Opinion
WERDEGAR, JIn 1990, petitioners Peter Sakarias and Tauno Waidla were each, in separate trials, convicted of first degree murder with special circumstances and sentenced to death in the killing of Viivi Piirisild. We affirmed each of their convictions and sentences on automatic appeal (People v. Sakarias (2000) 22 Cal.4th 596 [94 Cal.Rptr.2d 17, 995 P.2d 152] (Sakarias); People v. Waidla (2000) 22 Cal.4th 690 [94 Cal.Rptr.2d 396, 996 P.2d 46] (Waidla)), but issued orders to show cause in response to their petitions for writs of habeas corpus, on claims the prosecutor, in each trial, had presented factual theories inconsistent with those presented at the co-defendant’s trial. In response to Waidla’s petition, we also specified a procedural issue: whether claims of error under Miranda v. Arizona (1966) 384 U.S. 436 [16 L.Ed.2d 694, 86 S.Ct. 1602] (Miranda) are cognizable on habeas corpus. On receipt of the returns and traverses, we consolidated the two causes for consideration and decision and appointed a referee to hear evidence and make factual findings. The referee has now issued his report, and the parties have filed briefs on the merits.
The evidence at petitioners’ trials showed they both participated in the fatal attack on Viivi Piirisild, which was perpetrated with a hatchet and a knife. (See Sakarias, supra, 22 Cal.4th at pp. 611-613; Waidla, supra, 22 Cal.4th at p. 710.) But both petitioners contend their joint prosecutor, Los Angeles County Deputy District Attorney Steven Ipsen, inconsistently and falsely portrayed their respective roles in the attack, attributing to each, in their respective trials, a series of three blows struck to *145the victim’s head with the blade of the hatchet. Petitioners claim this prosecutorial inconsistency deprived them of due process, requiring that their death sentences be vacated. We agree with Sakarias that the prosecutor violated his due process rights by intentionally and without good faith justification arguing inconsistent and irreconcilable factual theories in the two trials, attributing to each petitioner in turn culpable acts that could have been committed by only one person. We also agree this violation prejudiced Sakarias, entitling him to relief. We do not decide whether the prosecutor’s conduct was a due process violation as to Waidla, as we conclude any such violation was harmless in his case.
Finally, we hold Miranda claims are cognizable on habeas corpus in California courts, but that such a claim is subject to denial on procedural grounds where, because it rests on facts in the appellate record, the claim was already raised and rejected, or could have been raised but was not, on direct appeal. (In re Harris (1993) 5 Cal.4th 813, 824-829 [21 Cal.Rptr.2d 373, 855 P.2d 391]; In re Waltreus (1965) 62 Cal.2d 218, 225 [42 Cal.Rptr. 9, 397 P.2d 1001]; In re Dixon (1953) 41 Cal.2d 756, 759 [264 P.2d 513].) Other procedural bars, of course, may also apply to a Miranda claim in a given case.
I. Factual and Procedural Background
A. Facts of the Crime
Waidla and Sakarias were both bom in Estonia while that nation was part of the Soviet Union. They met as conscripts in the Soviet Army, from which they defected together, coming in 1987 to Los Angeles. There, they were taken under the wing of an Estonian-American couple, Avo and Viivi Piirisild, who offered to help them obtain jobs and education. For a period in 1987 to 1988, Waidla lived in the Piirisilds’ guest house, performing remodeling work and other chores in exchange for his room and board. (Sakarias, supra, 22 Cal.4th at p. 609; Waidla, supra, 22 Cal.4th at pp. 705-707.)
Relations soon soured between petitioners and the Piirisilds. In May 1988, Waidla demanded the Piirisilds give him money or a sports car they had promised him for his work and threatened, otherwise, to report them for doing construction without a permit. When the Piirisilds told Waidla to leave their home, Waidla threatened to hurt or kill Avo. Later, Viivi received a postcard with a rattlesnake on it from Sakarias and Waidla, who were traveling together. Sakarias later told police he believed Viivi had been spreading harmful rumors about him and Waidla within the Estonian community, damaging their prospects for help from other Estonians around North America. (Sakarias, supra, 22 Cal.4th at pp. 610, 612; Waidla, supra, 22 Cal.4th at pp. 707-709.)
*146During early July 1988, petitioners broke into the Piirisilds’ unoccupied cabin in Crestline. They stayed for several days, leaving only when they ran out of food and taking with them various items of the Piirisilds’ property, including a hatchet. On July 12, angry, hungry, and in need of money, they went to the Piirisilds’ North Hollywood home and broke in through the back door. They ate food from the kitchen and took some jewelry while waiting for Viivi to return home. Sakarias later told the police he and Waidla were planning to get money for food and to confront Viivi and frighten her into giving them the sports car; he also said that having contemplated killing themselves because of their poor situation, they decided to kill Viivi first so “ ‘she is not gonna see my funeral’ ” or, with her husband, “ ‘laugh on us for the rest of their lives.’ ” (Sakarias, supra, 22 Cal.4th at pp. 610, 612-613; Waidla, supra, 22 Cal.4th at pp. 709-710.)
When Viivi entered the house through the front door, petitioners immediately attacked her, using a knife and the hatchet they had taken from the Crestline cabin. They bludgeoned her with the blunt end of the hatchet, stabbed her with the knife, and chopped at her with the hatchet blade. Overall, the medical examiner found five blunt force impacts to Viivi’s head (which fractured her skull and facial bones, knocked out her teeth, and broke her larynx), four stab wounds to her chest (two of which passed through vital organs), and three chopping wounds to her upper head. One of this last group of injuries, inflicted before death, was struck with “tremendous” force, penetrating Viivi’s skull completely. The other two chopping wounds were inflicted with somewhat less force, after or around the time of death. The medical examiner attributed Viivi’s death to the combination of wounds, several of which could have been fatal individually. After the attack in the entryway, petitioners dragged Viivi down the hall to a bedroom, where her body was found. According to the medical examiner’s testimony at Waidla’s trial, an abrasion on Viivi’s lower back, caused by rubbing of her skin against another surface (which could have been incurred when she was dragged to the bedroom), was inflicted after her death. (Sakarias, supra, 22 Cal.4th at pp. 611-613; Waidla, supra, 22 Cal.4th at p. 710.)
Sakarias told police that during the initial attack he wielded the knife while Waidla used the hatchet. Sometime later, at Waidla’s direction, he went to the bedroom and chopped Viivi’s head twice with the hatchet. (Sakarias, supra, 22 Cal.4th at p. 613.) Waidla gave a statement admitting only a single bludgeoning blow, with the back of the hatchet at the outset of the attack, and denying any memory of how the rest of the attack proceeded. He recanted even that confession at his trial, testifying he had left Los Angeles three days before Viivi Piirisild was killed. (Waidla, supra, 22 Cal.4th at p. 712.)
Petitioners sold the jewelry they took and used Viivi’s credit cards for airline tickets, telephone calls, and other purchases. They were arrested more *147than a month later near the United States-Canada border in New York State. (Sakarias, supra, 22 Cal.4th at p. 612; Waidla, supra, 22 Cal.4th at pp. 710-711.)
B. The Inconsistent Factual Theories
Petitioners were jointly charged with Viivi Piirisild’s murder, but their cases were severed after Sakarias was found incompetent to stand trial. (Waidla, supra, 22 Cal.4th at p. 704.) Waidla’s jury trial began with the prosecution’s opening statement on October 24, 1990; penalty arguments were made on January 2, 1991. Sakarias’s trial began on September 30, 1991, with penalty arguments on October 30, 1991.
As reflected in the summary above, the evidence at petitioners’ trials, taken as a whole, strongly suggests Waidla (who first wielded the hatchet, according to both petitioners’ statements) struck the first, antemortem blow with the hatchet blade in the entry way, while Sakarias (who admitted doing so) inflicted the two postmortem or perimortem chopping wounds in the bedroom. (There was no evidence in either trial to suggest the perpetrators switched weapons during the initial attack.) But the prosecutor, Ipsen, did not argue at either trial the version of the attack best supported by all the evidence. Instead, at each defendant’s trial he maintained the defendant on trial had inflicted all the chopping wounds.
In Waidla’s trial, Ipsen introduced Waidla’s admission that he, rather than Sakarias, had initially used the hatchet against Viivi Piirisild. (Sakarias’s confession to police, in which he admitted striking two blows with the hatchet in the bedroom, was not introduced at Waidla’s trial.) Although Waidla only admitted hitting Viivi with the back of the hatchet, Ipsen argued the jury should find Waidla actually used the hatchet throughout, “choosing ... the more devastating of the instruments,” while Sakarias “accepted]” the knife, “the lesser implement.” With the hatchet, Ipsen argued, Waidla first inflicted the blunt force injuries, then, “turning the hatchet blade so it was more effective . . . [he] was now able to chop through the top of her skull.” Ipsen suggested Waidla simply did not want to acknowledge his role in the attack, “his repeated striking of Viivi Piirisild, and swinging with the sharp end of the hatchet. . . until she was dead.” He emphasized the extended and repeated efforts both assailants made to ensure Viivi’s death, “as Mr. Waidla indicated, himself with the hatchet, Mr. Sakarias who came up later with the knife.” Waidla’s use of the hatchet blade continued, Ipsen argued, even after Viivi was dead: “ ‘[S]he’s alive, she’s alive, she’s alive.’ Sharp end, ‘she’s dead,’ and then further blows indicating further blows were struck after she was dead, the non-hemorrhagic chop wounds to the head.”
