I respectfully dissent.
The prosecution’s case was predicated solely upon circumstantial evidence because the defendant was the only survivor and thus the only witness to what occurred on January 2, 1981. Because of the presumption of innocence afforded a defendant in a criminal case, it is required that, where the People’s case rests exclusively or even substantially upon circumstantial evidence, that evidence must be irreconcilable with any rational conclusion other than defendant’s guilt. Further, when relying on circumstantial evidence, the People must prove beyond a reasonable doubt each fact essential to complete the chain of circumstances that establishes defendant’s guilt. (People v. Watson (1956) 46 Cal.2d 818, 830 [299 P.2d 243]; see Witkin, Cal. Evidence (2d ed. 1966) Introduction of Evidence at Trial, § 1136, pp. 1054-1055.)
It follows that defendant’s guilt may not lawfully be established by the mere expedient of raising doubts concerning defendant’s version of what occurred just prior to the drownings. A distinction crucial to the correct *398resolution of this case should be noted here. The prosecution undertook the burden of persuading the jury beyond a reasonable doubt that defendant murdered his wife and stepson. If such persuasion was not achieved, then the jury was required to find the defendant not guilty even though it doubted the defendant’s story and was not persuaded that he was innocent. (See 1 Jefferson, Cal. Evidence Benchbook (2d ed. 1982) Definitions and Distinctions, § 19.3, pp. 455-456.) The important distinction in criminal law between proof of guilt and proof of innocence was disregarded in this case. The People must prove defendant’s guilt, but defendant has no burden whatsoever to establish his innocence. A jury must be convinced beyond a reasonable doubt to find a defendant guilty, but it need not be at all convinced of a defendant’s innocence to find him not guilty. Stated differently, a finding of not guilty is not a finding of innocence.
It would appear from an overview of the trial in this case that the jury was subtly led to believe that, unless it found the defendant to be innocent, it was required to find him to be guilty. This result was attained by the erroneous reception of experiment evidence and compounded by a confusing and erroneous jury instruction and argument of counsel.
The evidence was uncontradicted that Verna and Douglas drowned, the first coroner who autopsied the bodies consistently maintaining that the deaths were accidental. The theory of the prosecution’s case was that prior to death both defendant’s wife and stepson suffered blows to their heads inflicted by the defendant and that these blows were the proximate cause of death. The People attempted to prove this theory by the introduction of three experiments and the so-called pattern evidence.
It is well established that in order for experiment evidence to be admissible, it must be shown that the evidence (1) is relevant, (2) was obtained under conditions substantially similar to those to which it is sought to be applied, and (3) will not cause undue delay in the trial or confuse the jury. (Culpepper v. Volkswagen of America, Inc. (1973) 33 Cal.App.3d 510, 521 [109 Cal.Rptr. 110].) The rollover, boy-rising, and dory-drop experiments were not relevant, were not conducted under substantially similar conditions, and unduly confused and influenced the jury. The pattern evidence not only failed to meet all three tests, but, as the majority concedes, did not even approach the level of reliable scientific evidence. (Maj. opn., ante, p. 391.)
Substantial Similarity
In contrast to the experiments conducted in Hasson v. Ford Motor Co. (1977) 19 Cal.3d 530 [138 Cal.Rptr. 705, 564 P.2d 857, 99 A.L.R.3d 158], *399Culpepper v. Volkswagen of America, Inc., supra, 33 Cal.App.3d 510, Andrews v. Barker Brothers Corp. (1968) 267 Cal.App.2d 530 [73 Cal.Rptr. 284], and People v. Lockheed Shipbuilding & Constr. Co. (1975) 50 Cal.App.3d Supp. 15 [123 Cal.Rptr. 778], none of the experiments commissioned by the People attempted to simulate defendant’s description of what occurred at Bird Rock. The rollover tests did not involve a puppy jumping overboard, a man and a boy attempting to retrieve it, and a woman bumping into the man with all three people in a state of panic. Rather, the participants in the rollover tests created their own maneuvers in an effort to overturn the dory. Further, the jury was instructed that it could not consider the rollover tests as suggesting any similarity to any event which occurred on the day of the drownings.
