Opinion
WALLIN, J.
A jury found appellant, Leland Dellinger, guilty of first degree murder of his two-year-old stepdaughter, Jaclyn. The trial court denied Dellinger’s motion for a new trial, but reduced the conviction to second degree murder. Since several of the errors pointed out by Dellinger prejudiced his ability to receive a fair trial, the conviction must be reversed.
On May 29, 1979, Dellinger picked up Jaclyn from a babysitter. When Diana Dellinger Pena (Pena), her mother, arrived home from work shortly after 6 p.m., Dellinger called for her to hurry. From the top of the stairs she saw Dellinger talking on the telephone and bending over Jaclyn, who was lying unconscious on the kitchen counter. A pan was on the stove and the gate above the stairs was open. Pena asked if Jaclyn was breathing. Dellinger began to cry, banged the counter with his fist, and said, “Jackie, why are you doing this to me?”
Paramedics arrived at the apartment at 6:13 p.m. Dellinger was administering mouth to mouth resuscitation. The paramedics rushed Jaclyn to the hospital where she died at 7:45 p.m.
[290]*290The police questioned Dellinger about Jaclyn’s injuries. He stated that after returning home with Jaclyn, he began preparing dinner and left her watching television. He heard a thud and rushed out of the kitchen to find Jaclyn lying on the first landing of the carpeted staircase. He immediately moved her to the kitchen counter and called the paramedics. He admitted giving her wine that afternoon. Pena testified that upon returning to the apartment, she saw a baby bottle “pretty much full” of wine. The toxicologist found no alcohol in Jaclyn’s blood.
After completing the first autopsy, Dr. Richard Fukumoto, a pathologist, determined the major cause of death was swelling of the brain due to blunt force trauma to the head, with a skull fracture. He ruled the cause of death was accidental. There was evidence of a spiral type fracture of the lower left leg probably caused by a twisting motion exerted by another person. Three months later toxicological results were received by the coroner showing the presence of cocaine in Jaclyn’s blood and large quantities in her liver and stomach. Dr. Fukumoto then determined that cocaine was probably a contributing cause of her death. He contacted Dr. Thomas Noguchi, Coroner for Los Angeles County, who consulted with Dr. Carley Ward, a biomedical engineer. Together they conducted a further investigation into Jaclyn’s death.
As a result of her investigation, Dr. Ward concluded the force of a fall down the stairs could not have caused the injuries Jaclyn suffered. Dr. Fukumoto thereafter recharacterized the cause of death as criminal and Dr. Noguchi concurred. At trial, Dr. Noguchi testified the injuries were not caused by a fall down the stairs. A second autopsy, in June of 1980, showed a contusion of the cervical spinal cord which probably resulted from the same blunt force causing the head injury.
The defense produced two experts to testify that Jaclyn could have received her injuries from a fall down the stairs: Dr. Werner Goldsmith, a mechanical engineer and impact specialist, and Dr. Irving Root, a pathologist for the San Bernardino County Coroner’s office.
Jaclyn’s babysitter was allowed to testify that one day in April 1979 she noticed a purplish handprint on Jaclyn’s lower back. When Dellinger picked up Jaclyn that afternoon, he said he had spanked Jaclyn during the night but did not realize how hard he hit her until he saw the bruise the next morning.
The babysitter also testified that three weeks before her death, Jaclyn could not walk. The babysitter took Jaclyn to the doctor and a cast was put on her left leg. Ten days later the cast was removed, but Jaclyn continued [291]*291having trouble walking. She had always been clumsy. Pena testified that Jaclyn could not walk because of new orthopedic shoes.
Pena and Dellinger were married five months before Jaclyn’s death and Dellinger had begun proceedings to adopt Jaclyn. Pena was allowed to testify Dellinger used cocaine before they were married but she had no knowledge of his using cocaine since moving into the apartment. She assumed he was no longer using it. She further testified that about two months before Jaclyn’s death, Dellinger gave Jaclyn wine to calm her down after she complained of pain in her leg. Pena did not use cocaine and denied ever giving Jaclyn narcotics or alcohol.
The apartment manager had not wanted the Dellingers to move into the apartment because the stairs were dangerous for a small child. However, he relented when Dellinger installed a latticework gate.
In the fall of 1979, Dellinger submitted to a polygraph examination at the Orange Police Department. The polygraph examiner determined Dellinger appeared truthful in his denial of the offense. After an evidentiary hearing, the trial court refused to admit the examiner’s testimony.
