decision in Roberson, Butler v. Aiken, 864 F. 2d 24, 25 (1988), it nevertheless determined that Butler was not entitled to the retroactive benefit of Roberson. According to the panel, the Edwards-Roberson limitations on police interrogation are only tangentially related to the truth-finding function. 864 F. 2d, at 25. They are viewed most accurately as part of the prophylactic protection of the Fifth Amendment right to counsel created to be “guidelines” for the law enforcement profession. Ibid, (citing Roberson, supra, at 680-682). The interrogation of Butler, while unquestionably contrary to present “guidelines,” was conducted in strict accordance with established law at the time. The panel, therefore, denied Butler’s petition for rehearing. A majority of the Circuit Judges denied, over a dissent, Butler’s petition for a rehearing en banc. We granted certiorari, 490 U. S. 1045 (1989), and now affirm.

Last Term in Penry v. Lynaugh, 492 U. S. 302 (1989), we held that in both capital and noncapital cases, “new rules will not be applied or announced in cases on collateral review unless they fall into one of two exceptions.” Id., at 313 (citing Teague v. Lane, 489 U. S., at 311-313; see infra, at 415-416 (discussing the exceptions and their inapplicability to the instant case). Referring to Teague, we reiterated that, in general, a case announces a “new rule” when it breaks new ground or imposes a new obligation on the States or the Federal Government. Penry, 492 U. S., at 314. Put differently, and, .indeed, more meaningfully for the majority of cases, a decision announces a new rule “ ‘if the result was not dictated by precedent existing at the time the defendant’s conviction became final.’” Ibid, (quoting Teague, supra, at 301) (emphasis in original).

A new decision that explicitly overrules an earlier holding obviously “breaks new ground” or “imposes a new obligation.” In the vast majority of cases, however, where the new decision is reached by an extension of the reasoning of *413previous cases, the inquiry will be more difficult. We said in Teague:

“ ‘The relevant frame of reference ... is not the purpose of the new rule whose benefit the [defendant] seeks, but instead the purposes for which the writ of habeas corpus is made available.’ Mackey[v. United States, 401 U. S. 667, 682 (1971) (Harlan, J., concurring in judgments in part and dissenting in part)]. . . . ‘The interest in leaving concluded litigation in a state of repose . . : may quite legitimately be found by those responsible for defining the scope of the writ to outweigh in some, many, or most instances the competing interest in readjudicating convictions according to all legal standards in effect when a habeas petition is filed.’ . . . Given the ‘broad scope of constitutional issues cognizable on habeas,’ ... it is ‘sounder, in adjudicating habeas petitions, generally to apply the law prevailing at the time a conviction became final than it is to seek to dispose of [habeas] cases on the basis of intervening changes in constitutional interpretation.’ . . . ‘[T]he threat of habeas serves as a necessary additional incentive for trial and appellate courts throughout the land to conduct their proceedings in a manner consistent with established constitutional standards. In order to perform this deterrence function, . . . the habeas court need only apply the constitutional standards that prevailed at the time the original proceedings took place.’” Teague, supra, at 306 (plurality opinion) (emphasis added; some brackets in original; some internal citations omitted).

Teague further observed:

“[I]n many ways the application of new rules to cases on collateral review may be more intrusive than the enjoining of [state] criminal prosecutions ... for it continually forces the States to marshal resources in order to keep in prison defendants whose trials and appeals conformed to *414then-existing constitutional standards. Furthermore, as we recognized in Engle v. Isaac,[456 U. S. 107, 128, n. 33 (1982),] ‘[s]tate courts are understandably frustrated when they faithfully apply existing constitutional law only to have a federal court discover, during a [habeas] proceeding, new constitutional commands.’ . . . See also Brown v. Allen, 344 U. S.[443], 534 [(1953)] (Jackson, J., concurring in result) (state courts cannot ‘anticipate, and so comply with, this Court’s due process requirements or ascertain any standards to which this Court will adhere in prescribing them’).” Teague, supra, at 310 (plurality opinion) (emphasis in original; some internal citations omitted).