*148Having elicited, in the Waidla trial, the opinion of Dr. James Ribe, the medical examiner, that the abrasion on Viivi’s lower back was incurred postmortem, Ipsen emphasized that the initial attack in the living room was fatal: “At the point that she was dragged into the back room, we know that Viivi Piirisild was already dead by the facts as the coroner testified. So, we know it was in that front room that the attack occurred, and that Viivi Piirisild was bludgeoned, chopped and stabbed until life left her body.” Finally, in penalty argument, Ipsen urged a death sentence, in part because Waidla, after hitting Viivi repeatedly with the hatchet’s blunt end, “chose to change the angle of the blade. . . . Although he felt her head and her flesh against the back of his hatchet numerous times, he knew his mission wasn’t accomplished, and that’s when he changed and switched and used the sharp edge of the hatchet to give that death blow.”
In Sakarias’s trial, the prosecutor asked the medical examiner, Dr. Ribe, about each stabbing, chopping, or blunt force injury shown in the autopsy photographs, in many instances asking whether the wounds were antemortem or postmortem, but he did not examine Dr. Ribe about the lower back abrasion at all. He thus avoided eliciting Dr. Ribe’s opinion, expressed in Waidla’s earlier trial, that the abrasion had occurred after death and could have been caused by dragging Viivi’s body along the carpeted hallway to the bedroom.
Due to this omission, no evidence was before Sakarias’s jury that Viivi Piirisild was dead by the time Sakarias, as he admitted, struck her with the hatchet in the bedroom. The prosecutor was thus able to, and did, argue that Sakarias had, in the bedroom, inflicted all three chopping injuries, including the first, antemortem one. Thus Ipsen, in his guilt phase argument, told the jury that Sakarias, in the bedroom, inflicted “three . . . sharp hatchet wounds to the top of Viivi’s head with a tremendous force. . . . ['][] . . . [][] We know that there are in fact three hatchet wounds; the first penetrating the top of the skull, and I know it was the first because it was a hemorrhagic wound, the one in the hairline, the one that chopped the top of her head completely off with the exception of some of the scalp that kept it completely on. [][] We know that when it’s hemorrhagic it means that Viivi, whether conscious or not, still suffered that blow while alive, and we know that the last two in the forehead area being non-hemorrhagic were at a time when her body had ceased to live, or unfortunately actually possibly that the blood flow was not great enough to cause hemorrhage. [!]... [][] Again, Mr. Sakarias indicates he believes he hit her two times with the hatchet when he used the hatchet. Again, by the evidence, he was off by only one blow.”
In the penalty phase argument at Sakarias’s trial, the prosecutor again portrayed Sakarias as having inflicted the antemortem hatchet-blade wound, *149which he characterized as finally causing Viivi’s death. Sakarias’s participation in the crime could not be considered minor, Ipsen argued; he was “as involved in the murder of Viivi Piirisild as one could ask, swinging what I suggest were the blows that actually ended her life.” Referring to Sakarias in the second person, Ipsen argued that if, after the attack in the living room, “you had called 911, realizing what you had done and attempted to save her fife, . . . perhaps you would deserve the pity, the sympathy, perhaps the scales of justice would lean in your direction. [][] . . . [f] If, when you walked back to the back room with that hatchet and thought Viivi Piirisild is still alive, and you must have, otherwise you wouldn’t have gone back there with that hatchet, and if you just simply didn’t chop the top of her head off, as the evidence indicated you did in that back room, thus finally ending her life.”
In addition to the prosecutorial arguments just recited, petitioners also complain of inconsistency in the prosecutor’s penalty phase arguments relating to domination. (See Pen. Code, § 190.3, factor (g) [“substantial domination” by another may be considered in mitigation].) At Waidla’s trial, Ipsen argued Waidla “is not one who is dominated by another, but instead the facts indicate that he was the dominate \sic\ person between himself and Mr. Sakarias, that he was the planner, he was the one who knew of the Piirisild home and knew of the facts surrounding the burglary, the robbery of Mrs. Piirisild.” At Sakarias’s trial, in contrast, Ipsen argued Sakarias was “in no way” dominated by Waidla: “They were separate individuals joined by a common plan, a common hatred, common goals.” Petitioners’ actions in killing Viivi and escaping were those of “a partnership like a right hand and a left hand,” with “absolutely no evidence of domination.”
C. The Habeas Corpus Proceedings
We issued orders to show cause based on petitioners’ allegations that the inconsistent factual presentations outlined above deprived them of fair trials on the question of penalty, in violation of the due process guarantee of the Fourteenth Amendment to the United States Constitution, the Eighth Amendment’s prohibition on cruel and unusual punishment, and article I, section 15 of the California Constitution. In their returns, the People, represented by the Attorney General, admitted Ipsen had argued inconsistent theories of the attack on Viivi Piirisild, but denied he did so intentionally, that he knowingly presented any false evidence or argument at either trial, that he manipulated the evidence to increase each petitioner’s culpability, or that any inconsistencies could have affected the verdicts. The People relied in part on a declaration by Ipsen, executed November 7, 2001, stating that he did not “knowingly present any false evidence or argument to either jury” and did not “intentionally” present inconsistent theories in argument. He did not, in Sakarias’s trial, intentionally omit evidence of the postmortem abrasion or its *150significance, and when he argued to Sakarias’s jury that Sakarias had inflicted the antemortem hatchet blow, Ipsen declared, he “did not remember that I had argued something different to the Waidla jury.”
Petitioners, by their traverses, placed at issue the truth of the People’s denials. We therefore appointed a referee to hear evidence and answer the following factual questions concerning Ipsen’s conduct of the two trials:
1. Was Prosecutor Steven Ipsen’s argument of inconsistent factual theories to the juries in the trials of petitioners Waidla and Sakarias intentional or inadvertent?
2. (a) Did Ipsen believe, at the time of Sakarias’s trial, that the murder victim, Viivi Piirisild, was already dead at the time she was dragged from the living room to the bedroom? (b) Did he have reason to believe Piirisild was dead when moved to the bedroom?
3. At Sakarias’s trial, did Ipsen deliberately refrain from asking the medical examiner, Dr. James Ribe, about a postmortem abrasion on the victim’s back?
4. At Waidla’s trial, did Ipsen refrain from seeking admission of Sakarias’s confession into evidence because it contradicted the factual theory he intended to argue to the Waidla jury?
The referee heard testimony from Ipsen and from the former head of Ipsen’s branch of the district attorney’s office. In addition, the referee admitted and reviewed numerous exhibits, including crime scene and autopsy photographs from the two trials, and took notice of the reporters’ transcripts of the trials. The referee summarized his findings on the submitted questions as follows:
“1. Ipsen’s argument of inconsistent factual theories to the juries in the trials of Waidla and Sakarias was an intentional strategic decision designed to fit the evidence Ipsen presented at the successive trials, to meet the proffered defense theories, and to maximize the portrayal of each defendant’s culpability.
“2. (a) At the time of the Sakarias trial, Ipsen did not believe that Piirisild was already dead when she was dragged from the living room to the bedroom, (b) At the time of the Sakarias trial, Ipsen had strong reason to believe that Piirisild was dead when she was dragged from the living room to the bedroom. Although Ipsen also had some lesser reason to believe she may have been alive, the great weight of the evidence did not support that view. Further, as explained below in Issue No. 3, Ipsen intentionally did not elicit *151testimony from Dr. Ribe about the postmortem abrasion on Piirisild’s back, because the most likely interpretation of the abrasion was inconsistent with the theory of the killing Ipsen presented at Sakarias’ trial.
“3. Ipsen deliberately refrained from asking Dr. Ribe about the postmortem abrasion on Piirisild’s back. He did so to tailor his evidentiary presentation to his changed theory of the hatchet wounds. The most likely explanation of that abrasion would have been inconsistent with the factual theory of the killing he presented in Sakarias’ trial.
“4. Ipsen believed that Sakarias’ confession was inadmissible at Waidla’s trial. For that reason, and not because it contradicted the factual theory he intended to argue to the Waidla jury, he did not offer it against Waidla.”
The parties (and the Los Angeles County District Attorney’s Office, as amicus curiae) have filed postreference briefs on the merits. They take no exception to the referee’s findings, but dispute the legal consequences of those findings.