I
Was a proper foundation established for admission of a biomedical engineer’s expert opinion on the amount of force needed to sustain the injury to the child’s head?
Dellinger objected to the introduction of Dr. Carley Ward’s expert opinion as a biomedical engineer. He contends the People failed to meet their threshold burden of proving the reliability of the underlying scientific techniques. After a lengthy evidentiary hearing the trial court found the field of biomechanics was not new and introduction of the testimony was supported somewhat by Pineda v. Los Angeles Turf Club, Inc. (1980) 112 Cal.App.3d 53 [169 Cal.Rptr. 66]. Dr. Ward used two techniques to determine the amount of force Jaclyn’s head could have sustained if she had fallen down the stairs: anthropomorphic dummy experiments and finite element analysis.
A. The Anthropomorphic Dummy Experiments
Dr. Ward obtained an anthropomorphic dummy similar to Jaclyn in height and weight. She went to the apartment where the injury occurred and had a police officer drop the dummy down the stairs from many different positions. Dr. Ward determined Jaclyn must have fallen backwards to incur her head injuries. She also concluded Jaclyn could not have fallen further than [292]*292the fifth step without being pushed. The fifth step was selected as the site of the worst impact. She did not stage any falls from the railing above the stairs.
The dummy had a special metal, metnite, on the back of its head. From the indentations in the metal, Dr. Ward determined the amount of force the head sustained on impact. She measured the indentations and then referred to a chart prepared by General Motors. The chart indicated the force from the test falls was 17.95 pounds per square inch (p.s.i.). However, relying on “known injury criteria,” she determined it would take 34 p.s.i. to cause the brain damage and skull fracture suffered by Jaclyn. She concluded from the dummy experiment that Jaclyn could not have sustained her injuries from a fall down the stairs.
B. The Finite Element Analysis
Dr. Ward verified her findings with another procedure, a finite element analysis. The characteristics of a structure (the brain and skull in this case) were fed into a computer, a force was mathematically applied, and the computer then predicted the pressures and stresses in the brain and skull. The force used in the analysis was the impact force developed by Ward at the apartment. Again Dr. Ward’s work indicated the stress Jaclyn would have received from a fall down the stairs could not have resulted in brain injury or skull fracture.
The seminal case in determining the foundational requirements for introduction of expert opinion based on new scientific techniques is People v. Kelly (1976) 17 Cal.3d 24 [130 Cal.Rptr. 144, 549 P.2d 1240],1 Kelly emphasizes it is not our task to determine the accuracy of the proffered scientific techniques, but only to assure that the People meet the burden of showing a scientific consensus supporting their use at this time, the qualification of Dr. Ward to give an opinion on the subject, and the scientific validity of the procedures used. {Id., at p. 30.)
The Attorney General, echoing the trial court’s finding, argues Kelly is . not applicable because the field of biomechanics is not new, relying on Pineda v. Los Angeles Turf Club, supra, 112 Cal.App.3d 53. However, neither the reliability of the field of biomechanics nor the qualifications of the expert were challenged in Pineda and the court never ruled on the foundational issues. The expert, who referred to principles of biomechanics, also based his opinion on the statements of five eyewitnesses and a video[293]*293tape of the accident. Pineda simply did not purport to accept the scientific validity of the general field of biomechanical engineering.
No one took home movies of Dellinger babysitting that afternoon nor were there any eyewitnesses to aid Dr. Ward in her evaluation. Consequently, we must apply the three-part Kelly test to determine the foundational reliability of both the anthropomorphic dummy experiments and the finite element analysis. The People failed to meet their burden of proof on each of the three foundational elements.
First, “the reliability of the method must be established, usually by expert testimony.” (People v. Kelly, supra, 17 Cal.3d at p. 30.) There are no reported California cases in which an anthropomorphic dummy was used in any situation. Dr. Ward testified she did not know of any experiments conducted using an anthropomorphic dummy in a fall situation and there was no corroborative testimony that this technique was accepted within the scientific community, Similarly, Dr. Ward was the only witness to testify to the applicability of the finite element computer analysis and the acceptability of using the technique for predicting forces that applied to the brain. “ [I]t [is] questionable whether the testimony of a single witness alone is ever sufficient to represent, or attest to, the views of an entire scientific community regarding the reliability of a new technique.” (Id., at p. 37.)