The “new rule” principle therefore validates reasonable, good-faith interpretations of existing precedents made by state courts even though they are shown to be contrary to later decisions. Cf. United States v. Leon, 468 U. S. 897, 918-919 (1984) (assuming the exclusionary rule “effectively deters some police misconduct and provides incentives for the law enforcement profession as a whole to conduct itself in accord with the Fourth Amendment, it cannot be expected, and should not be applied, to deter objectively reasonable law enforcement activity”).

Butler contends that Roberson did not establish a new rule and is, therefore, available to support his habeas petition. Butler argues that Roberson was merely an application of Edwards to a slightly different set of facts. Brief for Petitioner 9; Reply Brief for Petitioner 18. In support of his position, Butler points out that the majority had said that Roberson’s case was directly controlled by Edwards. Brief for Petitioner 10. At oral argument Butler’s counsel also pointed out.that the Roberson opinion had rejected Arizona’s request to create an “exception” to Edivards for interrogations concerning separate investigations. Tr. of Oral Arg. 4. According to counsel, the opinion in Roberson showed that *415the Court believed Roberson’s case to be within the “logical compass” of Edwards. Tr. of Oral Arg. passim.

But the fact that a court says that its decision is within the “logical compass” of an earlier decision, or indeed that it is “controlled” by a prior decision, is not conclusive for purposes of deciding whether the current decision is a “new rule” under Teague. Courts frequently view their decisions as being “controlled” or “governed” by prior opinions even when aware of reasonable contrary conclusions reached by other courts. In Roberson, for instance, the Court found Edwards controlling but acknowledged a significant difference of opinion on the part of several lower courts that had considered the question previously. 486 U. S., at 679, n. 3. That the outcome in Roberson was susceptible to debate among reasonable minds is evidenced further by the differing positions taken by the judges of the Courts of Appeals for the Fourth and Seventh Circuits noted previously. It would not have been an illogical or even a grudging application of Edwards to decide that it did not extend to the facts of Roberson. We hold, therefore, that Roberson announced a “new rule.”

The question remains whether the new rule in Roberson nevertheless comes within one of the two recognized exceptions under which a new rule is available on collateral review. Under the first exception, “a new rule should be applied retroactively if it places ‘certain kinds of primary, private individual conduct beyond the power of the criminal law-making authority to proscribe.’” Teague, 489 U. S., at 307 (plurality opinion) (quoting Mackey, 401 U. S. 667, 692 (1971) (Harlan, J., concurring in judgments in part and dissenting in part)). This exception is clearly inapplicable. The proscribed conduct in the instant case is capital murder, the prosecution of which is, to put it mildly, not prohibited by the rule in Roberson. Nor did Roberson address any “categorical guarantees accorded by the Constitution” such as a prohibition on the imposition of a particular punishment on a certain class of offenders. See Penry, 492 U. S., at 329.

*416Under the second exception, a new rule may be applied on collateral review “if it requires the observance of ‘those procedures that . . . are “implicit in the concept of ordered liberty.”’” Teague, supra, at 311 (plurality opinion) (quoting Mackey, supra, at 693 (Harlan, J., concurring in judgments in part and dissenting in part) (in turn quoting Palko v. Connecticut, 302 U. S. 319, 325 (1937) (Cardozo, J.))). Teague, it should be noted, however, discerned a latent danger in relying solely on this famous language from Palko:

“Were we to employ the Palko test without more, we would be doing little more than importing into a very different context the terms of the debate over incorporation. . . . Reviving the Palko test now, in this area of law, would be unnecessarily anachronistic. . . . [W]e believe that Justice Harlan’s concerns about the difficulty in identifying both the existence and the value of accuracy-enhancing procedural rules can be addressed by limiting the scope of the second exception to those new procedures without which the likelihood of an accurate conviction is seriously diminished.

