A California jury convicted Edgar M. Hendricks of two counts of first degree murder with use of a firearm; one count of robbery with use of a firearm; one count of burglary with use of a firearm; and one count of grand theft. The jury in the penalty phase found multiple murders and felony-murder special circumstances and returned a death verdict. After exhausting his state appeals, Hendricks sought habeas relief under 28 U.S.C. § 2254 , claiming that he received ineffective assistance of counsel in both the guilt and penalty phases of his state trial. 1 The district court granted the petition as to the penalty phase and denied the petition as to the guilt phase. The challenged order vacated the death sentence, leaving the state the option to retry the penalty phase. Hendricks appeals the denial of his petition challenging the guilt phase, arguing primarily that his counsel failed to investigate the possibility of a mental defense adequately. The state cross-appeals the order finding the pen *1036 alty phase constitutionally defective, contending that trial counsel presented enough mitigating evidence to pass constitutional muster.

Because trial counsel hired two mental health experts to look into mental defenses and both concluded after examining Hendricks that there was no basis for a mental defense, counsel’s decision not to pursue a mental defense was a reasonable strategic choice supported by an adequate investigation. However, because trial counsel conducted no investigation in preparation for the penalty phase and had no strategic reason for failing to do so, counsel’s performance in the penalty phase fell below the level of constitutional acceptability. As a result of counsel’s failure in the penalty phase, counsel did not make an effort to present mitigating evidence which was available, thereby prejudicing Hendricks. Accordingly, we affirm the district court order, granting the petition as to the penalty phase and denying the petition as to the guilt phase.

I. FACTS

The basic facts of this case are not disputed. For a thorough exposition of the underlying crimes, the conduct of counsel at both the guilt and penalty phases, as well as the testimony at the evidentiary hearing held on the habeas petition now under review, see the district court’s opinion, Hendricks v. Calderon, 864 F.Supp. 929 (N.D.Cal.1994). See also Hendricks v. Vasquez, 974 F.2d 1099 (9th Cir.1992) for this Court’s precis of the background facts.

II. THE LAW

This Court reviews habeas petitions de novo. Sanders v. Ratelle, 21 F.3d 1446, 1451 (9th Cir.1994). In a federal habeas action, a claim of ineffective assistance of counsel presents a mixed question of fact and law. Id. District court fact findings are accepted unless they are clearly erroneous. Id. at 1451-52 .

In order to prevail on a claim of ineffective assistance of counsel, the petitioner must establish that counsel’s actions were deficient and this shortcoming prejudiced the petitioner. Strickland v. Washington, 466 U.S. 668, 687-90 , 104 S.Ct. 2052, 2064-67 , 80 L.Ed.2d 674 (1984); Hendricks v. Vasquez, 974 F.2d at 1109 . “Although there is a strong presumption that counsel’s conduct falls within the wide range of reasonable professional assistance, and judicial scrutiny of counsel’s performance must be highly deferential, ... counsel must, at a minimum, conduct a reasonable investigation enabling him to make informed decisions about how best to represent his client.” Sanders v. Ratelle, 21 F.3d at 1456 (internal citation and quotations omitted). “[C]ounsel has a duty to make reasonable investigations or to make a reasonable decision that makes particular investigations unnecessary.” Strickland, 466 U.S. at 691 , 104 S.Ct. at 2066 . This Court has found counsel to be ineffective where an attorney neither conducted a reasonable investigation nor demonstrated a strategic reason for failing to do so. See Sanders, 21 F.3d at 1456 and cases cited therein.

In determining whether counsel’s conduct falls within the broad range of professionally acceptable conduct, this Court will not view counsel’s actions through “the distorting lens of hindsight.” Deutscher v. Whitley, 884 F.2d 1152, 1159 (9th Cir.1989), vacated on other grounds, Angelone v. Deutscher, 500 U.S. 901 , 111 S.Ct. 1678 , 114 L.Ed.2d 73 (1991). Rather, under the rule of contemporary assessment, an attorney’s actions must be examined according to what was known and reasonable at the time the attorney made his choices. Id.

