delivered the opinion of the Court.

This case calls for specific application of the standard of reasonable competence required on the part of defense counsel by the Sixth Amendment. We hold that even when a capital defendant’s family members and the defendant himself have suggested that no mitigating evidence is available, his lawyer is bound to make reasonable efforts to obtain and review material that counsel knows the prosecution will probably rely on as evidence of aggravation at the sentencing phase of trial.

I

On the morning of January 14, 1988, James Scanlon was discovered dead in a bar he ran in Allentown, Pennsylvania, his body having been stabbed repeatedly and set on fire. Ronald Rompilla was indicted for the murder and related offenses, and the Commonwealth gave notice of intent to ask *378for the death penalty. Two public defenders were assigned to the case.

The jury at the guilt phase of trial found Rompilla guilty on all counts, and during the ensuing penalty phase, the prosecutor sought to prove three aggravating factors to justify a death sentence: that the murder was committed in the course of another felony; that the murder was committed by torture; and that Rompilla had a significant history of felony convictions indicating the use or threat of violence. See 42 Pa. Cons. Stat. §§ 9711(d)(6), (8), (9) (2002). The Commonwealth presented evidence on all three aggravators, and the jury found all proven. Rompilla’s evidence in mitigation consisted of relatively brief testimony: five of his family members argued in effect for residual doubt, and beseeched the jury for mercy, saying that they believed Rompilla was innocent and a good man. Rompilla’s 14-year-old son testified that he loved his father and would visit him in prison. The jury acknowledged this evidence to the point of finding, as two factors in mitigation, that Rompilla’s son had testified on his behalf and that rehabilitation was possible. But the jurors assigned the greater weight to the aggravating factors, and sentenced Rompilla to death. The Supreme Court of Pennsylvania affirmed both conviction and sentence. Commonwealth v. Rompilla, 539 Pa. 499, 653 A. 2d 626 (1995).

In December 1995, with new lawyers, Rompilla filed claims under the Pennsylvania Post Conviction Relief Act, 42 Pa. Cons. Stat. § 9541 et seq. (2004), including ineffective assistance by trial counsel in failing to present significant mitigating evidence about Rompilla’s childhood, mental capacity and health, and alcoholism. The postconviction court found that trial counsel had done enough to investigate the possibilities of a mitigation case, and the Supreme Court of Pennsylvania affirmed the denial of relief. Commonwealth v. Rompilla, 554 Pa. 378, 721 A. 2d 786 (1998).

*379Rompilla then petitioned for a writ of habeas corpus under 28 U. S. C. §2254 in Federal District Court, raising claims that included inadequate representation. The District Court found that the State Supreme Court had unreasonably applied Strickland v. Washington, 466 U. S. 668 (1984), as to the penalty phase of the trial, and granted relief for ineffective assistance of counsel. The court found that in preparing the mitigation case the defense lawyers had failed to investigate “pretty obvious signs” that Rompilla had a troubled childhood and suffered from mental illness and alcoholism, and instead had relied unjustifiably on Rompilla’s own description of an unexceptional background. Rompilla v. Horn, No. CIV.A.99-737 (ED Pa., July 11, 2000), App. 1307-1308.

A divided Third Circuit panel reversed. Rompilla v. Horn, 355 F. 3d 233 (2004). The majority found nothing unreasonable in the state court’s application of Strickland, given defense counsel’s efforts to uncover mitigation material, which included interviewing Rompilla and certain family members, as well as consultation with three mental health experts. Although the majority noted that the lawyers did not unearth the “useful information” to be found in Rompilla’s “school, medical, police, and prison records,” it thought the lawyers were justified in failing to hunt through these records when their other efforts gave no reason to believe the search would yield anything helpful. 355 F. 3d, at 252. The panel thus distinguished Rompilla’s case from Wiggins v. Smith, 539 U. S. 510 (2003). Whereas Wiggins’s counsel failed to investigate adequately, to the point even of ignoring the leads their limited enquiry yielded, the Court of Appeals saw the Rompilla investigation as going far enough to leave counsel with reason for thinking further efforts would not be a wise use of the limited resources they had. But Judge Sloviter’s dissent stressed that trial counsel’s failure to obtain relevant records on Rompilla’s background was owing to the lawyers’ unreasonable reliance on *380family members and medical experts to tell them what records might be useful. The Third Circuit denied rehearing en banc by a vote of 6 to 5. Rompilla v. Horn, 359 F. 3d 310 (2004).

We granted certiorari, 542 U. S. 966 (2004), and now reverse.1

II

Under 28 U. S. C. § 2254, Rompilla’s entitlement to federal habeas relief turns on showing that the state court’s resolution of his claim of ineffective assistance of counsel under Strickland v. Washington, supra, “resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States,” § 2254(d)(1). An “unreasonable application” occurs when a state court “‘identifies the correct governing legal principle from this Court’s decisions but unreasonably applies that principle to the facts’ of petitioner’s case.” Wiggins v. Smith, supra, at 520 (quoting Williams v. Taylor, 529 U. S. 362, 413 (2000) (opinion of O’Connor, J.)). That is, “the state court’s decision must have been [not only] incorrect or erroneous [but] objectively unreasonable.” Wiggins v. Smith, supra, at 520-521 (quoting Williams v. Taylor, supra, at 409 (internal quotation marks omitted)).

Ineffective assistance under Strickland is deficient performance by counsel resulting in prejudice, 466 U. S., at 687, with performance being measured against an “objective standard of reasonableness,” id., at 688, “under prevailing professional norms,” ibid.; Wiggins v. Smith, supra, at 521. This case, like some others recently, looks to norms of adequate investigation in preparing for the sentencing phase of a capital trial, when defense counsel’s job is to counter the *381State’s evidence of aggravated culpability with evidence in mitigation. In judging the defense’s investigation, as in applying Strickland generally, hindsight is discounted by pegging adequacy to “counsel’s perspective at the time” investigative decisions are made, 466 U. S., at 689, and by giving a “heavy measure of deference to counsel’s judgments,” id., at 691.