II. Discussion
A. Review of Referee’s Findings
“The referee’s findings are not binding on us, but are entitled to great weight when supported by substantial evidence. (In re Ross (1995) 10 Cal.4th 184, 201 [40 Cal.Rptr.2d 544, 892 P.2d 1287]; In re Marquez (1992) 1 Cal.4th 584, 603 [3 Cal.Rptr.2d 727, 822 P.2d 435].) Deference to the referee is called for on factual questions, especially those requiring resolution of testimonial conflicts and assessment of witnesses’ credibility, because the referee has the opportunity to observe the witnesses’ demeanor and manner of testifying. (In re Ross, supra, 10 Cal.4th at p. 201; In re Jackson (1992) 3 Cal.4th 578, 585 [11 Cal.Rptr.2d 531, 835 P.2d 371].)” (In re Malone (1996) 12 Cal.4th 935, 946 [50 Cal.Rptr.2d 281, 911 P.2d 468].) Upon review, we find each of the referee’s findings supported by substantial evidence and, like the parties, we accept them.
1. Ipsen’s use of divergent factual theories was intentional
Ipsen testified at the reference hearing, as he stated in his earlier declaration, that his presentation of inconsistent theories was “not intentional.” He noted that in the interval between the trials he probably handled other cases and described himself as an “instinct[ive]” litigator who did not typically follow detailed notes or a script in his examination of witnesses. When he made his closing argument in the Sakarias case, he did not have in mind what *152he had argued to the Waidla jury: “the last thing I’m thinking about when I’m arguing in one trial is trying to remember what I argued in another trial.”
The referee found Ipsen’s claim of inadvertence “unconvincing”: “Despite a lapse of eight months between trials, it is unlikely that a competent and committed prosecutor like Ipsen, handling the severed trials of two defendants jointly charged with capital murder, would simply forget at the second trial what specific factual theory of the gruesome murder he presented at the first. . . . [T]he Waidla and Sakarias trials were Ipsen’s first murder cases, his first death penalty cases. He was depressed about the death verdict in Waidla for approximately two weeks.[1] It is improbable that his factual depiction of the killing in Waidla would have totally escaped his notice in Sakarias. Moreover, the assertion of inadvertence in presenting the inconsistent theories implies a level of carelessness that is simply not present in Ipsen’s prosecution of Sakarias.”
Substantial evidence supports the referee’s conclusion. Ipsen testified at the hearing that he “always,” including at the time of Waidla’s trial, “had a belief that Mr. Sakarias inflicted hatchet wounds in the back room.” Ipsen also testified he probably had the Sakarias statement, which contained that admission, before Waidla’s trial. In addition, Waidla’s statement to police, admitted at his trial, indicated that Sakarias had taken the hatchet into the bedroom after they dragged Viivi Piirisild there. Yet, in argument to the Waidla jury, Ipsen not only did not suggest Sakarias had ever used the hatchet, instead impliedly attributing all such blows to Waidla, but expressly argued Waidla had struck Viivi repeatedly with the blade, inflicting not only the hemorrhagic “death blow” but also the additional “non-hemorrhagic chop wounds to the head.” As Ipsen, according to his testimony, believed at the time that Sakarias had struck some of the hatchet blows, and as Waidla’s statement, which was in evidence, would have supported such an argument (or at least an argument that Sakarias might have struck Viivi with the hatchet in the bedroom), “the inference,” as the referee found, “is clear: Ipsen set aside that belief, and argued that Waidla inflicted all the hatchet wounds, thus enhancing the theory of Waidla’s culpability.”
At Sakarias’s trial, of course, the prosecutor introduced, and relied upon, Sakarias’s confession, which included his account of taking the hatchet into the bedroom and striking Viivi twice with it. But Ipsen also attributed to Sakarias the hemorrhagic, antemortem chopping wound, despite having proven and argued in Waidla’s trial, some months earlier, that Viivi was already dead when moved to the bedroom. As the referee found, on substantial evidence, Ipsen intentionally refrained from putting the same evidence *153before the Sakarias jury. The inference is therefore strong that his argument to the Sakarias jury, that Sakarias had inflicted the hemorrhagic chopping wound, was also intentional.
2. Ipsen had strong reason to believe, while prosecuting Sakarias, that the victim was already dead when moved to the bedroom
Although there were slight grounds for doubt, the referee found, “the great weight of the available evidence” supported the view that Viivi Piirisild died in the living room. The postmortem abrasion, in particular, was best explained as the result of Viivi’s body being dragged across the carpet to the bedroom. While the abrasion could conceivably have had other causes, “[t]he dragging explained the size, nature, and location of the abrasion” and was also consistent with the condition of Viivi’s clothing.
The finding that Ipsen had strong reason to believe Viivi was already dead when moved to the bedroom is supported by substantial evidence. The police detective testified in Sakarias’s trial that among the bloodstains on the living room carpet was a large concentration of blood “consistent with a body lying in that position bleeding for . . . an extended period of time . . . ,” possibly as long as 10 or 15 minutes. In contrast, the detective described no bloodstains on the floor of the bedroom and no large-volume stains at all; rather, the blood in the bedroom, spattered on walls and the ceiling, was in one area “minimal” in volume and in another had “enough substance to actually start to trickle down the wall” but “wasn’t a great amount of blood.” By far the most persuasive explanation for the abrasion on Viivi’s lower back, as the referee stated, was that it was caused by dragging her body to the bedroom. Since the medical examiner, as far as Ipsen knew, had not changed his opinion that the abrasion was incurred after death, Ipsen had no objective grounds on which to abandon his theory, which fit with all this evidence, that petitioners’ initial attack on Viivi Piirisild, in her living room, was fatal.
3. Ipsen deliberately refrained from asking Dr. Ribe about the postmortem abrasion in Sakarias’s trial
The referee, observing that in Sakarias’s trial Ipsen had introduced virtually all the same autopsy photographs as in Waidla’s trial but had omitted exhibit 59K, which showed the abrasion on Viivi Piirisild’s back, concluded Ipsen’s omissions of this exhibit and of questioning regarding the abrasion were deliberate, designed to avoid the presentation of evidence “inconvenient” to his new and different theory of the attack, evidence “much easier to omit than to explain.”
*154We conclude this finding is supported by substantial evidence. Though Ipsen testified he had “no memory of deciding not to introduce that [exhibit],” he agreed it was “odd” that “the one photograph is not in.” To be sure, Ipsen’s examination of Dr. Ribe in Sakarias’s trial was not as long as in Waidla’s; Ipsen testified he was less concerned with the medical and physical evidence in Sakarias’s trial, where Sakarias had confessed to the attack and raised a psychiatric defense, than in Waidla’s. But even in Sakarias’s trial, the examination of Dr. Ribe was methodical, covering each of Viivi Piirisild’s major injuries and including questions as to the hemorrhagic or nonhemorrhagic character of most. In an earlier examination of the police detective, moreover, Ipsen had already presented evidence that the abrasion on Viivi’s body and the condition of the carpeting corroborated Sakarias’s statement that petitioners had dragged her body to the bedroom. As the referee reasoned, “It is highly improbable that Ipsen recognized the significance of the abrasion as evidence that the body was dragged, but forgot the significance of the abrasion as evidence that Piirisild died in the living room before being dragged to the bedroom.” In light of this circumstantial evidence, and given Ipsen’s lack of specific memory as to how he came to omit examination on the postmortem character of the abrasion, the referee could reasonably conclude the omission was deliberate.
4. Ipsen did not offer Sakarias’s confession at Waidla’s trial because he believed it would be inadmissible
Ipsen testified he would have liked to introduce Sakarias’s confession, which implicated Waidla equally, in Waidla’s trial, but assumed it would be subject to a successful objection. “My understanding of the law at the time and still today, is that when I’m prosecuting Mr. Waidla and charging him with murder, I can’t use the statement of his accomplice against him.” At trial before a judge he knew to be highly experienced in criminal law, “If I had tried to get in evidence, which everyone knows is inadmissible and is wrong, I’d look like an idiot to say I’d like to offer the codefendant’s statement.”
The referee accepted Ipsen’s testimony on this point, stating the confession would have been inadmissible under People v. Aranda (1965) 63 Cal.2d 518 [47 Cal.Rptr. 353, 407 P.2d 265] and Bruton v. United States (1968) 391 U.S. 123 [20 L.Ed.2d 476, 88 S.Ct. 1620]. The prosecutor’s failure to offer the statement, the referee found, “did not relate to its inconsistency with the factual theory he intended to present at Waidla’s trial.”