To establish the prejudice prong of the Strickland standard, “[t]he defendant must show that there is a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different. A reasonable probability is a probability sufficient to undermine confidence in the outcome.” Strickland, 466 U.S. at 694 , 104 S.Ct. at 2068 . In the guilt phase, then, “the question is whether there is a reasonable probability that, absent the errors, the factfinder would have had a reasonable doubt respecting guilt.” Id. at 695 , 104 S.Ct. at 2068-69 . Likewise, in the sentencing phase, “the question is whether there is a reasonable probability that, absent the errors, the sentencer ... would have concluded *1037 that the balance of aggravating and mitigating circumstances did not warrant death.” Id.

III. COUNSEL’S PERFORMANCE IN THE GUILT PHASE

A. Failure to Investigate Hendricks’ Background

Hendricks argues that his attorney’s failure to investigate his background adequately, particularly his alleged history as a victim of sexual abuse and possible genetic predisposition to various psychiatric disorders, rendered counsel’s representation constitutionally deficient. Hendricks argues that the defective investigation provided an inadequate basis for counsel’s strategic decision not to pursue a mental defense. Hendricks further argues that regardless of the decision not to pursue a mental defense, the strategy pursued at trial, that is, corroborating the state of mind reflected by Hendricks’ confession, demanded an investigation into Hendricks’ family background, a task counsel never performed.

1. Decision Not to Pursue a Mental Defense

a. The Investigation Conducted

Hendricks’ chief trial counsel, Robert Ber-man, and co-counsel, Michael Gaines, the most experienced homicide lawyers in the San Francisco Public Defenders Office, suspected that Hendricks may have had psychiatric problems. Following his intuition, Ber-man hired mental health experts to examine Hendricks with an eye toward appraising any mental defenses that could be available. Berman first hired Dr. George Bach-y-Rita, a psychiatrist with whom Berman had worked before, and whom Berman believed to be a competent forensic mental expert. Dr. Bach-y-Rita met with Hendricks twice, about four and a half hours in total, to canvass any indication that Hendricks was insane or suffered from a diminished mental capacity. Dr. Bach-y-Rita found no evidence that Hendricks was insane, psychotic, or unable to control his behavior at the time of the crimes. In short, Bach-y-Rita found no promising basis for a mental defense. Bach-y-Rita diagnosed Hendricks as a sociopath and informed Berman that he could not assist in the defense.

Still, Bach-y-Rita thought further psychological testing might prove useful, so he suggested that Berman hire Dr. Linda Carson, a clinical psychologist with whom both Dr. Bach-y-Rita and Berman were acquainted. Dr. Carson interviewed Hendricks nine times, for a total of about fifteen hours. She ran several psychological tests on Hendricks, reviewed the taped confessions and police reports, as well as Hendricks’ life history, which she elicited from Hendricks in their talks. Like Dr. Bach-y-Rita, Dr. Carson concluded that Hendricks was sane at the time of the murders and found no evidence that Hendricks was insane or suffered from a diminished capacity.

Berman provided Dr. Carson and Dr. Bach-y-Rita with the limited information on Hendricks in his possession, that is, the confession and the police reports, before they reached any conclusions. Neither Dr. Bach-y-Rita nor Dr. Carson requested further background information on Hendricks from Berman or qualified their conclusions due to insufficient information on their patient. There is no evidence that the conclusions of the experts were tentative, snap judgments, or otherwise based on anything less than a full analysis of complete data. After hearing the conclusion of his two mental health experts that there was no basis for a mental defense, Berman decided that he would not pursue a mental defense and stopped his investigation down that path.

Not only did Hendricks’ own mental health experts fail to offer any grounds for a mental defense, but a mental defense could be pursued only at a substantial strategic cost. If Berman put on the testimony of Dr. Carson or Dr. Bach-y-Rita, the jury would have learned, during their cross-examination, of Hendricks’ other murders for which he was tried separately. Considering the conclusion of the mental health experts that Hendricks was sane, sociopathic, and did not suffer from a diminished capacity, along with the substantial negative risk of presenting a mental defense, the district court concluded, “trial counsel cannot be said to have made an *1038 uninformed or baseless choice when he decided not to explore an insanity or diminished capacity [defense] further.”

b. Legal Duty to Investigate

Berman’s investigation into Hendricks’ mental health and his reliance on the conclusion of his experts shields Berman from ineffective assistance of counsel claims under existing caselaw. In general, an attorney is entitled to rely on the opinions of mental health experts in deciding whether to pursue an insanity or diminished capacity defense. This Court did not condemn as incompetent an attorney’s decision not to pursue a mental defense where two experts concluded that the defendant was sane and a third expert could not reach a conclusion. Morgan v. Bunnell, 24 F.3d 49, 52 (9th Cir.1994). This Court has also held that even where there is a strong basis for a mental defense, an attorney may forego that defense where the attorney’s experts would be subject to cross-examination based on equally persuasive psychiatric opinions that reach a different conclusion. Harris v. Vasquez, 949 F.2d 1497, 1525 (9th Cir.1990), cert. denied, 503 U.S. 910 , 112 S.Ct. 1275 , 117 L.Ed.2d 501 (1992).

