I dissent.

It is a melancholy truth that as the final arbiter of the meaning and effect of the California Constitution, this court can with impunity deny our citizens rights guaranteed by that charter. What this court cannot do with impunity is *617 deny our citizens rights guaranteed by the United States Constitution. In the case at bar the majority attempt to do exactly that.

First, in the capital trial of this petitioner the prosecution failed to correct testimony by its key informant witnesses that it knew or should have known was false. Under decisions of the United States Supreme Court directly in point, this failure violated petitioner’s right to due process of law guaranteed by the Fifth Amendment to the federal Constitution, made applicable to the states through the Fourteenth Amendment. The standard of prejudice governing that constitutional violation is also a matter of federal law, declared and enforced by the United States Supreme Court and the federal courts. The majority purport to apply that standard to the facts of this case, and hold the federal constitutional violation harmless. As will appear, however, to reach that result the majority distort the meaning of the federal standard and misstate the record. Yet in neither respect will the majority have the last word. If petitioner seeks federal habeas corpus relief, the federal courts will both decide the correct meaning of the federal standard of harmless error and make an independent examination of this record to determine prejudice. I am confident that such examination will show the constitutional violation in this case to be prejudicial.

The same is true of the second and third grounds of relief at issue in this proceeding—the incompetence of trial counsel in failing to investigate the perjurious witnesses for impeachment evidence and in failing to investigate petitioner’s background for mitigating evidence. Each is a violation of petitioner’s right to the effective assistance of counsel guaranteed by the Sixth Amendment, and the standard of prejudice governing that violation is likewise a matter of federal law. Again the majority purport to apply that standard to the facts and hold the violations harmless, but reach that result by distorting the meaning of the standard and misstating the record. Again the federal courts, on an application for federal habeas corpus relief, will have the last word. And again I believe the violations will be held prejudicial.

In this opinion, therefore, I write primarily for the benefit of my federal colleagues. I deeply regret that petitioner must turn to another forum to vindicate his federal constitutional rights. When those rights are violated in a California state court, as they were here, the primary responsibility to redress the violation falls on the California judiciary. A majority of this court have twice had the opportunity to shoulder that responsibility in the case at bar (see also People v. Jackson (1980) 28 Cal.3d 264, 319-336 [ 168 Cal.Rptr. 603 , 618 P.2d 149 ] [dis. opn. of Mosk, J.] (hereafter Jackson I)), but have *618 twice failed in the task. Fortunately for petitioner, on the questions in issue this Supreme Court is neither infallible nor final. 1

I. Prosecution’s Knowing Use of Perjured Testimony

As part of its case-in-chief at petitioner’s trial the prosecution presented the testimony of two jailhouse informants, Mark Mikles and Ronald McFarland. On direct examination each of these witnesses testified to highly incriminating admissions that petitioner assertedly made to him while in custody. Each witness also acknowledged that he was currently under or facing sentence on prior felony convictions. On cross-examination defense counsel sought to impeach each witness on the ground of bias by asking questions suggesting that the witness expected to receive various sentencing benefits in exchange for his testimony against petitioner. On redirect examination the prosecutor sought to rehabilitate each witness by eliciting testimony in which each squarely denied that anyone in law enforcement had promised him anything for testifying against petitioner. The testimony denying such promises, it now appears, was false, and the prosecution either knew or should have known it was false.

A. The Mikles Testimony

The record on this issue is unusually clear; indeed, with respect to the perjury of the informant Mikles it verges on the overwhelming. Mikles testified in petitioner’s trial on December 27, 1978. As of that date, Mikles was in custody as a result of three separate criminal proceedings:

First, on March 22, 1978, Mikles was given concurrent sentences of six years each on two counts of robbery by then-Judge John A. Arguelles of the *619 Los Angeles Superior Court, sitting in Department South K in Long Beach (case No. A018325). 2

Second, on April 13, 1978, Mikles pleaded guilty to two counts of robbery before Judge William A. Munnell of the Los Angeles Superior Court, sitting in Department Southeast K in Norwalk (case No. A443866). 3 Mikles had not yet been sentenced on these charges when he testified eight months later in petitioner’s trial. The significance of this fact is explained below.

Third, at the time of petitioner’s trial a federal parole hold was in place against Mikles to enforce a pending 42-month sentence for federal parole violation.

At petitioner’s trial, defense counsel noted that Mikles had not yet been sentenced in the Norwalk case and suggested that the witness had a strong desire for the forthcoming sentence in that case to run concurrently with his six-year term in the Long Beach case. Mikles replied that is exactly what would happen, claiming it was “part of the plea bargain” he had made in April 1978. On redirect examination the prosecutor reinforced the point by asking Mikles whether it was true that his plea bargain in the Norwalk case “had no bearing” on his decision to testify against petitioner; Mikles answered, “That is correct.”

Apparently unsatisfied, defense counsel returned to the topic and expressed surprise that if Mikles truly knew what his sentence on the Norwalk charges would be he still had not been sentenced eight months after pleading guilty in that case. On further redirect examination the prosecutor sought to rehabilitate the witness by inviting Mikles to explain the long sentencing delay. In reply, Mikles testified that he had not yet been sentenced because the police had “new evidence” proving his innocence of the robbery charges in the Norwalk case. Driving the point home, the prosecutor then asked Mikles, “In other words, the reason you have not been sentenced in Norwalk on the Norwalk charges is not because of anything to do with your testifying in this case [i.e., in petitioner’s trial]?” Mikles reiterated, “It is new evidence in my robberies is what it is.”

This testimony was false: as will appear in more detail below, Mikles’s true reason for seeking to delay his sentencing in the Norwalk case was simply to give him more time to curry favor with law enforcement authorities—and through them, with judges—by testifying as an informant in petitioner’s trial and other trials as well.

*620 Mikles’s boldest lies, however, came in response to the prosecutor’s broader effort to insulate him against any inference that “anyone” ever promised him “anything” for testifying against petitioner. On redirect examination the following colloquy took place:

After the prosecutor established that he never personally promised Mikles any benefit for testifying, the following colloquy occurred:

The utter falsity of this testimony was exposed at the reference hearing, where Mikles told a very different version of the events. That version was corroborated by an unimpeachable law enforcement witness, by documentary exhibits, and by the actual dispositions of the cases against Mikles.

To begin with, Mikles admitted at the reference hearing that in March or April 1978, shortly after being sentenced in the Long Beach case and pleading guilty in the Norwalk case, he turned informant in the hope of ameliorating his punishment in those two matters. To this end he contacted *621 Los Angeles Deputy Sheriff Lavona Shea, who was known to work with informants, and offered to furnish information in several major cases, including petitioner’s. Mikles testified as follows:

According to Mikles, Deputy Shea promised him she would make sure that in the Long Beach and Norwalk cases Judges Arguelles and Munnell were told of his cooperation, provided he actually testified for the prosecution in the cases in which he had information. Mikles conceded that Shea’s promise was “one of the reasons” he subsequently testified against petitioner. Mikles then testified that Shea also agreed to assist him in seeking a reduction of the 42-month sentence he was facing for federal parole violation, and “that was another reason” why he testified against petitioner.

The terms of the deal were succinctly summarized in the following exchange:

As Mikles subsequently acknowledged, Deputy Shea “fulfilled all her promises” to him. Her assistance took several forms. Mikles’s first problem was the fact that in the Norwalk case he was scheduled to be sentenced on May 16, 1978, i.e., before he would have the opportunity to testify against petitioner and the defendants in two other murder cases. The solution was for Shea to help Mikles obtain continuances of that sentencing hearing until he could give such testimony. After Mikles acknowledged that Shea was present when Judge Munnell finally sentenced him, the following exchange took place:

With Deputy Shea’s assistance, the sentencing hearing in the Norwalk case was then continued for two months, and ultimately for a total of nine months until February 26, 1979. At the reference hearing Mikles agreed that the reason for seeking these continuances was to give him more time to *623 testify for the People in more cases, and thus “help them get more convictions” with the goal of “currying favor with law enforcement.”

According to Mikles, Deputy Shea also fulfilled her promises to make Judges Arguelles and Munnell aware of his services in testifying for the People. She repeatedly assured him that the fact and extent of his cooperation would be conveyed to the two judges both by letters and in person by her or by other law enforcement personnel. And she was as good as her word: he testified that Shea appeared both at the sentencing hearing in the Norwalk case and at a hearing on May 31, 1979, to modify Mildes’s sentence in the Long Beach case, and on each occasion she spoke to the court on Mikles’s behalf.

Finally, under questioning by counsel for petitioner Mikles admitted that at the time he turned informant in March and April 1978 he knew that “the more serious a case that [he] was going to testify in,” the more consideration he would receive from Judges Arguelles and Munnell, and he also knew of instances in which “informants had lied about what they claimed to have overheard and [had] gotten away with the lies and got out of jail, too.” Mikles further admitted that “one of the most important things in [his] entire life at that time was to get out of jail,” and that he would have done “whatever [he] had to do to get out of jail” short of killing someone. None of these dramatic admissions, of course, were heard by the jury in petitioner’s trial. 5

The version of the facts that Mikles told at the reference hearing was amply corroborated. First, Deputy Shea testified to essentially the same effect. 6 She testified that Mikles contacted her through the intermediary of another informant; that before meeting with him she checked his criminal record and learned he was under a six-year sentence in the Long Beach case, *624 was awaiting sentence in the Norwalk case, and was facing a forty-two-month federal sentence for parole violation; and that in her several meetings with Mikles he informed her of incriminating statements assertedly made to him by petitioner and by defendants in other murder cases. When asked what she promised Mikles in exchange for such information, she replied: “What I told Mr. Mikles I would do is if his information was productive information, and it could be used, and we could get a conviction, I would advise the district attorney that was handling his case of his assistance to us.” She explained that it was the practice for prosecutors to serve as “channels” of such information, i.e., they would convey it to the sentencing judges in the cases against Mikles.

But Deputy Shea did far more than that for Mikles. First, in the Norwalk case she helped him get the sentencing continuances he wanted: as she testified, “I probably contacted the district attorney’s office and asked them if they could continue the case and told them why. ... He was working as an informant, and I advised them that he was still working and it would help if they could get the case continued.” The prosecutors evidently complied with her requests.

Second, on January 19, 1979—i.e., only 16 working days after Mikles testified against petitioner and before the trial had even ended—Deputy Shea wrote a letter to the Los Angeles Probation Department “to supply additional and pertinent information for your consideration” in preparing Mikles’s probation report. In the letter Shea reported that Mikles had testified for the prosecution in petitioner’s trial and that “according to [Deputy District Attorney Paul] Marin”—the prosecutor in that trial—“Mikles’ testimony was extremely significant especially regarding the aggravation aspect which was the basis for the death penalty.” 7

Third, Deputy Shea acknowledged at the reference hearing that when Mikles was finally sentenced in the Norwalk case on February 26, 1979, “I testified on his behalf.” According to Shea, she told the court “what he had done over a period of a couple of years, the cases he had testified in, because most of them were major murder cases, as with your client. And I just provided the information from the witness stand that he had testified in these various cases. And that he had always been extremely reliable.”

Fourth, Deputy Shea testified that she also appeared at the hearing on April 5, 1979, on Mikles’s application for modification of the six-year *625 sentence that Judge Arguelles had imposed on him in the Long Beach case. The transcript of that proceeding was admitted into evidence at the reference hearing (petitioner’s exhibit 22) and hence is properly before us. It discloses the following sequence of events. The first five witnesses were law enforcement officers who briefly stated that in several prior cases Mikles had furnished reliable information or given testimony helpful to the prosecution. The sixth and final witness was Deputy Shea, who made a far more partisan presentation on Mikles’s behalf. First, Judge Arguelles acknowledged receipt of a copy of Shea’s above discussed letter of January 19, 1979, to the probation department. Shea then described to Judge Arguelles additional information that Mikles had furnished after she wrote that letter. 8 She next identified three investigators who were unable to attend the hearing but who, she asserted, “would testify” that Mikles gave them reliable information in still other cases.

Deputy Shea’s testimony then turned remarkably personal. She told Judge Arguelles that she was in contact with Mikles “on a weekly basis”; that Mikles’s wife and son had moved out of the area in the hope that he might receive “some consideration” (i.e., from Judge Arguelles) and upon his release might start a new life in a place free from criminal temptations; and that because Mikles had testified as an informant in the five cases she had mentioned, “it would be very difficult for him to go back into the criminal element.” 9

*626 Deputy Shea concluded by an apparent appeal to Judge Arguelles’s sympathy. She testified that to her knowledge Mikles “has a minimum of three contracts out on his life” and “his wife and child have been threatened.” And she went so far as to argue that Mikles had served the first year of his six-year sentence in county jail, which she asserted was “two or three times” more difficult a way to do time than in a state or federal prison, “and the reason he has done this time in county jail rather than a, shall we say, for lack for a better term, more luxurious situation, has been because he has been testifying in these five cases for the prosecution.”

After Deputy Shea testified, Judge Arguelles noted for the record that he had also received a number of letters and telephone calls from other law enforcement personnel on Mikles’s behalf. He concluded, “I have been virtually inundated with requests from persons, so they must feel very strongly about the defendant [Mikles] to come forward in the quantity and with the quality of recommendations that I have received. So all of this individually [and] cumulatively has to have an impact on me.” I discuss that impact below.

Mikles’s recanting testimony at the reference hearing is further corroborated by three documentary exhibits that the referee specifically relied on and appended to his supplemental report filed in this court on September 21, 1989. The first document (appen. A) is Deputy Shea’s above discussed letter of January 19, 1979, to the probation department. The second document (appen. B) is a letter that Judge Arguelles announced he would write at the end of the sentence modification hearing on April 5, 1979, asking the Director of Corrections to formally request him (Judge Arguelles) to recall his 1978 commitment of Mikles for the purpose of modifying the sentence. In the letter Judge Arguelles explained to the Director of Corrections that six law enforcement officers had testified that day on Mildes’s behalf and “The substance of the testimony was that the defendant [Mikles] had been of considerable value to all of their agencies as a prosecution witness in murder and robbery cases. Even this week he was testifying on their behalf.”

The third document (appen. C to referee’s supplemental rep.) is a letter from the Los Angeles Sheriff’s Office to federal parole authorities “for *627 presentation to the Federal Parole Hearing Board.” 10 The letter recited the details of Judge Arguelles’s modification of Mikles’s sentence (discussed below) and stated that although the trials in two additional murder cases in which Mikles testified (People v. Luna and People v. Humphries) resulted in mistrials because of hung juries, both were scheduled for retrial and Mikles would again testify for the prosecution. The letter ended, “Hopefully, this letter and information enclosed will be reviewed and consideration will be given in behalf of Mark S. Mikles.”

The final corroboration of Mikles’s recanting testimony is the dramatic sentencing benefits he received for his services as an informant. First, on February 26,1979, in the Norwalk case, Judge Munnell granted Mikles four years’ probation with no time in custody, as punishment for his conviction— before he testified against petitioner—on two counts of armed robbery.

Second, on May 31, 1979, in the Long Beach case, Judge Arguelles rescinded the sentence of two concurrent terms of six years each that he had imposed on Mikles as punishment for his conviction—before he testified against petitioner—on two counts of armed robbery. At the outset of the sentence modification hearing on April 5, 1979, Judge Arguelles had asked the prosecutor what lesser sentence his office intended to request; the prosecutor had replied that his office took the position that a reduction of two years, leaving a sentence of four years in prison, “would be fair.” Despite this recommendation by the prosecutor, Judge Arguelles granted Mities four years’ probation with no time in custody, and ordered him released forthwith.

Third, the referee found that the federal authorities thereafter removed their hold on Mikles and that he served no part of his previously pending 42-month sentence for federal parole violation. 11

In sum, the record establishes beyond question that Mikles lied under oath at petitioner’s trial when he denied that “anyone” in law enforcement promised him “anything” for testifying against petitioner. On the contrary, he was richly rewarded for that testimony, as he fully expected to be.

B. The McFarland Testimony

The second informant, McFarland, did not do so well, but it was not for want of trying. He, too, was promised certain benefits for testifying against petitioner, and he, too, lied about it under oath.

*628 When McFarland testified in petitioner’s trial (Dec. 29, 1978, and Jan. 3, 1979) he was in custody on a felony conviction: on May 26, 1978, he had been convicted of three counts of armed robbery in a trial before Judge D. Sterry Fagan of the Los Angeles Superior Court, sitting in Department South C in Long Beach, and on June 23, 1978, Judge Fagan had sentenced him to consecutive terms of five years to life on two of those counts and a concurrent term of five years to life on the third count. Six months later, at petitioner’s trial, defense counsel sought to learn precisely what happened at that sentencing hearing.

McFarland began by acknowledging that at the outset of the sentencing hearing his then-counsel, Deputy Public Defender Dennis Carroll, requested and obtained an order clearing the courtroom. When asked why the courtroom was thus cleared, McFarland testified it was because “I didn’t want nobody in there that I didn’t know," and specifically because he was in protective custody. 12 Defense counsel then suggested that the real reason why Carroll wanted the courtroom cleared was that he intended to urge leniency because McFarland had agreed to testify against petitioner. McFarland flatly denied that was the reason. Defense counsel then focused on Carroll’s ensuing argument to Judge Fagan, and suggested that the real reason Carroll gave for urging leniency was that McFarland had agreed to testify against petitioner. Again McFarland denied that was the reason, asserting that Carroll had stressed instead his youth and amenability to reform. Although McFarland finally and reluctantly admitted that Carroll “mentioned” he had agreed to testify against petitioner, he repeatedly testified that it was not a reason that Carroll gave for his plea of leniency. As will appear, this testimony was false.