Ipsen’s testimony constitutes substantial evidence supporting the referee’s finding. Though the Aranda/Bruton rule of exclusion applies only to statements of jointly tried codefendants (People v. Brown (2003) 31 Cal.4th 518, 537 [3 Cal.Rptr.3d 145, 73 P.3d 1137]), Ipsen could have reasonably assumed *155that most or all of Sakarias’s confession would be inadmissible in Waidla’s trial. Under California’s hearsay exception for declarations against penal interest (Evid. Code, § 1230), admissibility is limited to the “ ‘specifically disserving’ ” portions of the statement. (People v. Duarte (2000) 24 Cal.4th 603, 612 [101 Cal.Rptr.2d 701, 12 P.3d 1110]; see People v. Leach (1975) 15 Cal.3d 419, 441 [124 Cal.Rptr. 752, 541 P.2d 296].) Thus, Sakarias’s statements that Waidla had Initiated the attack on Viivi Piirisild, struck Viivi with the hatchet as she pleaded for him to stop, called for Sakarias to assist, and later directed him to strike Viivi with the hatchet in the bedroom (see Sakarias, supra, 22 Cal.4th at p. 613) could well have been held inadmissible as attempts to deflect culpability away from the declarant. (See People v. Duarte, supra, at pp. 612-613; see also id. at p. 626 (conc. opn. of Baxter, J.) [6th Amend, confrontation clause “may most often prohibit the use against an accused of directly incriminating statements against him that were made by a nontestifying accomplice while in police custody”].)2 Redaction of the confession to excise those portions would not necessarily have endowed the statement as a whole, made as it was under custodial questioning regarding Sakarias’s role in the homicide, with the indicia of reliability required under Evidence Code section 1230. (See People v. Duarte, supra, at pp. 614-618.) Finally, while the prosecutor might have overcome hearsay and confrontation clause objections to portions of the Sakarias statement solely implicating Sakarias, those portions would have been irrelevant in Waidla’s trial. With this background law, Ipsen’s testimony that he assumed the Sakarias statement would be inadmissible is believable and constituted substantial evidence to support the referee’s findings.
Like the parties, we therefore accept each of the referee’s findings as supported by substantial evidence.
B. The People’s Bad Faith Use of Inconsistent Theories Deprived Sakarias of Due Process, Requiring Vacation of His Death Sentence
Petitioners both claim Ipsen’s inconsistent attribution of the three hatchet-blade blows deprived them of due process. The Attorney General contends the use of inconsistent arguments at separate trials “is permissible provided a prosecutor does not argue something that the prosecutor knows to be false.” For reasons explained below, we conclude that fundamental fairness does not permit the People, without a good faith justification, to attribute to two defendants, in separate trials, a criminal act only one *156defendant could have committed. By doing so, the state necessarily urges conviction or an increase in culpability in one of the cases on a false factual basis, a result inconsistent with the goal of the criminal trial as a search for truth. At least where, as in Sakarias’s case, the change in theories between the two trials is achieved partly through deliberate manipulation of the evidence put before the jury, the use of such inconsistent and irreconcilable theories impermissibly undermines the reliability of the convictions or sentences thereby obtained. In short, in the absence of a good faith justification, “[c]ausing two defendants to be sentenced to death by presenting inconsistent arguments in separate proceedings . . . undermines the fairness of the judicial process and may precipitate inappropriate results.” (Poulin, Prosecutorial Inconsistency, Estoppel, and Due Process: Making the Prosecution Get Its Story Straight (2001) 89 Cal. L.Rev. 1423, 1425 (hereafter Prosecutorial Inconsistency).)
We also conclude, however, that where, as here, the available evidence points clearly to the truth of one theory and the falsity of the other, only the defendant against whom the false theory was used can show constitutionally significant prejudice. For that reason, we conclude that Sakarias, but not Waidla, is entitled to relief on his petition.
1. The People may not convict two individuals of a crime only one could have committed or obtain harsher sentences against two individuals by unjustifiably attributing to each a culpable act only one could have committed
Judicial disapproval of the state’s use of inconsistent and irreconcilable theories in separate trials for the same crimes was first articulated in opinions by individual Supreme Court and lower federal court judges. (See Jacobs v. Scott (1995) 513 U.S. 1067 [130 L.Ed.2d 618, 115 S.Ct. 711] (dis. opn. of Stevens, J., from denial of stay) [fundamentally unfair to execute a person “on the basis of a factual determination that the State has formally disavowed” in coperpetrator’s later trial]; Drake v. Kemp (11th Cir. 1985) 762 F.2d 1449, 1479 (conc. opn. of Clark, J.) [prosecutor’s “flip flopping of theories of the offense was inherently unfair”].) Drawing on these separate opinions, several federal courts have since held that a prosecutor’s inconsistent argument in two defendants’ separate trials attributing the same criminal or culpability-increasing act to each defendant denies the defendants fundamentally fair trials.
In Thompson v. Calderon (9th Cir. 1997) 120 F.3d 1045 (Thompson), reversed on other grounds sub nomine Calderon v. Thompson (1998) 523 U.S. 538 [140 L.Ed.2d 728, 118 S.Ct. 1489], a majority of the en banc court held that inconsistent prosecutorial theories may present a due process violation. *157There, Leitch and Thompson were both charged with raping and killing Ginger Fleischli. At their joint preliminary hearing and at Leitch’s trial, the prosecutor introduced and relied on evidence, including testimony by jailhouse informants recounting statements by Thompson, that indicated the two defendants had acted together, killing Fleischli because she was interfering with Leitch’s efforts to reconcile with his ex-wife. (Thompson, supra, at pp. 1055-1056 (plur. opn. of Fletcher, J.).) At Thompson’s trial (held before Leitch’s), however, the same prosecutor had introduced and relied upon other evidence, to the effect that Thompson alone killed Fleischli to prevent her reporting that he had raped her. (Id. at p. 1056.) The prosecutor thus “asserted as the truth before Thompson’s jury the story he subsequently labeled absurd and incredible in Leitch’s trial.” (Id. at p. 1057.)
The Thompson plurality concluded that “when no new significant evidence comes to light a prosecutor cannot, in order to convict two defendants at separate trials, offer inconsistent theories and facts regarding the same crime.” (Thompson, supra, 120 F.3d at p. 1058.) Three of the 11 judges participating fully joined with Judge Fletcher in her opinion on this point. (Id. at p. 1047.) Two more, in a concurring opinion by Judge Tashima, agreed that prosecutorial use of wholly inconsistent theories violates due process (id. at p. 1063), but believed that Thompson’s entitlement to relief depended on whether he was prejudiced, which in turn required a determination “which of the two inconsistent theories pursued by the prosecutor represents the tme facts and which is false” (id. at p. 1064).
In Smith v. Groose (8th Cir. 2000) 205 F.3d 1045 (Smith), members of two criminal groups who had separately burglarized the same house during overlapping periods of the same day were tried separately for the murder of the occupants. At the trials of both Cunningham, a member of the first group of burglars, and Smith, a member of the second, Lytle, also a member of the second group, testified the occupants were killed by Cunningham’s group. In one prior statement to police, Lytle had attributed the killings to a member of his own group, Bowman, while in another he said, consistently with his trial testimony, that Cunningham’s group had killed the occupants. At Smith’s trial, the prosecutor used Lytle’s prior statement implicating Bowman in the killings, arguing to the jury that Smith, Bowman’s accomplice in burglary, was guilty of felony murder. Later, at Cunningham’s trial, the prosecutor relied on Lytle’s testimony, introduced his prior consistent statement to police, did not introduce his prior inconsistent statement, and objected to defense efforts to impeach him. (Id. at pp. 1047-1048, 1050.) “In short, what the State claimed to be true in Smith’s case it rejected in Cunningham’s case, and vice versa.” (Id. at p. 1050.)
The Smith court concluded, “the use of inherently . . . contradictory theories violates the principles of due process” (Smith, supra, 205 F.3d at *158p. 1052), for “[t]he State’s duty to its citizens does not allow it to pursue as many convictions as possible without regard to fairness and the search for truth” (id. at p. 1051; see also United States v. Butner (W.D.Mo. 2000) 2000 WL 1842410, *15-17 [following Smith]).
Recently, the Sixth Circuit Court of Appeals reached the same conclusion in Stumpf v. Mitchell (6th Cir. 2004) 367 F.3d 594, certiorari granted sub nomine Mitchell v. Stumpf (2005) 543 U.S. 1042 [160 L.Ed.2d 610, 25 S.Ct. 824] (Stumpf). Stumpf and his accomplice, Wesley, robbed and killed a couple, the Stouts, in their home. That Stumpf shot Mr. Stout was undisputed, but whether he also shot Mrs. Stout or Wesley did so with Stumpf’s handgun was unclear. At Stumpf’s plea hearing, the prosecutor argued that since both victims were shot with the same weapon, the evidence showed Stumpf must have killed both victims “ ‘in order not to leave anyone available to identify him.’ ” (Id. at p. 613.) But at Wesley’s later trial, the prosecutor introduced a jail informant’s testimony that Wesley had confessed to picking up Stumpf s handgun and shooting Mrs. Stout. On that basis, the prosecutor argued Stumpf had left the room after shooting Mr. Stout, whereupon Wesley, “ ‘whose own gun was jammed, picked that chrome colored Raven up and as Mrs. Stout sat helplessly on her bed, shot her four times in order to leave no witnesses to the crime.’ ” (Ibid.; id. at pp. 596-598.)
The appellate court concluded, “the use of inconsistent, irreconcilable theories to convict two defendants for the same crime is a due process violation.” (Stumpf, supra, 367 F.3d at p. 611.) The vice rests in the fact that of two inconsistent and irreconcilable theories, one must be false: “Because inconsistent theories render convictions unreliable, they constitute a violation of the due process rights of any defendant in whose trial they are used.” (Id. at p. 613.) In Stumpf, the state had clearly used such irreconcilable theories, for “[a]t each proceeding, the prosecutor argued that the defendant had been the one to pull the trigger, resulting in the fatal shots to [Mrs.] Stout.” (Ibid.)