However, despite Berman’s inquiry into Hendricks’ mental condition and clear reasons for foregoing a mental defense, Hendricks contends that Berman’s failure to investigate Hendricks’ social history independently constitutes deficient representation. Hendricks argues “that by 1981 [the time of trial] it was the recognized duty of defense counsel in capital cases to obtain social history evidence relevant to a client’s mental condition where there was an indication of mental disturbance and to provide such evidence to an expert in order for the client’s condition to be properly evaluated.” Under Hendricks’ 1995 view, Berman was dutybound to interview Hendricks’ family members and friends, to obtain medical, school, and employment records, to otherwise verify Hendricks’ autobiography and pass this information on to his mental health experts. “Without such background material,” Hendricks continues, “the expert’s diagnosis of the client’s condition could not be meaningful.” Hendricks’ argument, then, is not that examination by two mental health experts was per se inadequate. Rather, Hendricks argues that trial counsel had a duty to provide the experts with his client’s family social history, even absent any request from the experts. He argues that counsel’s failure to do so undermines the expert’s conclusions to such an extent that they cannot provide an adequate basis on which the attorney could make strategic choices.

Neither reason nor the existing authority would lead one to conclude that an attorney in 1980 had such an affirmative constitutional duty. Although this Court has not specifically addressed this question, the Eleventh Circuit and the Central District of California have studied the matter. Facing the identical issue this case presents, the Eleventh Circuit concluded that an attorney “was not obligated to track down every record that might possibly relate to [the defendant’s] mental health and could affect a diagnosis.” Card v. Dugger, 911 F.2d 1494, 1512 (11th Cir.1990). In that case, as here, there was “no indication that the experts felt incapable of basing their conclusions on the information they obtained through their own testing and examinations” and the experts unanimously concluded that the defendant was sane. Id. Similarly, the Central District of California has held, “In the absence of a specific request, an attorney is not responsible for gathering background material that might be helpful to a psychiatrist evaluating his client.” Bloom v. Vasquez, 840 F.Supp. 1362, 1370 (C.D.Cal.1993). Hindsight now teaches that prudent counsel should enlarge their investigative scope in capital cases. But hindsight is a weak standard for fixing constitutional minima.

To now impose a duty on attorneys to acquire sufficient background material on which an expert can base reliable psychiatric conclusions, independent of any request for information from an expert, would defeat the whole aim of having experts participate in the investigation. An integral part of an expert’s specialized skill at analyzing information is an understanding of what information is relevant to reaching a conclusion. Experts are valuable to an attorney’s investiga *1039 tion, then, not only because they have special abilities to process the information gathered by the attorney, but because they also are able to guide the attorney’s efforts toward collecting relevant evidence. To require an attorney, without interdisciplinary guidance, to provide a psychiatric expert with all information necessary to reach a mental health diagnosis demands that an attorney already be possessed of the skill and knowledge of the expert. Hendricks’ own psychiatric expert, Dr. Lisak, conceded that a psychological report reflects what the expert thinks is important. If an attorney has the burden of reviewing the trustworthiness of a qualified expert’s conclusion before the attorney is entitled to make decisions based on that conclusion, the role of the expert becomes superfluous.

To validate the contrary result, Hendricks holds up Berman’s and Gaines’ confession of their own incompetence as evidence that a duty to provide mental health experts with background information existed. As the government points out, the latter-day emergence of Berman’s and Gaines’ belief in their own incompetence runs afoul of the rule of contemporary assessment. Certainly, in 1981, Hendricks’ attorneys did not believe they had any duty to investigate Hendricks’ social history in the face of the unanimous opinions of their own experts that there was no basis for a mental defense.