“Because we operate from the premise that such procedures would be so central to an accurate determination of innocence or guilt, we believe it unlikely that many such components of basic due process have yet to emerge.” Teague, supra, at 312-313 (plurality opinion).

Because a violation of Roberson’s added restrictions on police investigatory procedures would not seriously diminish the likelihood of obtaining an accurate determination — indeed, it may increase that likelihood — we conclude that Roberson did not establish any principle that would come within the second exception.

The judgment of the Court of Appeals is therefore

Affirmed.

Last Term in Teague v. Lane, 489 U. S. 288 (1989), this Court manifested its growing hostility toward Congress’ decision to authorize federal collateral review of state criminal convictions,1 curtailing the writ of habeas corpus by dramatically restructuring retroactivity doctrine. The plurality declared that a federal court entertaining a state prisoner’s habeas petition generally may not reach the merits of the legal claim unless the court determines, as a threshold matter, that a favorable ruling on the claim would flow from the application of legal standards “ ‘prevailing at the time [the petitioner’s] conviction became final.’” Id., at 306 (quoting Mackey v. United States, 401 U. S. 667, 689 (1971) (Harlan, J., concurring in judgment in part and dissenting in part)). Thus, with two narrow exceptions, Teague, supra, at 307, 311-313, “new” rules of law provide no basis for habeas relief. The plurality stated that a ruling qualifies as “new” “if the result was not dictated by precedent existing at the time the defendant’s conviction became final.” 489 U. S., at 301 (emphasis in original).

Today, under the guise of fine-tuning the definition of “new rule,” the Court strips state prisoners of virtually any meaningful federal review of the constitutionality of their incarceration. A legal ruling sought by a federal habeas petitioner is now deemed “new” as long as the correctness of the rule, based on precedent existing when the petitioner’s conviction became final, is “susceptible to debate among reasonable minds.” Ante, at 415. Put another way, a state prisoner can secure habeas relief only by showing that the state *418court’s rejection of the constitutional challenge was so clearly invalid under then-prevailing legal standards that the decision could not be defended by any reasonable jurist. With this requirement, the Court has finally succeeded in its thinly veiled crusade to eviscerate Congress’ habeas corpus regime.

I

Because constitutional interpretation is an evolutionary process, the analytical distinction between legal rules “prevailing” at the time of conviction and “new” legal rules is far from sharp. This distinction must be drawn carefully, with reference to the nature of adjudication in general and the purposes served by habeas corpus in particular. But while the Court purports to draw guidance from the retroactivity analysis advanced by Justice Harlan, see ante, at 413 (quoting Teague, supra), the Court simply ignores Justice Harlan’s admonition that “[t]he theory that the habeas petitioner is entitled to the law prevailing at the time of his conviction is . . . more complex than the Court has seemingly recognized.” Desist v. United States, 394 U. S. 244, 263 (1969) (Harlan, J., dissenting). Instead, the Court embraces a virtually all-encompassing definition of “new rule” without pausing to articulate any justification therefor. Result, not reason, propels the Court today.

A

The Court’s preclusion of federal habeas review for all but the most indefensible state-court rejections of constitutional challenges is made manifest by the Court’s conclusion that our recent holding in Arizona v. Roberson, 486 U. S. 675 (1988), qualifies as establishing a “new rule.” Long before Roberson, this Court recognized the presumptively coercive nature of custodial interrogations and held that an interrogation must cease if and when a suspect requests an attorney. “If the interrogation continues without the presence of an attorney and a statement is taken, a heavy burden rests upon the government to demonstrate that the defendant know*419ingly and intelligently waived his privilege against self-incrimination and his right to retained or appointed counsel.” Miranda v. Arizona, 384 U. S. 436, 475 (1966). In Edwards v. Arizona, 451 U. S. 477 (1981), we applied this rule to a situation where detectives renewed an interrogation of the accused about a series of offenses, after he had requested counsel during an earlier interrogation concerning the same offenses. We “reconfirm[ed] these views [expressed in Miranda, sitpra] and, to lend them substance, emphasize[d] that it is inconsistent with Miranda and its progeny for the authorities, at their instance, to reinterrogate an accused in custody if he has clearly asserted his right to counsel.” 451 U. S., at 485. We carefully considered the circumstances under which a suspect, who has requested counsel during a custodial interrogation, may be deemed to have validly waived his right to counsel prior to the resumption of interrogation. We concluded 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 interrogation even if he has been advised of his rights. ” Id., at 484. As a result, we established a bright-line rule: a suspect who has “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.” Id., at 484-485.