These courts and judges have found a prosecutor’s 180-degree change in theory “deeply troubling” (Jacobs v. Scott, supra, 513 U.S. at p. 1069), in part because by taking a formal position inconsistent with the guilt or culpability of at least one convicted defendant, the government, through the prosecutor, has cast doubt on the factual basis for the conviction. “If the prosecutor’s statements at the Hogan trial were correct, then Jacobs is innocent of capital murder.” (Ibid.) “The conclusion seems inescapable that the prosecutor obtained Henry Drake’s conviction through the use of testimony he did not believe . . . .” (Drake v. Kemp, supra, 762 F.2d at p. 1479.) “The prosecutor ... at Leitch’s trial essentially ridiculed the theory he had used to obtain a conviction and death sentence at Thompson’s trial.” (Thompson, supra, 120 F.3d at p. 1057.) As both of two irreconcilable *159theories of guilt cannot be true, “inconsistent theories render convictions unreliable.” (Stumpf, supra, 367 F.3d at p. 613.)
Because it undermines the reliability of the convictions or sentences, the prosecution’s use of inconsistent and irreconcilable theories has also been criticized as inconsistent with the principles of public prosecution and the integrity of the criminal trial system. A criminal prosecutor’s function “is not merely to prosecute crimes, but also to make certain that the truth is honored to the fullest extent possible during the course of the criminal prosecution and trial.” (United States v. Kattar (1st Cir. 1988) 840 F.2d 118, 127.) His or her goal must be “not simply to obtain a conviction, but to obtain a fair conviction.” (Brown v. Borg (9th Cir. 1991) 951 F.2d 1011, 1015, italics omitted.) “Although the prosecutor must prosecute with earnestness and vigor and ‘may strike hard blows, he is not at liberty to strike foul ones.’ ” (Smith, supra, 205 F.3d at p. 1049, quoting Berger v. United States (1935) 295 U.S. 78, 88 [79 L.Ed. 1314, 55 S.Ct. 629]; see also ABA Model Code Prof. Responsibility, EC 7-13 [“The responsibility of a public prosecutor differs from that of the usual advocate; his duty is to seek justice, not merely to convict”].)
For the government’s representative, in the grave matter of a criminal trial, to “chang[e] his theory of what happened to suit the state” is unseemly at best. (Drake v. Kemp, supra, 762 F.2d at p. 1479.) “The state cannot divide and conquer in this manner. Such actions reduce criminal trials to mere gamesmanship and rob them of their supposed purpose of a search for truth.” (Ibid.) Thus, even a court that did not believe inconsistent positions, by themselves, to be constitutional error found it “disturbing to see the Justice Department change the color of its stripes to such a significant degree . . . depending on the strategic necessities of the separate litigations.” (United States v. Kattar, supra, 840 F.2d at p. 127; see also Thompson, supra, 120 F.3d at p. 1072 (dis. opn. of Kozinski, J.) [prosecutor’s use of inconsistent factual theories “surely does not inspire public confidence in our criminal justice system”].)
We have previously indicated that an inconsistent prosecutorial argument “made in bad faith” could be misconduct, and conversely that such argument was not improper if “based on the record and made in good faith” (People v. Farmer (1989) 47 Cal.3d 888, 923 [254 Cal.Rptr. 508, 765 P.2d 940] (Farmer)), though we did not have occasion in that case to deal more definitively with the problem. With the issue more squarely before us here, we hold that the People’s use of irreconcilable theories of guilt or culpability, unjustified by a good faith justification for the inconsistency, is fundamentally unfair, for it necessarily creates the potential for—and, where prejudicial, actually achieves—a false conviction or increased punishment on a false *160factual basis for one of the accuseds.“The criminal trial should be viewed not as an adversarial sporting contest, but as a quest for truth.” (United States v. Kattar, supra, 840 F.2d at p. 127.)
By intentionally and in bad faith seeking a conviction or death sentence for two defendants on the basis of culpable acts for which only one could be responsible, the People violate “the due process requirement that the government prosecute fairly in a search for truth . . . .” (Smith, supra, 205 F.3d at p. 1053.) In such circumstances, the People’s conduct gives rise to a due process claim (under both the United States and California Constitutions) similar to a claim of factual innocence. Just as it would be impermissible for the state to punish a person factually innocent of the charged crime, so too does it violate due process to base criminal punishment on unjustified attribution of the same criminal or culpability-increasing acts to two different persons when only one could have committed them. In that situation, we know that someone is factually innocent of the culpable acts attributed to both. (See Prosecutorial Inconsistency, supra, 89 Cal. L.Rev. at p. 1425 [“When the prosecution advances a position in the trial of one defendant and then adopts an inconsistent position in the trial of another on the same facts, the prosecution is relying on a known falsity”].)
2. The People unjustifiably used inconsistent and irreconcilable theories to obtain a death sentence against Sakarias
Prosecutor Ipsen attributed first to Waidla alone and later to Sakarias alone, in their respective trials, a series of blows to the victim’s head with the hatchet blade. These two theories are irreconcilable; that Waidla alone inflicted each of these wounds, as the prosecutor maintained at his trial, and that Sakarias alone also did so, as the prosecutor maintained at his trial, is not possible. One or the other theory (or both, if each man inflicted some but not all of the wounds) must be false.
The acts attributed to both Waidla and Sakarias in turn were not necessary to establish their guilt of first degree murder (Pen. Code, § 189) or the truth of the charged robbery- and burglary-murder special circumstances (id., § 190.2, subd. (a)(17)). But the prosecutor attributed the three hatchet-edge blows to each defendant in turn in order to establish an aggravating circumstance of the crime (id., § 190.3) on the basis of which the jury was urged to sentence each defendant to death. At least where the punishment involved is death, due process is as offended by the People’s inconsistent and irreconcilable attribution of culpability-increasing acts as by the inconsistent and irreconcilable attribution of crimes. (See Jacobs v. Scott, supra, 513 U.S. at p. 1070 [heightened need for reliability in capital cases “underscores the gravity” of prosecutorial inconsistency]; Prosecutorial Inconsistency, supra, *16189 Cal. L.Rev. at pp. 1468-1470 [consistency as to both defendants’ guilt of crime does not justify inconsistency as to culpability, where pertinent to capital sentence].) We cannot accept the dissent’s apparent view (see conc. & dis. opn. of Baxter, J., post, at p. 179) that the state may seek and obtain death sentences for two defendants using inconsistent or irreconcilable factual theories that it could not use to obtain convictions against the same two.
Because Ipsen used different attributions of the chopping wounds to argue each petitioner should receive the death penalty, his factual theories were significantly inconsistent and irreconcilable. The present case is thus critically distinguishable from those in which the prosecutor’s theories were held fundamentally consistent because any variation did not concern a fact used to convict the defendant or increase his or her punishment. (See Nguyen v. Lindsey (9th Cir. 2000) 232 F.3d 1236, 1240-1241 [variation in prosecutorial argument as to which of two gangs fired the first shot in a gun battle that killed a bystander not significant where prosecutor at both trials pursued the same “underlying theory” that all participants in the gang battle were equally responsible for the death].)3 Unlike those cases, here Ipsen’s underlying theory of why petitioners each deserved the death penalty was not the same in the two cases; in Waidla’s case, it included Waidla’s supposed striking of all three blows with the hatchet blade, while in Sakarias’s case it included Sakarias’s striking the same three blows.4
*162We turn to the question of justification. As observed earlier, in Farmer, supra, 47 Cal.3d at page 923, we suggested inconsistent prosecutorial argument was not improper if made in “good faith.” We did not, however, explicate in Farmer the concepts of good and bad faith in the context of prosecutorial inconsistency.