The legal authority Hendricks cites offers meager support for his argument. Hendricks’ cited authority merely discusses the basic duty under Strickland to investigate potential defenses before making strategic decisions. Sec ABA Standards Relating to the Administration of Criminal Justice, Standard 4-4.1; U.S. v. DeCoster, 487 F.2d 1197 (D.C.Cir.1973); Rummel v. Estelle, 590 F.2d 103 (5th Cir.1979). Hendricks’ authority that touches upon the duty to investigate mental defenses for the most part deals with an attorney’s failure to investigate the possibility of a mental defense. See People v. Frierson, 25 Cal.3d 142, 162 , 158 Cal.Rptr. 281 , 599 P.2d 587 (1979); People v. Pope, 23 Cal.3d 412, 428 , 152 Cal.Rptr. 732 , 590 P.2d 859 (1979); Deutscher v. Whitley, 884 F.2d at 1160 . At most, these cases establish a duty to seek out psychiatric evaluation of a client where the grounds of a mental defense are apparent, a duty Berman clearly discharged. Hendricks’ cases make no comment on an attorney’s alleged duty to investigate material relevant to a mental defense in the face of expert advice that there is no basis for such a defense. See also Evans v. Lewis, 855 F.2d 631, 637 (9th Cir.1988) (failure to investigate defendant’s mental condition when there is evidence of impairment constitutes deficient performance, and is prejudicial when it hampers later presentation of evidence of mental impairment).

Hendricks’ attorneys certainly fell within the broad range of presumptively acceptable conduct by hiring two mental health professionals to investigate potential mental defenses and then relying on their shared, unqualified conclusion that there was no basis for a mental defense. To hold otherwise would raise the Sixth Amendment hurdle well above the floor of minimal competence, requiring attorneys to have the specialized knowledge to evaluate an expert’s conclusions before relying upon them in making strategic choices. By forcing lawyers to second-guess their experts, the position Hendricks argues would effectively eliminate the legitimate role experts play in guiding and narrowing an attorney’s investigation.

Because the performance of Hendricks’ attorneys was not deficient, we do not reach the prejudice prong of the Strickland test on the guilt phase of the case.

2. The Decision Not to Put on Dr. Carson

Hendricks also argues that Berman should have investigated Hendricks’ background to corroborate Dr. Carson’s testimony that Hendricks killed his victims in a rage and that robbery was an afterthought. Hendricks contends that if Berman had performed such an investigation, the corroboration would have strengthened the impact of Dr. Carson’s conclusions so much that Ber-man would have decided to put Dr. Carson on the stand despite the cost of informing the jurors of Hendricks’ previous murder convictions. To support this conclusion, Hendricks cites Berman’s admission at the evidentiary hearing that he would have presented Dr. *1040 Carson’s testimony in the guilt phase if he had done a more thorough investigation.

What decision Berman may have made if he had more information at the time is exactly the sort of Monday-morning quarterbacking the contemporary assessment rule forbids. It is meaningless, after more than a decade, for Berman now to claim that he would have done things differently if only he had more information. With more information, Benjamin Franklin might have invented television.

The inquiry relevant to an ineffective assistance of counsel claim on this matter is whether Berman had a duty to search for evidence corroborating Dr. Carson’s testimony prior to deciding that the risks of Carson’s testimony outweighed the benefits. There was no such duty.

Without the aid of hindsight, it is incredible to argue that informing the jury of Hendricks’ other murders was the only reasonable choice or even that absent more information, choosing to keep this evidence from the jury was unreasonable. Once Berman made the reasonable strategic decision to keep evidence of the other murders out, and with it Dr. Carson’s testimony, there was no need to investigate evidence corroborating that testimony.

3. Failure to Corroborate Hendricks’ Confession

Next, Hendricks argues that even the strategy pursued, that is, negating premeditation and the predicate felony convictions with Hendricks’ confession statements regarding his state of mind, required an unperformed investigation into Hendricks’ personal background. Hendricks asserts that an investigation into Hendricks’ life history would have been valuable “in establishing the truth of Hendricks’ confession statements regarding his life history, thus enhancing the credibility of petitioner’s statements regarding his state of mind at the time of the killings.”