On redirect examination the prosecutor sought to rehabilitate McFarland, as he had done with Mikles, by eliciting testimony to the effect that “no one” promised him “anything” for testifying against petitioner.

“Q. [by the prosecutor] Mr. McFarland, is it true that you saw me for the first time last week?

“A. [by McFarland] Yes, sir.

“Q. And is it true that no one promised you anything in exchange for your testimony against Mr. Jackson?

“A. Yes, that’s right.

*629 “Q. Now, when you were up in this court last June 23, 1978, when you were convicted by way of being sentenced on the robbery charges of which this Court found you guilty, as you indicated, was I here in court?

“A. No, sir.

“Q. Was there a District Attorney present at the time you were sentenced and convicted on that date of June 23, ’78?

“A. Yes, sir.

“Q. And did that District Attorney make any statements in your presence to the Court that the sentence should be lenient because you were going to testify in the Jackson case?

“A. No, sir.”

At the reference hearing, however, a much different picture emerged. Counsel for petitioner confronted McFarland with the transcript of his sentencing hearing before Judge Fagan on June 23, 1978. That transcript reported Carroll as advising Judge Fagan that the reason he asked for the courtroom to be cleared was because McFarland was cooperating with the prosecution in the Jackson case and had agreed to testify concerning statements that petitioner assertedly made to him. McFarland reluctantly acknowledged the truth of what the transcript reported. 13 Counsel for petitioner asked whether anyone in law enforcement had promised or told him of any benefit he would receive for his testimony, and McFarland replied, “I was promised nothing at all or told nothing.” Counsel then confronted the witness with another portion of the transcript of the sentencing hearing, in which Carroll advised Judge Fagan that McFarland “is asking the court [to] consider retaining jurisdiction in this case for this reason: Mr. Marin, the prosecutor in the Jackson case, indicated he may write a letter to the sentencing authorities at the completion of that case in support of perhaps a review at that time of Mr. McFarland’s sentence.” To make such a review possible, Carroll also asked Judge Fagan to retain jurisdiction over McFarland’s case.

McFarland acknowledged that Carroll made the quoted statement to Judge Fagan on his behalf. And when asked again whether he knew before his sentencing that Judge Fagan would be made aware of his agreement to testify against petitioner, McFarland reluctantly admitted that “I asked my *630 attorney to—I believe I asked him to mention it to the judge if, you know, maybe it would help my sentence out a little.”

Attorney Carroll also testified at the reference hearing, and acknowledged making the above quoted statement to Judge Fagan at McFarland’s sentencing hearing. Carroll testified to two letters that Deputy District Attorney Marin agreed, prior to McFarland’s sentencing, to write on McFarland’s behalf. First, Carroll testified that on April 6, 1978, he met with Marin and another deputy district attorney to negotiate a plea bargain for McFarland, but that both prosecutors declined to recommend the lighter sentence he proposed. 14 Carroll testified that after failing to obtain his bargain he asked Marin at least to write a letter to “the prison authorities” requesting that McFarland be transferred to Arizona to serve his time in that state, because he had family there and because his life could be at risk in a California prison after he testified against petitioner. According to Carroll, Marin replied that he did not know if such a transfer to Arizona were possible, but “he had no objection” to writing such a letter.

Second, Carroll testified about the letter referred to in his above quoted statement to Judge Fagan, i.e., a letter that Marin would write to Judge Fagan after petitioner’s trial, advising the judge that McFarland had indeed testified against petitioner. Carroll testified that although he did not know whether Marin actually wrote a letter to Judge Fagan about McFarland’s testimony against petitioner, “He agreed to write such a letter.” Indeed, Carroll testified that he thereupon told McFarland “what Mr. Marin had promised to do, i.e., write a letter to the judge after the Jackson case was over.”

Nothing in the record contradicts the foregoing testimony. 15 In my view the testimony is ample to support the referee’s finding that Marin agreed he would write two distinct letters on McFarland’s behalf after the latter had testified against petitioner. It follows that McFarland, like Mildes, lied under oath at petitioner’s trial when he claimed that “no one” promised him “anything” in exchange for his testimony against petitioner.

*631 C. The Governing Federal Law

It is a fundamental rule of federal constitutional jurisprudence that the due process clause of the Fourteenth Amendment “cannot tolerate a state criminal conviction obtained by the knowing use of false evidence.” (Miller v. Pate (1967) 386 U.S. 1, 7 [ 17 L.Ed.2d 690, 694 , 87 S.Ct. 785 ].) The rule has its roots in Mooney v. Holohan (1935) 294 U.S. 103 [ 79 L.Ed. 791 , 55 S.Ct. 340 , 98 A.L.R. 406 ]. There a petitioner convicted of murder in the California courts alleged that the sole basis of his conviction “was perjured testimony, which was knowingly used by the prosecuting authorities in order to obtain that conviction, and also that these authorities deliberately suppressed evidence which would have impeached and refuted the testimony thus given against him.” (Id. at p. 110 [ 79 L.Ed. at p. 793 ].) On these allegations he claimed a violation of the due process clause of the Fourteenth Amendment and sought federal habeas corpus. In his return—in the nature of a demurrer—the California Attorney General argued that due process required only notice and opportunity to be heard. The United States Supreme Court held that the petitioner’s allegations, if true, would warrant relief from his conviction. Rejecting the Attorney General’s position, the high court reasoned: “we are unable to approve this narrow view of the requirement of due process. That requirement, in safeguarding the liberty of the citizen against deprivation through the action of the State, embodies the fundamental conceptions of justice which lie at the base of our civil and political institutions. [Citation.] It is a requirement that cannot be deemed to be satisfied by mere notice and hearing if a State has contrived a conviction through the pretense of a trial which in truth is but used as a means of depriving a defendant of liberty through a deliberate deception of court and jury by the presentation of testimony known to be perjured. Such a contrivance by a State to procure the conviction and imprisonment of a defendant is as inconsistent with the rudimentary demands of justice as is the obtaining of a like result by intimidation.” (Id. at p. 112 [ 79 L.Ed. at p. 794 ].)

It is true that in the case at bar the prosecution did not solicit the false testimony but simply failed to correct it when it appeared; the prosecutor who presented the perjurious witnesses did not personally know their testimony was false; and the false testimony did not go to an element of the prosecution’s case but merely to the witnesses’ credibility. Yet in two cases that are “the leading Supreme Court rulings in this area” (Zeigler v. Callahan (1st Cir. 1981) 659 F.2d 254, 263 ) the high court dismissed these facts as distinctions without a difference. The two cases, as will appear, are directly in point.

The first is Napue v. Illinois (1959) 360 U.S. 264 [ 3 L.Ed.2d 1217 , 79 S.Ct. 1173 ] (hereafter Napue). There, one Hamer testified for the prosecution *632 and identified Napue as a coparticipant in the charged robbery-murder. The jury had been told that Hamer was serving a long sentence on his earlier conviction for the same crime. On cross-examination for bias, Hamer claimed only that an unnamed public defender had said he would try to “do what he could” for him. (Id. at p. 268, fn. 3 [ 3 L.Ed.2d at p. 1220 ].) When defense counsel asked directly whether anyone had promised any benefit for testifying, Hamer replied, “There ain’t nobody promised me anything.” On redirect examination the prosecutor drove the point home by asking the witness, “Have I promised you that I would recommend any reduction of sentence to anybody?” Again Hamer replied, “You did not.” (Id. at p. 267, fn. 2 [3 L.Ed.2d at pp. 1219-1220].)

The testimony was false. After Hamer testified against Napue and another participant in the same murder, the prosecutor in Napue’s trial filed a coram nobis petition on Hamer’s behalf in which he alleged that Hamer had been reluctant to testify against Napue without an assurance that the prosecution would recommend a reduction of his sentence, and the prosecutor had therefore represented to Hamer that if he cooperated in the prosecution of Napue and others “a recommendation for a reduction of his [Hamer’s] sentence would be made and, if possible, effectuated.” The prosecutor then asked the court to effect a “consummation of the compact” he had thus entered into with Hamer. (Napue, supra, 360 U.S. at p. 266 [ 3 L.Ed.2d at p. 1219 ].)

When Napue learned of Hamer’s perjury he sought postconviction relief, but was unsuccessful in the state courts. The United States Supreme Court unanimously reversed, holding that the prosecutor’s failure to correct the testimony of Hamer that he knew to be false denied Napue due process of law in violation of the Fourteenth Amendment.

The high court first reasoned that “a conviction obtained through use of false evidence, known to be such by representatives of the State, must fall under the Fourteenth Amendment, Mooney v. Holohan, 294 U.S. 103 .... The same result obtains when the State, although not soliciting false evidence, allows it to go uncorrected when it appears." (Napue, supra, 360 U.S. at p. 269 [3 L.Ed.2d at pp. 1220-1221].)

Next the high court addressed the question of the nature of the false testimony: “The principle that a State may not knowingly use false evidence, including false testimony, to obtain a tainted conviction, implicit in any concept of ordered liberty, does not cease to apply merely because the false testimony goes only to the credibility of the witness. The jury’s estimate of the truthfulness and reliability of a given witness may well be determinative *633 of guilt or innocence, and it is upon such subtle factors as the possible interest of the witness in testifying falsely that a defendant’s life or liberty may depend.” (Napue, supra, 360 U.S. at p. 269 [3 L.Ed.2d at pp. 1220-1221].)

Turning to the question of prejudice, the high court denied that it was bound by the determination of the state supreme court that Hamer’s false testimony could not in any reasonable likelihood have affected the verdict: “The duty of this Court to make its own independent examination of the record when federal constitutional deprivations are alleged is clear, resting, as it does, on our solemn responsibility for maintaining the Constitution inviolate.” (Napue, supra, 360 U.S. at p. 271 [ 3 L.Ed.2d at p. 1222 ].)

The state supreme court had found the constitutional violation harmless because the jury had other grounds for disbelieving Hamer—it knew that an unnamed public defender had told Hamer he would “do what he could” for him. The high court rejected that conclusion, reasoning that if the jury had also known of the promise made to Hamer by the prosecutor, “it might well have concluded that Hamer had fabricated testimony in order to curry the favor of the very representative of the State who was prosecuting the case in which Hamer was testifying, for Hamer might have believed that such a representative was in a position to implement (as he ultimately attempted to do) any promise of consideration.” (Napue, supra, 360 U.S. at p. 270 [3 L.Ed.2d at pp. 1221-1222].)

The second leading case is Giglio v. United States (1972) 405 U.S. 150 [ 31 L.Ed.2d 104 , 92 S.Ct. 763 ] (hereafter Giglio). There the high court extended the rule of Napue, supra, 360 U.S. 264 , to the situation in which the prosecutor who presented the perjurious witness did not know that the witness had in fact been promised a benefit for testifying. In Giglio , one Taliento testified for the prosecution and identified the petitioner as a coparticipant in the crime; although Taliento was named as a coconspirator, he was not indicted. On cross-examination for bias, defense counsel asked Taliento whether “anybody” told him “at any time” that if he testified he would not be prosecuted; Taliento replied, “Nobody told me I wouldn’t be prosecuted.” (Id. at p. 151 [31 L.Ed.2d at pp. 106-107].) In closing argument to the jury, the assistant United States attorney emphasized that Taliento “received no promises that he would not be indicted.” (Id. at p. 152 [31 L.Ed.2d at pp. 107-108].)

The testimony was false, but the assistant United States attorney who tried the case did not know it. Prior to trial another assistant United States *634 attorney—who had presented the case to the grand jury—had indeed promised Taliento that if he testified at trial he would not be prosecuted. The United States Supreme Court unanimously reversed the ensuing judgment of conviction, relying principally on Napue, supra, 360 U.S. 264 . The court explained that the failure to disclose the promise to the prosecutor did not require a different result: “whether the nondisclosure was a result of negligence or design, it is the responsibility of the prosecutor. The prosecutor’s office is an entity and as such it is the spokesman for the Government. A promise made by one attorney must be attributed, for these purposes, to the Government.” (Giglio, supra, 405 U.S. at p. 154 [31 L.Ed.2d at pp. 108-109].)

In both Napue and Giglio the testimony of the perjurious witness was crucial to the prosecution’s case. 16 In such circumstances it is not surprising that the high court spent little time on the question of prejudice. The rule that emerged from the two decisions, however, was that a conviction obtained by the knowing use of false testimony must be set aside “if there is any reasonable likelihood that the false testimony could have affected the judgment of the jury.” (United States v. Agurs (1976) 427 U.S. 97, 103 [ 49 L.Ed.2d 342, 349-350 , 96 S.Ct. 2392 ], citing Napue, supra, 360 U.S. at p. 271 [ 3 L.Ed.2d at p. 1222 ], and Giglio, supra, 405 U.S. at p. 154 [31 L.Ed.2d at pp. 108-109].)

As the majority herein point out (maj. opn., ante, pp. 597-598), this standard of prejudice “generally has been equated with the ‘harmless beyond a reasonable doubt’ standard of Chapman v. California (1967) 386 U.S. 18 [ 17 L.Ed.2d 705 , 87 S.Ct. 824 , 24 A.L.R.3d 1065 ].” In a footnote at this point (fn. 10) the majority quote Justice Blackmun’s analysis in United States v. Bagley (1985) 473 U.S. 667, 679, 680 , footnote 9 [ 87 L.Ed.2d 481, 492-493 , 105 S.Ct. 3375 ]: “The Court in Chapman noted that there was little, if any, difference between a rule formulated, as in Napue , in terms of ‘ “whether there is a reasonable possibility that the evidence complained of might have contributed to the conviction,” ’ and a rule ‘ “requiring the *635 beneficiary of a constitutional error to prove beyond a reasonable doubt that the error complained of did not contribute to the verdict obtained.” ’ [Citation.] It is therefore clear, as indeed the Government concedes, . . . that this Court’s precedents indicate that the standard of review applicable to the knowing use of perjured testimony is equivalent to the Chapman harmless-error standard.”

Whether formulated as a rule requiring a reversal if the court finds “any” reasonable likelihood that the constitutional violation “could have affected” the verdict, or as a rule requiring reversal unless the beneficiary of the violation “proves beyond a reasonable doubt” that the violation did not “contribute” to the verdict, this is a high standard of prejudice—indeed, the highest that the law requires the state to meet in order to save a tainted conviction. But the need for such a standard is clear: as explained in United States v. Agurs, supra, 427 U.S. 97, 104 [ 49 L.Ed.2d 342, 350 ], the high court applies this standard in cases of knowing use of perjured testimony “not just because they involve prosecutorial misconduct, but more importantly because they involve a corruption of the truth-seeking function of the trial process.” In short, they affect ‘the honor and integrity of the sovereign’s administration of justice . . . because they are subversive of the institution of fair trial itself.’ ” (Link v. United States (8th Cir. 1965) 352 F.2d 207, 211 .)

It has fallen primarily to the federal appellate courts to implement the rule of Napue, supra, 360 U.S. 264 , and Giglio, supra, 405 U.S. 150 . Under their authority, numerous federal appellate courts have reversed convictions of federal crimes because the government violated the due process clause of the Fifth Amendment by failing to correct testimony of its witnesses that it knew or should have known was false—most often, as here, false testimony denying that the witnesses had been promised benefits for cooperating with the prosecution. (United States v. Iverson (D.C. Cir. 1980) 637 F.2d 799 ; United States v. Barham (5th Cir. 1979) 595 F.2d 231 ; United States v. Butler (9th Cir. 1978) 567 F.2d 885 ; United States v. Sanfilippo (5th Cir. 1977) 564 F.2d 176 ; United States v. Pope (9th Cir. 1976) 529 F.2d 112 ; United States v. Kaplan (7th Cir. 1972) 470 F.2d 100 ; see also U.S. v. Rivera Pedin (11th Cir. 1988) 861 F.2d 1522, 1529-1530 [witness falsely denied talking with friend of defendant about being paid not to testify for prosecution]; United States v. Sutton (4th Cir. 1976) 542 F.2d 1239 [no false testimony, but prosecutor falsely assured the jury there had been no threats to his witness].)

Because the same rule applies to the states through the due process clause of the Fourteenth Amendment, other federal appellate courts have granted federal habeas corpus relief to petitioners convicted of state crimes under *636 similar circumstances (Brown v. Wainwright (11th Cir. 1986) 785 F.2d 1457 ; Campbell v. Reed (4th Cir. 1979) 594 F.2d 4 ; United States ex rel. Wilson v. Warden Cannon, etc. (7th Cir. 1976) 538 F.2d 1272 ; Boone v. Paderick (4th Cir. 1976) 541 F.2d 447 ; Hamric v. Bailey (4th Cir. 1967) 386 F.2d 390, 393-395 ) or have found sufficient grounds to remand the matter to the federal district court for an evidentiary hearing (Blankenship v. Estelle (5th Cir. 1977) 545 F.2d 510 ; Reickauer v. Cunningham (4th Cir. 1962) 299 F.2d 170 ).

D. Application of the Governing Federal Law to Petitioner’s Case

On the record in this case it is clear that petitioner’s right to due process of law under the Fourteenth Amendment was violated when the prosecution failed to correct the testimony of the informants Mikles and McFarland falsely denying that anyone had promised them any sentencing benefits for their cooperation. The majority so conclude (maj. opn., ante, p. 597), and I agree with them on this point. 17

I do not agree, however, with the majority’s further conclusion that this patent violation of petitioner’s federal due process rights was harmless. As the foregoing cases illustrate, in this context the correct procedure is for the appellate court to answer two questions in sequence: First, if the prosecutor had performed his constitutional duty to disclose to the jury that his witness falsely denied any promise of benefits, would the jury have disbelieved the remainder of the witness’s testimony? Second, if the jury had disbelieved the remainder of the witness’s testimony, would its rejection of that testimony have affected its verdict?