In Roberson, supra, the State of Arizona “ask[ed] us to craft an exception to that rule.” 486 U. S., at 677. Noting that Edivards involved two interrogations concerning the same offenses, the State of Arizona sought an exception “for cases in which the police want to interrogate a suspect about an offense that is unrelated to the subject of their initial interrogation.” 486 U. S., at 677. We declined, finding “unavailing” the State’s “attempts at distinguishing the factual setting here from that in Edwards.” Id., at 685. We ex*420plained that the rule articulated in Edwards reflected our concern that “if a suspect believes that he is not capable of undergoing [custodial] questioning without advice of counsel, then it is presumed that any subsequent waiver that has come at the authorities’ behest, and not at the suspect’s own instigation, is itself the product of the ‘inherently compelling pressures’ and not the purely voluntary choice of the suspect.” 486 U. S., at 681. That Roberson’s second interrogation concerned a different subject than his first in no way assuaged our concern that Roberson’s initial request for counsel reflected an inability to deal with police questioning without legal advice. Therefore Edwards’ application did not turn on the subject matter of the two interrogations. 486 U. S., at 682; see id., at 684 (“[T]here is no reason to assume that a suspect’s state of mind is in any way investigation-specific”). We likewise “attach[ed] no significance to the fact that the officer who conducted the second interrogation did not know that [Roberson] had made a request for counsel,” because “Edivards focuses on the state of mind of the suspect and not of the police.” Id., at 687.

B

It is clear from our opinion in Roberson that we would have reached the identical conclusion had that case reached us in 1983 when Butler’s conviction became final. In Roberson, we simply applied the legal principle established in Miranda and reconfirmed in Edivards to a set of facts that was not dissimilar in aiiy salient way. We did not articulate any new principles of Fifth Amendment jurisprudence that were not already established in 1983.

Yet today the Court classifies Roberson as a “new rule” notwithstanding the above, characterizing the “outcome in Roberson [as] susceptible to debate among reasonable minds.” Ante, at 415. For this conclusion, the majority appears to rely solely on the fact that the court below and several state courts had incorrectly predicted the outcome in Roberson by *421holding that the Edwards rule ought not apply where the second interrogation involves different subject matter. Ibid. But this reliance is perplexing. The majority might mean to suggest that a particular result is reasonable so long as a certain number of courts reach the same result. But this would be an odd criterion for “reasonableness.” Its application would be ad hoc, both because there appears to be no principled basis for choosing any particular number of courts whose agreement is required before the result is deemed “reasonable,” and because the criterion ultimately rests on a bootstrap to the extent that the later courts reaching the result simply rely on the earlier courts’ having done the same.

On the other hand, the majority might mean that the lower court decisions foreshadowing the dissent’s position in Roberson, though ultimately erroneous, were nevertheless “reasonable” according to some objective criterion of adjudication.2 But the Court does not purport to identify any such criterion or explain its application in this case. Instead, the Court announces in peremptory fashion that “[i]t would not have been an illogical or even a grudging application of Edwards to decide that it did not extend to the facts of Roberson.” Ante, at 415. This characterization is mystifying, given our explanation in Roberson that the result was clearly dictated by Edwards. See supra, at 419-420.