An attorney “must ... provide factual support for [the] defense where such corroboration is available.” U.S. v. Tucker, 716 F.2d 576, 594 (9th Cir.1983). Failure to pursue such corroborating evidence with an adequate pretrial investigation may establish constitutionally deficient performance. Id. Yet, “the duty to investigate and prepare a defense is not limitless: it does not necessarily require that every conceivable witness be interviewed____” U.S. v. Tucker, 716 F.2d at 584 . The type of corroborating evidence Hendricks claims an investigation would have unearthed is so tenuously related to the defense that it belongs in this latter category of evidence in which the decision to search for it or neglect it is left to the discretion of the attorney. Failure to look for witnesses to bolster Hendricks’ credibility in these circumstances cannot alone constitute ineffective assistance.

Where this Court has found a failure to locate potential witnesses to constitute incompetent lawyering, the attorney refused to perform any investigation into leads directly related and of potentially great benefit to the defense. See, Id.; Sanders v. Ratelle, 21 F.3d at 1456-57 . In the present case, the evidence Hendricks claims his attorney had a duty to find is only tangentially relevant to the defense. Witnesses that could verify Hendricks’ childhood history obviously could not attest directly to Hendricks’ state of mind at the time of each of the killings. Without the explanatory testimony of a psychology expert, what happened in Hendricks’ childhood does not corroborate Hendricks’ statement that he killed in a rage or out of fear for his own safety.

As Hendricks himself argues, the uncalled background witnesses would have aided the defense only to the extent that they would have corroborated Hendricks’ own account of his past. Such evidence could have bolstered his credibility when he spoke of his states of mind at the times of the murders. While the Sixth Amendment requires an attorney to look for evidence that corroborates the defense he pursues, the Sixth Amendment has not been expanded to require an attorney to hunt down such marginally relevant and indirectly beneficial evidence. Even assuming the existence of a duty to find evidence that bolsters a key witness’ credibility by corroborating irrelevant facts in his testimony, such evidence is so loosely tied to the defense that *1041 its absence would not undermine confidence in the outcome. Strickland, 466 U.S. at 694 , 104 S.Ct. at 2068 .

B. Failure to Bring the Motion In Limine Before Trial

After the close of the prosecution’s case-in-chief, Berman brought and lost a motion in limine to preclude cross-examination of Dr. Carson regarding the Los Angeles murders. Believing that keeping word of the other murders from the jury was of paramount importance, Berman scrapped his planned defense. Consequently, Berman presented no defense evidence and the jury learned of the other murders for the first time in the penalty phase. That was precisely the time when the prejudicial impact of such hews would be the greatest.

Hendricks argues that Berman’s failure to bring the motion in limine prior to jury selection was incompetent. Hendricks (and the now contrite Berman) suggest that if Berman had known from the start that Dr. Carson could be cross-examined on the other murders, he may have used Dr. Carson in the guilt phase anyway and inoculated the jurors during voir dire to the information that Hendricks had killed others. These suggestions, of course, are retrospective speculation. Also, as the prosecution points out, the strategy that Hendricks now protests was the only sensible course, that is, informing the jury of the uncharged murders, in fact, could equally have exposed Berman to an ineffective assistance charge following an unfavorable verdict.

With the clarity of hindsight, it is difficult to imagine what strategic concern could justify the decision to wait until the close of the government’s ease to bring the motion. At a minimum, knowing the scope of cross-examination from the start would have been an important factor in deciding what defense to prepare and eliminated the devastating element of surprise in the judge’s ruling.

Even though it would have been desirable to bring the motion in limine sooner, however, Hendricks has not shown what alternative defense could possibly have been offered had the motion been timely. Because he has not shown “a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different,” he has not met the prejudice prong of the Strickland test. Strickland, 466 U.S. at 694 , 104 S.Ct. at 2068 .

C. Failure to Present a Coherent Defense

Hendricks argues that counsel’s failure to present a coherent defense constituted ineffective assistance of counsel. The alleged incompetence Hendricks complains of has two facets. First, Hendricks asserts that Berman’s failure to pursue a strategy that integrated the guilt and penalty phases was deficient performance. We have been cited no authority holding that a California attorney must treat the guilt and penalty phases of a death penalty trial as an organic whole and pursue the same or even complementary strategies in both phases. The Constitution does not require an integrated strategy any more than it requires an attorney to strive for a not guilty verdict in the guilt phase or orient the entire case around mitigating the crime. The choice to pursue a particular strategy in one phase at the cost of lowering the likelihood of success in the other phase is a tactical decision for which there is no correct answer, but only second guesses.