A

A standard of reasonableness applied as if one stood in counsel’s shoes spawns few hard-edged rules, and the merits of a number of counsel’s choices in this case are subject to fair debate. This is not a case in which defense counsel simply ignored their obligation to find mitigating evidence, and their workload as busy public defenders did not keep them from making a number of efforts, including interviews with Rompilla and some members of his family, and examinations of reports by three mental health experts who gave opinions at the guilt phase. None of the sources proved particularly helpful.

Rompilla’s own contributions to any mitigation case were minimal. Counsel found him uninterested in helping, as on their visit to his prison to go over a proposed mitigation strategy, when Rompilla told them he was “bored being here listening” and returned to his cell. App. 668. To questions about childhood and schooling, his answers indicated they had been normal, ibid., save for quitting school in the ninth grade, id., at 677. There were times when Rompilla was even actively obstructive by sending counsel off on false leads. Id., at 663-664.

The lawyers also spoke with five members of Rompilla’s family (his former wife, two brothers, a sister-in-law, and his son), id., at 494, and counsel testified that they developed a good relationship with the family in the course of their representation, id., at 669, 729. The state postconvietion court found that counsel spoke to the relatives in a “detailed manner,” attempting to unearth mitigating information, id., at 264, although the weight of this finding is qualified by the *382lawyers’ concession that “the overwhelming response from the family was that they didn’t really feel as though they knew him all that well since he had spent the majority of his adult years and some of his childhood years in custody,” id., at 495; see also id., at 669. Defense counsel also said that because the family was “coming from the position that [Rom-pilla] was innocent. . . they weren’t looking for reasons for why he might have done this.” Id., at 494.

The third and final source tapped for mitigating material was the cadre of three mental health witnesses who were asked to look into Rompilla’s mental state as of the time of the offense and his competency to stand trial. Id., at 473-474, 476. But their reports revealed “nothing useful” to Rompilla’s ease, id., at 1358, and the lawyers consequently did not go to any other historical source that might have cast light on Rompilla’s mental condition.

When new counsel entered the case to raise Rompilla’s postconviction claims, however, they identified a number of likely avenues the trial lawyers could fruitfully have followed in building a mitigation case. School records are one example, which trial counsel never examined in spite of the professed unfamiliarity of the several family members with Rompilla’s childhood, and despite counsel’s knowledge that Rompilla left school after the ninth grade. Id., at 677. Other examples are records of Rompilla’s juvenile and adult incarcerations, which counsel did not consult, although they were aware of their client’s criminal record. And while counsel knew from police reports provided in pretrial discovery that Rompilla had been drinking heavily at the time of his offense, Lodging to App. 111-120 (hereinafter Lodging), and although one of the mental health experts reported that Rompilla’s troubles with alcohol merited further investigation, App. 723-724, counsel did not look for evidence of a history of dependence on alcohol that might have extenuating significance.

Before us, trial counsel and the Commonwealth respond to these unexplored possibilities by emphasizing this Court’s *383recognition that the duty to investigate does not force defense lawyers to scour the globe on the off chance something will turn up; reasonably diligent counsel may draw a line when they have good reason to think further investigation would be a waste. See Wiggins v. Smith, 539 U. S., at 525 (further investigation excusable where counsel has evidence suggesting it would be fruitless); Strickland v. Washington, supra, at 699 (counsel could “reasonably surmise . . . that character and psychological evidence would be of little help”); Burger v. Kemp, 483 U. S. 776, 794 (1987) (limited investigation reasonable because all witnesses brought to counsel’s attention provided predominantly harmful information). The Commonwealth argues that the information trial counsel gathered from Rompilla and the other sources gave them sound reason to think it would have been pointless to spend time and money on the additional investigation espoused by postconviction counsel, and we can say that there is room for debate about trial counsel’s obligation to follow at least some of those potential lines of enquiry. There is no need to say more, however, for a further point is clear and dispositive: the lawyers were deficient in failing to examine the court file on Rompilla’s prior conviction.

B

There is an obvious reason that the failure to examine Rompilla’s prior conviction file fell below the level of reasonable performance. Counsel knew that the Commonwealth intended to seek the death penalty by proving Rompilla had a significant history of felony convictions indicating the use or threat of violence, an aggravator under state law. Counsel further knew that the Commonwealth would attempt to establish this history by proving Rompilla’s prior conviction for rape and assault, and would emphasize his violent character by introducing a transcript of the rape victim’s testimony given in that earlier trial. App. 665-666. There is no question that defense counsel were on notice, since they acknowledge that a “plea letter,” written by one of them four days *384prior to trial, mentioned the prosecutor’s plans. Ibid. It is also undisputed that the prior conviction file was a public document, readily available for the asking at the very courthouse where Rompilla was to be tried.

It is clear, however, that defense counsel did not look at any part of that file, including the transcript, until warned by the prosecution a second time. In a colloquy the day before the evidentiary sentencing phase began, the prosecutor again said he would present the transcript of the victim’s testimony to establish the prior conviction.

“[DEFENSE]: I would also like to review whatever he’s going to read from.

“[PROSECUTOR]: Well, I told you that I was going to do this a long time ago. You certainly had the opportunity to review the Transcript.

“[DEFENSE]: Well, I would like a copy of this.

“[PROSECUTOR]: I don’t think that’s my duty to provide you with a copy. That’s a public record, and you could have gotten that Transcript at any time prior to this Trial. I made one copy for myself, and I’d like to have it now.