The Thompson plurality suggested, a prosecutor’s change in theories could be justified where “new significant evidence comes to light” between the trials. (Thompson, supra, 120 F.3d at p. 1058.) We agree a significant change in the available evidence might, under some circumstances, warrant the use of an inconsistent prosecutorial theory in a subsequent trial.5 Here, one difference in evidence between the two trials was the introduction at Waidla’s trial, but not at Sakarias’s, of Dr. Ribe’s testimony that the abrasion on the victim’s back was nonhemorrhagic and therefore appeared to have been inflicted after death. But Ipsen’s deliberate strategic choice in Sakarias’s trial not to examine Dr. Ribe regarding the abrasion on the victim’s lower back, as he had done a few months earlier at Waidla’s trial, plainly cannot establish Ipsen’s good faith or otherwise justify the use of irreconcilable theories. To the contrary, such manipulation of the evidence for the purpose of pursuing inconsistent theories establishes the prosecutor’s bad faith. Indeed, as a commentator has remarked, cases in which a prosecutor’s use of inconsistent theories in successive trials reflects a deliberate change in the evidence presented are particularly clear violations: “In extreme cases, the prosecutorial inconsistency signals intentional manipulation by the prosecution and is readily characterized as a violation of due process.” (Prosecutorial Inconsistency, supra, 89 Cal. L.Rev. at p. 1474; see, e.g., Smith, supra, 205 F.3d at p. 1051 [prosecution’s “manipulation of the evidence” enabled it to use factually contradictory theories].) Ipsen’s deliberate omission of evidence for the purpose of making possible his use of inconsistent and irreconcilable theories makes Sakarias’s trial such an “extreme case[].” (Prosecutorial Inconsistency, supra, 89 Cal. L.Rev. at p. 1474; cf. ABA Model Code Prof. Responsibility, EC 7-13 [“Further, a prosecutor should not intentionally avoid pursuit of evidence merely because he believes it will damage the prosecution’s case or aid the accused”].)6
*163The referee found, on substantial evidence, that Ipsen omitted questioning about the back abrasion in order to avoid presenting evidence “inconvenient” to his new theory that Sakarias had inflicted all three chopping wounds to the victim’s head. Such intentional manipulation of the evidence was instrumental to, and cannot justify, the prosecutor’s use of irreconcilable theories. The inconsistent argument in Sakarias’s trial was not made with the “good faith” to which we alluded in Farmer, supra, 47 Cal.3d at page 923.
The dissenting opinion finds no indication of bad faith in Ipsen’s conduct because, where the information available to the prosecutor is of public record or has been disclosed to the defense, “the People would not generally be required to introduce, in their own case, evidence helpful to the defense.” (Conc. & dis. opn. of Baxter, J., post, at p. 172.) We agree no due process violation arises simply from a prosecutor’s failure to introduce evidence favorable to the defense. But where, as here, a prosecutor who seeks convictions or death sentences against two individuals through inconsistent and irreconcilable factual theories deliberately omits in one trial evidence used in the other, so as to make possible the argument of the inconsistent theories, the prosecutor’s manipulation of evidence does show that the inconsistent theories were not pursued in good faith. The People, therefore, deprived Sakarias of due process by unjustifiably using inconsistent and irreconcilable factual theories to obtain a death sentence against him. Whether that conduct was prejudicial must still be determined. (See pt. II.B.3, post.)7
Whether the prosecutor can be said to have manipulated the evidence at Waidla’s trial is less clear. Sakarias’s confession was not introduced at Waidla’s trial because Ipsen assumed it would be ruled inadmissible—a realistic assumption. At oral argument, counsel for Waidla asserted that Ipsen had deliberately failed to introduce at Waidla’s trial crime scene evidence, which he did introduce at Sakarias’s later trial, regarding blood spatters in the bedroom. But this evidence did not, in light of the other physical evidence, strongly suggest the victim was killed in the bedroom, and the record does not establish whether Ipsen omitted it at Waidla’s trial in order to avoid such an implication. We need not decide whether the prosecutor acted in bad faith *164at Waidla’s trial, however, because, as we conclude below, the probably false aspects of the argument Ipsen made in seeking the death penalty against Waidla were not reasonably likely to have influenced the verdict. (See pt. II.B.4, post)
3. The People’s use of inconsistent and irreconcilable theories was prejudicial as to Sakarias
The prejudice question is in these circumstances a complex one, involving two questions as to each petitioner and each culpability-increasing act inconsistently attributed to petitioners: for each petitioner we must ask, first, whether the People’s attribution of the act to the petitioner is, according to all the available evidence, probably false or probably true, and, second, whether any probably false attribution of a culpability-increasing act to the petitioner could reasonably have affected the penalty verdict.
As previously explained, the prosecutor’s unjustified use of inconsistent and irreconcilable factual theories to convict two people of a crime only one could have committed, or to obtain harsher sentences for both on the basis of an.act only one could have committed, violates due process because in those circumstances the state has necessarily convicted or sentenced a person on a false factual basis. It follows that where the probable truth of the situation can be determined—where we are able to say which of the prosecution theories was likely true and which false—only the defendant prejudiced by the false attribution is entitled to relief. (See Thompson, supra, 120 F.3d at p. 1064 (conc. opn. of Tashima, J.) [“To reach a conclusion of prejudice or no prejudice as to Thompson first requires a finding of which of the two inconsistent theories pursued by the prosecutor represents the true facts and which is false”]; Prosecutorial Inconsistency, supra, 89 Cal. L.Rev. at p. 1478 [in assessing entitlement to relief for inconsistent prosecutorial theories, “courts should ask whether the correct position on the evidence can be determined from the evidence available. If the correct position is ascertainable, the court should rule accordingly”].) Only as to the defendant convicted or sentenced by use of the probably false theory can it be said the prosecution has presented a materially false picture of the defendant’s culpability.
We need not decide here what result obtains when the likely truth of the prosecutor’s inconsistent theories cannot be determined, for the case at bench is not one of ambiguous or inconclusive evidence.8 As the referee found, the *165great weight of available evidence indicates that Viivi Piirisild was dead or near death when dragged into the bedroom and thus that Waidla, rather than Sakarias, struck the antemortem, hemorrhagic hatchet-blade blow. True, some evidence—blood spatters in the bedroom—suggests the possibility of an antemortem blow being struck in that room. But that evidence is overwhelmed by other evidence that Viivi’s hemorrhagic wounds, including the hemorrhagic chopping wound Ipsen characterized as the “death blow,” were inflicted during petitioners’ initial attack on her in the living room, including the large pool of blood in the living room, the minimum quantity of blood on the bedroom walls and ceiling, the lack of blood on the bedroom floor, Sakarias’s statement that petitioners felled Viivi in the living room before eventually dragging her into the bedroom, and the nonhemorrhagic character of the abrasion on Viivi’s back. As both petitioners’ statements have Waidla using the hatchet during that initial attack and Sakarias the knife, and as no evidence at all suggests the two exchanged weapons during the initial attack, the inescapable inference from all the available evidence is that Waidla inflicted the hemorrhagic chopping wound to Viivi’s head.
To the extent the false attribution of the antemortem hatchet-blade blow to Sakarias was potentially material to the penalty decision, it deprived Sakarias of a fair penalty trial and entitles him to relief. Sakarias and the Attorney General agree this aspect of prejudice should be tested on the “reasonable likelihood” standard applicable to the knowing presentation of false evidence, which is equivalent to the “harmless beyond a reasonable doubt” test of Chapman v. California (1967) 386 U.S. 18 [17 L.Ed.2d 705, 87 S.Ct. 824], (See United States v. Agurs (1976) 427 U.S. 97, 103 [49 L.Ed.2d 342, 96 S.Ct. 2392]; In re Malone, supra, 12 Cal.4th at p. 977; In re Jackson, supra, 3 Cal.4th at pp. 597-598.) Because the prosecutor intentionally used an inconsistent and probably false theory to obtain a death sentence against Sakarias, we agree with the parties that Sakarias is entitled to relief if he can show a reasonable likelihood the prosecutor’s use of the tainted factual theory affected the penalty verdict. (Accord, United States v. Kattar, supra, 840 F.2d at p. 128; Prosecutorial Inconsistency, supra, 89 Cal. L.Rev. at p. 1471.)
Aside from attributing the hemorrhagic chopping wound to Sakarias, the prosecutor introduced and relied upon other significant aggravating circumstances. Sakarias undisputedly played a direct role in the brutal, unprovoked killing of Viivi Piirisild. The uncontroverted evidence showéd that Sakarias stabbed Viivi four times in the chest, including two potentially fatal wounds passing through vital organs, and that he later took the hatchet, went to the bedroom, and struck her at least twice in the head with the hatchet blade. Sakarias had a loaded handgun when arrested and later was found in *166possession of shanks in the county jail (for use, he said, against gang members who had robbed him). He also made statements during trial indicating a lack of remorse for killing Viivi and suggesting that he and Waidla had intended to kill Avo Piirisild as well. (Sakarias, supra, 22 Cal.4th at pp. 614—616.)
Other considerations, however, make it impossible for us to conclude beyond a reasonable doubt that the prosecutorial argument that Sakarias struck all the hatchet-blade blows, including the first, antemortem one, played no role in the penalty decision. The first hatchet-blade wound was especially severe and was described in gruesome detail by the medical examiner on direct examination by the prosecutor. The path of the blade was parallel to the top of the head, straight up and down as if the victim was lying on the floor. The blade entered on the upper forehead, penetrated through the scalp and the skull bone, then hit the inside of the skull. Dr. Ribe believed the blade then “bounced” off the bone and continued to the rear and right of the victim’s head, leaving another incision. The blow’s force not only penetrated the front of the skull but fractured the back as well, pushing it backward. As a result, a portion of the upper skull and scalp were partially detached from the rest of the head, forming a flap that could be easily folded back. Because of the strength of an adult human’s skull, Dr. Ribe believed a “tremendous amount of force,” as much as an average man could exert swinging the hatchet “very hard,” was needed to cause this wound.