Next, Hendricks argues that the defense pursued was so ill-conceived that the choice to pursue it cannot be considered a strategic decision, but was itself incompetence. Hendricks now contends: “A mental impairment defense was ... the only viable defense under the circumstances. It was not simply an option counsel should have considered ...; diminished capacity or insanity were the only theories that made any sense in terms of (1) avoiding a penalty phase altogether, or (2) assuming the case went into a penalty phase, arousing enough sympathy in the minds of the jurors to avoid a sentence of death.”

Hendricks’ claim that a mental defense was the only hope of arousing enough sympathy in the jurors to avoid a death sentence is speculation of the most imaginative order. Even looking backward, one cannot say that evidence of Hendricks’ possible psychiatric *1042 problems would have had greater impact than Berman’s impassioned plea for mercy. At a minimum, the choice to put on a mental defense in the penalty phase was not so clear and obvious that any other decision is necessarily professionally deficient.

Hendricks’ argument regarding the theoretical cogency of the defense in the guilt phase is more persuasive, but fails on the facts here as well. When Hendricks calls the defense that was pursued a farce, he is not exaggerating. As Hendricks points out, even if successful, the defense pursued could not avoid conviction in the guilt phase. The strategy to corroborate Hendricks’ confession statements that he did not intend to rob his victims prior to killing them and that he killed because he believed he was in danger could not refute the multiple murder special circumstance. Given that the strategy pursued, even if completely successful, doomed Hendricks to a guilty verdict and advancement to the penalty hearing, Hendricks asserts that foregoing a mental defense, the only defense that could avoid the penalty phase and maybe even conviction altogether, was not a reasonable tactical choice, but incompetence.

This argument would have more appeal had not both of his own mental health experts believed that there was no basis for a mental defense. Dr. Carson and Dr. Bach-y-Rita reported that what Hendricks holds up as the only plausible defense was not at all plausible, but was baseless. While it is true that the defense chosen was also lacking in efficacy, the alternative was equally undesirable.

What Hendricks’ protest over the cogency of his defense really shows is that not every murder case can be won by the defense. Berman had to defend a generally unre-morseful client who confessed on tape to several murders, and when on the stand confessed to yet another uncharged murder, and loudly recounted his merciless crimes for the jury over his attorney’s vain attempt to beg for mercy. In many cases, the law and facts will be so overwhelmingly in favor of the government that defense counsel can do little more than try to poke holes in the government’s case in cross-examination. The hopelessness of some cases may even relegate the most competent defense counsel to the role of official hand-holder. The Sixth Amendment does not hold an attorney responsible for the difficulty of the case he inherits. The choice to pursue a bad strategy makes no comment on an attorney’s judgment where no better choice exists.

D. Other Claimed Deficiencies

Hendricks alleges three other investigative failures as establishing the deficient performance required for an ineffective assistance of counsel claim. Hendricks asserts that Berman failed to 1) investigate the circumstances of victim Haynes’ arrest for robbery and assault; 2) determine whether Nelson, the roommate of victim Parmer, had a reputation for not paying prostitution debts; and 3) whether victim Parmer owned a gun. Presumably, investigation into these three areas would have aided Hendricks’ defense by supporting the confession statements that Hendricks killed Parmer as part of an effort to collect a prostitution debt and that Hendricks killed both Parmer and Haynes out of a fear for his own safety. The district court correctly rejected similar arguments “because Hendricks identifies no specific information that trial counsel could have discovered about the victims ... had it conducted further investigation at the time of trial.” Absent an account of what beneficial evidence investigation into any of these issues would have turned up, Hendricks cannot meet the prejudice prong of the Strickland test. James v. Borg, 24 F.3d 20, 26 (9th Cir.1994).

The allusion to Haynes’ arrest for robbery and assault is arguably a specific enough allegation that evidence showing Haynes’ violence exists. However, the relationship between evidence of Haynes’ violence and Hendricks’ defense is so tenuous and the benefit to the defense so small that failure to investigate here cannot constitute ineffective assistance. Unless Hendricks knew of Haynes’ past resort to violence, it is unclear how Haynes’ arrest record reinforces Hendricks’ claim that he subjectively feared for his own safety.