Curiously, the majority neither ask nor answer the first question: their prejudice analysis appears to assume that if the jury had known of the false testimony of Mikles and McFarland it would have rejected their testimony in to to. While I agree with that assumption in this case, it may not be true in other cases. The referee made a specific finding on the point, and it deserves a few words in the interest of clarity.

*637 Disclosure that a witness lied under oath when he denied that anyone promised him anything in exchange for his testimony undermines the remainder of his testimony in two separate but cumulative ways. First, it discloses that the witness had a specific incentive to lie: the jury could well conclude that a witness who had been promised valuable sentencing benefits if he testified for the prosecution would be likely to color or even fabricate his testimony in an effort to please the prosecutor and “earn” the benefits. As the referee in the case at bar observed in his original report, “The informer witness has every reason to believe that the stronger his testimony may be against a defendant, irrespective of its truth, the more likely the law enforcement officers’ promises of assistance will result in leniency from the judges capable of providing leniency.” Indeed, the United States Supreme Court recognized in Napue, supra, 360 U.S. at page 269 [3 L.Ed.2d at pages 1220-1221], that “it is upon such subtle factors as the possible interest of the witness in testifying falsely that a defendant’s life or liberty may depend.” 18

Second and no less important, the disclosure demonstrates that the witness has not only the incentive but also the ability and willingness to lie: the jury could well conclude that a witness who actually lied under oath in one part of his testimony—when he denied any promise of benefits—has shown that he would be able and willing to lie in the remainder of it. This is the psychological truth that is captured in the maxim, falsas in uno, falsus in omnibus. 19

In the case at bar, therefore, disclosure that Mi kies and McFarland falsely denied any promise of benefits would have been devastating to their credibility, because it would have demonstrated that they had both the incentive to lie under oath and a history of doing so in this very trial. Coupled with the content of their testimony at the reference hearing, moreover, is the manner in which they testified. On the basis of that experience the referee found as follows in his original report: “Having listened to their testimony and observed their demeanor at the Reference Hearing, it is the Referee’s conclusion that, had the impeachment evidence been presented at petitioner-defendant Jackson’s trial, the jury, in all lik[e]lihood, would have considered the testimony of McFarland and Mikles as being untrue . . . .” This is a purely factual finding supported by ample substantial evidence, and we *638 should therefore adopt it in answering the first question in the prejudice analysis. 20

The second question in the prejudice analysis is as follows: assuming, as do the majority, that the jury would have disbelieved the testimony of Mikles and McFarland in its entirety, is there “any reasonable likelihood” that its rejection of that testimony “could have affected” its verdict? The majority conclude there is no such possibility, but the record and the federal decisions are clearly to the contrary.

1. The Special Circumstance Findings

I begin with the first contested issue—the effect of the error on the special circumstance findings. 21

Petitioner was tried under the 1977 death penalty law. (Pen. Code, former § 190 et seq.; Stats. 1977, ch. 316, p. 1256.) The information charged petitioner with the murders of Mrs. Curtis and Mrs. Ott, but the prosecution sought the death penalty only for the Ott murder: the amended information charged petitioner with two special circumstances relating to that crime, viz., that the Ott murder was willful, deliberate, and premeditated and was committed during the course of a burglary (Pen. Code, former § 190.2, subd. (c)(3)(v)) and that petitioner committed an additional first degree murder by killing Mrs. Curtis (id., subd. (c)(5)). The jury found both allegations true. As the majority acknowledge, however, under the 1977 death penalty law neither of the special circumstances alleged in this case could be found true unless the jury also found that petitioner was personally present during the commission of the act causing Mrs. Ott’s death and that he physically aided or committed such acts “with intent to cause death.” (Id., subd. (c).) The amended information therefore alleged these matters, and the jury also found such allegations true. It is the latter findings that petitioner now attacks, particularly the finding that he aided or committed the fatal act “with intent to cause [Mrs. Ott’s] death.”

In his original report the referee found that if the jury had known the truth about Mikles and McFarland it would have not only rejected their testimony *639 as unbelievable but would also have deemed “the remaining evidence of the prosecution insufficient to sustain these special circumstance allegations.” I recognize that the latter finding of the referee resolves a mixed question of law and fact and hence is subject to our independent review. (In re Marquez, supra, 1 Cal.4th 584, 603 .) As will appear, however, a fair reading of the record compels us to reach the same conclusion as the referee on this question.

The majority find no prejudice because of “the strength of the evidence provided by [petitioner’s] own statement to the police and by the numerous noninformant witnesses describing [petitioner’s] statements and demeanor shortly after the crimes” (maj. opn., ante, p. 599). In other words, the majority disregard the testimony of Mikles and McFarland, weigh the remaining testimony, and conclude it has such “strength” that there is no “reasonable likelihood” the error “could have affected” the jury’s determination of the special circumstance allegations. Or, to use the alternate wording of the same standard of prejudice, the majority in effect conclude that the noninformant evidence is of such “strength” that it “proves beyond a reasonable doubt” that the error did not “contribute” to that determination. I cannot agree. The evidence relied on by the majority for this purpose is far too weak to support the heavy burden placed on it by the controlling federal law.

The majority first invoke petitioner’s own statements to the police. The crimes against Mrs. Curtis occurred on August 29, 1977, and she died four days later; the crimes against Mrs. Ott occurred on September 7, 1977. Petitioner voluntarily surrendered to the police at 4:30 p.m. on September 17, 1977; the police arrested, fingerprinted, and identified him, and began taking a tape-recorded statement from him shortly after 5 p.m. In that statement petitioner denied any involvement in the crimes against either victim, and the police soon terminated the interrogation. The police told petitioner they did not believe his story, and he agreed to tell the truth. The police then brought in the homicide detectives who had been assigned to the cases, and a second tape-recorded interrogation, lasting half an hour, began shortly after 6 p.m.

In the second statement petitioner made a number of incriminating admissions. Insofar as relevant here, petitioner admitted the elements of his guilt of felony murder in each case: he acknowledged that he and one or more confederates broke into the apartment of each victim at night with intent to steal, and that during the commission of each burglary the victim received a fatal beating. With respect to the Ott murder, petitioner also admitted two elements required for proof of the special circumstance allegations: he *640 admitted that he was personally present during the commission of the acts causing Mrs. Ott’s death—i.e., the infliction of the fatal beating—and that he physically aided the commission of those acts by striking one blow himself; as the majority point out, that blow helped to bring about the submission of Mrs. Ott to the will of her assailants.

As the majority also concede, however, nothing in petitioner’s statements to the police concerning the Ott murder constituted an admission of the third—and equally necessary—element of the charged special circumstances, i.e., that he acted with intent to cause Mrs. Ott’s death. On the contrary, insofar as disclosed by petitioner’s statements his intent was simply to steal. 22 When asked why he entered Mrs. Ott’s apartment, petitioner replied, “To find the money; that’s all.” He added, “That’s all we was gonna do is [find the money] and leave.” When asked what he was looking for when he proceeded to search the apartment, he replied, “For money; that’s all.” Petitioner said that when Mrs. Ott awoke and a confederate seized her, “I went to her and I asked her where was the money at . . . .” And petitioner said that when he thereafter shook Mrs. Ott two or three times, “All I was trying [to do] was to make her tell where the money was, but she wouldn’t tell. So, you know, we left. ... We just got the T.V. and stuff and left.”

Finding no evidence of intent to cause death—because there is none—in petitioner’s statements describing the burglary of Mrs. Ott’s apartment, the majority stress that petitioner admitted to the police that when he participated in the assault on Mrs. Ott he knew Mrs. Curtis had died a few days earlier from the effects of a similar assault in which he had also participated. But there is less here than meets the eye. It is true the jury could reasonably infer from petitioner’s knowledge of the death of Mrs. Curtis that in participating in a similar assault on Mrs. Ott, a person of similar age, he acted recklessly or with gross negligence. But the statute (Pen. Code, former § 190.2, subd. (c)) required more: it required proof beyond a reasonable doubt that petitioner specifically intended to kill Mrs. Ott, which is a wholly distinct state of mind. There is a continuum from innocent inattention to culpable negligence to wanton recklessness; but the difference between the *641 latter and actual intent to kill is not simply a further difference in degree—it is a difference in kind. In plain words, it is the difference between doing an act without caring what its consequences may be, and consciously intending that the act kill a specific human being.

Thus the mere fact that petitioner knew Mrs. Curtis had recently died in a similar assault was not sufficient to support an inference that he specifically intended to kill Mrs. Ott; something more was needed. For example, if the prosecution had proved that petitioner intentionally killed Mrs. Curtis as part of a modus operandi to prevent his burglary victims from identifying him, then the fact that a few days later a similar victim died at his hands in a similar fashion would have supported an inference that the latter killing was also intentional. But the prosecution had essentially no evidence at all that petitioner intended to cause Mrs. Curtis’s death—even Mikles and McFarland were no help on the point. This fact is confirmed by two salient features of the prosecution’s case. First, the prosecutor proceeded only on a felony-murder theory as to Mrs. Curtis, expressly advising the jury not to concern itself with whether he had proved premeditation and deliberation. Second, as to the Curtis murder the prosecutor made no effort to allege and prove a special circumstance and seek the death penalty; at the time, of course, all special circumstances required proof of intent to cause the victim’s death.

The majority next rely on three noninformant witnesses who testified to certain remarks that petitioner assertedly made in their presence shortly after the crimes. Whether taken singly or collectively, however, these remarks also fail to furnish—as the majority claim—“strong” evidence of the missing element of the prosecution’s case, i.e., that petitioner aided in the commission of the fatal acts with intent to cause Mrs. Ott’s death.

The majority first stress a remark attributed to petitioner by llena Gaines, an upstairs neighbor of Mrs. Curtis. Gaines testified that on August 29,1977, while Mrs. Curtis was being removed from her apartment on a stretcher, petitioner came up the street, stopped on the sidewalk, smiled or laughed, and said that “he was the one who did that.” But even if Gaines’s testimony were trustworthy, which it is not, 23 the remark she attributed to petitioner is weak and inconclusive evidence on the issue before us. First, even if *642 petitioner had been referring to the death of Mrs. Ott rather than Mrs. Curtis, an admission that he “did that” would have established, if believed, only that he was a cause in fact of her death; it would not have established that he intended to cause the death. The same remark, obviously, could also be made by one who caused the death by accident, by negligence, or by recklessness —or by any unintended killing in the course of a burglary. Second, in assertedly making the quoted remark petitioner was in any event not referring to the death of Mrs. Ott—who was still alive—but to the death of Mrs. Curtis, a death that even the prosecution did not contend that petitioner intended to cause.

The majority next assert that “Another acquaintance testified” to certain similar remarks by petitioner. (Maj. opn., ante, p. 599.) The majority’s assertion, however, overstates the record and misleads the reader. The “acquaintance” to whom the majority refer was Debria Lewis, and she “testified” at the trial in only a very limited sense of that word. Lewis was one of three prosecution witnesses in this relatively brief trial who did not in fact appear in person, but whose testimony at the preliminary hearing was read into the trial record by the prosecutor. The jury was thus deprived of the opportunity to observe the demeanor of these witnesses as they testified, and they were not subjected to the searching cross-examination typical of a murder trial; instead, they “testified” through the mouth of the prosecutor, and their story was tested only by the minimal, pro forma cross-examination typical of a preliminary hearing, before no jury at all. Moreover, as to two of these witnesses—Larry Rushing and Debria Lewis—the specific reasons why they were excused further undermined confidence in their credibility. 24 Over a defense objection, the preliminary hearing testimony of Rushing and Lewis was admitted on the ground they were “unavailable” to appear in person. (Evid. Code, § 1291.) In ruling that the prosecution had used reasonable diligence in trying to find them, the court stressed that Rushing “is a fugitive from justice” and Lewis “is apparently a prostitute, at least part-time prostitute, who goes by many aliases and would be extremely difficult to locate if, in fact, she didn’t want to be located.”

The majority nevertheless rely heavily on the preliminary hearing testimony of the witness Lewis. On September 11, 1977, Lewis had a visit from petitioner and his cousin Debra Hall, who was also Lewis’s neighbor. Lewis testified that as petitioner was looking at a copy of the evening newspaper in her apartment he saw an article about the death of Mrs. Curtis and said, “ ‘This is what I done.’ ” According to Lewis, when she told him he should not have done it he replied, “if she [Mrs. Curtis] had just been still—had been still and given him the money, that she would have been walking *643 around today.” Contrary to the majority’s position, however, the quoted remarks are weak evidence on the issue before us for the same reasons as the above discussed remark attributed to petitioner by the witness Gaines. Here, too, petitioner’s remarks to Lewis might establish that he was a cause in fact of death or even that he was guilty of felony murder, but they do not establish that he intended to cause the death. And in any event, here also the remark referred to the wrong victim—not to Mrs. Ott but to Mrs. Curtis, whose death the prosecution did not contend that petitioner intended to cause.

Next the majority assert that Lewis “testified . . . that [petitioner] had further described the two victims as ‘two old bags [who] were a nuisance and . . . got what they deserved.' ” (Maj. opn., ante, p. 599, italics added by majority.) The majority make much of this remark: they quote it no less than three times in the course of their opinion, and twice italicize its last four words. Unfortunately, in so doing they take their stand on very shaky ground: when the record is closely examined, its seeming support for the quoted remark dissolves into quicksand.

To begin with, it is highly misleading to say, as do the majority, that Lewis “testified” that petitioner “described” the two victims in the manner quoted. In fact, the entire thrust of Lewis’s testimony was to the contrary. When the prosecutor first raised the question on direct examination he specifically asked her, “At any time during that conversation that you had with Mr. Jackson, did you say anything about two old bags?” The prosecutor demanded a “Yes or no” answer, and Lewis squarely answered, “No.” Apparently seeking to impeach his own witness by a prior inconsistent statement, the prosecutor next established that two days later, on September 13, 1977, the police questioned Lewis on the subject of petitioner’s visit. Although she testified that she “didn’t really remember” what she told the police, the prosecutor again asked, “During that conversation that you had with the police officer or officers two days later after the conversation with Jackson, did you tell a police officer during the conversation you had with them that Jackson told you, that is, he, Jackson, used the terms ‘them,’ ‘her,’ and ‘two old bags’?” Again Lewis squarely answered, “No.” Literally unwilling to take no for an answer, the prosecutor asked still again, “At any time during that conversation that you had with the police officers, did you tell them that Jackson stated that the ‘two old bags were a nuisance and they got what they deserved’?” This time the prosecutor got the answer he wanted—the witness replied, “Yes”—and he dropped the matter in favor of other topics.

At that point the record stood as follows: (1) Lewis had directly testified that petitioner had not made any remark to her describing the victims as “two *644 old bags”; (2) when the prosecutor sought to impeach her by asking whether she thereafter told the police that petitioner had made such a remark, she testified that she had not done so; but (3) when the prosecutor persisted and asked essentially the same question again, she testified to the contrary. The record thus cried out for an examination by defense counsel to rehabilitate the witness, or at the very least to clarify the ambiguity created by her apparently inconsistent answers. But this was the preliminary hearing, not the trial, and the deputy public defender assigned to represent petitioner in that proceeding simply said, “No questions.” 25

This factual ambiguity was not resolved in the testimony at the preliminary hearing or at trial, and was not addressed by the referee, yet the majority now improperly resolve it in this court. First, they resolve in favor of the prosecution Lewis’s two contrary answers to the prosecutor’s impeaching question, impliedly finding the first to be false and the second to be true; next, they take the answer they deem to be true not only as impeaching Lewis’s initial testimony directly to the contrary but also as establishing the truth of the matter asserted (see Evid. Code, § 1235). The majority thus impliedly find that Lewis not only admitted she told the police on September 13, 1977, that petitioner made the “two old bags” remark to her on September 11,1977, but also that she “testified” that petitioner did in fact make that remark on September 11 even though she had directly testified he did not.

In any event, the majority achieve very little by this effort to rewrite the record. Even if there had been direct, unimpeached, and unambiguous testimony at the trial that petitioner in fact made the quoted remark, it would have remained weak evidence on the crucial issue of whether he aided in the commission of the fatal acts with intent to cause Mrs. Ott’s death. Standing alone, the quoted remark does not even establish that the speaker was a cause in fact of death: at some point in our lives we have all said of the death or punishment of another—even though we did not personally inflict it—that “he got what he deserved.” And even when taken together with an admission that the speaker “did that,” the remark may well be callous but it does not establish that the speaker intended to cause the death referred to: again, the same remark could be made by one who caused the death by accident, by negligence, by recklessness—or by any unintended killing in the course of a burglary. 26 *645 The third witness cited by the majority was petitioner’s cousin Debra Hall, who was present during petitioner’s above mentioned visit to Lewis’s apartment on September 11, 1977. Hall’s testimony, however, was so full of self-contradictions, denials, and claimed memory lapses that the prosecutor was allowed to treat her as a hostile witness and ask her leading questions “for the purpose of impeachment.” In so doing, he asked her whether it was true that petitioner pointed to a newspaper article about the deaths of Mrs. Curtis and Mrs. Ott and said, “This is what I did” and “It was because I needed some money.” Hall replied in the affirmative. The majority rely on this testimony, but it suffers from the same weakness as the essentially identical testimony of Lewis. 27 It, too, might establish that petitioner was a cause in fact of death or even that he was guilty of felony murder, but it does not establish that he intended to cause the death of Mrs. Ott.