In the guilt phase argument, the prosecutor discussed in detail Sakarias’s attack on Viivi Piirisild with the hatchet, stating he went into the bedroom “to strike a few more blows, to make sure that Viivi was dead in case the stabbing and the bludgeoning weren’t enough. [j[] We know that there were three, in this series of blows, sharp hatchet wounds to the top of Viivi’s head with a tremendous force. ... [1] And it was with this strength that Peter Sakarias swung this hatchet to penetrate this skull, to reach that most vital organ . . . . . [j[] We know that there are in fact three hatchet wounds; the first penetrating the top of the skull, and I know it was the first because it was a hemorrhagic wound, the one in the hairline, the one that chopped the top of her head completely off with the exception of some of the scalp that kept it completely on. [][] . . . [f] We know that this last series of chop wounds . . . was consistent with the last three blows she received.” In the penalty argument, the prosecutor twice again asserted that Sakarias had inflicted all the chopping wounds, “swinging what I suggest were the blows that actually ended her life.” Sakarias, according to Ipsen, “simply . . . chop[ped] the top of her head off, as the evidence indicated you did in that back room, thus finally ending her life.”
As to mitigation, Sakarias was young (21 years old) at the time of the offense, had no record of violence, and had suffered persecution in the Soviet *167Army. He was diagnosed with schizo-affective disorder, characterized by paranoia and bipolar affect; the jury was informed that he had been found incompetent to stand trial in May 1990 and spent almost a year in a state hospital. The defense also played tapes of Sakarias’s mother, father, and friends, recorded in Estonia, describing his childhood and youth. (Sakarias, supra, 22 Cal.4th at pp. 614—616.)
Some aspect or aspects of the case evidently gave one or more jurors considerable pause in the sentencing decision, as the penalty jury deliberated for more than 10 hours over three days and, at one point, declared itself unable to reach a unanimous verdict, before finally returning a verdict of death.
In light of the prominence the prosecutor gave the antemortem chopping wound, treating it as the final, fatal wound, and the likely impact the medical examiner’s description of that wound and the force necessary to inflict it would have had, that the prosecutor’s attribution of that blow to Sakarias had an effect on the penalty verdict is reasonably likely. (United States v. Agurs, supra, 427 U.S. at p. 103.) Though Sakarias’s undisputed conduct in stabbing Viivi Piirisild and stealing her property played, no doubt, a major role in the jury’s decision, we cannot conclude beyond a reasonable doubt that the jury’s decision would have been the same had it not also been told that Sakarias finally ended Viivi’s life by swinging a hatchet with all his strength, nearly cutting off the top of her head. The prosecutor’s intentional and unjustified argument, inconsistent with the factual theory and evidence he had presented at Waidla’s trial, that Sakarias struck the antemortem hatchet-blade blow was therefore prejudicial to Sakarias as to penalty.9
4. The People’s use of inconsistent and irreconcilable theories was harmless as to Waidla
Our conclusion is necessarily different as to Waidla. As discussed earlier, and as the referee found, the great weight of the evidence available—the statements of both petitioners, the physical crime scene evidence, and the medical examiner’s expert testimony—tended to show that Waidla wielded the hatchet in the initial attack, that the first chopping wound was inflicted before Viivi Piirisild’s death, and that Viivi died in her living room from the initial attack before being dragged to the back bedroom. Ipsen’s argument in Waidla’s trial that Waidla struck the first, antemortem blow with the hatchet blade, therefore, was likely true.
*168Waidla points to Sakarias’s statement that he struck Viivi with the hatchet in the bedroom. But Sakarias confessed to hitting Viivi with the hatchet only twice, which left open the possibility—a probability under the other evidence—that the first of the three chopping wounds was inflicted by Waidla during the initial attack. To be sure, in his arguments to the Waidla jury Ipsen suggested that Waidla had inflicted all the chop wounds, including the two postmortem or perimortem wounds that Ipsen had strong reason to believe were actually attributable to Sakarias. But this apparently false argument, inconsistent with the arguments made in Sakarias’s trial, was not prejudicial to Waidla as to penalty, even under the reasonable likelihood standard of United States v. Agurs, supra, 427 U.S. at page 103. To the extent he focused on particular blows, the impact of the prosecutor’s arguments derived from the hemorrhagic hatchet-blade wound, the chop “through the top of her skull” or “death blow,” rather than from the fact that “further blows were struck after she was dead.”
Waidla admitted that after burglarizing the home of Viivi Piirisild, an older woman who had been his benefactor, he hit her with the back of a hatchet he had stolen from her vacation cabin. The medical examiner opined that these blunt force wounds, which fractured several bones and knocked out Viivi’s teeth, contributed to her death. The evidence, moreover, was strong that Waidla, during this attack, turned the hatchet around and struck Viivi with the sharp blade with such force as to penetrate her skull and cut a flap of skull and scalp from the top of her head. Even balanced against Waidla’s youth, lack of a violent criminal record, and personal history of brutalization in the Soviet Army (Waidla, supra, 22 Cal.4th at pp. 706, 712), the circumstances of the crime offered a compelling case in aggravation. More to the point, the case would not have been made significantly less compelling by a prosecutorial concession that Sakarias may have inflicted the two later chop wounds, after Viivi had expired. As the Attorney General remarks, “Under the circumstances, it would have made no difference to the jury whether Waidla inflicted the nonhemorrhagic chopping wounds or handed the weapon to Sakarias so he could inflict them, particularly since the wounds were likely postmortem.”10 We conclude beyond a reasonable doubt that the prosecutor’s attribution of these two wounds to Waidla, though likely false, did not affect Waidla’s penalty and does not entitle him to relief.11
*169C. Claims o/Miranda Error Are Cognizable on Habeas Corpus
In his petition, Waidla claims error in the trial court’s denial of his motion to suppress his custodial statements to police on grounds they were taken in violation of Miranda and related decisions. (Miranda, supra, 384 U.S. 436.) We included the question of the cognizability of such claims on habeas corpus in our order to show cause.
The Attorney General posits two reasons Miranda claims should be held noncognizable on habeas corpus. First, he urges that “since Miranda claims are based on the appellate record, they should be litigated on direct appeal, not on habeas corpus.” We agree a Miranda claim presented on habeas corpus but based solely on the appellate record should generally be denied on procedural grounds. Where the claim was already raised and rejected on the direct appeal, we will ordinarily decline to examine it again. (In re Harris, supra, 5 Cal.4th at p. 825, citing In re Waltreus, supra, 62 Cal.2d at p. 225.) When the issue could have been, but was not, raised on appeal, the unjustified failure to present it on appeal generally precludes its consideration on habeas corpus. (In re Harris, supra, at p. 829, citing In re Dixon, supra, 41 Cal.2d at p. 759.) But a Miranda claim based substantially on facts outside the appellate record would not fall within these procedural rules, as it could not have been adequately presented on direct appeal. The Attorney General’s first argument, therefore, does not support a general rule of noncognizability.
Second, the Attorney General maintains Miranda challenges should be limited to direct appeal because the cost to the state interest in finality of judgments incurred by allowing a collateral challenge is not balanced by any gain in “assuring trustworthy evidence.” The premise of this argument is that the Miranda rule, like the Fourth Amendment exclusionary rule, serves only to deter unconstitutional government conduct and not to protect the innocent from being falsely convicted. (See In re Sterling (1965) 63 Cal.2d 486, 487 [47 Cal.Rptr. 205, 407 P.2d 5] [refusing to entertain search and seizure issue on habeas corpus, in part because “ ‘the use of illegally seized evidence carries with it no risk of convicting an innocent person’ ”]; cf. Stone v. Powell (1976) 428 U.S. 465, 489-495 [49 L.Ed.2d 1067, 96 S.Ct. 3037] [holding state prisoners may not raise search and seizure issues in federal habeas corpus petitions, in part because the cost, in loss of trustworthy evidence, of applying exclusionary rule would not be balanced by slight increase in deterrence from allowing collateral attack].)
*170We find the analogy inapt. As the Supreme Court explained in rejecting the same argument, Miranda, unlike the Fourth Amendment exclusionary rule, “safeguards ‘a fundamental trial right.’ ” (Withrow v. Williams (1993) 507 U.S. 680, 691 [123 L.Ed.2d 407, 113 S.Ct. 1745].) “By bracing against ‘the possibility of unreliable statements in every instance of in-custody interrogation,’ Miranda serves to guard against ‘the use of unreliable statements at trial.’ ” (Id. at p. 692.) While a statement taken in violation of Miranda is not necessarily involuntary, the presence or absence of Miranda advisements and waivers is one circumstance to be considered in evaluating voluntariness. (Id. at pp. 693-694; see, e.g., People v. Memro (1995) 11 Cal.4th 786, 827 [47 Cal.Rptr.2d 219, 905 P.2d 1305].) Other circumstances being equal, a custodial confession taken without Miranda advisements or despite the suspect’s invocation of his or her Miranda rights is less likely to be trustworthy than one preceded by advisements and knowing waivers.