The only conclusion discernible from the majority’s discussion is that the majority would label “new” any rule of law favoring a state prisoner that can be distinguished from prior precedent on any conceivable basis, legal or factual.3 The *422converse of this conclusion is that, in the majority’s view, adjudication according to “prevailing” law must consist solely of applying binding precedents to factual disputes that cannot be distinguished from prior cases in any imaginable way. Because after Teague v. Lane, 489 U. S. 288 (1989), a federal court may entertain a habeas petition on the merits only if the petitioner seeks application of “prevailing” law as so narrowly defined, the majority today limits federal courts’ habeas corpus function to reviewing state courts’ legal analysis under the equivalent of a “clearly erroneous” standard of review. A federal court may no longer consider the’merits of the petitioner’s claim based on its best interpretation and application of the law prevailing at the time her conviction became final; rather, it must defer to the state court’s decision rejecting the claim unless that decision.is patently unreasonable.4

II

The Court’s exceedingly broad definition of “new rule”— and conversely its narrow definition of- “prevailing” law— betrays a vision of adjudication fundamentally at odds with any this Court has previously recognized. According to Justice Harlan, whose retroactivity jurisprudence undergirds *423Teague and its progeny: “One need not be a rigid partisan of Blackstone to recognize that many, though not all, of this Court’s constitutional decisions are grounded upon fundamental principles whose content does not change dramatically from year to year, but whose meanings are altered slowly and subtly as generation succeeds generation.” Desist, 394 U. S., at 263 (Harlan, J., dissenting). As every first-year law student learns, adjudication according to prevailing law means far more than obeying precedent by perfunctorily applying holdings in previous cases to virtually identical fact patterns. Rather, such adjudicátion requires a judge to evaluate both the content of previously enunciated legal rules and the breadth of their application. A judge must thereby discern whether the principles applied to specific fact patterns in prior cases fairly extend to govern analogous factual patterns. In Justice Harlan’s view, adjudication according to preyailing law demands that a court exhibit “conceptual faithfulness” to the principles underlying prior precedents, not just “decisional obedience” to precise holdings based upon their unique factual patterns. Id., at 266, n. 5 (employing Justice Fortas’ terminology). The inability of lower courts to predict significant reformulations by this Court of the principles underlying prior precedents does not excuse them from the obligation to draw reasoned conclusions from principles that are well established at the time of their decisions.

The majority suggests obliquely that adoption of a “‘new rule’ principle [that] validates reasonable, good-faith interpretations of existing precedents,” ante, at 414 — which in turn means that adjudication according to “prevailing” law requires only strict “decisional obedience” to existing precedents — would still serve the “deterrence function” animating federal habeas review. Ibid, (emphasis in original). But this claim begs a central question: deterrence of what? Under the definition of “prevailing” law embraced today, federal courts may not entertain habeas petitions challenging state-court rejections of constitutional claims unless those *424state decisions are clearly erroneoús. So at best, the threat of habeas review will deter state courts only from completely indefensible rejections of federal claims. State courts essentially are told today that, save for outright “illogical” defiance of a binding precedent precisely on point, their interpretations of federal constitutional guarantees — no matter how cramped and unfaithful to the principles underlying existing precedent — will no longer be subject to oversight through the federal habeas system. State prosecutors surely will offer every conceivable basis in each case for distinguishing our prior precedents, and state courts will be free to “‘disregard the plain purport of our decisions and to adopt a let’s-wait-until-it’s-decided [by the Supreme Court] approach.’” United States v. Johnson, 457 U. S. 537, 561 (1982) (quoting Desist, supra, at 277 (Fortas, J., dissenting)); cf. Johnson, supra, at 561 (rejecting contention that “all rulings resolving unsettled Fourth Amendment questions should be nonretroactive [to cases pending upon direct review because otherwise,] in close cases, law enforcement officials would have little incentive to err on the side of constitutional behavior”).5