Moreover, defendant’s wife and stepson died on January 2, 1981, the middle of winter. The dory tests were conducted on July 17, 1981, the middle of summer. At the motion to exclude the tests, the trial court ruled that the weather conditions on the two days were substantially similar. However, at trial Mr. Strange, a meteorologist, testified that the two days were much different as far as the characteristics of the waves were concerned. In addition, there was uncontradicted evidence that on January 2 there was a wave train generated by a Pacific storm 1,900 miles from Bird Rock which caused long period swells with far more power than locally generated waves. No such condition was present when the rollover tests were conducted.
The court instructed the jury that it could not consider the rollover testing as suggesting that the weather and sea conditions were in any way similar on the two days. Yet substantial similarity in weather conditions was a preliminary fact which had to be established before the experiments were admissible. Since this key element was missing, it was error to allow the evidence to be presented to the jury.
Even more importantly, when the evidence of the rollover experiments was initially admitted, the experiments were ruled to have taken place under “the worst possible conditions,” but there was not one shred of evidence offered at trial to support this contention. In spite of that absence of evidence, the prosecutor used this phrase throughout the trial and in both his opening statement and closing argument. This led the jury to believe that, if the dory would not turn over under “the worst possible conditions,” it could not have capsized on January 2, 1981, in the manner in which defendant said it had.
The other two experiments were similarly defective. The boy-rising tests were conducted not in the ocean but in a saltwater tank on the University *400of California Santa Barbara campus. The dory-drop tests consisted of dropping the dory on a dummy’s head while it was floating in salt water. Neither of these tests included the slightest element of confusion or panic.
Although the experiment evidence failed to meet the criterion of substantial similarity and hence was improperly admitted, a major portion of the prosecution’s case in chief consisted of testimony concerning the results of the various experiments. The error was not cured by the limiting instruction because, although not admitted for this purpose, the experiment evidence was specifically used to disprove defendant’s version of the dory trip.
Relevancy
Relevant evidence is defined as evidence “. . . having any tendency in reason to prove or disprove any disputed fact that is of consequence to the determination of the action.” (Evid. Code, § 210.) The authorities I have reviewed in which experiment evidence was held to be admissible involved in every instance evidence that was offered to prove a disputed fact at issue in the lawsuit. (Hasson v. Ford Motor Co., supra, 19 Cal.3d 530, 549-550 [tests of brake-fluid temperatures admitted to contribute to search for accident’s cause]; Culpepper v. Volkswagen of America, Inc., supra, 33 Cal.App.3d 510, 521 [tests admitted to show the speed at which a Volkswagen would roll over]; Andrews v. Barker Brothers Corp., supra, 267 Cal.App.2d 530, 537-538 [evidence admitted to show stability of chair that had collapsed]; People v. Lockheed Shipbuilding & Constr. Co., supra, 50 Cal.App.3d Supp. 15, 36 [evidence supporting theory of cause of explosion].) In sum, experiment evidence is admissible only if it has any tendency in reason to prove or disprove a material fact upon which the outcome of the lawsuit depends. (Culpepper v. Volkswagen of America, Inc., supra, 33 Cal.App.3d 510, 522.) Here the experiments were specifically not admitted for this purpose.
The Rollover
Prior to trial, the court ruled the rollover tests admissible for the limited purpose of demonstrating the stability of the dory under various circumstances including the “worst possible conditions.” The court expressly stated that the tests would not be admitted “for the purpose of demonstrating the occurrence or nonoccurrence of the events at Little Scorpion Anchorage, or how they did occur, or did not occur . . . . ” Moreover, the court later instructed the jury that the rollover tests were admitted to show the speed with which the dory would capsize “under the circumstances of the experiment. ” (Italics added.) It further admonished the jury that the tests were not admitted to demonstrate or suggest any similarity to any event which *401occurred on January 2, 1981. Yet the determination of what occurred on January 2, 1981, was the ultimate issue at trial. It is thus clear that the rollover tests were not relevant and should not have been admitted.
There is, however, a more fundamental relevance problem with the rollover experiments. Through expert opinion based on the rollover experiments, the prosecution was allowed to adduce testimony that the cause of death could not have been accidental drowning.