“[DEFENSE]: Well, Judge, then I’m going to need to get a copy of it. I’m going to need to get a copy of it.” Id., at 32, 36.2

*385At the postconviction evidentiary hearing, Rompilla’s lawyer confirmed that she had not seen the transcript before the hearing in which this exchange took place, id., at 506-507, and crucially, even after obtaining the transcript of the victim’s testimony on the eve of the sentencing hearing, counsel apparently examined none of the other material in the file.3

With every effort to view the facts as a defense lawyer would have done at the time, it is difficult to see how counsel could have failed to realize that without examining the readily available file they were seriously compromising their opportunity to respond to a case for aggravation. The prosecution was going to use the dramatic facts of a similar prior offense, and Rompilla’s counsel had a duty to make all reasonable efforts to learn what they could about the offense. Reasonable efforts certainly included obtaining the Commonwealth’s own readily available file on the prior conviction to learn what the Commonwealth knew about the crime, to discover any mitigating evidence the Commonwealth would downplay, and to anticipate the details of the aggravating *386evidence the Commonwealth would emphasize.4 Without making reasonable efforts to review the file, defense counsel could have had no hope of knowing whether the prosecution was quoting selectively from the transcript, or whether there were circumstances extenuating the behavior described by the victim. The obligation to get the file was particularly pressing here owing to the similarity of the violent prior offense to the crime charged and Rompilla’s sentencing strategy stressing residual doubt. Without making efforts to learn the details and rebut the relevance of the earlier crime, a convincing argument for residual doubt was certainly beyond any hope.5

*387The notion that defense counsel must obtain information that the State has and will use against the defendant is not simply a matter of common sense. As the District Court points out, the American Bar Association Standards for Criminal Justice in circulation at the time of Rompilla’s trial describes the obligation in terms no one could misunderstand in the circumstances of a case like this one:

“It is the duty of the lawyer to conduct a prompt investigation of the circumstances of the case and to explore all avenues leading to facts relevant to the merits of the case and the penalty in the event of conviction. The investigation should always include efforts to secure information in the possession of the prosecution and law enforcement authorities. The duty to investigate exists regardless of the accused’s admissions or statements to the lawyer of facts constituting guilt or the accused’s stated desire to plead guilty.” 1 ABA Standards for Criminal Justice 4-4.1 (2d ed. 1982 Supp.).6

“[W]e long have referred [to these ABA Standards] as ‘guides to determining what is reasonable.’” Wiggins v. Smith, 539 U. S., at 524 (quoting Strickland v. Washington, 466 U. S., at 688), and the Commonwealth has come up with no reason to think the quoted standard impertinent here.7

*388At argument the most that Pennsylvania (and the United States as amicus) could say was that defense counsel’s efforts to find mitigating evidence by other means excused them from looking at the prior conviction file. Tr. of Oral Arg. 37-39, 45-46. And that, of course, is the position taken by the state postconviction courts. Without specifically discussing the prior case file, they too found that defense eoun-*389sel’s efforts were enough to free them from any obligation to enquire further. Commonwealth v. Rompilla, No. 682/ 1988 (Pa. Ct. Common Pleas, Aug. 28, 1996), App. 263-264, 272-273.

We think this conclusion of the state court fails to answer the considerations we have set out, to the point of being an objectively unreasonable conclusion. It flouts prudence to deny that a defense lawyer should try to look at a file he knows the prosecution will cull for aggravating evidence, let alone when the file is sitting in the trial courthouse, open for the asking. No reasonable lawyer would forgo examination of the file thinking he could do as well by asking the defendant or family relations whether they recalled anything helpful or damaging in the prior victim’s testimony. Nor would a reasonable lawyer compare possible searches for school reports, juvenile records, and evidence of drinking habits to the opportunity to take a look at a file disclosing what the prosecutor knows and even plans to read from in his case. Questioning a few more family members and searching for old records can promise less than looking for a needle in a haystack, when a lawyer truly has reason to doubt there is any needle there. E.g., Strickland, supra, at 699. But looking at a file the prosecution says it will use is a sure bet: whatever may be in that file is going to tell defense counsel something about what the prosecution can produce.

The dissent thinks this analysis creates a “rigid, per se” rule that requires defense counsel to do a complete review of the file on any prior conviction introduced, post, at 404 (opinion of Kennedy, J.), but that is a mistake. Counsel fell short here because they failed to make reasonable efforts to review the prior conviction file, despite knowing that the prosecution intended to introduce Rompilla’s prior conviction not merely by entering a notice of conviction into evidence but by quoting damaging testimony of the rape victim in that case. The unreasonableness of attempting no more than they did was heightened by the easy availability of the file *390at the trial courthouse, and the great risk that testimony about a similar violent crime would hamstring counsel’s chosen defense of residual doubt. It is owing to these circumstances that the state courts were objectively unreasonable in concluding that counsel could reasonably decline to make any effort to review the file. Other situations, where a defense lawyer is not charged with knowledge that the prosecutor intends to use a prior conviction in this way, might well warrant a different assessment.

C

Since counsel’s failure to look at the file fell below the line of reasonable practice, there is a further question about prejudice, that is, whether “there is a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.” 466 U. S., at 694. Because the state courts found the representation adequate, they never reached the issue of prejudice, App. 265, 272-273, and so we examine this element of the Strickland claim de novo, Wiggins v. Smith, 539 U. S., at 534, and agree with the dissent in the Court of Appeals. We think Rompilla has shown beyond any doubt that counsel’s lapse was prejudicial; Pennsylvania, indeed, does not even contest the claim of prejudice.

