Two people know who killed Amanda Ortiz. One of them, Elizabeth Ruiz, Amanda’s mother, swore after defendant’s trial that she killed the child; defendant has always sworn he did not do so. The jury heard only defendant’s denial, and found him guilty of second degree murder. A new jury should consider the case now that Ruiz has confessed.
I.
I recognize that a trial court enjoys wide discretion to grant or deny a new trial motion (People v. Williams (1988) 45 Cal.3d 1268, 1318 [248 Cal.Rptr. 834, 756 P.2d 221]), that a nonparty’s posttrial confession is to be viewed with some suspicion (People v. McGaughran (1961) 197 Cal.App.2d 6, 17 [17 Cal.Rptr. 121]), and that in deciding whether a confession or recantation warrants a new trial on the ground of newly discovered evidence, the trial court should consider whether a jury could find the newly discovered statement credible (People v. Minnick (1989) 214 Cal.App.3d 1478, 1482 [263 Cal.Rptr. 316]). The trial court in this case declared Ruiz unworthy of belief—thereby signaling its view that a jury would also be unpersuaded— and further stated that it was convinced beyond a reasonable doubt that defendant killed Amanda.
Nevertheless, “ ‘in determining whether there has been a proper exercise of discretion on [a new trial] motion [on the ground of newly discovered *333evidence], each case must be judged from its own factual background.’” (People v. Dyer (1988) 45 Cal.3d 26, 52 [246 Cal.Rptr. 209, 753 P.2d 1].) The trial court commented that “possibly more than in most trials I have presided over, it was a search for the truth. And it was a search for the truth occasioned at least in large part by Elizabeth Ruiz and her many varying statements constantly during the trial.”
In such circumstances, twelve minds may be better than one, no matter how learned in the law and steeped in trial experience that one mind might be. I therefore conclude that the trial court abused its discretion despite a conscientious review of the facts and the law.
To conclude that defendant murdered Amanda, the trier of fact had to find, in light of the instructions given, that defendant intended to and did unlawfully kill (Pen. Code, §§ 187, subd. (a), 188), or that, with conscious disregard for human life, he intended to and did commit an act the natural consequences of which were dangerous thereto. (People v. Nieto Benitez (1992) 4 Cal.4th 91, 112 [13 Cal.Rptr.2d 864, 840 P.2d 969] (conc. opn. of Mosk, J.).) Both mental states embody malice, the first express, the second implied. (Pen. Code, § 188.) At closing argument the prosecution stepped aside from any consideration of either first degree murder or express malice and instead urged a verdict of second degree murder on an implied malice theory. The jury then returned a verdict of guilt of second degree murder.
It is difficult to square that verdict with this record. The prosecution’s theory—that defendant killed Amanda to ensure quiet in the apartment so he could have uninterrupted sex with Ruiz—strains against the facts. There was evidence that defendant was in tears while at the police station after he learned of the murder; he asked the police not to leave him alone while in custody, presumably so that he would not harm himself. At trial defendant gave a coherent and articulate account of the events of the night Amanda was killed. And the record contains many epistolary attestations to defendant’s good character. Most striking, there was substantial evidence that Ruiz stayed passionately loyal to defendant, even though he was charged with her daughter’s murder—one of her many visits to defense counsel’s office took place on the day of the child’s funeral.
To be sure: character references are no assurance that a person has not killed; a murderer can speak smoothly in his own defense; defendant’s conduct at the police station could be explained by fear of punishment rather than anguish over Amanda’s death; and Ruiz’s visit to defendant’s lawyer the day of her daughter’s funeral may simply have been a macabre epilogue to the child’s life story.
*334But the prosecution’s case was less than overwhelming. The People offered the testimony of two percipient witnesses to the killing. Neither was helpful. The record raises significant questions whether the testimony of five-year-old Johnnie Ortiz was influenced by previous interviews, and the testimony he gave was, as far as I can discern from the record, not particularly useful in any event. The majority somewhat misunderstand the testimony of nine-year-old David Cano Yáñez, and Cano’s account was less probative than the majority imply. Young Cano could only vaguely describe having perceived certain banging noises, the presence of a woman, and a man saying “shhh” at the time the prosecution theorized defendant killed Amanda. Part of the exchange that the majority state Cano overheard in fact occurred earlier that evening, not when defendant allegedly was alone with Amanda.