The evidence relied by the majority pales into insignificance, moreover,, when it is compared with the truly devastating testimony of Mikles and McFarland. Because the majority quote Mi kies’s testimony in extenso (maj. opn., ante, p. 588, fn. 4) I need not repeat it verbatim, but certain passages deserve to be highlighted. Thus Mikles testified that in February 1978, while in the holding tank awaiting transportation from court back to county jail, he struck up a conversation with petitioner about the murder charges then pending against the latter. When Mikles asked how he had turned “a burglary into a murder,” petitioner assertedly gave him a detailed description of what he actually did to Mrs. Ott during the burglary of her apartment. According to Mikles, petitioner admitted that when Mrs. Ott first emerged from her bedroom during the burglary “he went over and he hit her a couple of times, and I guess she was backing up into the bedroom, and he kept firing on her until she—he knocked her out on the bed.” Again according to Mikles, petitioner admitted that after further searching the apartment he beat Mrs. Ott into unconsciousness a second time: “he is in the bedroom, and the lady is making noise. She is waking up. She is screaming or something, so he beats on her a couple of more times, you know. How many times I really couldn’t say, till . . . he knocked her out again. She fell down on the bed.” Finally, Mikles testified that when he asked petitioner why he had “pounced” *646 so much on an elderly woman, petitioner “said that when she woke up, you know, he just went off. It just pissed him off so bad ... he beat on her to the point, when he knocked her out again, he told me that he was so hot at her that there was a bottle on the stand next to the bed and he took the bottle and he fucked her in her pussy with it, is his exact words he told me.” 28

It is preposterous to hold, as the majority do, that a federal due process violation that led the jury to give credence to the foregoing testimony of Mikles and McFarland was harmless under the high standard of prejudice governing such violations. If the jury had disbelieved the testimony of Mikles and McFarland in its entirety—as the referee found and the majority assume that it would have done but for the violation—the only remaining evidence directly bearing on the question whether petitioner intended to cause Mrs. Ott’s death would have been his statement to the police that he struck her once and shook her two or three times in an effort to learn where the money was. Although such testimony could have been persuasive evidence of an intent to steal, a properly informed and instructed jury would have had great difficulty in finding that it also proved an intent to kill. The due process violation, however, made that task substantially easier: the jury was told not just that petitioner struck Mrs. Ott one blow, but that he rained blows on her—“he kept firing on her”—until she fell unconscious; and the jury was told not just that he shook her two or three times, but that when she came to her senses it so infuriated him—“It just pissed him off so bad”—that he beat her into unconsciousness a second time and sexually assaulted her with a wine bottle. Given Mrs. Ott’s advanced age—she was 90 at the time of the crimes—a jury that had been led to believe this devastating testimony would have had no difficulty in finding in it an intent to kill sufficient to support the charged special circumstances. To say, as do the majority, that on this record there is no “reasonable likelihood” that the due process violation “could have affected” the verdict is to shamelessly distort the plain meaning of those words.

It is also a vain attempt to rewrite history. The experienced prosecutor who tried this case—Deputy District Attorney Paul Marin—knew the importance of the testimony of Mikles and McFarland, and he made quite sure the *647 jury knew it as well. Thus in his argument to the jury he focused on the witnesses who testified concerning petitioner’s damaging statements, and warned that “I am sure defense counsel is going to argue, “Oh, you can’t believe them because they have some interest. They are biased or are prejudiced and they just want to point the finger at and accuse Jackson and so forth and you can’t believe them.” Seeking to forestall that argument, Marin then told the jury that Mikles and McFarland, in particular, “are independent one of the other, seeing and talking to the defendant at different times and different places and the defendant telling both of them in substance pretty much the same thing, neither one of them apparently knowing anything at all about these two burglaries mentioned of Mrs. Curtis and Mrs. Ott.” So important was their testimony to the prosecution that Marin told the jury that if the only witnesses he had were Mikles, McFarland, and llena Gaines, 29 ‘that would be sufficient, I would submit to you, to prove to you beyond a reasonable doubt that the Defendant Jackson committed that killing and that burglary of Mrs. Curtis’ apartment.”

Turning to the Ott murder, prosecutor Marin again relied on the testimony of Mikles and McFarland, and argued to the jury that it was corroborated by the fact that although each told the same story, “There is no connection between Mikles and McFarland. They obviously, apparently had no conversation among themselves or any meetings whatsoever. They were in different locations at different times, but coincidentally both of them were at different times in the same location as the defendant and the defendant made the same—substantially the same confession to Mr. McFarland as he made to Mr. Mikles

Finally, prosecutor Marin once more sought to preempt any defense effort to impeach Mikles and McFarland on the ground of interest or bias: referring to defense counsel, Marin told the jurors, “Don’t let him fool you about trying to show some kind of bias or prejudice on the part of Mikles or McFarland and also trying to get you, I submit, to in some way dislike either Mikles or McFarland, so that you won’t pay too much attention to their testimony.” Marin acknowledged that Mikles and McFarland had prior felony convictions, but argued, “That doesn’t mean they have lied or haven’t told you the truth; as a matter of fact, the evidence indicates they have told you the truth.”

The record contains still other persuasive expressions of the importance of the testimony of Mikles and McFarland. Thus, as noted above, on January 19, 1979, Los Angeles Deputy Sheriff Shea wrote a letter to the Los Angeles *648 Probation Department in which she reported that Mikles had testified for the prosecution in petitioner’s trial and that “according to Marin, Mikles’ testimony was extremely significant . . . .” Next, at the hearing on April 5, 1979, on Mikles’s application for modification of his robbery sentence, one of the witnesses called by the People was William Collette, a Long Beach homicide detective who had investigated the Curtis and Ott crimes and had taken Mikles’s information about petitioner’s admissions; when asked to characterize the result of Mikles’s testimony against petitioner, Collette declared, “The testimony was crucial.” And at the reference hearing on the present petition for habeas corpus, petitioner’s trial counsel testified with regard to the testimony of Mikles and McFarland, “in my opinion that was the guts of the Jackson case, the testimony of those two witnesses.” More specifically, trial counsel testified that “the case itself was basically even, except for those two witnesses. . . . The only thing that put it into the area of special circumstances were those two witnesses, in my opinion.” 30

On this issue the prosecutor, the homicide detective, and trial counsel had one undeniable advantage over the members of this court: they were there and we were not. They were there, in the courtroom, when Mikles and McFarland told their story on the stand. They saw and heard the impact of that testimony and the prosecutor’s argument on the jurors. We, by contrast, can only peruse a cold, stale transcript more than 13 years after the trial, and guess at that impact. The majority’s guess is both implausible on its face and, as shown above, demonstrably refuted by the record.

It is also contrary to governing federal law. It will be instructive to note, first, what this case is not. “This is not a case in which the witness’ bias becomes irrelevant because the witness’ testimony is fully corroborated, nor is this a case in which the witness’ testimony has been thoroughly impeached and proof of his bias would be merely cumulative.” (U.S. v. Rivera Pedin, supra, 861 F.2d 1522, 1530 .) Thus the prosecution’s failure to correct false testimony in violation of Napue, supra, 360 U.S. 264 , and Giglio, supra, 405 U.S. 150 (hereafter referred to as a Napue-Giglio violation) has been held harmless when that testimony was, inter alia, “merely corroboratory.” (United States v. Ramos Algarin (1st Cir. 1978) 584 F.2d 562, 567, fn. 7 .) Here the testimony of Mikles and McFarland was not merely corroboratory—it added a whole new dimension to the prosecution’s case, furnishing highly specific, detailed, and damaging evidence that was neither stated nor fairly implied in the testimony of any other witness.

*649 A Napue-Giglio violation has also been held harmless when the perjurious witness was exposed to “massive impeachment” on other grounds. (United States v. Antone (5th Cir. 1979) 603 F.2d 566, 570 .) 31 Here the sole impeachment evidence heard by the jury was the fact that Mikles and McFarland had prior felony convictions. But such convictions are almost by definition part of the personal history of a jailhouse informant, so that additional evidence of a more specific ground of impeachment—e.g., that the witness lied to the jury in denying inducements to testify—is not merely cumulative. Thus in United States v. Sanfilippo, supra, 564 F.2d 176, 178 , the court rejected the government’s argument that such inducements to testify would be cumulative to the witness’s prior convictions: “The fact that the history of a witness shows that he might be dishonest does not render cumulative evidence that the prosecution promised immunity for testimony. A jury may very well give great weight to a precise reason to doubt credibility when the witness has been shown to be the kind of person who might perjure himself.” (Accord, Brown v. Wainwright, supra, 785 F.2d 1457, 1466 [“the truth would have introduced a new source of potential bias”].) 32

Nor does this case fall within a third class of decisions that have held a Napue-Giglio violation harmless because the untainted evidence on the contested issue—usually the defendant’s guilt—was “overwhelming.” (United States v. Ramos Algarin, supra, 584 F.2d 562, 565 ; accord, Breest v. Perrin (1st Cir. 1980) 624 F.2d 1112, 1116 [untainted evidence “point[s] strongly toward guilt”]; United States v. Anderson (5th Cir. 1978) 574 F.2d 1347, 1356 [“strong, independent evidence of guilt”].) Here, far from being overwhelming, the untainted evidence on the contested issue of intent to cause death, as explained in detail above, was weak and inconclusive.

I turn now to federal appellate decisions holding a Napue-Giglio violation to be prejudicial. Although on this question each case turns to some extent on its facts, a number of federal decisions nevertheless find prejudice despite *650 the presence of untainted evidence similar to—if not stronger than—the evidence relied on by the majority in the case at bar. Thus in United States v. Barham, supra, 595 F.2d 231 , the defendant was convicted of counterfeiting on the testimony of six principal government witnesses. Defendant and his witnesses told a contrary story, but the jury evidently did not find it credible. During trial, however, three of the government witnesses (Joey Shaver and Diane and Jerry Beech) gave false and misleading testimony denying that anyone had promised them any benefits for testifying against the defendant; in fact, the prosecutor had promised them that if they gave “truthful testimony” against the defendant, Shaver and Jerry Beech would be prosecuted on only one count and their cooperation would be made known to the sentencing judge, and Diane Beech would not be prosecuted at all. The prosecutor not only failed to correct the false testimony of these witnesses, he wittingly or unwittingly reiterated and reinforced it by his choice of questions.

Stressing that witness credibility was important because the two sides told contrary stories, the federal appellate court declared that the defendant “was entitled to a jury that, before deciding which story to credit, was truthfully apprised of any possible interest of any Government witness in testifying falsely.” (United States v. Barham, supra, 595 F.2d at p. 243 , italics in original.) The court then observed that knowledge of the government’s promises to the three witnesses in question would have given the jury both “a concrete reason to believe” that the witnesses might have fabricated their testimony to collect on those promises, and “another, more persuasive reason to doubt their testimony—the very fact that they had attempted to give the jury a false impression concerning promises from the Government.” (Ibid.) The court explained (id. at p. 242) that the “reasonable likelihood” standard of proof for judging a Napue-Giglio violation “is a brother, if not a twin, of the standard (‘harmless beyond a reasonable doubt’) for determining whether constitutional error can be held harmless,” citing Chapman v. California, supra, 386 U.S. 18, 24 . Accordingly, the court reversed the conviction because “we cannot conclude that the jury, had it been given a specific reason to discredit the testimony of these key Government witnesses, would still have found that the Government’s case and [the defendant’s] guilt had been established beyond a reasonable doubt.” ( 595 F.2d at p. 243 .)

The court so held, however, in face of the fact that the testimony of the three untainted government witnesses (Sandra Simon, Charles Fowler, and Willie Vereen) furnished substantial independent evidence of the defendant’s guilt. All three were confederates of the defendant in his counterfeiting enterprise: at the defendant’s request, Fowler printed the bogus bills, Simon colored and dried them, and Vereen distributed them. (United States v. *651 Barham, supra, 595 F.2d at pp. 234-235.) Despite their unobjectionable testimony plainly implicating the defendant, the court did not hold the Napue-Giglio violation harmless. Rather, the court reasoned, “There is no doubt that the evidence in this case was sufficient to support a verdict of guilty. But the fact that we would sustain a conviction untainted by the false evidence is not the question. After all, we are not the body which, under the Constitution, is given the responsibility of deciding guilt or innocence. The jury is that body, and, again under the Constitution, the defendant is entitled to a jury that is not laboring under a Government-sanctioned false impression of material evidence when it decides the question of guilt or innocence with all its ramifications.” (Id. at p. 242.)

The foregoing reasoning is not made less applicable to the case at bar simply because petitioner here did not tell his story on the stand or call witnesses on his behalf: a criminal defendant, of course, cannot be penalized for remaining silent. Nor is a different result compelled by the fact that petitioner admitted to the police that he did strike Mrs. Ott once. Federal appellate decisions have held a Napue-Giglio violation to be prejudicial even in the face of a full confession by the defendant to all the elements of the crime charged—far more damaging than this petitioner’s single admission. Thus in United States v. Iverson, supra, 637 F.2d 799 , the defendant was convicted of aiding and abetting one Susan Johnson in cashing a stolen and forged government check. There was no question that Johnson stole and forged the check, and the teller who cashed it at the defendant’s request established the defendant’s role in facilitating the cashing. The sole issue was whether the defendant had the requisite criminal intent, or whether, as she claimed, she did not know the check was stolen. Johnson testified that the defendant had full knowledge of their criminal scheme. But on cross-examination for bias Johnson also testified that she had already been sentenced for her part in the affair, thus leading the jury to believe she had nothing to gain by testifying against the defendant. The latter testimony was false, but the prosecutor, who knew or should have known the truth, did not correct it.

In reversing the judgment, the federal appellate court held that the Napue-Giglio violation was not rendered harmless by the fact that the evidence also included a full confession by the defendant: the court reasoned that the confession could support the conviction only if corroborated by independent evidence, and “The teller’s testimony did not provide the necessary corroboration, for it did not bear on [the defendant’s] intent” (United States v. Iverson, supra, 637 F.2d at p. 805 ). That left only the tainted testimony of Johnson to prove the defendant’s intent. Applying the governing federal standard of prejudice, the court concluded there was “a reasonable likelihood that the testimony the witness gave could have affected the judgment of the *652 jury. Since that testimony was false in a critical respect, the conviction cannot stand.” (Id. at p. 806; accord, Campbell v. Reed, supra, 594 F.2d 4 .) 33

Indeed, even if the untainted testimony in this case could be viewed, as the majority claim, as circumstantial evidence from which the jury might have found tiie necessary intent to cause Mrs. Ott’s death, it remains true that the dramatic testimony of Mikles and McFarland made the jury’s task substantially easier. In similar circumstances federal appellate decisions have consistently held a Napue-Giglio violation to be prejudicial. (United States v. Sutton, supra, 542 F.2d 1239, 1242-1243 ; Boone v. Paderick, supra, 541 F.2d 447, 453 ; United States ex rel. Wilson v. Warden Cannon, etc., supra, 538 F.2d 1272, 1277-1278 .)

Finally, it will be remembered that in his argument to the jury the prosecutor repeatedly asserted that the testimony of Mikles and McFarland had not been impeached for bias or interest, and urged the jury to rely on it. By so arguing, however, the prosecutor not only vouched for the importance of that testimony but compounded the due process violation that occurred when he failed to correct the testimony as it was being given. The same compounding took place in Giglio, supra, 405 U.S. at page 152 [31 L.Ed.2d at pages 107-108], and has been stressed in numerous subsequent federal decisions as a factor pointing to prejudice. (Boone v. Paderick, supra, 541 F.2d 447,450 ; United States ex rel. Wilson v. Warden Cannon, etc., supra, 538 F.2d 1272, 1277 ; United States v. Pope, supra, 529 F.2d 112, 114 .) As the court held in United States v. Sanfilippo, supra, 564 F.2d 176, 179 , “Thus the Government not only permitted false testimony of one of its witnesses to go to the jury, but argued it as a relevant matter for the jury to consider. Whether either instance alone would merit reversal, we need not decide, for together they do.”

For all the foregoing reasons I conclude that the federal due process violation shown on this record was prejudicial as to the special circumstance findings, and hence that the petition for habeas corpus should be granted and the judgment vacated insofar as it finds the special circumstance allegations to be true.

*653 2. The Penalty Verdict

If the judgment is vacated as to the special circumstance findings, of course, it will ipso facto be vacated with respect to penalty. Nevertheless, because the majority reject petitioner’s due process claim as to the special circumstance findings and proceed to discuss—and also reject—his contention that the due process violation was separately prejudicial with respect to the penalty decision, I must address that issue as well. As will appear, I would hold the due process violation no less prejudicial as to penalty.

It is undisputed that Mi kies and McFarland were the only witnesses who testified that petitioner admitted sexually assaulting Mrs. Ott with a wine bottle. The majority first claim there is no reasonable likelihood that such testimony could have affected the penalty verdict because, in the majority’s opinion, “it is far more likely” that petitioner’s participation in the two murders within a ten-day period and his lack of remorse after the crimes “played the key role” in the jury’s penalty decision. (Maj. opn., ante, p. 600.)