The Attorney General credits Ward’s work with widespread acceptance. However, this acceptance was reported solely by Ward herself. She testified the skull model had been used by several universities. Additionally, she compared her conclusions to various studies and consulted with those more knowledgeable about injuries to children. Her diligence in demonstrating scientific acceptance is insufficient for admissibility because the testimony was both self-serving and uncorroborated.
Dr. Ward was the only expert to assure the court of the reliability of her techniques. Other testimony at trial and in declarations attached to the motion for a new trial criticized both the methods and assumptions she used. Even the studies she relied upon contradicted her conclusions. In a study conducted by the University of Michigan Highway Safety Research Institute, the authors warned, “an instrumented anthropomorphic dummy did not seem to be a good simulator since orientation at impact would be difficult to control in a free fall situation and since dummy reactions are often not human like.” (Insurance Institute for Highway Safety, Final Rep., Dec. 5, 1977, Impact Tolerance Through Free-Fall Investigation.) In the same study it was emphasized that “. . . the tolerance limits obtained in this and other studies involving relatively limited samples, should not be misinterpreted as minimum values that can be experienced by most of the population with[294]*294out serious consequences.” (Ibid.) Nevertheless, Dr. Ward’s conclusions were predicated on such limits of human tolerance.
Dr. Ward verified her conclusions from the dummy experiment with her finite element analysis. However, the data obtained from the dummy experiment was used in these calculations. Dr. Ward agreed that if this data was wrong, the result of the finite element analysis would be similarly flawed. Moreover, there was substantial evidence that her application of the finite element analysis was not generally accepted within the scientific community.
The final step in both the dummy experiment and the finite element analysis was to compare the amount of force the child would have sustained from the fall to a chart depicting “known injury criteria.” The “known injury criteria” indicated that 34 p.s.i. (pounds per square inch) were necessary to cause brain damage. However, Dr. Ward testified only that the criteria were determined by unnamed “researchers and different universities.” Dr. Ward did not duplicate any experimentation to test the validity of the criteria and no other scientist attested to its general acceptance. There were no adaptations made for injury criteria in children. Again, it was Dr. Ward alone who verified the acceptance of her own work. Consequently, the first prong of the Kelly analysis, the reliability of the method, was not sufficiently established by expert testimony.
Nor did the People establish the second prong of Kelly that “the witness furnishing such testimony must be properly qualified as an expert to give an opinion on the subject.” (People v. Kelly, supra, 17 Cal.3d at p. 30.) Dr. Ward’s testimony was based on her research and experience with animals and cadavers. However, the University of Michigan study stated “extrapolation of injury and tolerance data from animals to humans has been subject to controversy; the scaling techniques have not been perfected.” (Insurance Institute for Highway Safety, op. cit., supra.) She had no medical background or experience with the variances attributable to the brain and skull of a child. She had no expertise in fall-related injuries.
The Attorney General points out her credentials are impressive. She has a Ph.D. in biomedical engineering and has worked in the field at several universities and with the Department of Transportation, the Los Angeles County Coroner and the Los Angeles Police Department. However, “ ‘[t]he competency of an expert is relative to the topic and fields of knowledge about which the person is asked to make a statement. In considering whether a person qualifies as an expert, the field of expertise must be carefully distinguished and limited.’ [Citation.]” (People v. Kelly, supra, 17 [295]*295Cal.3d at p. 39.) The deficiencies noted above are significant. She may be well versed on head impact studies involving adult cadavers and animals but her inexperience with fall-related injuries and children merits close scrutiny. The expertise deficiency alone may not have been fatal. However, without corroboration on the reliability of the techniques or the acceptability of the procedures, her application of general principles of bio-mechanics to this specific situation involving a child’s fall is insufficient for admissibility.
Third, Kelly requires “the proponent of the evidence [to] demonstrate that correct scientific procedures were used in the particular case.” (People v. Kelly, supra, 17 Cal.3d at p. 30.) There was an insufficient showing that Dr. Ward used acceptable procedures for either of her techniques. She arrived at the conclusion the victim could not have fallen beyond the fifth step by instructing a police officer to drop a dummy down the stairs from different positions. However, there was evidence received at trial and in declarations attached to the motion for a new trial stating a dummy would not execute the same trajectory as a child because of its lack of musculature reflexes. Yet the fifth step conclusion was used in her finite element analysis as well.