This Court has never endorsed such a cramped view of the deterrent purpose of habeas review: we have always expected the threat of "habeas to encourage state courts to adjudicate federal claims “correctly,” not just “reasonably.” See, e. g., Teague, supra, at 306-307 (deterrence rationale requires “[r]eview on habeas to determine that the conviction rests upon correct application of the law in effect at the time of the conviction”) (emphasis added) (quoting Solem v. Stumes, 465 U. S. 638, 653 (1984) (Powell, J., concurring *425in judgment)). And, as explained above, “correct” adjudication has always been thought to require courts to exhibit “conceptual faithfulness” to the principles underlying our precedents and thereby to anticipate reasonably foreseeable applications of those principles. See, e. g., Johnson, supra, at 549 (“When a decision of this Court merely has applied settled precedents to new and different factual situations, no real question [of retroactivity] has arisen .... In such cases, it has been a foregone conclusion that the rule of the later case applies in earlier cases, because the later decision has not in fact altered that rule in any material way”).6

*426Indeed, even Justice Harlan, the chief proponent of the view that federal habeas is designed merely to deter erroneous state-court rejections of constitutional claims, believed that federal review is appropriate when a state court fails to presage reasonably foreseeable applications of established constitutional principles beyond the precise factual settings of prior precedent. Justice Harlan would have held state courts responsible for “applying] a well-established constitutional principle to govern a case which is closely analogous to those which have been previously considered in the prior case law.” Desist, 394 U. S., at 263. In the context of this case, Justice Harlan would not have held the rule in Roberson to be “new” today unless he could “say with . . . assurance that this Court would have ruled differently” (i. e., in the State’s favor) at the time Butler’s conviction became final. 394 U. S., at 264 (emphasis added). In contrast, the majority embraces the opposite presumption; it holds Roberson’s rule to be “new” because it cannot say with assurance that the Court could not have ruled in favor of the State at that time.7 Thus the Court’s holding today is unfaithful even to the purported progenitor of its position.

*427Most significantly, the limited scope of the deterrence promised by the Court’s holding today defeats Congress’ purpose in establishing the scheme of federal habeas review of state criminal proceedings. Congress established such review because it perceived a potential “inadequacy of state procedures to raise and preserve federal claims [and was] concerned] that state judges may be unsympathetic to federally created rights.” Kaufman v. United States, 394 U. S. 217, 225-226 (1969).8 Congress intended the “[t]hreat of habeas” both to “serv[e] as a necessary additional incentive for [state] trial and appellate courts throughout the land to conduct their proceedings in a manner consistent with established constitutional standards,” ante, at 413 (citations omitted), and to provide petitioners a remedy for unlawful state deprivations of their liberty interests through a fresh and full review of their claims by an Article III court. As we recognized in Fay v. Noia, 372 U. S. 391, 424 (1963), “the manifest federal policy [underlying § 2254(a) is] that federal constitutional rights of personal liberty shall not be denied without the fullest opportunity for plenary federal judicial review” (emphasis added). See also, e. g., Kaufman, supra, at 228 (“Congress has determined that the full protection of . . . constitutional rights requires the availability of a mechanism for collateral attack. The right then is not merely to a fed*428eral forum but to full and fair consideration of constitutional claims”). It has long been established, therefore, that federal habeas proceedings ought not accord any deference to the state court’s constitutional ruling under collateral attack.9 Instead, the federal court must determine for itself the proper scope of constitutional principles and their application to the particular factual circumstances. As explained by Justice Frankfurter, “[t]he congressional requirement is [that the] State court cannot have the last say when it, though on fair consideration and what procedurally may be deemed fairness, may have misconceived a federal constitutional right.” Brown v. Allen, 344 U. S. 443, 508 (1953). Congress thus entitled petitioners to de novo review of their federal claims in federal habeas proceedings.10 But the *429Court’s decision today denies federal courts the role on habeas review that Congress envisioned because it limits them to remedying only clearly unreasonable state-court applications of federal law, rather than all erroneous ones.