Specifically, Dr. Hunter testified that a review of the film of the rollover tests led him to conclude that it is “entirely unlikely” that both of Verna’s injuries were caused by the dory and that it was “very unlikely” an accident involving the dory caused Verna to drown. Dr. Petty testified that Douglas’ injuries could not have been inflicted by the dory rolling over as he “saw it in the film and the videotape”; Criminalist Mauzey opined that, based on the films of the rollover tests, one but not two of the “sets” of marks on defendant’s stepson could have been inflicted by the dory; and Dr. Davis, focusing his attention on the videotape that he saw of the dory-turning-over tests, stated that the boat did not inflict the injuries sustained by defendant’s stepson. Thus, notwithstanding the fact that the rollover tests were not admitted to show what occurred at Bird Rock, they were expressly used for this purpose. The error in the admission of this evidence was twofold: (1) it placed upon the defendant the unconstitutional burden of proving his innocence and (2) it relieved the prosecution of the duty of proving defendant’s guilt beyond a reasonable doubt. This violated two of the most basic tenets of our criminal justice system.
The Dory-Drop and Boy-Rising Tests
The boy-rising tests were admitted by the court for the sole purpose of determining the acceleration rate produced between a boy rising in salt water and the dory closing on him. Similarly, the dory drop conducted at Minicars was admitted for the purpose of attempting to ascertain the maximum force and acceleration generated by the impact of a dory gunwale and a dummy head floating in salt water. However, the issues at trial did not involve collisions between floating dummies and the dory or ranges of closing velocities. The issue was whether on January 2, 1981, defendant unlawfully killed his wife and stepson with malice aforethought. Therefore, in order to be relevant, the evidence must have tended to show either that he did or that he did not. Since these tests were specifically neither offered nor admitted for this purpose, they were clearly irrelevant.
Moreover, the trial court itself instructed the jury that neither the videotapes of the experiments nor any testimony based on them could be consid*402ered “to demonstrate or suggest any similarity to any event which occurred on January 2, 1981.”
However, once again, these tests were indeed used for the purpose of showing what did or did not occur at Bird Rock.
Specifically, Criminalist Mauzey testified that the combination of the rollover and dory-drop tests convinced him that the dory could not have caused the injuries sustained by defendant’s stepson. This testimony was allowed notwithstanding the fact that, as the majority points out (maj. opn., ante, at pp. 388-389), the court had previously ruled that only qualified medical experts could testify concerning the injuries to human heads. The effect of the admission of these experiments and Criminalist Mauzey’s testimony was to require defendant to prove his innocence rather than to require the People to prove his guilt. (See People v. Briggs (1967) 255 Cal.App.2d 497, 500 [63 Cal.Rptr. 111].)
The Pattern Evidence
Criminalist Mauzey used autopsy photographs to trace a parallelogram “pattern” of marks on defendant’s stepson’s scalp, after which Los Angeles Coroner Thomas Noguchi knocked the carbon-paper-swathed head of a dummy against arbitrarily selected portions of the dory, rescue craft, Bird Rock, and Perseverance. From this exercise, Mauzey testified that the marks could only have been caused by someone hitting defendant’s stepson twice on an isolated spot inside the dory.
People v. Kelly (1976) 17 Cal.3d 24 [130 Cal.Rptr. 144, 549 P.2d 1240] requires the proponent of scientific evidence involving a new process to establish the reliability of the method, the qualifications of the expert to give an opinion based on the experiment, and a demonstration that correct scientific procedures were used. (Id., at p. 30.) As the majority concedes, the tests and conclusions reached by Mauzey failed to meet the above qualifications and should not have been admitted. (Maj. opn., ante, p. 391.)
However, contrary to the majority’s assertion, the admission of this evidence was highly prejudicial. The jury was told that, because Mauzey had “approximated” and “estimated” a single place on the dory which could reproduce this one of countless possible patterns, murder was established beyond a reasonable doubt. Such “logic” hardly seems sufficient to keep a case in court, let alone to warrant a guilty verdict—and it is logic of the type our Supreme Court has firmly rejected. (People v. Collins (1968) 68 Cal.2d 319, 330-331 [66 Cal.Rptr. 497, 438 P.2d 33, 36 A.L.R.3d 1176].) This was an extremely close case, based solely on circumstantial evidence. *403The jury deliberated a full week before reaching a verdict and relied heavily on the experiment evidence in doing so as demonstrated by its questions. The pattern evidence may have seemed to point strongly to defendant’s guilt and was heavily relied upon by the People in argument. It was therefore prejudicial error to admit this evidence (see People v. Hannon (1977) 19 Cal.3d 588, 603 [138 Cal.Rptr. 885, 564 P.2d 1203]) because, contrary to the majority’s position, it is extremely likely that “the absence of this evidence would have changed the result reached by the jury.” (Maj. opn., ante, p. 391.)