The truth is that no one—not even the majority—can know what evidence “played the key role” in this jury’s penalty decision. In fact, no one can even know if there was any piece or pieces of evidence that played such a role. It is equally possible that the jury obeyed the court’s instructions to consider, in fixing the penalty, “all of the evidence” admitted during the trial and “all the applicable factors” in aggravation and mitigation, and that the penalty verdict was the product of the cumulative effect of all such evidence and factors. Indeed, it was not even required that all such evidence and factors affect all the jurors equally: the law permits some jurors to give greater weight to some evidence and factors and other jurors to give greater weight to other evidence and factors—they need agree only on the bottom line, the penalty itself.

Fortunately, the standard of prejudice that governs the federal due process violation shown in this case does not ask a reviewing court to do the impossible—to read the minds of the jurors. As explained above, our obligation under the relevant United States Supreme Court decisions is far less speculative: we must set aside the judgment as to penalty if, on the whole record, we find “any” reasonable likelihood that the violation “could have affected” the penalty verdict; or, in the alternate wording of the same standard, we must set aside the penalty unless the beneficiary of the violation “proves beyond a reasonable doubt” that the violation did not “contribute” to the verdict. With that standard in mind I turn to the facts of this case.

The most striking aspect of the record in this regard is the sustained effort of prosecutor Marin to put before the jury the evidence that petitioner *654 sexually assaulted Mrs. Ott with a wine bottle, and to make sure the jury appreciated the importance of that evidence. Thus during his case-in-chief the prosecutor called as a witness Gary Lane, a police officer who had responded to the report of the Ott murder. Lane described in detail the appearance of the victim’s battered corpse and ransacked apartment, and identified numerous photographs of both. On redirect examination, the prosecutor first attempted to raise the matter of the sexual assault. He asked the witness, “Did you see any bottle anywhere in that particular bedroom upon your initial entry?” Lane said that he saw a bottle standing on the floor next to Mrs. Ott’s bed. The prosecutor then asked whether the witness saw any “material” on the bottle, and Lane said that he did. When the prosecutor asked what kind of “material” it was, defense counsel objected on the ground that the prejudicial effect of the evidence would substantially outweigh any probative value it might have. (Evid. Code, § 352.) In his ensuing offer of proof the prosecutor said he expected that Lane would testify he saw a small hair on the bottle, and that other witnesses would testify it appeared similar to the victim’s pubic hair; the prosecutor also announced that he intended to ask Lane whether he saw any injury to the victim’s pubic area. The prosecutor argued that such evidence was relevant to show “the manner and means” by which the victim was killed, and added, “It also, of course, would have relevancy upon the penalty phase,” because in that phase the jury must consider all the “circumstances of the crime” shown in the guilt phase.

The trial court ruled against the prosecutor. Finding that the proposed evidence lacked probative value, the court pointed out that any injury to the victim’s pubic area was “not the cause of death” and “does not establish the burglary.” By contrast, the court declared that such evidence would be “highly inflammatory,” and hence concluded that its prejudicial impact outweighed its probative value. The court left open the possibility of admitting the evidence in the penalty phase, but ruled that “because it is extremely prejudicial” it would be excluded in the guilt phase under Evidence Code section 352.

The prosecutor nevertheless raised the matter again in the guilt phase, while questioning the physician who performed the autopsy on Mrs. Ott. Although the witness described extensive injuries consistent with the victim’s having received at least eight or ten blows with a blunt object, he was of the opinion that death was actually caused by strangulation. The witness also described a laceration in the victim’s vagina and volunteered the opinion that it was caused by a foreign object such as “a piece of wood.” The prosecutor asked, however, whether the injury could have been “caused by a bottle,” and the witness replied, “it could have been.”

Later in the guilt phase, as we have seen, the prosecutor once more brought the same matter before the jury when he put on the testimony of *655 Mildes and McFarland in which those witnesses recounted in graphic detail how petitioner admitted that he sexually assaulted Mrs. Ott with a wine bottle.

Finally, the prosecutor not only returned to the same topic in the penalty phase—the penalty phase was about nothing else. First the prosecutor called Sergeant Daniel Sallmen, an investigating officer who had examined Mrs. Ott’s apartment on the night of the crimes. Sallmen testified that on arriving at the scene he learned that Toni McDowell, Mrs. Ott’s daughter, had earlier found and touched a bottle in her mother’s bedroom, and when he entered the bedroom he saw the same bottle standing on the floor next to the bed. The prosecutor asked whether he made an examination of the bottle, and Sallmen said he had, observing two hairs and a reddish-brown dried substance on the neck and cap of the bottle. When asked his opinion of the nature of the dried substance, the witness identified it as dried blood. Lastly, the prosecutor asked whether he had compared the two hairs on the bottle with the victim’s pubic hair, and the witness said they were similar.

The testimony of Toni McDowell, the only other witness at the penalty phase, also focused exclusively on the matter of the bottle. McDowell was the first person to arrive on the scene of the crimes. The prosecutor asked her how soon she saw the bottle after entering her mother’s bedroom, and she replied, “it was a little while, because I stood there and screamed for a minute . . . .” The witness went on to explain that as she approached the bed to check her mother’s pulse, her foot struck the bottle lying on the floor and she reached down and set it upright. When asked whether she had previously seen that bottle in her mother’s apartment, McDowell replied, “Yes. She always kept it on hand for her friends,” and identified its contents as muscatel.

Prosecutor Marin thus exploited every opportunity to put before the jury his evidence that petitioner sexually assaulted Mrs. Ott with a wine bottle. Indeed, testimony on this topic was the last testimony the jury heard before beginning its penalty deliberations. 34 Next, as he did in the guilt phase, Marin made sure the jury gave great weight to this evidence by stressing it in his argument urging the death penalty.

Prosecutor Marin began his argument by reviewing the various aggravating and mitigating factors permitted by law—from the circumstances of the *656 crimes to petitioner’s character, background, history, and mental and physical condition. (Pen. Code, former § 190.3.) As to each of these, Marin argued that on the evidence before the jury it was a factor in aggravation rather than mitigation. Such an argument, of course, is perfectly proper. But at the end of his argument Marin turned to the evidence of the assault with the wine bottle—which he had not previously mentioned in his remarks—and made the most of it. Thus he urged that petitioner deserved to die “because his acts were so savagely cruel—they really were. As you know, they were. When one even thinks a little more deeply into the facts of this case, one thing that I have left out up to this point—I have considered with you the beatings and the other circumstances of the crimes that this defendant committed and his bragging to other people and laughing about his victims dying and being beaten, but I have left out one very cruel, savage, inhumane act that the defendant took another step in inflicting upon his victim, Mrs. Ott, and you know what that was.

“You can imagine, after beating to death that helpless old lady, he took a wine bottle and shoved it up the private parts of his victim, Mrs. Ott, and bragged about it. What do you think about that as far as this defendant’s character, his background, his history, his mental and his physical condition to do that to his victim? Does justice cry out for the death penalty in this case? Is this the kind of man that should get the death penalty? I submit to you that it is.” (Italics added.)

This was the final point that prosecutor Marin made to the jurors, the last reason he gave them for returning a verdict of death. From the frequency and timing of Marin’s presentation of this evidence and his dramatic use of it in his argument to the jury, it is clear this experienced prosecutor believed the evidence would carry great weight in the penalty deliberation. And even an elementary knowledge of human nature tells us the prosecutor was correct: the evidence in question was so shocking that no reasonable juror could fail to be moved by it. In these circumstances, the majority’s claim that it is “far more likely” that some other evidence “played the key role” in the penalty decision cannot be taken seriously: it is at best wishful thinking, and at worst a deliberate distortion of the record.

Equally untenable is the majority’s second ground for finding the due process violation to be harmless as to penalty. The majority argue in effect that even if the evidence of the wine bottle incident did contribute to the penalty verdict, the testimony of Mikles and McFarland did not contribute to that evidence. To state the argument is to refute it. The majority concede that in his statements to the police petitioner did not admit he sexually assaulted Mrs. Ott with the bottle; yet the majority nevertheless claim that the jury *657 “would have drawn” that same conclusion even without the testimony of Mildes and McFarland, on the basis of two other items of evidence. (Maj. opn., ante, p. 600.) Again the record is otherwise.

The first item of evidence relied on by the majority is the fact that when the police informed petitioner they had found his fingerprints in the Ott apartment, he “volunteered the information” that he had touched a wine bottle on a table next to Mrs. Ott’s bed. (Maj. opn., ante, p. 600.) Seen in its context, however, petitioner’s remark is not as inculpatory as the majority would have us believe.

The transcript of petitioner’s two successive statements to the police on September 17, 1977, reveals that near the beginning of the first interrogation the police asked petitioner, “Why would your fingerprints be in the ladies’ house?” Petitioner first replied that the door to the Ott apartment was open and he just walked in; that he saw a bottle and touched it; that when he realized the condition of the apartment, he panicked and left. After the police expressed disbelief in that story, petitioner said he would tell the truth and began to recount how he and his companions burglarized both apartments. As he was describing places he searched for money in the Ott apartment, the police asked, “What items inside her house did you handle?” Petitioner answered that he handled the lamp, probably a few doorknobs, and the aforementioned bottle. After further discussion of the Ott burglary the police asked, “Did you ever touch a little metal box?” Petitioner replied, “Was it in the bedroom? I think all of us touched that box.”

The interrogation then shifted to the Curtis burglary and followed a similar line of inquiry. Petitioner said he entered the Curtis apartment through a side window; the police said his fingerprints were found on the rear window, and petitioner conceded he might have touched that window too. In describing his search of the Curtis apartment, petitioner volunteered the fact that “I did touch some boxes in the bedroom on the dresser.” Near the end of the interrogation the police asked, “Did you wipe anything off in the apartment so there wouldn’t be fingerprints?” Petitioner replied that he did wipe off “a few door knobs,” using a handkerchief he found in the apartment.

In a burglary-murder investigation it is not unusual for the police to ask, and a suspect to answer, questions seeking to identify objects on the premises that the suspect may have touched; the purpose, of course, is to provide leads for fingerprinting and to confirm fingerprints already found. The record shows that is what happened here; and in that context petitioner’s admission he touched the wine bottle would have appeared entirely unremarkable to the jury, as the bottle was simply one of a number of household objects that he admitted touching in the course of these burglaries.

*658 The second item relied on by the majority in this regard is another example of their overstatement of the record. The majority invoke “forensic evidence” assertedly indicating that the wine bottle had in fact been used to sexually assault Mrs. Ott. (Maj. opn., ante, p. 600.) The majority do not specify who furnished this “forensic evidence,” but only two witnesses come to mind. Upon examination, the testimony of each falls far short of the majority’s goal.

The first witness, the autopsy physician, need not detain us long. As noted above, he testified only—and in response to a suggestion by the prosecutor—that the injury to Mrs. Ott’s vagina “could have been” caused by a bottle. The majority claim the “forensic evidence” showed the injury had in fact been so caused, and by this very bottle.

The only other relevant witness was Sergeant Sallmen, the officer who testified that he observed two hairs and a dried substance on the neck and cap of the bottle, and gave his opinions that the substance was blood and the hairs were pubic hairs of the victim. It is a gross exaggeration, however, to call Sallmen’s opinions “forensic evidence.” After eliciting the foregoing description of the bottle from Sallmen, the prosecutor asked whether, based on his experience, he had formed an opinion as to the nature of the dried substance, and the witness replied that he had. When the prosecutor asked for that opinion, however, defense counsel objected on the ground that tiie witness was not qualified to express it. The prosecutor hastily explained, “We offer this as a lay opinion and not as an expert opinion,” and the witness said only that “It appeared dried blood to me.” When the prosecutor subsequently asked Sallmen whether he had compared the hairs on the bottle with the pubic hair of the victim, defense counsel voiced the same objection and the witness said only that “They appeared to be similar to me.”

The prosecutor’s concession that he was not offering Sergeant Sallmen’s testimony as expert opinion was correct, because the witness was a policeman, not a trained criminalist. But for the same reason, Sallmen’s thumbnail personal opinions of what the items looked like to him do not amount to “forensic evidence” in the proper meaning of that phrase. 35

*659 In sharp contrast to the foregoing indirect and inconclusive evidence, the testimony of Mikles and McFarland, if believed, directly and firmly established (1) that the injury to Mrs. Ott’s vagina was caused by someone forcibly inserting therein the wine bottle found in her apartment, and (2) petitioner was that someone. Each witness made the point in a few powerful words: Mikles testified that petitioner admitted “he took the bottle and he fucked her in her pussy with it,” and McFarland testified that petitioner admitted “he seen a wine bottle . . . and stuck it into her vagina.”

As it was in the case of the special circumstances findings, it is preposterous to hold that a federal due process violation that led the jury to give credence to the foregoing testimony of Mikles and McFarland was harmless under the high standard of prejudice governing such violations. We need not speculate whether the jury might have imposed the death penalty even without this testimony; what is certain is that the testimony of Mikles and McFarland made it substantially easier for the jury to find that petitioner committed the act of sexually assaulting this 90-year-old victim with a bottle, and to agree with prosecutor Marin that a man who committed such a “very cruel, savage, and inhumane act” in the course of a killing deserved to die. More accurate than the majority’s view of this evidence is the contemporaneous report of Los Angeles Deputy Sheriff Shea, who stated in her above quoted letter of January 19,1979, that “according to Marin,” Mildes’s testimony was extremely significant, “especially regarding the aggravation aspect which was the basis for the death penalty.” In these circumstances the People have utterly failed to prove beyond a reasonable doubt that this federal due process violation did not even contribute to the penalty verdict.

For all the foregoing reasons I conclude that the federal due process violation shown on this record was also prejudicial as to the jury’s penalty decision, and hence that the petition for habeas corpus should be granted and the judgment vacated insofar as it imposes the penalty of death.

II. Incompetence of Counsel in Failing to Investigate the Jailhouse Informants for Impeachment Evidence

As a separate and distinct ground for vacating the judgment insofar as it finds the special circumstance allegations to be true, the referee further *660 found that counsel’s performance was constitutionally deficient because he failed to investigate Mildes and McFarland before trial, and that because such an investigation would have disclosed substantial impeachment evidence, the deficiency prejudiced petitioner in defending against the special circumstance allegations. The majority decline to reach the first of these issues, and hold on the second that petitioner has not demonstrated the requisite degree of prejudice. I would reach each issue, and hold that the record amply supports both petitioner’s claim of ineffective assistance of counsel and his claim of prejudice.

A. Ineffective Assistance of Counsel

The first issue is surely not difficult. This is not the usual case in which counsel undertook some degree of investigation and the question is whether it was enough, or in which the shortcomings in counsel’s investigation were unintended and the question is whether they were excusable. Here petitioner’s trial counsel, Theodore P. Veganes, candidly admitted at the reference hearing that he undertook no pretrial investigation of Mikles and McFarland at all, and that he did so deliberately.

Veganes was appointed counsel for petitioner on December 8, 1977, and trial began one year later, on December 8, 1978. It was to be his first capital trial. Although he could not remember exactly when in the pretrial period he learned of the proposed testimony of Mikles and McFarland, Veganes testified at the reference hearing that “I was aware of it way before the case came to jury.” He further admitted that at no time during the pretrial period did he retain an investigator to inquire into the background of Mikles or McFarland and in particular to learn whether the prosecution had promised them any benefits for their testimony, nor did he personally investigate any of these matters himself. Indeed, Veganes conceded that “I didn’t know much about Mr. Mikles” when cross-examining him, and his only prior contact with Mikles was to speak with him in the courthouse “About five minutes before he testified . . . ,” 36

Veganes further testified at the reference hearing that his failure to investigate Mikles and McFarland was a deliberate tactical decision. Although *661 Veganes knew from other communications that prosecutor Marin intended to call Mikles and McFarland as witnesses, Marin inadvertently omitted their names from a handwritten list of 32 proposed witnesses that he gave Veganes before trial. 37 Veganes thereupon conceived of the “tactic” of deliberately keeping himself as ignorant as possible about Mikles and McFarland in the hope that when they were called to the stand he could persuade the court to bar their testimony in its entirety on the theory that he had “detrimentally relied” on Marin’s incomplete list in failing to investigate them.

As the referee correctly ruled in his original report, “For a tactical decision to constitute effective assistance of counsel, it must be a decision which a reasonably competent and experienced criminal defense attorney would make.” (See, e.g., In re Marquez, supra, 1 Cal.4th 584, 602 , and cases cited.) The referee found that Veganes did not make a reasonably competent decision in refraining from investigating Mikles and McFarland on the ground stated. Although this finding resolves a mixed question of fact and law and thus is subject to our independent review (id. at p. 603), the record compels us to reach the same conclusion as the referee.

To begin with, Veganes’s own testimony at the reference hearing undermines his claim that his decision not to investigate was reasonable. He admitted that even before petitioner’s trial he realized from his experience both as a prosecutor and as a defense attorney that Mikles and McFarland “were potentially very powerful witnesses for the People.” As noted above, he believed their testimony was “the guts of the Jackson case.” He was also aware that Mr. Marin had a reputation for being “an extremely thorough and powerful prosecutor”; he knew that Marin would not “leave out crucial evidence. ... I knew that he would try to use them,” i.e., Mikles and McFarland.

Not only did Veganes fail to investigate Mikles and McFarland either personally or by means of private investigators, he admitted that he did not *662 use available discovery procedures for the purpose. He testified that he knew from his experience as a prosecutor that if he had filed a formal discovery motion, or even a written “informal” discovery request under local court rules, “it would have put a duty on the prosecutor to gather up all information known to law enforcement concerning the background of Mildes and McFarland,” and that he “would have gotten the cooperation of Mr. Marin and his investigating officers and they would have provided [him] with the background information that they had.”