The Attorney General cites People v. Hill (1973) 9 Cal.3d 784 [109 Cal.Rptr. 93, 512 P.2d 317], in which we rejected a collateral attack (presented on appeal from a penalty retrial) on the guilt judgment, made on the basis that admission of the defendant’s confession violated Escobedo v. Illinois (1964) 378 U.S. 478 [12 L.Ed.2d 977, 84 S.Ct. 1758] (deprivation of right to counsel at interrogation). We observed, first, that the claim had already been rejected in the defendant’s previous appeal. (People v. Hill, supra, at p. 786; see People v. Hill (1967) 66 Cal.2d 536, 552-554 [58 Cal.Rptr. 340, 426 P.2d 908].) We went on to say that where the defendant had a full opportunity to litigate the issue and the judgment was long final we would not entertain a collateral attack “based upon an issue which does not relate to the guilt or innocence of the defendant.” (People v. Hill, supra, 9 Cal.3d at p. 787.)
As already explained, the rule on habeas corpus is consistent with our reasoning in People v. Hill, supra, 9 Cal.3d at page 786, in that we ordinarily will not entertain a habeas corpus claim already raised and rejected on direct appeal. (In re Harris, supra, 5 Cal.4th at p. 825.) To the extent our further characterization of a claim under Escobedo v. Illinois, supra, 378 U.S. 478, as unrelated to guilt or innocence is inconsistent with the high court’s (Withrow v. Williams, supra, 507 U.S. at p. 692) and our own characterization of a Miranda claim as one relating to the reliability of the evidence of guilt, we disapprove our former remark.
We therefore reject the Attorney General’s suggestion of a blanket rule of noncognizability for Miranda claims on habeas corpus, but observe that where such claims are based on the appellate record they will ordinarily be denied on the procedural ground that they were, or could have been, presented on direct appeal. (In re Harris, supra, 5 Cal.4th at pp. 824-829.) In *171addition, a Miranda claim that could not be presented on appeal because the defendant did not raise it at trial , would also ordinarily be barred on habeas corpus. (See In re Seaton (2004) 34 Cal.4th 193, 199-200 [17 Cal.Rptr.3d 633, 95 P.3d 896] [claim forfeited on appeal because it was not raised at trial is also barred on habeas corpus, unless facts essential to the claim were not and could not reasonably have been known at trial].)
III. Disposition
Our orders to show cause were limited to the issues discussed in this opinion; petitioners’ other claims will be resolved by separate orders, as is our practice. (See In re Scott (2003) 29 Cal.4th 783, 829 [129 Cal.Rptr.2d 605, 61 P.3d 402].)
The order to show cause as to Waidla’s petition is discharged.
Sakarias’s petition for a writ of habeas corpus is granted insofar as it claims prosecutorial inconsistency material to the penalty verdict, and the judgment of the Los Angeles County Superior Court in People v. Peter Sakarias, No. A711340, therefore is vacated insofar as it imposes a sentence of death. Upon finality of our opinion, the Clerk of the Supreme Court shall remit a certified copy of the opinion and order to the Los Angeles County Superior Court for filing, and respondent Attorney General shall serve another copy thereof on the prosecuting attorney in conformity with Penal Code section 1382, subdivision (a)(2). (See In re Jones (1996) 13 Cal.4th 552, 588 [54 Cal.Rptr.2d 52, 917 P.2d 1175].)
George, C. 1, Kennard, 1, Chin, 1, Brown L, and Moreno, L, concurred.
1 Ipsen testified he was depressed “that this whole thing had to happen . . . [t]he destruction of the families and everything.”
Recently, admission in criminal cases of “testimonial” statements by declarants not subject to cross-examination has been held generally violative of the confrontation clause (Crawford v. Washington (2004) 541 U.S. 36 [158 L.Ed.2d 177, 124 S.Ct. 1354, 1364-1365]), but that broad rule had not yet been recognized in 1990 when Waidla was tried.
See also, e.g., People v. Turner (1994) 8 Cal.4th 137, 194 [32 Cal.Rptr.2d 762, 878 P.2d 521] (claim of prosecutorial misconduct by making inconsistent statements regarding whether an accomplice shared Turner’s intent to kill rejected, in part, on ground that prosecutor’s theory in both cases was that Turner was actual killer); Haynes v. Cupp (9th Cir. 1987) 827 F.2d 435, 439 (“variations in emphasis” not cause for reversal where “underlying theory of the case” is consistent); Nichols v. Scott (5th Cir. 1995) 69 F.3d 1255, 1271, fn. 32 (verdicts and judgments in two cases where prosecutor made varying arguments as to who fired fatal shot “are not inconsistent” and did not entitle Nichols to relief from capital murder conviction because neither the guilt nor the punishment verdict depended on that fact).
On the other hand, we can find no due process or Eighth Amendment violation in Ipsen’s allegedly inconsistent arguments regarding domination. Though Ipsen’s emphasis in each trial was different, his basic factual position, which was largely based on the same evidence in the two cases, was consistent: neither defendant could claim mitigation because of his “substantial domination” (Pen. Code, § 190.3, factor (g)) by the other. (See State v. Lavalais (La. 1996) 685 So.2d 1048, 1056-1057 [no due process violation where prosecutor argued in successive trials that Lavalais was and was not under the domination and control of coperpetrator: the fundamental facts the prosecutor presented did not vary, and the appearance of inconsistency merely reflected the fact that “the state’s emphasis as to culpability was different in the two trials”].) Semantically, Ipsen’s argument in Waidla’s trial that Waidla was “the dominate [sic] person between himself and Mr. Sakarias” may have been inconsistent with his argument in Sakarias’s trial that there was “no evidence of domination.” Conceptually, however, it was not contradictory for Ipsen to point out in Waidla’s case that he had the greater motivation (because the Piirisilds assertedly owed him money or a car) and greater knowledge of the house (and thus might have taken a lead role in planning the burglary), while observing in *162Sakarias’s case that the evidence showed Sakarias had taken his friend’s grievance as his own and participated cooperatively in the crimes.
In discussing this type of justification, we do not mean to imply no others are theoretically possible. But unlike the dissent (conc. & dis. opn. of Baxter, J., post, at p. 171) we do not believe the lack of settled law on the subject of inconsistent factual theories establishes a prosecutor’s good faith in using such theories.
The prosecutorial manipulation of evidence here also makes Sakarias’s due process claim significantly stronger than that of the petitioner in Stumpf, supra, 367 F.3d 594. Evidence the Stumpf prosecution later presented in the trial of Stumpf s accomplice, Wesley—a jail informant’s testimony that Wesley had confessed to the fatal shooting—“was not available” at the time of Stumpf s plea hearing because at that time Wesley “was still in Texas, fighting extradition” to Ohio. (Id. at pp. 610, 613.) As the dissenting judge in Stumpf explained: “It is *163undisputed that the prosecution did not know of [the informant’s] statement at the time of Stumpf s conviction and sentencing. Nothing indicates that the prosecution cherry-picked facts in order to confirm Stumpf s guilty plea in the evidentiary hearing.” (Id. at p. 620 (dis. opn. of Boggs, C. J.).) Unlike Ipsen’s conduct in Sakarias’s trial, therefore, Stumpf s prosecutor “did not manipulate evidence.” (Id. at p. 621.)
Because Ipsen’s bad faith in Sakarias’s trial is clearly demonstrated by his deliberate manipulation of the evidence, we need not decide whether the burden of showing good or bad faith lies with the People or the petitioner. Nor need we address what obligation, if any, the People may have to correct inconsistent judgments obtained through inconsistent arguments made by the prosecutor in good faith.
Where the evidence is highly ambiguous as to each accused perpetrator’s role, some courts have relied on “the uncertainty of the evidence” to justify the prosecutor’s use of “alternate theories" in separate cases. (Parker v. Singletary (11th Cir. 1992) 974 F.2d 1562, 1578; Littlejohn v. State (Okla.Crim.App. 1998) 1998 OKCR 75 [989 P.2d 901, 909]; but see Prosecutorial Inconsistency, supra, 89 Cal. L.Rev. at p. 1477 [“If . . . the court cannot *165determine which of the two versions is true, the prosecution should lose the benefit of both positions’’].)
Because the penalty judgment must be vacated on this ground, we need not reach Sakarias’s further contentions that the prosecutor’s knowing use of false argument or evidence violated due process, that the use of inconsistent theories violated the Eighth Amendment to the United States Constitution, or that Ipsen was judicially estopped from changing positions between the trials.
As in Sakarias, the Waidla penalty jury deliberated for several days and at one point declared itself deadlocked. While this certainly tends to show the penalty question was close for one or more jurors, it does not show that the attribution of two nonhemorrhagic chop wounds, a relatively insignificant point under any theory of the killing, was likely to have made a difference.
For the same reason, Waidla is not entitled to relief on his claim that the People knowingly employed false evidence or argument against him. The apparently false attribution of the two nonhemorrhagic chopping wounds to Waidla was not material, even under the *169reasonable likelihood standard (United States v. Agurs, supra, 427 U.S. at p. 103) for knowing use of false evidence. Nor did that attribution of blows make Waidla’s penalty determination unreliable in violation of the Eighth Amendment to the United States Constitution.