If the defense lawyers had looked in the file on Rompilla’s prior conviction, it is uncontested they would have found a range of mitigation leads that no other source had opened up. In the same file with the transcript of the prior trial were the records of Rompilla’s imprisonment on the earlier conviction, App. 508, 571, 631, which defense counsel testified she had never seen, id., at 508. The prison files pictured Rompilla’s childhood and mental health very differently from anything defense counsel had seen or heard. An evaluation by a corrections counselor states that Rompilla was “reared in the slum environment of Allentown, Pa. vicinity. He early came to [the] attention of juvenile authorities, quit *391school at 16, [and] started a series of incarcerations in and out Penna. often of assaultive nature and commonly related to over-indulgence in alcoholic beverages.” Lodging 40. The same file discloses test results that the defense’s mental health experts would have viewed as pointing to schizophrenia and other disorders, and test scores showing a third grade level of cognition after nine years of schooling. Id., at 32-35.8

The accumulated entries would have destroyed the benign conception of Rompilla’s upbringing and mental capacity defense counsel had formed from talking with Rompilla himself and some of his family members, and from the reports of the mental health experts. With this information, counsel would have become skeptical of the impression given by the five family members and would unquestionably have gone further to build a mitigation case. Further effort would presumably have unearthed much of the material postconviction counsel found, including testimony from several members of Rompilla’s family, whom trial counsel did not interview. Judge Sloviter summarized this evidence:

“Rompilla’s parents were both severe alcoholics who drank constantly. His mother drank during her preg*392nancy with Rompilla, and he and his brothers eventually developed serious drinking problems. His father, who had a vicious temper, frequently beat Rompilla’s mother, leaving her bruised and black-eyed, and bragged about his cheating on her. His parents fought violently, and on at least one occasion his mother stabbed his father. He was abused by his father who beat him when he was young with his hands, fists, leather straps, belts and sticks. All of the children lived in terror. There were no expressions of parental love, affection or approval. Instead, he was subjected to yelling and verbal abuse. His father locked Rompilla and his brother Richard in a small wire mesh dog pen that was filthy and excrement filled. He had an isolated background, and was not allowed to visit other children or to speak to anyone on the phone. They had no indoor plumbing in the house, he slept in the attic with no heat, and the children were not given clothes and attended school in rags.” 355 F. 3d, at 279 (dissenting opinion) (citations omitted).

The jury never heard any of this and neither did the mental health experts who examined Rompilla before trial. While they found “nothing helpful to [Rompilla’s] case,” Rompilla, 554 Pa., at 385, 721 A. 2d, at 790, their postconviction counterparts, alerted by information from school, medical, and prison records that trial counsel never saw, found plenty of “‘red flags’” pointing up a need to test further. 355 F. 3d, at 279 (Sloviter, J., dissenting). When they tested, they found that Rompilla “suffers from organic brain damage, an extreme mental disturbance significantly impairing several of his cognitive functions.” Ibid. They also said that “Rompilla’s problems relate back to his childhood, and were likely caused by fetal alcohol syndrome [and that] Rom-pilla’s capacity to appreciate the criminality of his conduct or to conform his conduct to the law was substantially impaired at the time of the offense.” Id., at 280 (Sloviter, J., dissenting).

*393These findings in turn would probably have prompted a look at school and juvénile records, all of them easy to get, showing, for example, that when Rompilla was 16 his mother “was missing from home frequently for a period of one or several weeks at a time.” Lodging 44. The same report noted that his mother “has been reported . . . frequently under the influence of alcoholic beverages, with the result that the children have always been poorly kept and on the filthy side which was also the condition of the home at all times.” Ibid. School records showed Rompilla’s IQ was in the mentally retarded range. Id., at 11, 13, 15.

This evidence adds up to a mitigation case that bears no relation to the few naked pleas for mercy actually put before the jury, and although we suppose it is possible that a jury could have heard it all and still have decided on the death penalty, that is not the test. It goes without saying that the undiscovered “mitigating evidence, taken as a whole, ‘might well have influenced the jury’s appraisal’ of [Rompilla’s] culpability,” Wiggins v. Smith, 539 U. S., at 538 (quoting Williams v. Taylor, 529 U. S., at 398), and the likelihood of a different result if the evidence had gone in is “sufficient to undermine confidence in the outcome” actually reached at sentencing, Strickland, 466 U. S., at 694.

The judgment of the Third Circuit is reversed, and Pennsylvania must either retry the case on penalty or stipulate to a life sentence.

It is so ordered.

concurring.

I write separately to put to rest one concern. The dissent worries that the Court’s opinion “imposes on defense counsel a rigid requirement to review all documents in what it calls the ‘case file’ of any prior conviction that the prosecution might rely on at trial.” Post, at 396 (opinion of KENNEDY, J.). But the Court’s opinion imposes no such rule. See ante, at 389-390. Rather, today’s decision simply ap*394plies our longstanding case-by-case approach to determining whether an attorney’s performance was unconstitutionally deficient under Strickland v. Washington, 466 U. S. 668 (1984). Trial counsel’s performance in Rompilla’s case falls short under that standard, because the attorneys’ behavior was not “reasonable considering all the circumstances.” Id., at 688. In particular, there were three circumstances which made the attorneys’ failure to examine Rompilla’s prior conviction file unreasonable.

First, Rompilla’s attorneys knew that their client’s prior conviction would be at the very heart of the prosecution’s case. The prior conviction went not to a collateral matter, but rather to one of the aggravating circumstances making Rompilla eligible for the death penalty. The prosecutors intended not merely to mention the fact of prior conviction, but to read testimony about the details of the crime. That crime, besides being quite violent in its own right, was very, similar to the murder for which Rompilla was on trial, and Rompilla had committed the murder at issue a mere three months after his release from prison on the earlier conviction. In other words, the prosecutor clearly planned to use details of the prior crime as powerful evidence that Rompilla was a dangerous man for whom the death penalty would be both appropriate punishment and a necessary means of incapacitation. Cf. App. 165-166 (prosecutor’s penalty-phase argument). This was evidence the defense should , have been prepared to meet: A reasonable defense lawyer would have attached a high importance to obtaining the record of the prior trial, in order to anticipate and find ways of deflecting the prosecutor’s aggravation argument.