Moreover, Congress’ insistence that “federal courts have the ‘last say’ with respect to questions of federal law” raised during state criminal proceedings, Kaufman, supra, at 225, cannot be satisfied by this Court’s jurisdiction to review state proceedings directly. State courts are well aware that the “Supreme Court’s burden and responsibility are too great to permit it to review and correct every misstep made by the lower courts in the application of accepted principles. Hence the Court generally will not grant certiorari just because the decision below may be erroneous.” R. Stern, E. Gressman, & S. Shapiro, Supreme Court Practice § 4.17, p. 221 (6th ed. 1986).11 We have long recognized that Congress’ decision in 1867 to “exten[d] to state prisoners . . . the federal habeas corpus remedy bespoke congressional unwillingness to trust direct appellate review of state court decisions by the Supreme Court as the lone avenue of vindication of the new constitutional strictures” of the Fourteenth Amendment. *430Brennan, Federal Habeas Corpus and State Prisoners: An Exercise in Federalism, 7 Utah L. Rev. 423, 426 (1961). But today’s decision, essentially foreclosing habeas review as an alternative “avenue of vindication,” overrides Congress’ will, and leaves federal judicial protection of fundamental constitutional rights during the state criminal process solely to this Court upon direct review. I share Congress’ lack of confidence in such a regime. After today, despite constitutional defects in the state processes leading to their conviction or sentencing, state prisoners will languish in jail — and others like Butler will die — because state courts were reasonable, even though wrong.12

*431The majority apparently finds such injustice acceptable based upon an asserted “ ‘interest in leaving concluded litigation in a state of repose. ’ ” Ante, at 413 (quoting Teague, 489 U. S., at 306). This will not do. It is one thing to preclude federal habeas petitioners from asserting claims based on legal principles contrary to or at least significantly dissimilar from those in existence at the time their convictions became final; such a basis for habeas relief engenders the possibility of “ ‘continually forcing] the States to marshal resources in order to keep in prison defendants whose trials and appeals conformed to then-existing constitutional standards. ’ ” Ante, at 413-414 (quoting Teague, supra, at 310). It is a far different thing to say that concerns for repose and resource scarcity justify today’s judicial decision to protect States from the consequences of retrying or resentencing defendants whose trials and appeals did not conform to then-existing constitutional standards but are viewed as suffering from only “reasonable” defects. “This Court has never held . . . that finality, standing alone, provides a sufficient reason for federal courts to compromise their protection of constitutional rights under §2254.” Reed v. Ross, 468 U. S. 1, 15 (1984). Until today.

Ill

It is Congress and not this Court who is “ ‘responsible for defining the scope of the writ.’ ” Ante, at 413 (citations omit*432ted).13 Yet the majority, whose Members often pride themselves on their reluctance to play an “activist” judicial role by infringing upon legislative prerogatives, does hot hesitate today to dismantle Congress’ extension of federal habeas to state prisoners. Hereafter, federal habeas relief will be available in only the most egregious cases, in which state courts have flouted applicable Supreme Court precedent that cannot be distinguished on any arguable basis. I must dissent from this curtailment of the writ’s capacity for securing individual liberty. “For surely it is an abuse to deal too casually and too lightly with fights guaranteed by the Federal Constitution, even though they involve limitations upon State power and may be invoked by those morally unworthy. Under the guise of fashioning a procedural rule, we are not justified in wiping out the practical efficacy of a jurisdiction conferred by Congress on the District Courts.” Brown, 344 U. S., at 498-499 (opinion of Frankfurter, J.).

IV

Even if I did not believe that petitioner is entitled in this habeas proceeding to claim the protections of the Fifth Amendment as defined by this Court in Roberson, I would vacate his death sentence. I adhere to my view that the death penalty is in all circumstances cruel and unusual punishment. Gregg v. Georgia, 428 U. S. 153, 227 (1976).