Veganes also acknowledged the value of such information. He agreed that “it would have hurt [Mikles’s] credibility” if he had been able to prove, through the testimony of Deputy Shea, that in exchange for testifying against petitioner Mildes expected to receive substantial sentencing benefits in the Long Beach case, in the Norwalk case, and in the federal parole matter.

Finally, Veganes conceded that his tactic was unprecedented. He knew of no appellate decision, in California or elsewhere, holding that a prosecutor’s inadvertent omission of a potential witness’s name from such a list constitutes grounds for barring the witness from testifying. Nor could Veganes recall any serious felony case that he had participated in as prosecutor or defense counsel in which a witness was barred from testifying “because the prosecutor simply failed to put his name on a witness list.”

Veganes’s testimony on these points was confirmed by two experienced criminal defense attorneys, Steven H. Hough and Charles Gessler, whom petitioner called as expert witnesses. 38 In the form of a hypothetical question, Hough was asked for his opinion of Veganes’s tactical decision to refrain from investigating Mikles and McFarland in the hope of excluding the testimony on the theory that he had “detrimentally relied” on the prosecutor’s inadvertent admission of their names from a witness list. Hough gave as his opinion that such a tactic would exhibit “very bad judgment.” The witness likened the tactic to invited error, stressed that to use it would be “taking a very big gamble,” and concluded that “it is a very, very bad tactic” for a number of reasons, especially when used in a capital case. Hough further testified that in his personal experience he could not recall any defense attorney using such a tactic, and he knew of no appellate decision in *663 any jurisdiction that has allowed a witness to be barred from testifying on this ground. 39

Petitioner’s second expert witness, Charles Gessler, testified to the same effect. In addition, Gessler gave two principal reasons for his opinion that Veganes’s tactic was unsound and did not constitute reasonably effective assistance of counsel. One ground was that evidence impeaching a witness, particularly a jailhouse informant, may be lost unless counsel preserves it by investigating the witness as soon as possible. The other ground of Gessler’s opinion was his disbelief that a trial judge would bar a witness from testifying merely because the prosecutor inadvertently omitted his name from an informal witness list. Rather, Gessler testified, “I think probably a judge would give a reasonable continuance to counsel, if he based his failure to investigate on that omission from the list, but not the sanction of preventing the witness from testifying.”

That is exactly what happened at the trial, but for petitioner it came too late. When prosecutor Marin began to tell the jury, in his opening statement 11 days after trial began, what the testimony of Mikles and McFarland would be, Veganes invoked his “tactic”: he objected to any reference to Mikles and McFarland because he had “detrimentally relied” on the fact that their names were not on the witness list that Marin had given him before trial. After a lengthy hearing in chambers, however, the court ruled that it would not bar the prosecutor from calling Mikles and McFarland but would grant the defense a short continuance to investigate them and would appoint an investigator for the purpose if Veganes desired it; at Veganes’s request, the court thereafter appointed such an investigator. Trial resumed a week after the court made its ruling, and Mikles took the stand on the following day. In the time available—the continuance included the Christmas holiday —the investigator apparently did not develop the impeachment evidence discussed in part I of this opinion, as Veganes made no mention of it in his ensuing cross-examination of Mikles and McFarland.

On this record there can be no doubt that we should adopt as our own the conclusion reached by the referee in his original report, viz., “In light of the crucial nature of the testimony of Mikles and McFarland on the special circumstance allegations and the obviously inadvertent nature of the prosecutor’s omission of their names on the witness list furnished Mr. Veganes, *664 and the failure of Mr. Veganes to seek formal discovery from the prosecutor, it is the Referee’s conclusion that no reasonably competent and experienced criminal defense lawyer would consider this failure to investigate a reasonable tactical decision.”

B. Prejudice

The majority hold that whether or not Veganes’s failure to investigate Mikles and McFarland denied petitioner his constitutional right to reasonably effective assistance of counsel, such denial did not prejudice petitioner in his defense against the special circumstance allegations. I cannot agree.

In light of what counsel failed to do in this case—to investigate Mikles and McFarland and thereby to discover impeachment evidence against them—there are, once again, two questions to be answered in determining prejudice. First, if counsel had provided reasonably effective assistance and hence had discovered—and presumably introduced—such impeachment evidence, would the jury have disbelieved the remainder of the testimony of Mikles and McFarland? Second, if the jury had disbelieved the remainder of their testimony, would its rejection of that testimony have affected its verdict?

Although the majority appear to assume that the answer to the first question is yes, a brief discussion is nevertheless in order. In contrast to the constitutional violation discussed in part I of this opinion—the prosecutor’s failure to correct false testimony of Mikles and McFarland denying any consideration for serving as witnesses against petitioner—the present violation demonstrates only that the witnesses had an incentive to lie; it does not also demonstrate that they did in fact lie, i.e., when they denied that incentive on the stand. 40 Nevertheless, the record shows two additional reasons why the jury would have disbelieved the remainder of the informants’ testimony if counsel had undertaken the missing investigation.

First, the record discloses that petitioner is Black and both his victims were White. The referee found that an investigation into Mikles’s background would have disclosed the fact—which Mikles admitted at the reference hearing—that he was a past member of the Aryan Brotherhood, a White supremacist prison organization. Both of petitioner’s expert witnesses stressed the effect such evidence would have on a jury in the circumstances *665 of this case: Hough expressed the opinion that if such a jury were told the informant was a former member of the Aryan Brotherhood, it would be “powerful impeachment material,” and Gessler was of the opinion that “This should be extremely effective impeachment to a jury,” i.e., when a past member of the Aryan Brotherhood “is now coming forward and testifying to a statement that was supposedly made to him by a Black defendant. . . .”

In addition to showing Mikles’s possible bias against petitioner on racial grounds, the fact that he was a past member of the Aryan Brotherhood could well have had other effects on the jury. As the majority correctly observe (maj. opn., ante, p. 603), “This information also might have raised possible doubts as to the veracity of Mikles’s general claim that [petitioner] had confided in him and the veracity of Mikles’s specific testimony that [petitioner] had made a racist comment concerning the victims in this case.” 41 Second, as noted above, the prosecutor repeatedly argued to the jury that Mikles and McFarland were “independent one of the other,” that there was “no connection” between them, and that they had “no conversation among themselves or any meetings whatsoever.” The referee found, on the contrary, that an investigation would have disclosed that Mikles and McFarland “were not acting completely independently of each other” in testifying against petitioner; rather, each had been a trusty in the same section of the jail before trial, and each was aware that the other intended to testify against petitioner. The referee further found the jury could have concluded from such information that Mikles and McFarland had the opportunity to tailor their testimony so as to remove any inconsistencies and thus improve their chances of obtaining the promised sentencing benefits. The record amply supports both those findings, and the majority impliedly agree.

I conclude, as the referee expressly found and the majority apparently assume, that if counsel had performed competently and investigated Mikles *666 and McFarland the impeachment evidence he would have discovered would have led the jury to disbelieve their testimony in toto.

I turn to the second question in the prejudice analysis. As noted above, the referee found that if the jury had heard the impeachment evidence against Mikles and McFarland it would have not only rejected their testimony in its entirety but would also have deemed the remaining prosecution evidence insufficient to support the special circumstance allegations. The majority reach a contrary conclusion on pure a fortiori reasoning. The majority’s major premise is, as they hold earlier in their opinion, that there is no “reasonable likelihood” the prosecutor’s failure to correct the informants’ false testimony “could have” affected the jury’s determinations as to special circumstances and penalty, because of the asserted “strength” of the noninformant evidence. Their minor premise is that to prevail on the claim now in issue—counsel’s failure to investigate the informants—petitioner must show a significantly greater likelihood of prejudice, i.e., a “reasonable probability” that the error “would have” affected the jury’s determinations. The majority conclude that because the noninformant evidence precluded petitioner from meeting the former—and lesser—standard of prejudice, a fortiori the same evidence precludes petitioner from meeting the latter—and greater— standard.

The majority’s argument fails, however, because its major premise is unsound. As demonstrated in part I of this opinion, on the precise question before us—whether petitioner specifically intended to cause Mrs. Ott’s death—the noninformant evidence is not strong but weak, and it is substantially outweighed by the truly devastating testimony of Mikles and McFarland. It follows that there is indeed a “reasonable likelihood” that the prosecutor’s failure to correct the informants’ false testimony “could have” affected the jury’s determination of the special circumstance allegations.

Thus the question whether counsel’s failure to investigate the informants was prejudicial cannot be resolved by a facile a fortiori argument. Rather, we must squarely apply to the facts of this case the standard of prejudice governing claims of constitutionally ineffective assistance of counsel. The United States Supreme Court declared that standard, of course, in the landmark case of Strickland v. Washington (1984) 466 U.S. 668, 694 [ 80 L.Ed.2d 674, 697-698 , 104 S.Ct. 2052 ] (hereafter Strickland): “The 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.” The majority pay lip service to this rule but ignore its real meaning. Throughout the Strickland opinion, the high court takes pains to explain what it means in this context by the phrase “reasonable probability” of prejudice: *667 as will appear, it means that a denial of effective assistance of counsel will be deemed prejudicial if it reasonably undermines “confidence” in the “reliability” of the judgment.

Thus the high court begins by explaining in Strickland that the Sixth Amendment guaranty of counsel exists to protect the fundamental right to a fair trial, which the court defines as “a trial whose result is reliable.” ( 466 U.S. at p. 687 [ 80 L.Ed.2d at p. 693 ].) The court discusses and rejects as inappropriate two other standards of prejudice. First, the mere possibility of prejudice: “It is not enough for the defendant to show that the errors had some conceivable effect on the outcome of the proceeding.” (Id. at p. 693 [ 80 L.Ed.2d at p. 697 ].) That standard is so minimal that it would be impractical as a tool for judging the wide range of acts or omissions of counsel that might be deemed ineffective assistance. Equally inappropriate as a test is “probability” of prejudice in the traditional preponderance-of-evidence sense: “a defendant need not show that counsel’s deficient conduct more likely than not altered the outcome in the case.” (Ibid.) Although that standard is familiar in other contexts and promotes the finality of criminal judgments, it is too high for present purposes: “An ineffective assistance claim asserts the absence of one of the crucial assurances that the result of the proceeding is reliable, so finality concerns are somewhat weaker and the appropriate standard of prejudice should be somewhat lower. The result of a proceeding can be rendered unreliable, and hence the proceeding itself unfair, even if the errors of counsel cannot be shown by a preponderance of the evidence to have determined the outcome.” (Id. at p. 694 [80 L.Ed.2d at pp. 697-698], italics added.)

Accordingly, the high court holds in Strickland that when the error is denial of effective assistance of counsel the standard of prejudice should be higher than a mere possibility of a different result but lower than a reasonable probability that a different result would have been more likely than not. The court reaches the desired level by giving the phrase “reasonable probability” a special meaning in this context, incorporating into it the concept of “confidence” in the “reliability” of the judgment. Thus the high court declares that “The benchmark for judging any claim of ineffectiveness must be whether counsel’s conduct so undermined the proper functioning of the adversarial process that the trial cannot be relied on as having produced a just result.” (Strickland, supra, 466 U.S. at p. 686 [80 L.Ed.2d at pp. 692-693], italics added.) In short, a “reasonable probability” within the meaning of Strickland is a probability sufficient to “undermine confidence in the outcome.” (I d. at p. 694 [80 L.Ed.2d at pp. 697-698].)

When the governing standard of prejudice is thus properly understood, its application to the facts of the case at bar becomes clear. Here, counsel’s *668 professional error in failing to investigate Mikles and McFarland led the jury to give credence to the highly incriminating testimony of those two witnesses. The question is the effect of that error on the jury’s finding on the special circumstance allegation that petitioner specifically intended to kill Mrs. Ott. For the reasons given in part I of this opinion, the testimony of Mikles and McFarland outweighs the noninformant evidence on this issue to the point that it is “reasonably likely” that the error “could have” affected that finding. But it does more. Viewed dispassionately, the testimony of Mikles and McFarland also outweighs the noninformant evidence to the point that it is “reasonably probable”—within the special meaning of Strick land.—that counsel’s error “would have” affected the finding. This does not mean it is more likely than not that counsel’s error actually changed the result, but Strickland does not require such a showing. It means only that on the narrow issue of intent to kill, counsel’s error injected into the jury’s deliberations evidence of such disproportionate impact that it made the finding at least “unreliable.” On this record, it would be absurd to hold that a denial of constitutionally effective assistance of counsel which led the jury to believe that petitioner mercilessly and repeatedly beat his 90-year-old victim into unconsciousness and sexually assaulted her with a bottle does not at least “undermine confidence in the outcome” of the jury’s deliberations on this crucial issue. (Strickland, supra, 466 U.S. at p. 694 [80 L.Ed.2d at pp. 697-698].)

For all the foregoing reasons I conclude that the denial of constitutionally effective assistance of counsel in failing to investigate the informants for impeachment evidence was prejudicial as to the special circumstance findings, and hence that on this ground as well the petition for habeas corpus should be granted and the judgment vacated insofar as it finds the special circumstance allegations to be true. 42

III. Incompetence of Counsel in Failing to Investigate Petitioner’s Background for Mitigating Evidence

As a separate and distinct ground for vacating the judgment insofar as it imposes the penalty of death, the referee further found that Veganes’s performance was constitutionally deficient because he failed to investigate petitioner’s background before trial, and that because such investigation would have disclosed substantial mitigating evidence, the deficiency prejudiced petitioner in presenting his case at the penalty phase for a lesser punishment than death.

*669 The relevant facts are not in dispute. Veganes undertook no pretrial investigation for the purpose of discovering mitigating evidence, and in the penalty phase of the trial he presented no evidence on petitioner’s behalf. At the reference hearing, however, petitioner called as witnesses five close family members who could have furnished substantial mitigating evidence. These witnesses testified in detail to appalling conditions of neglect and abuse that petitioner suffered throughout his childhood. The witnesses further stated they would have testified on petitioner’s behalf in the penalty phase if Veganes had but asked. And as the Attorney General concedes, a reasonable investigation would have produced all five of these witnesses: Veganes had spoken briefly with two of them before trial (petitioner’s father and aunt), and through them could readily have learned of the other three (the first aunt’s husband, a second aunt, and petitioner’s half brother).

The majority fully and fairly summarize the testimony of each of these witnesses. (Maj. opn., ante, pp. 606-609.) But the majority go hopelessly astray in their application of the rules of Strickland, supra, 466 U.S. 668 , to the foregoing facts. They are apparently misled by yet another claim of Veganes that his failure to act was a “tactic”: Veganes testified at the reference hearing that he did not investigate petitioner’s background because he did not intend to present any evidence in mitigation, and he did not intend to present any such evidence in order to bar the prosecution from introducing, in rebuttal, evidence of certain prior violent conduct by petitioner.

The majority first reject the Attorney General’s argument that Veganes’s asserted tactic justifies his failure to investigate petitioner’s background, reasoning that a defense counsel cannot rationally decide to present no mitigating evidence without first discovering what mitigating evidence is available. Accordingly, the majority hold that Veganes rendered constitutionally ineffective assistance of counsel by failing to investigate petitioner’s background “to enable him to make an informed decision as to the best manner of proceeding at the penalty phase” (maj. opn., ante, p. 612).

The majority next purport to determine prejudice, but they do not simply answer the question whether on this record Veganes’s deficient performance “undermine[s] confidence in the outcome” within the meaning of Strickland, supra, 466 U.S. at page 694 [880 L.Ed.2d at pages 697-698], Instead, the majority proceed to revive the issue of the constitutional effectiveness of Veganes’s representation, but with a different slant. The Attorney General had argued, as a fallback position, that even if Veganes had investigated petitioner’s background and had discovered the missing mitigating evidence, his same “tactic” would have justified a decision by Veganes not to introduce that evidence at the penalty phase; ergo, his failure to investigate was *670 harmless after all. The majority now accept that argument, and declare that the question whether Veganes’s constitutionally deficient representation was prejudicial turns on the question whether his performance “would have been constitutionally deficient if, for the tactical purpose expressed at the reference hearing, counsel had refrained from introducing the evidence that would have been discovered by a reasonable investigation.” (Maj. opn., ante, p. 613.) The majority’s answer to this new question is that in the final analysis Veganes’s tactical decision did not—mirabile dictu—render his representation constitutionally deficient. (Id. at p. 615.)

To sum up, the majority first hold that Veganes’s performance was deficient because his tactical decision did not justify his failure to investigate for the purpose of discovering what mitigating evidence was available, but the majority then hold that this deficient performance was not prejudicial because his performance was not deficient after all because the very same tactical decision did justify his failure to introduce the mitigating evidence that he would have discovered if he had not performed deficiently by failing to investigate in the first place. I do not exaggerate: that is in essence the reasoning of the majority. (See maj. opn., ante, pp. 609-615.)

It is prudent to be suspicious of any reasoning, like the theory of cycles and epicycles that Ptolemy devised to explain the apparently irregular motions of the planets, that is ingenious but unnecessarily complicated. The majority’s reasoning is both. And when it is tested against the facts, the majority’s reasoning, like Ptolemy’s theory, is exposed as fallacious.