*1043 IV. INEFFECTIVE ASSISTANCE OF COUNSEL IN THE PENALTY PHASE

The district court held that Hendricks was denied effective assistance of counsel in the penalty phase and the government cross-appeals. The district court found Berman’s failure to conduct any investigation into mitigating evidence without any strategic reason justifying the lack of preparation to be deficient performance. The district court held that Hendricks was prejudiced by counsel’s failure to present readily available mitigating evidence, to corroborate Dr. Carson’s penalty phase testimony, and to connect Hendricks’ penalty phase evidence to the statutory mitigating factors. We affirm here as well.

A. Deficient Performance

The district court’s factual findings greatly simplify the analysis of counsel’s penalty phase performance. The district court found that neither trial counsel nor his investigators conducted any investigation directed at developing mitigating evidence. The court further concluded that counsel did not make a strategic choice to forego a penalty phase investigation and that on the facts of this case, no discernible strategy could justify counsel’s neglect.

These factual findings are not clearly erroneous and therefore must be accepted by this Court. The record is clear, and it is undisputed, that Berman did not investigate mitigating evidence for the penalty phase. In keeping with his belief that the penalty phase was a “strange blip” at the end of the trial, not substantially different from the off-the-rack sentencing hearing, Berman did not prepare for the penalty phase in any significant way. Likewise, there is no evidence that Berman made a strategic choice that obviated the need to investigate. Berman, in his own words, described the “strategy” he pursued: “Let’s put on whatever we have and beg for mercy.” Berman decided to plead for mercy, not because presenting the mitigation evidence available would open the door to damning rebuttal, but because pleading for mercy was the strategy pursued in the only other penalty hearing in which he participated.

Likewise, it is difficult to imagine a reasonable strategic choice not to put on mitigation evidence here. The prosecution asserts that the decision not to present mitigating evidence represents a reasonable choice to keep the “big picture” in front of the jury by foregoing the minutiae of Hendricks’ past and its relation to specific mitigating factors. This argument is a stretch at best. The choice that must be defended as strategic is not a decision about how best to present mitigating evidence, but one about whether to investigate mitigating evidence at all. The record does not show, and the prosecution does not sufficiently explain, how presenting evidence of Hendricks’ nightmarish upbringing and mental problems could prejudice him.

Once these facts are established, the legal analysis falls into place. Begging for mercy is not incompetence per se. See Deutscher v. Whitley, 884 F.2d 1152, 1159 (9th Cir.1989) and cases cited therein. However, where counsel is on notice that his client may be mentally impaired, counsel’s failure to investigate his client’s mental condition as a mitigating factor in a penalty phase hearing, without a supporting strategic reason, constitutes deficient performance. Id.; Evans v. Lewis, 855 F.2d 681, 686-37 (9th Cir.1988). Accordingly, Berman’s performance here fell below constitutionally acceptable standards.

The prosecution takes an all or nothing approach, contending that if the investigation into mental defenses in the guilt phase was sufficient to foreclose further investigation in the guilt phase, it must be sufficient to foreclose investigation into mitigation based on mental condition in the penalty phase as well. The differing legal standards in the guilt and penalty phases render this argument a non-sequitur. Evidence of mental problems may be offered to show mitigating factors in the penalty phase, even though it is insufficient to establish a legal defense to conviction in the guilt phase. See Cal.Penal Code § 190.3(d), (h). So, it does not follow that an investigation sufficient to foreclose the possibility of a mental defense necessarily forecloses the possibility of presenting evidence *1044 of mental impairment as mitigation in the penalty phase.

Moreover, the broad latitude given defendants to present mitigating evidence encompasses more than evidence relating to mental health. Cal.Penal Code § 190.3(k) (Any other circumstance which extenuates the gravity of the crime even though it is not a legal excuse for the crime [is a relevant mitigating factor]). Accordingly, even if Berman’s investigation into a mental defense in the guilt phase was sufficient to extinguish any duty to look into mental impairment as a mitigating factor (which it is not), Berman would still have to investigate Hendricks’ hard childhood, drug problems, etc., as they are relevant mitigating factors independent of Hendricks’ psychiatric problems.

B. Prejudice

The district court found that Hendricks was prejudiced by Berman’s ineptitude in several ways: readily available mitigating evidence was not presented due to the lack of investigation; Dr. Carson’s account of Hendricks’ past was not corroborated, undermining the value of her testimony; and the little evidence that Berman did put on was not connected to the statutory mitigating factors.