Undue Confusion
It is virtually impossible to discern the overall effect of the misuse of the experiment evidence. It was ubiquitous as it was presented to the jury at every turn.
Experiment evidence should not be admitted where it may be given undue weight. (See People v. Kelly, supra, 17 Cal.3d 24, 31-32.) That this jury was confused by and gave undue weight to these experiments is obvious from the fact that it asked to review them during deliberations and requested permission to conduct its own experiments.
These experiments involved very technical, difficult concepts not generally within the knowledge of laypersons. Moreover, although the jury was admonished as to the limited purpose for which the experiments were admitted, the prosecutor was allowed to roam far afield in his use of them during trial and during argument. Thus, it was prejudicial error to admit the experiment evidence because it is reasonably probable that the jury would have reached a result more favorable to the defendant had the tests been excluded. (People v. Watson, supra, 46 Cal.2d 818, 836.
The confusion engendered by the admission of this evidence is perhaps best demonstrated by the fact that the majority itself relies on the experiment evidence to establish what occurred at Bird Rock. (Maj. opn., ante, at p. 387.)
CALJIC No. 2.62
It was error for the court to give CALJIC No. 2.62.
As this court has emphasized, “[o]f primary importance to the application of CALJIC No. 2.62 is whether the facts or evidence that defendant allegedly fails to explain or deny are within defendant’s knowledge. [Citations.]” (People v. De Larco (1983) 142 Cal.App.3d 294, 309 [190 Cal.Rptr. 757].)
*404Here there was no evidence in the record that defendant did fail to explain or deny any fact which was exclusively within his knowledge. If defendant’s statement concerning the atmosphere of panic and confusion surrounding the capsizing of the dory were believed, it would not be reasonable to expect him to be able to explain how the dory capsized and the bruises occurred, especially since the experts themselves could not agree as to whether the bruises were premortem or postmortem. “ ‘No inference can be drawn if defendant does not have the knowledge necessary to explain or deny the evidence against him.’ [Citations.]” (People v. Saddler (1979) 24 Cal.3d 671, 680 [156 Cal.Rptr. 871, 597 P.2d 130].) The fact that the prosecution disagreed with defendant’s explanation of how he intended to make the premium payments did not warrant the instruction. The crucial fact is that defendant did offer an explanation. “[A] contradiction is not a failure to explain or deny.” (Id., at p. 682.)
Further, here, unlike in Saddler, the error was prejudicial. In Saddler there was an eyewitness and the jury deliberated for a mere hour and a half before reaching a verdict.
By contrast, here defendant was the only witness to the evénts at Bird Rock. The prosecution’s case consisted entirely of circumstantial evidence buttressed by the improperly admitted experiments. Prejudice is shown by the following facts: the jury was sequestered and began deliberations on Monday morning; on Friday afternoon, before the Mother’s Day weekend, having heard the instruction twice, having reviewed films of the various experiments, and having sought (unsuccessfully) to conduct experiments of its own, the jury finally asked the trial court to reread that portion of defendant’s testimony concerning the dory trip. The request was made at 2:08 p.m. The 45 minutes of testimony were immediately reread, and the jury returned its verdicts at 4:10 that afternoon.
In other words, after deliberating for more than four days, the jury heard defendant’s description of the events—the testimony upon which this instruction dwelled—and decided defendant’s guilt in approximately one hour.
Conclusion
“[I]n a review of error, the crucial question is not whether there is substantial evidence to support the judgment, but whether error affected the judgment.” (Traynor, The Riddle of Harmless Error (1970) p. 28.) This was an extremely close case in which the properly admitted evidence was just barely sufficient to support the verdicts. Under these circumstances reversal is required because it is extremely likely that the erroneous instruc*405tion “tip[ped] the delicate balance of evidence against the defendant.” (See People v. Saddler, supra, 24 Cal.3d 671, 686 (dis. opn. of Mosk, J.).)
I would reverse and remand.
A petition for a rehearing was denied May 22, 1985, and the opinion was modified to read as printed above. Appellant’s petition for review by the Supreme Court was denied August 19, 1985. Bird, C. J., Mosk, J., and Kaus, J., were of the opinion that the petition should be granted.