It is time to return to fundamentals. The majority appear to have lost sight of the crucial importance of taking evidence of a defendant’s personal history and background in the penalty phase of a capital case. “There are two reasons for this rule. First, because of the drastic nature of the penalty it is critical that the sentencing authority have full, accurate, and reliable information about the person to be punished .... In a noncapital case ‘where sentencing discretion is granted, it generally has been agreed that the sentencing judge’s “possession of the fullest information possible concerning the defendant’s life and characteristics” is “[hjighly relevant—if not essen tial—[to the] selection of an appropriate sentence. . . .” ’ (Lockett [v. Ohio (1978)] 438 U.S. at pp. 602-603 [ 57 L.Ed.2d at p. 988 , 98 S.Ct. 2954 ].) Yet ‘If an experienced trial judge, who daily faces the difficult task of imposing sentences, has a vital need for accurate information about a defendant and the crime he committed in order to be able to impose a rational sentence in the typical criminal case, then accurate sentencing information is an indispensable prerequisite to a reasoned determination of whether a defendant shall live or die by a jury of people who may never before have made a *671 sentencing decision.’ (Italics added.) (Gregg v. Georgia (1976) 428 U.S. 153, 190 [ 49 L.Ed.2d 859, 884 , 96 S.Ct. 2909 ] (plur. opn. of Stewart, J., Powell, J., and Stevens, J.).)

“Secondly, and even more fundamentally, ‘The basic concept underlying the Eighth Amendment is nothing less than the dignity of man.’ (Trop v. Dulles (1958) 356 U.S. 86, 100 [ 2 L.Ed.2d 630, 642 , 78 S.Ct. 590 ] (plur. opn. of Warren, C. J.); accord, People v. Anderson (1972) 6 Cal.3d 628, 650 [ 100 Cal.Rptr. 152 , 493 P.2d 880 ].) To impose a sentence of death in ignorance of ‘the defendant’s character, background, [and] history’ is to deny that dignity and treat the defendant as something less than a human being: ‘A process that accords no significance to relevant facets of the character and record of the individual offender or the circumstances of the particular offense excludes from consideration in fixing the ultimate punishment of death the possibility of compassionate or mitigating factors stemming from the diverse frailties of humankind. It treats all persons convicted of a designated offense not as uniquely individual human beings, but as members of a faceless, undifferentiated mass to be subjected to the blind infliction of the penalty of death.’ (Woodson [v. North Carolina (1976)] 428 U.S. at p. 304 [49 L.Ed.2d at p. 961].) It follows that ‘in capital cases the fundamental respect for humanity underlying the Eighth Amendment, [citation], requires consideration of the character and record of the individual offender and the circumstances of the particular offense as a constitutionally indispensable part of the process of inflicting the penalty of death.' (Italics added; ibid.) (Accord, Lockett, 438 U.S. at p. 605 [ 57 L.Ed.2d at p. 990 ]; Rockwell v. Superior Court (1976) 18 Cal.3d 420, 428 [ 134 Cal.Rptr. 650 , 556 P.2d 1101 ].)” (Jackson I, supra, 28 Cal.3d 264, 321-322 (dis. opn. of Mosk, J.).)

In the case at bar Veganes categorically admitted at the reference hearing that he would not have called any witness for petitioner in the penalty phase no matter how valuable his or her testimony:

Surely this is a startling admission by an attorney charged with defending a man on trial for his life. And in light of the value the Constitution places on *672 the very evidence that Veganes refused to seek out and present, his reason for that refusal will need to be persuasive indeed. Regrettably, it is not.

In his testimony at the reference hearing Veganes described his reasoning process as follows: (1) he believed that petitioner’s best chance to avoid the death penalty was to convince the jury to entertain a lingering doubt as to the degree of his participation in the Ott murder, and (2) he believed that any such doubt would be eliminated if the jury learned that petitioner had previously committed a similar assault on two women in Berkeley. As will appear, the record shows that both beliefs were patently unreasonable.

First, after the guilt phase had concluded no competent defense attorney could reasonably have believed that petitioner’s best chance in the penalty phase would be to convince this jury to entertain a doubt as to the degree of petitioner’s participation in the fatal beating of Mrs. Ott. To begin with, any such doubt would have been laid to rest by the testimony of Mikles and McFarland graphically describing how petitioner admitted that he twice personally beat Mrs. Ott—“he kept firing on her”—until she was senseless. And as we have seen, the prosecutor repeatedly emphasized that testimony in his argument to the jury.

Moreover, the jury made it clear that it entertained no such doubts. In his own presentation to the jury, Veganes argued that the prosecution had failed to prove the special circumstance allegations because the evidence did not show that it was petitioner rather than one of his companions who inflicted the fatal blows. 43 The jury impliedly rejected that argument, however, when it returned its verdict finding “That the defendant, Earl Lloyd Jackson, with the intent to cause death, physically aided or committed the act or acts causing [the] death of Mrs. Gladys Ott . . . .” Under its instructions, the jury could not have so found if it doubted that petitioner personally participated in the fatal beating.

Secondly, no competent defense attorney could reasonably have believed that the evidence of the Berkeley incident was more important than any mitigating evidence he could have presented through petitioner’s family *673 members, and in particular the mitigating evidence of the five close relatives that a reasonable investigation would have produced. At the outset it bears noting that Veganes did not know much about the Berkeley incident prior to trial because he again kept himself deliberately ignorant of the facts. He testified at the reference hearing that at “the very beginning of the case” the prosecution gave him a copy of a rap sheet reciting that petitioner had been arrested on assault charges in Berkeley approximately one year before the crimes here in issue. But rap sheets do not show the details of the charges they list, and Veganes admitted that at no time during the pretrial period did he send an investigator to follow up on the matter by interviewing either the responsible police officer or the alleged victims. When asked why he had not done so, Veganes testified that “I thought my best chances of success were to forgo any knowledge” of the offenses so that he could later object to their admission at the penalty phase. But Veganes could have so objected even if he had not kept himself in ignorance of the facts: the statutory bar to admission of such other-crimes evidence is not triggered by defense counsel’s mere knowledge of the nature of such evidence, but by the prosecution’s failure to give counsel express and timely notice of an intent to introduce the evidence in its case-in-chief. (Pen. Code, § 190.3, 4th par.) 44

If Veganes had investigated the Berkeley incident when he first learned of it, he would have discovered that its resemblance to the present crimes was superficial at best, and that its effect would have been far outweighed by the benefits of the mitigating evidence he could have developed from petitioner’s close family members.

Although evidence of the Berkeley incident was not introduced at trial, the police report of the event will tell us enough about it for our limited needs. 45 According to the police report, on the night of August 30, 1976, petitioner and two fellow teenage boys became involved in an altercation with four other teenage boys in the schoolyard of the Willard Junior High School in Berkeley, and the dispute soon expanded to include two passing teenage girls. Not surprisingly, the participants told the police differing versions, of the incident. According to petitioner, he and his friends were approached by two of the other boys, both fifteen years old, who unsuccessfully tried to sell them marijuana. Petitioner told the police that “During this conversation an argument ensued and a fight began. Jackson also states that the two girls who had walked through the courtyard had said something to them and one *674 of them attacked him from behind and he turned around and backhanded one, only to defend himself.”

In turn, the companions of the first two boys told the police that the latter started talking with the petitioner and his friends “and for some reason an argument ensued” and fists began to fly. These witnesses also told the police that a shouting match then developed between the two girls and petitioner and his friends, “and after one of the [girls] said something back to the suspect Jackson, Jackson backhanded her and a short scuffle ensued between those two.”

As described by the two girls, Roberts and Slifer, the fight was more serious than a “short scuffle.” These witnesses told the police that after Roberts and one of petitioner’s friends had exchanged comments, petitioner entered the conversation and asked the girls, “Don’t you think you’re a little bit young to be out this late?” 46 Mimicking his words, Slifer retorted to petitioner, “Don’t you think you’re a little bit young to be out this late?” 47 The remark so angered petitioner that he punched Slifer in the face, and also punched Roberts when she intervened in the fight. According to the two girls, petitioner hit them several more times and knocked them both to the ground before leaving the scene. 48 Finally, the police report notes that upon complaints by the girls or their parents petitioner was charged with two counts of simple battery, and an arrest warrant issued on October 7, 1976; 49 the report contains no record, however, of the disposition of the charges.

Even assuming that the Long Beach prosecutor in the present case could have located these juvenile witnesses in Berkeley two years after the event and they had told the same story under oath in a capital trial, nevertheless it borders on the grotesque to compare this schoolyard altercation, triggered by typical teenage taunts, with the savage and unprovoked beatings during the Curtis and Ott burglaries that left one elderly and defenseless woman dead and another dying. No reasonably competent attorney would have seen more than superficial similarities between these events.

Such an attorney would also have seen that any effect that the Berkeley incident would have had on the jury would be trivial compared with the powerful impact that the testimony recounting petitioner’s disastrous childhood would have made. That testimony would have told the jury an unrelieved tale of physical and psychological neglect and abuse that no child *675 could have experienced without suffering deep emotional scars. 50 At the reference hearing petitioner’s expert witness Gessler testified that “given what I’ve read and heard of Mr. Jackson’s background, I would go to great lengths to see that the jury got that. I think it’s one of the most sympathetic backgrounds that I have come across as to showing why by age 14 Mr. Jackson went the way that he did.” When asked whether his opinion would be different if the introduction of that mitigating evidence would have allowed the prosecution to call witnesses to the Berkeley incident, Gessler adhered to his position, explaining that “I don’t see that as having such devastating impact that it would outweigh the defense evidence of Mr. Jackson’s childhood and development and background.” In accord with the expert testimony, the referee found that the Veganes’s refusal to investigate for fear of “opening the door” to rebuttal evidence of the Berkeley incident was not a tactical decision that reasonably competent defense attorney would have made in the circumstances. Although this finding is subject to our independent review, a fair reading of the record again compels us to reach the same conclusion as the referee.

The majority concede that “another attorney reasonably might have made a different tactical decision," yet the majority “cannot say” that Veganes’s decision was unreasonable. (Maj. opn. ante, p. 615.) Impliedly finding this to be a close question, the majority tilt the scales against petitioner by invoking a presumption in favor of counsel: the majority rely on the rule that in this context courts should “accord great deference” to counsel’s tactical decisions and should “avoid second-guessing counsel on the basis of hindsight.” (Ibid.) As will appear, in this case the presumption does not justify the result that the majority want to reach.

I do not dispute that there is, and should be, a presumption in favor of the reasonableness of decisions by trial counsel that are truly tactical in nature. Strickland says as much ( 466 U.S. at p. 689 [880 L.Ed.2d at pp. 694-695]), and for good reason. But the presumption is not conclusive, and when necessary it must yield to the constitutional imperative of adequate legal assistance. Thus even before Strickland a leading decision of this court explained that “Reviewing courts should avoid second-guessing counsel’s informed choice among tactical alternatives, but a defense attorney’s freedom to make such decisions is not without limits. Every person accused of a criminal offense is entitled to constitutionally adequate legal assistance. [Citation.] That right is denied if trial counsel makes a critical tactical decision which would not be made by diligent, ordinarily prudent lawyers in *676 criminal cases.” (People v. Pope (1979) 23 Cal.3d 412, 424 [ 152 Cal.Rptr. 732 , 590 P.2d 859 , 2 A.L.R.4th 1 ].) More recently, we emphasized that “deferential scrutiny of counsel’s performance is limited in extent and indeed in certain cases may be altogether unjustified. ‘[Djeference is not abdication’ (People v. McDonald (1984) 37 Cal.3d 351 , 377 . . .); it must never be used to insulate counsel’s performance from meaningful scrutiny and thereby automatically validate challenged acts or omissions. Otherwise, the constitutional right to the effective assistance of counsel would be reduced to form without substance.” (People v. Ledesma (1987) 43 Cal.3d 171, 217 [ 233 Cal.Rptr. 404 [ 729 P.2d 839 ].) On the facts before us, deference would indeed amount to abdication.

The record demonstrates that this is simply not a close case on the question whether a competent defense counsel could reasonably believe that evidence of the Berkeley-incident would outweigh any mitigating evidence he could have presented through petitioner’s family members, “regardless of how compelling their testimony was.” The referee concluded that the evidence establishes “without question” that Veganes’s performance was deficient in this respect. Surely this conclusion is sound. Knowing how important it is for the penalty jury to have all the information it needs for a rational and individualized sentencing decision, competent counsel would be far more likely to investigate both the Berkeley incident and the circumstances of petitioner’s background, and having discovered the truth about each, conclude that any risk that the prosecution might be able to prove the former is outweighed by the certain benefits of informing the penalty jury about the latter. In these circumstances, petitioner has met and rebutted the presumption that Veganes’s “tactic” was professionally adequate. Rather, the “tactic” stands revealed for what it is—another example of Veganes’s mistaken priorities and misunderstanding of even his most basic duties of investigation and preparation in a capital case. (See also this opn., fn. 36, ante.) Once again counsel’s “tactic” does not justify his inaction or save his performance from constitutional inadequacy.

I reach, therefore, the issue of prejudice. Under the governing standard of Strickland , “the question is whether there is a reasonable probability that, absent the errors, the sentencer . . . would have concluded that the balance of aggravating and mitigating circumstances did not warrant death.” ( 466 U.S. at p. 695 [80 L.Ed.2d at pp. 698-699].) Directly responsive to this question is the answer given by the experienced referee in our case: “It is reasonably probable that, at the penalty phase of the trial, the mitigating evidence would have evoked the response in the jury to give petitioner-defendant Jackson a life sentence rather than the death penalty.” I agree, as must anyone who reads with an open mind the majority’s own summary of *677 the missing testimony of the five close family members describing the appalling conditions in which petitioner was forced to spend his formative years. (Maj. opn. ante, pp. 606-609.) On this record, it would be absurd to hold that a denial of constitutionally effective assistance of counsel which deprived the jury of that crucial testimony does not at least “undermine confidence in the outcome” of the jury’s penalty deliberations within the meaning of Strickland, 466 U.S. at page 694 [80 L.Ed.2d at pages 698-699].

Thus viewed, the case at bar is functionally indistinguishable from our recent decision in In re Marquez, supra, 1 Cal.4th 584 . There, too, counsel for a capital defendant undertook virtually no investigation into the circumstances of his client’s childhood and background because he feared it might turn up aggravating evidence. (Id. at pp. 599-602.) We reviewed four federal decisions on similar facts (Armstrong v. Dugger (11th Cir. 1987) 833 F.2d 1430, 1433-1434 ; Thomas v. Kemp (11th Cir. 1986) 796 F.2d 1322, 1324-1325 ; Blake v. Kemp (11th Cir. 1985) 758 F.2d 523, 533-535 [82 A.L.R.Fed. 889]; Pickens v. Lockhart (8th Cir. 1983) 714 F.2d 1455,1465-1468 ), and we concluded:

There is, I submit, no principled way of distinguishing the case at bar from Marquez, from the federal authorities on which it relies, and from even more recent federal decisions that also vacate state death sentences on federal habeas corpus because counsel failed without reasonable tactical excuse to *678 investigate the petitioner’s troubled childhood and background for mitigating evidence. (Brewer v. Aiken (7th Cir. 1991) 935 F.2d 850, 857-859 ; Kenley v. Armontrout (8th Cir. 1991) 937 F.2d 1298, 1304-1309 .)

For all the foregoing reasons I conclude that the denial of constitutionally effective assistance of counsel in failing to investigate petitioner’s background for mitigating evidence was prejudicial as to the penalty verdict, and hence that on this ground as well the petition for habeas corpus should be granted and the judgment vacated insofar as it imposes the penalty of death.

Kennard, J., concurred.

Petitioner’s application for a rehearing was denied October 28, 1992, and the opinion was modified to read as printed above. Mosk, J., and Kennard, J., were of the opinion that the application should be granted.

At the outset I pause to identify the distinguished referee who assisted us in this case. Although the majority do not name him, he is not the typical retired superior court judge whom we ordinarily appoint. He is, rather, the Honorable Bernard S. Jefferson, retired Presiding Justice of the Court of Appeal, Second Appellate District, Division One. In addition to his experience on that court, he served as an Associate Justice pro tempore on this court in a dozen cases, he has taught for many years in both law schools and the California Judicial College, and he is the author of numerous scholarly articles and the well-known California Evidence Benchbook (2d ed. Cont.Ed.Bar 1982), a major treatise on the law of evidence that has been cited in more than 300 published decisions of our courts. I mention these facts to urge that although in this proceeding we independently review mixed questions of law and fact—such as issues of prejudice—Justice Jefferson’s long experience in deciding those very questions should lead us to accord particular respect to the conclusions he reaches in his reports to this court in the case at bar.

Judge Arguelles, of course, later served as an Associate Justice of this court. The parties often refer to the proceeding before Judge Arguelles as the Long Beach case, and I shall do the same.

The parties often refer to this proceeding as the Norwalk case, and I shall do the same.

In the argot of the subculture of informants, he was the “snitch” and she was his “juice lady.”

Nor did the jury learn that Mikles’s testimony against petitioner was part of his modus operandi: he admitted at the reference hearing that during the period surrounding petitioner’s trial he testified as an informant for the prosecution in no less than six additional cases, four of which—like petitioner’s—were murder trials.

Indeed, it was a habit that Mikles apparently found hard to break, for it was still his first line of defense almost a decade later: at the session of the reference hearing held on March 28, 1988, Deputy Shea testified that approximately one year earlier Mikles had been arrested for a new crime, “as a matter of fact, at the station where I work. And it happened to be on my days off. And he called me to tell me he was in custody, and he had some information. And I basically told him I didn’t have time, you know, could he please contact somebody else.” (Italics added, paragraphing omitted.)

Deputy Shea’s credentials as a witness were impressive: at the time of her testimony she was in her 20th year of service with the Los Angeles Sheriff’s Office, had been promoted to the rank of sergeant, and as a detective had regularly worked with informants for some 16 years.