Second, the prosecutor’s planned use of the prior conviction threatened to eviscerate one of the defense’s primary mitigation arguments. Rompilla was convicted on the basis of strong circumstantial evidence. His lawyers structured the entire mitigation argument around the hope of convincing the jury that residual doubt about Rompilla’s guilt made *395it inappropriate to impose the death penalty. In announcing an intention to introduce testimony about Rompilla’s similar prior offense, the prosecutor put Rompilla’s attorneys on notice that the prospective defense on mitigation likely would be ineffective and counterproductive. The similarities between the two crimes, combined with the timing and the already strong circumstantial evidence, raised a strong likelihood that the jury would reject Rompilla’s residual doubt argument. Rompilla’s attorneys’ reliance on this transparently weak argument risked damaging their credibility. Such a scenario called for further investigation, to determine whether circumstances of the prior case gave any hope of saving the residual doubt argument, or whether the best strategy instead would be to jettison that argument so as to focus on other, more promising issues. Cf. Yarborough v. Gentry, 540 U. S. 1, 7 (2003) (per curiam); Bell v. Cone, 535 U. S. 685, 700 (2002) (noting that sound tactical judgment may sometimes call for omitting certain defense evidence or arguments).

Third, the attorneys’ decision not to obtain Rompilla’s prior conviction file was not the result of an informed tactical decision about how the lawyers’ time would best be spent. Although Rompilla’s attorneys had ample warning that the details of Rompilla’s prior conviction would be critical to their case, their failure to obtain that file would not necessarily have been deficient if it had resulted from the lawyers’ careful exercise of judgment about how best to marshal their time and serve their client. But Rompilla’s attorneys did not ignore the prior case file in order to spend their time on other crucial leads. They did not determine that the file was so inaccessible or so large that examining it would necessarily divert them from other trial-preparation tasks they thought more promising. They did not learn at the 11th hour about the prosecution’s intent to use the prior conviction, when it was too late for them to change plans. Rather, their failure to obtain the crucial file “was the result of in*396attention, not reasoned strategic judgment.” Wiggins v. Smith, 539 U. S. 510, 534 (2003). As a result, their conduct fell below constitutionally required standards. See id., at 533 (“ ‘[Strategic choices made after less than complete investigation are reasonable’ only to the extent that ‘reasonable professional judgments support the limitations on investigation’ ” (quoting Strickland, 466 U. S., at 690-691)).

In the particular circumstances of this case, the attorneys’ failure to obtain and review the case file from their client’s prior conviction did not meet standards of “reasonable professional judgmen[t].” Id., at 691. Because the Court’s opinion is consistent with the “ ‘case-by-case examination of the evidence’ ” called for under our cases, Williams v. Taylor, 529 U. S. 362, 391 (2000), I join the opinion.

dissenting.

Today the Court brands two committed criminal defense attorneys as ineffective — “outside the wide range of professionally competent assistance,” Strickland v. Washington, 466 U. S. 668, 690 (1984)—because they did not look in an old ease file and stumble upon something they had not set out to find. By implication the Court also labels incompetent the work done by the three mental health professionals who examined Ronald Rompilla. To reach this result, the majority imposes on defense counsel a rigid requirement to review all documents in what it calls the “case file” of any prior conviction that the prosecution might rely on at trial. The Court’s holding, a mistake under any standard of review, is all the more troubling because this case arises under the Antiter-rorism and Effective Death Penalty Act of 1996. In order to grant Rompilla habeas relief the Court must say, and indeed does say, that the Pennsylvania Supreme Court was objectively unreasonable in failing to anticipate today’s new case file rule.

*397In my respectful submission it is this Court, not the state court, which is unreasonable. The majority’s holding has no place in our Sixth Amendment jurisprudence and, if followed, often will result in less effective counsel by diverting limited defense resources from other important tasks in order to satisfy the Court’s new per se rule. Finally, even if the Court could justify its distortion of Strickland, Rompilla still would not be entitled to relief. The Court is able to conclude otherwise only by ignoring the established principle that it is the defendant, not the State, who has the burden of demonstrating that he was prejudiced by any deficiency in his attorneys’ performance.

These are the reasons for my dissent.

I

Under any standard of review the investigation performed by Rompilla’s counsel in preparation for sentencing was not only adequate but also conscientious.

Rompilla’s attorneys recognized from the outset that building an effective mitigation case was crucial to helping their client avoid the death penalty. App. 516, 576. Rom-pilla stood accused of a brutal crime. In January 1988, James Scanlon was murdered while he was closing the Cozy Corner Cafe, a bar he owned in Allentown, Pennsylvania. Scanlon’s body was discovered later the next morning, lying in a pool of blood. Scanlon had been stabbed multiple times, including 16 wounds around the neck and head. Scanlon also had been beaten with a blunt object, and his face had been gashed, possibly with shards from broken liquor and beer bottles found at the scene of the crime. After Scanlon was stabbed to death his body had been set on fire.

Substantial evidence linked Rompilla to the crime. See generally Commonwealth v. Rompilla, 539 Pa. 499, 505-506, 653 A. 2d 626, 629-630 (1995). He was at the Cozy Corner Cafe near closing time on the night of the murder and was observed going to the bathroom approximately 10 times dur*398ing a 1-hour period. A window in that bathroom, the police later determined, was the probable point of entry used by Scanlon’s assailant. A pair of Rompilla’s sneakers seized by the police matched a bloody footprint found near the victim’s body, and blood on the sneakers matched the victim’s blood type. Rompilla’s fingerprint was found on one of the two knives used to commit the murder. Sometime after leaving the bar on the night of the murder, Rompilla checked into a nearby motel under a false name. Although he told the police he left the bar with only two dollars, Rompilla had paid cash for the room and flashed a large amount of money to the desk clerks. The victim’s wallet was discovered in the bushes just outside of Rompilla’s motel room. When the police questioned Rompilla about the murder, his version of events was inconsistent with the testimony of other witnesses.