There were flagrant loopholes in the acceptability of the procedures and calculations used in the finite element analysis. In her calculations Dr. Ward neglected to consider several important factors, including the mass of the child’s body and the compressibility of the carpet, skull, and scalp. These factors were integral to the calculation of the force of the impact of Jaclyn’s head on the stair. Dr. Ward tried to calculate the velocity of the head hitting the fifth step, but in doing so, she used only the mass of the head, not the body. Dr. Ward admitted that the inclusion of the body mass would have altered her calculations and results. Furthermore, the figure used in the analysis representing the mass of the head came from a chart, not from measurements of Jaclyn’s head.
Similarly, although Dr. Ward took measurements on the compressibility of the carpet, she did not use them. In addition, she did not have data on the compressibility of the victim’s skull and scalp. She assumed an average compressibility measurement, although conceding that if this were decreased, the force to the head would be greater. Even a slight change in this initial assumption could have significantly distorted the results by as much as 200 percent. She failed to offer a rational or scientific basis for her decision not to use the actual compressibility of the carpet, skull, and scalp. We conclude these foundational omissions invalidated the scientific conclusions offered to the jury. We are not convinced that “correct scientific procedures were used.” (People v. Kelly, supra, 17 Cal.3d at p. 30.)
[296]*296Although a sufficient Kelly foundation for admission of the biomedical engineer’s expert opinion on the amount of force needed to sustain the injury was not presented, this finding does not foreclose another attempt to establish the reliability of the techniques.2 “We simply circumscribe, carefully and deliberately, the admission of evidence born of new techniques until the time when there is demonstrated solid scientific approval and support of the new methods.” (People v. Kelly, supra, 17 Cal.3d at p. 41.)
Finally, the Attorney General argues that the admission of Dr. Ward’s testimony, even if erroneous, was not prejudicial. It was merely corroborative of Dr. Noguchi’s testimony. In light of the entire record, this argument must fail. Dr. Fukumoto, from the Orange County Coroner, based his conclusions on Dr. Ward’s findings. He testified he would reclassify the death as accidental if her conclusion were erroneous or disregarded. Dr. Noguchi’s testimony was short and straightforward, in his opinion the absence of bruises established that Jaclyn did not fall down the stairs. But another pathologist reached the opposite conclusion. As a result, the People’s case rested on “the aura of certainty” enveloping Dr. Ward’s scientific findings. (See e.g., People v. Kelly, supra, 17 Cal.3d at pp. 31-32.) The jury could have been unduly prejudiced by the admission of the scientific findings of Dr. Ward and, therefore, the error was not harmless. (People v. Watson (1956) 46 Cal.2d 818, 836 [299 P.2d 243].) Although admission of the scientific testimony is reversible error, we address the remainder of the issues raised by Dellinger because of the likelihood of their recurrence on retrial.
II
Was evidence of defendant’s past conduct and the victim’s prior injury properly admitted?
Dellinger objects to the admission of testimony on (1) his prior use of cocaine; (2) Jaclyn’s prior leg injury; and (3) the handprint-shaped bruise on Jaclyn’s lower back and his admission of having spanked her in the middle of the night.3
[297]*297Evidence of other criminal acts or misconduct of a defendant is inadmissible to prove the accused had the propensity or disposition to commit the crime charged. (Evid. Code, § 1101, subd. (a); People v. Guerrero (1976) 16 Cal.3d 719, 724 [129 Cal.Rptr. 166, 548 P.2d 366]; People v. Gibson (1976) 56 Cal.App.3d 119, 127 [128 Cal.Rptr. 302].) “The reason for this rule is not that such evidence is never relevant; to the contrary, the evidence is excluded because it has too much probative value. ‘The natural and inevitable tendency of the [jury] ... is to give excessive weight to the vicious record of crime thus exhibited, and either to allow it to bear too strongly on the present charge, or to take the proof of it as justifying a condemnation irrespective of guilt of the present charge.’ [Citations.]” (People v. Guerrero, supra, 16 Cal.3d at p. 724.)
However, evidence of defendant’s prior conduct is admissible to prove some fact other than mere propensity such as motive, opportunity, intent, preparation, plan, knowledge, identity, or absence of mistake or accident. (Evid. Code, § 1101, subd. (b).) Evidence admissible under subdivision (b) of Evidence Code section 1101 remains subject to exclusion under Evidence Code section 352. “The courts have recognized that whenever [this] evidence is offered . . . there is always the potential for great prejudice to a defendant because of its possible misuse by the jury as character trait or propensity evidence.” (People v. Gibson, supra, 56 Cal.App.3d at p. 128.)