In the same letter Deputy Shea also told the probation department that in a pending murder trial (People v. Luna) law enforcement personnel believed that Mikles testimony will be “extremely significant and that he will be an important and credible witness,” and that in a third murder case (People v. Gomez) law enforcement personnel were of the view that “Mikles was a very credible and cooperative witness and that his testimony was significant in this conviction.”

Deputy Shea argued that because Mikles furnished that information after the date had been set to hear his motion to modify his sentence, “it certainly couldn’t be interpreted as attempting to further influence” Judge Arguelles. The non sequitur is obvious.

Deputy Shea’s argument that if Judge Arguelles would give Mikles “some consideration” he would abandon his life of crime proved singularly ill-founded. The hearing before Judge Arguelles took place in 1979. Nine years later, on March 24, 1988, Mikles testified as follows at the reference hearing:

On cross-examination Mikles specified that he had been convicted, inter alia, of robbery in 1982 and of receiving stolen property in 1985. Indeed, in seeking to show that he does not always turn informant when charged with or convicted of a crime, Mikles acknowledged that *626 “I’ve done a few good stretches since then”—i.e., since the time of petitioner’s trial. (See also fn. 5, ante.)

With this hindsight it is fair to infer that when Deputy Shea predicted to Judge Arguelles that Mikles would remain law-abiding if given the chance, she was subordinating her professional judgment of his character to the deal she had made to help him if he helped her.

Although undated, internal evidence shows the letter was written after May 31, 1979, the date on which Judge Arguelles modified Mikles’s sentence in the Long Beach case.

Finally, the referee also found that because of Mikles’s testimony against petitioner and others, approximately 15 additional robbery charges pending against him in Long Beach were dropped.

McFarland claimed he was in protective custody because he had telephoned a local television news “hotline” from jail and reported that a fellow prisoner had some information about the “Hillside Strangler.”

The transcript was admitted into evidence at the reference hearing (respondent’s exhibit B), and hence is before us.

Carroll testified, “I wanted one count with five years to life, and they wanted two counts with consecutive five-years-to-life sentences."

In particular, prosecutor Marin did not deny that he agreed to write a letter to Judge Fagan on McFarland’s behalf. Rather, Marin testified at the reference hearing that when he used informants it was “my general practice ... to tell them that I could not promise them anything, however, if they were to testify crucially in, let’s say, a case that I was handling, as a People’s witness, that I would relay that information to—whether that is the sentencing judge or the officers; that’s all. . . . Now, I may have told that to Mr. McFarland. I may have told that to Mr. Mikles and ... I would have then followed that up by making such a statement.” When specifically asked whether he agreed to write to Judge Fagan about McFarland’s testimony against petitioner, Marin testified that he could not recall but “I may have and I may even have sent such a letter.”

Thus in Napue, supra, 360 U.S. 264, 265-266 [ 3 L.Ed.2d 1217, 1218-1219 ], the high court characterized Hamer as “the principal witness for the State.” It asserted that “Hamer’s testimony was extremely important” in identifying his confederate Napue as one of the participants in the crime. (Id. at p. 266 [ 3 L.Ed.2d at p. 1219 ].) It concluded that “On the basis of the evidence presented, which consisted largely of Hamer’s testimony, the jury returned a guilty verdict.” (Ibid.)

In Giglio, supra, 405 U.S. 150,151 [ 31 L.Ed.2d 104, 106-107 ], the high court characterized Taliento as the prosecution’s “key witness” and noted that he was “the only witness linking petitioner with the crime.” It concluded, “Here the Government’s case depended almost entirely on Taliento’s testimony; without it there could have been no indictment and no evidence to carry the case to the jury.” (Id. at p. 154 [31 L.Ed.2d at pp. 108-109].)

I cannot join in the majority’s refusal to “resolve the factual question” (maj. opn., ante, p. 593, fn. 8) whether prosecutor Marin agreed that he would write a letter to Judge Fagan on McFarland’s behalf after the latter had testified against petitioner. As I explain above (pt. I.B., ante), there is substantial evidence to support the referee’s purely factual finding that Marin agreed to do so. As the majority acknowledge, we review such a finding under a highly deferential standard, giving it great weight. (In re Marquez (1992) 1 Cal.4th 584, 603 [ 3 Cal.Rptr.2d 727 , 822 P.2d 435 ], and cases cited.) The majority cite no evidence sufficient to overcome that heavy presumption.

This ground for disbelieving a witness is embodied in a standard jury instruction. (CALJIC No. 2.20 (5th ed. 1988); see also Evid. Code, § 780, subd. (f).)

This ground for disbelieving a witness is embodied in a different jury instruction. (CALJIC No. 2.21.2 (5th ed. 1988).)

I recognize that the majority opinion quotes this finding, although in a different context. (Maj. opn., ante, p. 603.)

The majority begin instead by holding the error harmless as to guilt, reasoning that petitioner’s admissions to the police, corroborated in this respect by noninformant testimony, “unquestionably established’’ his guilt of (1) the burglaries of the apartments of Mrs. Curtis and Mrs. Ott and (2) the first degree felony murders of the same victims, who received fatal beatings during the commission of those burglaries. (Maj. opn., ante, p. 598.) The point is pure dictum, however, because petitioner concedes it ah initio, contesting only the effect of the error on the special circumstance findings and the penalty verdict.

The tape recordings of the two statements were played to the jury, and the court reporter simultaneously transcribed them into the record to the best of her ability. The resulting transcript, however, is replete with passages that the reporter found “inaudible.” As the trial court explained at a pretrial hearing to suppress the same tape-recorded statements, “It is generally very difficult for a reporter to take the tape down, because the acoustics are not generally too good” in the courtroom. Accordingly, the court admitted into evidence a transcription of the tapes prepared before trial, which reproduced the statements more completely and more accurately. (People’s exhibit 8A.) The quotations that follow are taken from that transcription.

In a massive understatement, the majority admit that Gaines’s credibility was “somewhat” impeached when she admitted on cross-examination that “at the time of the offenses, she had been living with Elton Boyd, who participated with [petitioner] in the burglary of Mrs. Curtis’s apartment and who, in a separate trial, had been convicted of the murder of Mrs. Curtis. The questioning raised the possibility that Gaines may have been biased against [petitioner] because [petitioner] had informed the police of Boyd’s participation in the Curtis offense.” (Maj. opn., ante, p. 587, fn. 3.) Indeed, at the time of these events Gaines was only 19 years old, had been “going with” Boyd for about a year and a half, and had a child by him.

The third witness died before trial.

Counsel for then-codefendant Boyd, Attorney Ronald Slick, did conduct a brief cross-examination but asked no questions on the point.

I note that the present majority’s belief in the great significance of Lewis’s preliminary hearing testimony was not shared by the justices who affirmed petitioner’s conviction on his automatic appeal to this court. Justice Richardson, in an opinion joined in this regard by Justices Clark, Manuel, and Newman, rejected a claim that the prosecutor failed to use due *645 diligence to find the witnesses Rushing and Lewis, and reasoned as follows: “even were we to hold that the prosecutor failed to exercise due diligence, any error in admitting the former testimony of witnesses Rushing and Lewis would be harmless beyond a reasonable doubt [citing, inter alia, Chapman v. California (1967) 386 U.S. 18, 24 ( 17 L.Ed.2d 705, 710-711 , 87 S.Ct. 824 , 24 A.L.R.3d 1065 )], because neither witness provided significant, noncumulative testimony. Rushing testified merely that defendant had admitted hitting an elderly lady during the course of a burglary; Lewis was one of several witnesses (Gaines, Hall, Mikles, McFarland) in whom defendant confided regarding his role in the murders.” (Jackson I supra, 28 Cal.3d 264, 313 , italics added.) Justice Richardson’s reading of the record was correct.

In one of the few discrepancies between them, Lewis insisted that the newspaper article was about the death of Mrs. Curtis only.

Although in less colorful language, McFarland told essentially the same story. According to McFarland, while in custody in the Los Angeles County jail in October 1977 he had a conversation with petitioner in which the latter admitted that he and some companions broke into Mrs. Ott’s apartment “and they started ransacking her house and then the lady came out of her bedroom. And Jackson said he seen her and pushed her back onto the bed and started beating her, and then they continued to ransacking. And then the old lady woke up and started hollering and Jackson started beating her then she was unconscious. This is what he told me.

“And then he said he seen a wine bottle or a long-necked bottle of some kind and stuck it into her vagina.”

Gaines, it will be remembered, testified only that as Mrs. Curtis was being removed on a stretcher petitioner remarked that it was he who “did that.”

The referee agreed with this assessment, finding in his original report that “The testimony of Mikles and McFarland as to the incriminating statements made by petitioner-defendant Jackson constituted the prosecution’s most significant and compelling evidence to establish the special circumstances alleged . . . .”

In United States v. Antone, supra, 603 F.2d 566, 570 , the fact withheld from the jury was simply that the witness’s lawyer had been paid from government funds. By contrast, after two full days of impeachment testimony the jury knew many other facts about the witness that were far more damaging to his credibility: “He admitted to having used addictive drugs, to having been treated for mental illness, and to having made prior inconsistent statements about matters which were the subject of his trial testimony. He admitted that he lied in the past in order to protect himself and that he had a personal bias against at least one of the defendants. The particulars of his plea bargains with the Government were admitted and he conceded that he was testifying in order to avoid the electric chair.” (Ibid.) In the case at bar, of course, the jury heard no such evidence.

Both of the foregoing grounds for finding lack of prejudice in a Napue-Giglio violation— ample corroboration and substantial impeachment—were present in the California case relied on by the majority, People v. Morris (1988) 46 Cal.3d 1, 34-35 [ 249 Cal.Rptr. 119 , 756 P.2d 843 ],

In Campbell v. Reed, supra, 594 F.2d 4 , the defendant was convicted of burglary in state court after his accomplice testified against him. The latter, however, falsely testified that he had not been promised any benefits for testifying against the defendant, and the prosecutor did not correct the perjury. In ordering federal habeas corpus relief, the court held that the Napue-Giglio violation was not rendered harmless by the facts that “The other evidence introduced at trial consisted of a confession signed by [the defendant] and the testimony of a police detective that [the defendant] pointed out other residences that he allegedly had broken into.” (Id. at p. 8.) The court concluded (ibid.), “Viewing the record as whole, we conclude that the jury’s verdict might have been different” had it known the true nature of the accomplice’s motivation to testify against the defendant.

In Napue, supra, 360 U.S. 264 , the United States Supreme Court specifically noted that the testimony of Hamer falsely denying that he had been promised sentencing benefits “was the last testimony of Hamer’s heard by the jury . . . .” (Id. at p. 270 [3 L.Ed.2d at pp. 1221-1222].)

Sergeant Sallmen’s testimony, moreover, was vigorously challenged on cross-examination. Under questioning by defense counsel, Sallmen conceded the following facts: He had not made an official report that would confirm the testimony he gave on direct examination; his partner, Sergeant Luman, made the report of their investigation; Luman’s report, however, mentioned only one hair on the bottle and said nothing at all about dried blood; and although he (Sallmen) read Luman’s report, he did not object to its description of the condition of the bottle. Sallmen further conceded that although there was a police photographer on the scene, he did not direct that a photograph be taken specifically showing the appearance of the neck *659 and cap of the bottle. Finally, although a criminalist at the Long Beach Police Department crime laboratory thereafter examined the bottle and prepared a report on its condition, Sallmen testified that the criminalist was no longer with the Long Beach Police Department and the prosecutor did not call him as a witness. The jury therefore remained ignorant of the contents of that report, and there is in fact no forensic evidence in the record establishing the condition of the bottle or the presence of petitioner’s—rather than McDowell’s—fingerprints on it.

Veganes’s failure to inquire into the background of Mikles and McFarland was symptomatic of his entire approach to preparing—or not preparing—for this trial. Thus in the course of his testimony at the reference hearing Veganes also admitted that (1) he did not ask the trial court for funds to pay investigators pursuant to Penal Code section 987.9, (2) he did not retain any such investigators for any purpose, (3) he did not file any formal discovery motions, (4) he did not file any written “informal” discovery requests under local court rules, (5) he served no subpoenas on anyone to testify for petitioner in any phase of the trial, and (6) although the prosecution furnished him with a list of 32 witnesses it intended to call, Veganes interviewed none of them except petitioner’s cousin Debra Hall, whom he consistently referred to as petitioner’s “aunt.” He apparently chanced on the latter witness in the Long Beach courthouse jail where she was being held on unrelated charges; he spoke with her for perhaps half an *661 hour, but admitted “She didn’t say too much.” Finally, as will appear (pt. Ill, post), Veganes undertook no investigation to find potential mitigating evidence to present at the penalty phase.

In short, Veganes underscored what he did not do by describing what he did do: “The investigation occurred by myself, and it was limited to my reading of the police report and subsequent follow-up of reading of the transcripts of the other trial [i.e., of Elton Boyd, who had previously been convicted of the Curtis murder].

“That’s it.”

The list was not a formal compilation of all possible prosecution witnesses prepared in reply to a demand presented as part of a discovery motion, because Veganes made no such motion. Rather, it appears to have been a more informal list of the probable prosecution witnesses intended primarily for the trial court to read to the prospective jurors at the outset of voir dire, in order to learn if any jurors were related to or acquainted with a prosecution witness. Marin apparently gave a handwritten copy of this list to Veganes as a courtesy.

At the time of the reference hearing Hough was a head deputy in the Los Angeles Public Defender’s Office. He had been a public defender since 1966, a certified criminal law specialist since the program was instituted in 1973, and had tried some 200 criminal jury cases.

Gessler was the coordinator for capital cases in the Los Angeles Public Defender’s Office. He had been a public defender since 1965, a certified criminal law specialist since 1973, and had tried more than 50 murder cases.

Hough also testified that in his opinion it would not constitute reasonably effective assistance of counsel for an attorney in a capital case who had been given a list of 32 prosecution witnesses, including 2 informants, several months before trial, to limit his pretrial investigation to a 20- to 30-minute interview with only 1 such witness—not an informant— found in the courthouse jail. The witness further declared it would not constitute reasonably effective assistance of counsel for such attorney to fail to present either a formal discovery motion or a written “informal” discovery request under local court rules.

Veganes*s failure to investigate Mikles and McFarland would have been ineffective assistance of counsel, of course, even if the prosecutor had not asked them—and they had not falsely answered—the question whether they had been promised any benefits for testifying against petitioner.

In the comment referred to, Mikles claimed that he overheard petitioner say to other Black inmates, “ ‘So what if I did kill . . . those two old White bitches?’ ”

The majority raise but do not decide the question of what if any effect the recent decision in Dawson v. Delaware (1992) 503 U.S._[ 117 L.Ed.2d 309 , 112 S.Ct. 1093 ], has on the admissibility of impeachment evidence that Mikles was a past member of the Aryan Brotherhood. The short answer is none: Dawson is plainly distinguishable. It does not hold that evidence of membership in the Aryan Brotherhood is inadmissible to impeach a witness; on the contrary, it reaffirms (id. at p._[17 L.Ed.2d at pp. 316-317, 112 S.Ct. at p. 1097 ]) a decision allowing such impeachment evidence (United States v. Abel (1984) 469 U.S. 45, 52-53 [ 83 L.Ed.2d 450, 457-458 , 105 S.Ct. 465 ]). Instead, Dawson holds that a stipulation concerning the Aryan Brotherhood that was admitted in the penalty phase of the case before it was so narrow that it was “totally without relevance” to the issues in that proceeding. (503 U.S. at p--[17 L.Ed.2d at pp. 316-317, 112 S.Ct. at p. 1097 ].) For example, “the murder victim was white, as is Dawson; elements of racial hatred were therefore not involved in the killing.” (Id. at p__[17 L.Ed.2d at p. 318, 112 S.Ct. at p. 1098 ].) The facts of our case are otherwise, of course, and the evidence would not have been irrelevant here.

The majority also hold without further explanation that the denial of effective assistance of counsel in failing to investigate the informants was not prejudicial as to the penalty determination. (Maj. opn., ante, p. 605.) I would find the error prejudicial as to penalty for the reasons stated above and in part I.D.2. of this opinion.

Thus Veganes argued, “Based on the evidence that you have here, there is no indication that Mr. Jackson was the one that did the choking. Based on the evidence that you heard here, there is no indication that Mr. Jackson is the one that did the hitting, with intent to kill. Based on the evidence you heard here, you know there were other individuals that were involved that were doing hitting, but you don’t know to what degree they were hitting, when they were hitting, how they were hitting and who was present when they were hitting. All you have is speculation that—if you believe the facts to be proven, that Mr. Jackson was involved, but to what degree of his involvement, pursuant to the rules—you don’t have enough. You may want to find enough, because of the nature of the offense, but you do not have enough, not beyond a reasonable doubt.”

Indeed, when the matter came up at trial Veganes did object on the ground of lack of notice, and the court advised the prosecutor that such evidence would be excluded if he offered it at the penalty phase.

The police report was admitted into evidence at the reference hearing (petitioner’s exhibit 12), and hence is before us.

Roberts and Slifer had recently turned 16; the hour was a few minutes before midnight.

Petitioner was just short of 19 years old.

The fight was apparently not all one-sided: the girls told the police that as Roberts was getting back on her feet, “Slifer then kicked Jackson in the groin.”

Petitioner had no prior arrest history.

At the beginning of their opinion the majority liken this case to Dickens’s Bleak House; on this issue I liken it rather to Oliver Twist.