Rompilla was represented at trial by Fredrick Charles, the chief public defender for Lehigh County at the time, and Maria Dantos, an assistant public defender. Charles and Dantos were assisted by John Whispell, an investigator in the public defender’s office. Rompilla’s defense team sought to develop mitigating evidence from various sources. First, they questioned Rompilla extensively about his upbringing and background. App. 668-669. To make these conversations more productive they provided Rompilla with a list of the mitigating circumstances recognized by Pennsylvania law. Id., at 657. Cf. Strickland, supra, at 691 (“[W]hen a defendant has given counsel reason to believe that pursuing certain investigations would be fruitless or even harmful, counsel’s failure to pursue those investigations may not later be challenged as unreasonable”). Second, Charles and Dantos arranged for Rompilla to be examined by three experienced mental health professionals, experts described by Charles as “the best forensic psychiatrist around here, [another] tremendous psychiatrist and a fabulous forensic psychologist.” App. 672. Finally, Rompilla’s attorneys ques*399tioned his family extensively in search of any information that might help spare Rompilla the death penalty. Id., at 493-494, 557-558, 669-670, 729-730. Dantos, in particular, developed a “very close” relationship with Rompilla’s family, which was a “constant source of information.” Id., at 557, 729. Indeed, after trial Rompilla’s wife sent Dantos a letter expressing her gratitude. Id., at 733. The letter referred to Charles and Dantos as “superb human beings” who “fought and felt everything [Rompilla’s] family did.” Ibid.

The Court acknowledges the steps taken by Rompilla’s attorneys in preparation for sentencing but finds fault nonetheless. “[T]he lawyers were deficient,” the Court says, “in failing to examine the court file on Rompilla’s prior conviction.” Ante, at 383.

The prior conviction the Court refers to is Rompilla’s 1974 conviction for rape, burglary, and theft. See Commonwealth v. Rompilla, 250 Pa. Super. 139, 378 A. 2d 865 (1977). Before the sentencing phase of the capital case, the Commonwealth informed Rompilla’s attorneys that it intended to use these prior crimes to prove one of the statutory aggravating circumstances — namely, that Rompilla had a “significant history of felony convictions involving the use or threat of violence to the person.” 42 Pa. Cons. Stat. § 9711(d)(9) (2002). Rompilla’s attorneys were on notice of the Commonwealth’s plans, and from this the Court concludes that effective assistance of counsel required a review of the prior conviction case file.

A per se rule requiring counsel in every case to review the records of prior convictions used by the State as aggravation evidence is a radical departure from Strickland and its progeny. We have warned in the past against the creation of “specific guidelines” or “checklist^] for judicial evaluation of attorney performance.” 466 U. S., at 688. See also Wiggins v. Smith, 539 U. S. 510, 521 (2003); Roe v. Flores-Ortega, 528 U. S. 470, 477 (2000). “No particular set of detailed rules for counsel’s conduct can satisfactorily take account of *400the variety of circumstances faced by defense counsel or the range of legitimate decisions regarding how best to represent a criminal defendant. Any such set of rules would interfere with the constitutionally protected independence of counsel and restrict the wide latitude counsel must have in making tactical decisions. Indeed, the existence of detailed guidelines for representation could distract counsel from the overriding mission of vigorous advocacy of the defendant’s cause.” Strickland, 466 U. S., at 688-689 (citations omitted). For this reason, while we have referred to the ABA Standards for Criminal Justice as a useful point of reference, we have been careful to say these standards “are only guides” and do not establish the constitutional baseline for effective assistance of counsel. Ibid. The majority, by parsing the guidelines as if they were binding statutory text, ignores this admonition.

The majority’s analysis contains barely a mention of Strickland and makes little effort to square today’s holding with our traditional reluctance to impose rigid requirements on defense counsel. While the Court disclaims any intention to create a bright-line rule, ante, at 389-390; see also ante, at 393-394 (O’Connor, J., concurring), this affords little comfort. The Court’s opinion makes clear it has imposed on counsel a broad obligation to review prior conviction case files where those priors are used in aggravation — and to review every document in those files if not every single page of every document, regardless of the prosecution’s proposed use for the prior conviction. Infra, at 403, 407-408. One Member of the majority tries to limit the Court’s new rule by arguing that counsel’s decision here was “not the result of an informed tactical decision,” ante, at 395 (O’Connor, J., concurring), but the record gives no support for this notion. The Court also protests that the exceptional weight Rompil-la’s attorneys at sentencing placed on residual doubt required them to review the prior conviction file, ante, at 389-390; ante, at 394-395 (O’Connor, J., concurring). In fact, *401residual doubt was not central to Rompilla’s mitigation case. Rompilla’s family members did testify at sentencing that they thought he was innocent, but Dantos tried to draw attention away from this point and instead use the family’s testimony to humanize Rompilla and ask for mercy. App. 123-149.

The majority also disregards the sound strategic calculation supporting the decisions made by Rompilla’s attorneys. Charles and Dantos were “aware of [Rompilla’s] priors” and “aware of the circumstances” surrounding these convictions. Id., at 507. At the postconviction hearing, Dantos also indicated that she had reviewed documents relating to the prior conviction. Ibid. Based on this information, as well as their numerous conversations with Rompilla and his family, Charles and Dantos reasonably could conclude that reviewing the full prior conviction case file was not the best allocation of resources.

The majority concludes otherwise only by ignoring Strickland’s command that “[j]udieial scrutiny of counsel’s performance must be highly deferential.” 466 U. S., at 689. According to the Court, the Constitution required nothing less than a full review of the prior conviction case file by Rompilla’s attorneys. Even with the benefit of hindsight the Court struggles to explain how the file would have proved helpful, offering only the vague speculation that Rompilla’s attorneys might have discovered “circumstances extenuating the behavior described by the [rape] victim.” Ante, at 386. What the Court means by “circumstances” is a mystery. If the Court is referring to details on Rompilla’s mental fitness or upbringing, surely Rompilla’s attorneys were more likely to discover such information through the sources they consulted: Rompilla, his family, and the three mental health experts that examined him.