“The Constitution prohibits imposition of the death penalty without adequate consideration of factors which might evoke mercy.” Deutscher v. Whitley, 884 F.2d at 1161 . It is undisputed that the storehouse of information chronicling Hendricks’ miserable life story detailed in Dr. Lisak’s declaration was readily available and would have been relevant to several mitigating factors. Ber-man’s failure to present this information sufficiently undermines confidence in the fairness of the death verdict to satisfy the prejudice prong of Strickland .

The government asserts that the evidence against Hendricks was so overwhelming that the presentation of mitigating evidence could not have had an impact on the sentence. The determination of whether to impose a death sentence is not an ordinary legal determination which turns on the establishment of hard facts. The statutory factors give the jury broad latitude to consider amorphous human factors, in effect, to weigh the worth of one’s life against his culpability. Presumably the imposition of a death sentence is entrusted to a jury because it is a uniquely moral decision in which bright line rules have a limited place. In light of the whole record, and despite the substantial evidence of aggravation, we conclude that the failure of Berman to present mitigating evidence rendered the sentencing hearing neither fair nor reliable. See Lockhart v. Fretwell, 506 U.S. 364, 368-70 , 113 S.Ct. 838, 842-43 , 122 L.Ed.2d 180 (1993). But for Berman’s ineffective performance, there is a reasonable probability that the death sentence would not have been imposed. See id. at 373-75 , 113 S.Ct. at 845 (O’Connor, J., concurring); see also Clabourne v. Lewis, 64 F.3d 1373, 1378 (1995). Given the law’s sensitivity to failures in the penalty phase of a death penalty case, Berman’s failure to present Hendricks’ history of sexual and physical abuse was prejudicial.

The prosecution counters that the mitigating evidence was adequately presented to the jury. However, the three witnesses Berman presented in the penalty phase did little to aid in Hendricks’ fight for his life. Rowena Bates’ testimony was extremely limited, barely touching on Hendricks’ rape and disappointment over the rape of Teal, Bates’ son. Hendricks himself did little more than echo his taped confession and embellish it with confession of yet another murder. Dr. Carson testified at length regarding Hendricks’ past, but as this Court has already noted, the lack of corroborating evidence combined with the prosecution’s devastating cross-examination may have left the jury with the impression that Hendricks’ unfortunate history was all a fabrication. Hendricks v. Vasquez, 974 F.2d at 1110 . Given the effectiveness of the state’s cross-examination of Dr. Carson, as the district court concluded, this case is more akin to those cases in which no mitigating evidence was put on than those cases in which cumulative evidence was put on. Without some corroboration, Hendricks’ word, whether received through Dr. Carson or Hendricks himself, probably counted for little to the jurors.

*1045 Finally, the prosecution asserts that there is no requirement that an attorney link the mitigating evidence to the specific statutory factors. Faulting counsel for this failure would probably narrow the boundaries of acceptable argument style too much. However, Berman’s failure to present the abundant and available background information on Hendricks in any credible form fully satisfies the prejudice required under Strickland .

V. CONCLUSION

In the guilt phase, Hendricks’ counsel was entitled to rely on the shared and unqualified opinion of his two mental health experts that his client was sane. Absent a request from the experts, counsel had no duty to investigate the social history or any other bases of the expert’s psychiatric diagnosis.

In the penalty phase, Hendricks’s counsel knew of his client’s difficult life, had the opportunity to present mitigating evidence, and squandered it. Counsel’s failure here, in light of the costs of error, fell below the Strickland standard.

For these reasons, we affirm the district court Order in all respects.

. This is Hendricks' third appeal to this court. In 1990, Hendricks sought habeas relief which the district court summarily denied. On appeal, this Court remanded for further proceedings. Hendricks v. Vasquez, 908 F.2d 490 (9th Cir.1990). On remand, Hendricks filed an amended petition along with a motion for an evidentiary hearing. The district court dismissed the petition and denied the motion for an evidentiary hearing. On appeal, this Court again remanded for an evidentiary hearing on Hendricks’ ineffective assistance of counsel claims. Hendricks v. Vasquez, 974 F.2d 1099 (9th Cir.1992). After the case was reassigned, the court issued the order now under review.