Small wonder that, after the evidently exhausted jurors returned their verdict following some three and one-half days of deliberations, the trial court observed, “It’s obvious in looking at you that it’s been stressful for you.”
All these factors give rise to substantial doubt about defendant’s role in Amanda’s death. As the Court of Appeal observed, “it is inherently improbable that a mother would continue a romantic interest in the killer of her baby.” After the trial Ruiz not only confessed to killing Amanda—a statement that, if believed, would subject her to liability for murder and other offenses—she also told a friend, Victoria Diaz, that she had killed the child. And against the prosecution’s theory that defendant was impatient for sex, there was evidence suggesting that Ruiz’s lust for defendant was equal to or greater than his for her. Considering the strong hints scattered throughout the record of Ruiz’s overall neglect of her children, I would not be surprised if on retrial a jury might weigh whether it was Ruiz who beat Amanda, while defendant sat in the living room, so that she could have sex with him straightaway.
Notwithstanding the foregoing observations, I draw no ultimate conclusion about defendant’s innocence or guilt. But given the difficult facts of this case, I would err on the side of caution and uphold the Court of Appeal’s determination that a new trial is warranted.
II.
A subject that has intrigued commentators for centuries is whether, with all of the safeguards of our Anglo-American system of jurisprudence, an innocent person can be convicted of homicide.
*335A recent study produces a revealing and disturbing answer. Professor Michael L. Radelet, Professor Hugo Adam Bedau, and Constance E. Putnam have published an account of 400 defendants who were erroneously charged with and convicted of homicide, and, were later proved to be innocent. Of that number, 26 were tried in California; and, of those, 6 were sentenced to death. (See Radelet et al., In Spite of Innocence (1992).)
What this study suggests is that we cannot uncritically accept a defendant’s conviction of homicide and cavalierly disregard a subsequent confession to that offense by another person who was undeniably present on the occasion. How reliable the subsequent confession may be is debatable. But here it was not considered by a jury, the ultimate trier of fact. To preserve judicial integrity, I believe that the opportunity should not be denied.
This court unanimously held, in In re Hall (1981) 30 Cal.3d 408 [179 Cal.Rptr. 223, 637 P.2d 690], that the defendant had been improperly convicted of first degree murder. There it was a suggestive lineup that resulted in tainted identification. We said, “We need not decide, of course, whether the allegations of suggestiveness are true or whether, if true, they constitute a denial of due process. It is sufficient for the present purpose to observe that the defense was potentially meritorious, and that petitioner was denied an adjudication on the matter . . . .” (Id. at p. 434, italics omitted.) The key words, applicable to the instant case, are “potentially meritorious” and “denied an adjudication.”
III.
Even if I agreed with the result in this case, I could not subscribe to part of the majority’s discussion.
Given their premise that the trial court’s determination about Ruiz’s credibility deserves complete deference, the majority are on safe ground when they strongly imply that, for reasons of credibility alone, the new trial motion was properly denied. If complete deference on the issue of credibility was warranted, then the trial court’s view that Ruiz was unworthy of belief was the alpha and omega of the question whether a new trial should have been granted. (See People v. McGaughran, supra, 197 Cal.App.2d 6, 17.)
The remainder of the discussion, however, is misleading dictum.
The majority come perilously close to concluding that Ruiz’s confession was cumulative. It was not. Unless a materially similar version of a confession or recantation is introduced at trial and rejected by the jury, such *336evidence can never be cumulative, for it would always “contradict[] the strongest evidence introduced against the defendant” (People v. Martinez (1984) 36 Cal.3d 816, 823 [96 Cal.Rptr. 142]).1
In sum, cumulativeness does not qualify as a reason to deny the new trial motion. The majority’s discussion is unnecessary to the decision they reach in this case. Reasonable minds can differ on whether the trial court’s determination should be sustained on the stated ground that Ruiz’s confession was unworthy of belief. If so, then that ground alone suffices. I am, of course, of a different view from that of the majority. I would affirm the judgment of the Court of Appeal. A jury—not the trial judge, not this court—should weigh the new facts.
The majority do well to avoid charging defendant with lack of “diligence” in producing Ruiz’s confession at trial. The law of causality prevents introducing evidence before it comes into being. I know of no power by which defendant could have obtained Ruiz’s confession before she was prepared to give it; his authority does not extend to extracting information in shuttered rooms with bright lights and a rubber hose. (See People v. Hairgrove (1971) 18 Cal.App.3d 606, 610 [96 Cal.Rptr. 142].)