Perhaps the circumstances to which the majority refers are the details of Rompilla’s 1974 crimes. Charles and Dantos, however, had enough information about the prior *402convictions to determine that reviewing the case file was not the most effective use of their time. Rompilla had been convicted of breaking into the residence of Josephine Macrenna, who lived in an apartment above the bar she owned. App. 56-89. After Macrenna gave him the bar’s receipts for the night, Rompilla demanded that she disrobe. When she initially resisted, Rompilla slashed her left breast with a knife. Rompilla then held Macrenna at knifepoint while he raped her for over an hour. Charles and Dantos were aware of these circumstances of the prior conviction and the brutality of the crime. Id., at 507. It did not take a review of the case file to know that quibbling with the Commonwealth’s version of events was a dubious trial strategy. At sentencing Dantos fought vigorously to prevent the Commonwealth from introducing the details of the 1974 crimes, id., at 16-40, but once the transcript was admitted there was nothing that could be done. Rompilla was unlikely to endear himself to the jury by arguing that his prior conviction for burglary, theft, and rape really was not as bad as the Commonwealth was making it out to be. Recognizing this, Rompilla’s attorneys instead devoted their limited time and resources to developing a mitigation case. That those efforts turned up little useful evidence does not make the ex ante strategic calculation of Rompilla’s attorneys constitutionally deficient.

One of the primary reasons this Court has rejected a checklist approach to effective assistance of counsel is that each new requirement risks distracting attorneys from the real objective of providing vigorous advocacy as dictated by the facts and circumstances in the particular case. The Court’s rigid requirement that counsel always review the case files of convictions the prosecution seeks to use at trial will be just such a distraction. Capital defendants often have a history of crime. For example, as of 2003,64 percent of inmates on death row had prior felony convictions. U. S. Dept, of Justice, Bureau of Justice Statistics, T. Bonczar & T. Snell, Capital Punishment, 2003, p. 8 (Nov. 2004), available *403at http://www.ojp.usdoj.gov/bjs/pub/pdf/cp03.pdf (as visited June 16,2005, and available in Clerk of Court’s case file). If the prosecution relies on these convictions as aggravators, the Court has now obligated defense attorneys to review the boxes of documents that come with them.

In imposing this new rule, the Court states that counsel in this case could review the “entire file” with “ease.” Ante, at 386, n. 4. There is simply no support in the record for this assumption. Case files often comprise numerous boxes. The file may contain, among other things, witness statements, forensic evidence, arrest reports, grand jury transcripts, testimony and exhibits relating to any pretrial suppression hearings, trial transcripts, trial exhibits, post-trial motions, and presentence reports. Full review of even a single prior conviction case file could be time consuming, and many of the documents in a file are duplicative or irrelevant. The Court, recognizing the flaw in its analysis, suggests that cases involving “warehouses of records” “will call for greater subtlety.” Ibid. Yet for all we know, this is such a case. As to the time component, the Court tells us nothing as to the number of hours counsel had available to prepare for sentencing or why the decisions they made in allocating their time were so flawed as to constitute deficient performance under Strickland.

Today’s decision will not increase the resources committed to capital defense. (At the time of Rompilla’s trial, the Le-high County Public Defender’s Office had two investigators for 2,000 cases. App. 662.) If defense attorneys dutifully comply with the Court’s new rule, they will have to divert resources from other tasks. The net effect of today’s holding in many cases — instances where trial counsel reasonably can conclude that reviewing old case files is not an effective use of time — will be to diminish the quality of representation. We have “consistently declined to impose mechanical rules on counsel — even when those rules might lead to better representation,” Roe v. Flores-Ortega, 528 U. S., at 481; I see *404no occasion to depart from this approach in order to impose a requirement that might well lead to worse representation.

It is quite possible defense attorneys, recognizing the absurdity of a one-size-fits-all approach to effective advocacy, will simply ignore the Court’s new requirement and continue to exercise their best judgment about how to allocate time and resources in preparation for trial. While this decision would be understandable — and might even be required by state ethical rules, cf. Pa. Rules of Professional Conduct, Preamble, and Rule 1.1 (2005) — it leaves open the possibility that a defendant will seek to overturn his conviction based on something in a prior conviction case file that went unre-viewed. This elevation of needle-in-a-haystack claims to the status of constitutional violations will benefit undeserving defendants and saddle States with the considerable costs of retrial and/or resentencing.

Today’s decision is wrong under any standard, but the Court’s error is compounded by the fact that this case arises on federal habeas. The Pennsylvania Supreme Court adjudicated Rompilla’s ineffective-assistance-of-eounsel claim on the merits, and this means 28 U. S. C. §2254(d)’s deferential standard of review applies. Rompilla must show that the Pennsylvania Supreme Court decision was not just “incorrect or erroneous,” but “objectively unreasonable.” Lockyer v. Andrade, 538 U. S. 63, 75 (2003) (citing Williams v. Taylor, 529 U. S. 362, 410, 412 (2000)). He cannot do so.

The Court pays lipservice to the Williams standard, but it proceeds to adopt a rigid, per se obligation that binds counsel in every case and finds little support in our precedents. Indeed, Strickland, the case the Court purports to apply, is directly to the contrary: “Most important, in adjudicating a claim of actual ineffectiveness of counsel, a court should keep in mind that the principles we have stated do not establish mechanical rules.” 466 U. S., at 696. The Pennsylvania Supreme Court gave careful consideration to Rompilla’s Sixth Amendment claim and concluded that “counsel reasonably *405relied upon their discussions with [Rompilla] and upon their experts to determine the records needed to evaluate his mental health and other potential mitigating circumstances.” Commonwealth v. Rompilla, 554 Pa. 378, 385-386, 721 A. 2d 786, 790 (1998). This decision was far from unreasonable. The Pennsylvania courts can hardly be faulted for failing to anticipate today’s abrupt departure from Strickland.

We have reminded federal courts often of the need to show the requisite level of deference to state-court judgments under 28 U. S. C. § 2254(d). Holland v. Jackson, 542 U. S. 649 (2004) (per curiam); Middleton v. McNeil, 541 U. S. 433 (2004) (per curiam); Yarborough v. Gentry, 540 U. S. 1 (2003) (per curiam); Mitchell v. Esparza, 540 U. S. 12 (2003) (per curiam); Early v. Packer, 537 U. S. 3 (2002) (per curiam); Woodford v. Visciotti, 537 U. S. 19 (2002) (per curiam). By ignoring our own admonition today, the Court adopts a do-as-we-say, not-as-we-do approach to federal habeas review.

II

Even accepting the Court’s misguided analysis of the adequacy of representation by Rompilla’s trial counsel, Rompilla is still not entitled to habeas relief. Strickland assigns the defendant the burden of demonstrating prejudice, 466 U. S., at 692. Rompilla cannot satisfy this standard, and only through a remarkable leap can the Court conclude otherwise.

The Court’s theory of prejudice rests on serendipity. Nothing in the old case file diminishes the aggravating nature of the prior conviction. The only way Rompilla’s attorneys could have minimized the aggravating force of the earlier rape conviction was through Dantos’ forceful, but ultimately unsuccessful, fight to exclude the transcript at sentencing. The Court, recognizing this problem, instead finds prejudice through chance. If Rompilla’s attorneys had reviewed the case file of his prior rape and burglary conviction, the Court says, they would have stumbled across “a range of mitigation leads.” Ante, at 390.

*406The range of leads to which the Court refers is in fact a handful of notations within a single 10-page document. The document, an “Initial Transfer Petition,” appears to have been prepared by the Pennsylvania Department of Corrections after Rompilla’s conviction to facilitate his initial assignment to one of the Commonwealth’s maximum-security prisons. Lodging to App. 31-40.

Rompilla cannot demonstrate prejudice because nothing in the record indicates that Rompilla’s trial attorneys would have discovered the transfer petition, or the clues contained in it, if they had reviewed the old file. The majority faults Rompilla’s attorneys for failing to “learn what the Commonwealth knew about the crime,” “discover any mitigating evidence the Commonwealth would downplay,” and “anticipate the details of the aggravating evidence the Commonwealth would emphasize.” Ante, at 385-386. Yet if Rompilla’s attorneys had reviewed the case file with these purposes in mind, they almost surely would have attributed no significance to the transfer petition following only a cursory review. The petition, after all, was prepared by the Department of Corrections after Rompilla’s conviction for the purpose of determining Rompilla’s initial prison assignment. It contained no details regarding the circumstances of the conviction. Reviewing the prior conviction file for information to counter the Commonwealth, counsel would have looked first at the transcript of the trial testimony, and perhaps then to probative exhibits or forensic evidence. There would have been no reason for counsel to read, or even to skim, this obscure document.

The Court claims that the transfer petition would have been discovered because it was in the “same file” with the transcript, ante, at 391, but this characterization is misleading and the conclusion the Court draws from it is accordingly fallacious. The record indicates only that the transfer petition was a part of the same case file, but Rompilla provides no indication of the size of the file, which for all we know *407originally comprised several boxes of documents. App. 508, 571, 631. By the time of Rompilla’s state postconviction hearing, moreover, the transfer petition was not stored in any “file” at all — it had been transferred to microfilm. Id., at 461. The Court implies in a footnote that prejudice can be presumed because “Pennsylvania has conspicuously failed to contest Rompilla’s” inevitable-discovery argument. Ante, at 391, n. 8. The Commonwealth’s strategy is unsurprising given that discussion of the prior conviction case file takes up only one paragraph of Rompilla’s argument, Brief for Petitioner 35-36, but it is also irrelevant. It is well established that Rompilla, not the Commonwealth, has the burden of establishing prejudice. Strickland, supra, at 694.

The majority thus finds itself in a bind. If counsel’s alleged deficiency lies in the failure to review the file for the purposes the majority has identified, then there is no prejudice: for there is no reasonable probability that review of the file for those purposes would have led counsel to accord the transfer petition enough attention to discover the leads the majority cites. Prejudice could only be demonstrated if the deficiency in counsel’s performance were to be described not as the failure to perform a purposive review of the file, but instead as the failure to accord intense scrutiny to every single page of every single document in that file, regardless of the purpose motivating the review. At times, the Court hints that its new obligation on counsel sweeps this broadly. See ante, at 386, n. 4 (“[t]he ease with which counsel eould examine the entire file . . . ”); ante, at 386, n. 5 (“[Cjounsel had no way of knowing the context of the transcript and the details of the prior conviction without looking at the file as a whole”). Surely, however, the Court would not require defense counsel to look at every document, no matter how tangential, included in the prior conviction file on the off chance that some notation therein might provide a lead, which in turn might result in the discovery of useful information. The Constitution does not mandate that defense attor*408neys perform busy work. This rigid requirement would divert counsel’s limited time and energy away from more important tasks. In this way, it would ultimately disserve the rationale underlying the Court’s new rule, which is to ensure that defense counsel counter the State’s aggravation case effectively.

If the Court does intend to impose on counsel a constitutional obligation to review every page of every document included in the case file of a prior conviction, then today’s holding is even more misguided than I imagined.

* * *

Strickland anticipated the temptation “to second-guess counsel’s assistance after conviction or adverse sentence” and cautioned that “[a] fair assessment of attorney performance requires that every effort be made to eliminate the distorting effects of hindsight, to reconstruct the circumstances of counsel’s challenged conduct, and to evaluate the conduct from counsel’s perspective at the time.” 466 U. S., at 689. Today, the Court succumbs to the very temptation that Strickland warned against. In the process, the majority imposes on defense attorneys a rigid requirement that finds no support in our cases or common sense.

I would affirm the judgment of the Court of Appeals.