To minimize the potential for prejudice, the courts have placed other restrictions on the admissibility of section 1101, subdivision (b), evidence. The proffered evidence must be relevant to an ultimate fact “ ‘actually in dispute.’” (People v. Tassell, supra, 36 Cal.3d 77, 84.) Moreover, the probative value of defendant’s past conduct must be substantial. (Id., at p. 88; People v. Thompson (1980) 27 Cal.3d 303, 318 [165 Cal.Rptr. 289, 611 P.2d 883].) “If evidence is ‘merely cumulative with respect to other evidence which the People may use to prove the same issue,’ it is excluded under a rule of necessity. [Citations.] ... If there is any doubt, the evidence should be excluded.” (People v. Thompson, supra, 27 Cal.3d at p. 318.) Guided by these general principles, the admissibility of the challenged evidence is now considered.
A. Prior Use of Cocaine
Pena testified that Dellinger had used cocaine before they were married but to her knowledge he had discontinued the practice. They had been married five months at the time of Jaclyn’s death. The Attorney General claims the prior use of cocaine shows Dellinger had the knowledge to commit the crime of murder by poison. However, “admission of other crimes [298]*298evidence cannot be justified merely by asserting an admissible purpose.” (People v. Guerrero, supra, 16 Cal.3d at p. 724.)
Generally, knowledge, intent, or state of mind evidence is admissible if there is no doubt that defendant has committed the act, but there is some question as to his or her mental state at the time. (Id., at p. 726.) Here Dellinger vehemently denied furnishing cocaine to Jaclyn or leaving it accessible to her. In essence, the Attorney General claims that use or familiarity with a narcotic is indicative of the mental state necessary to kill another with that substance. If adopted, this position would emasculate the policy of excluding other crimes evidence unless the probative value is substantial. As ingestion of cocaine was a contributory cause of death, there was a tremendous potential for prejudicing the jury with evidence of Del-linger’s prior use of the substance. Yet his prior use had only marginal probative value to the issues “actually in dispute.” (People v. Tassell, supra, 36 Cal.3d at p. 84.)
“ ‘It has frequently been recognized . . . that because of the sound reasons behind the general rule of exclusion, the relevancy of evidence of other crimes, and therefore its admissibility, must be examined with care. [Citation.] The evidence should be received with “extreme caution,” and if its connection with the crime charged is not clearly perceived, the doubt should be resolved in favor of the accused. [Citations.]’ ” (People v. Guerrero, supra, 16 Cal.3d at p. 724.) The trial court failed to resolve any doubts in favor of the accused and abused its discretion under Evidence Code section 352 by admitting the evidence which was substantially more prejudicial than probative. (People v. Gibson, supra, 56 Cal.App.3d 119, 131.)
B. The Victim’s Fractured Leg
Jaclyn was taken by her babysitter to a doctor two weeks before her death because she was having great difficulty walking. The doctor put a cast on her leg. The coroner testified there was evidence of a spiral fracture caused by someone twisting the leg.
The Attorney General maintains the injury was relevant to Dellinger’s mental state in relation to the theory of murder in the commission of felony child endangering. The spiral fracture undoubtedly is relevant to a finding that Jaclyn was abused before her death. However, there is no evidence that Dellinger brutally twisted Jaclyn’s leg until it broke. “Evidence of other similar crimes linked to no one at all is clearly inadmissible to prove any element of the crime charged against a defendant, even though the crime occurred within reasonable proximity of time and [299]*299place. [Citations.]” {People v. Jackson (1967) 254 Cal.App.2d 655, 658 [62 Cal.Rptr. 208].) Since someone other than Dellinger may have injured the child and the chance of prejudice was so great, the trial court abused its discretion by admitting the evidence of the fracture.
C. The Handprint-shaped Bruise on the Lower Back
Jaclyn’s babysitter testified she had noticed a bruise in the shape of a handprint on Jaclyn’s lower back. Dellinger admitted spanking the child in the middle of the night. He expressed remorse and said he did not realize how hard he had hit her. This evidence of Dellinger’s past conduct was remarkably different from his use of cocaine and the leg fracture.
In this episode, Dellinger admitted striking the child with excessive force. The People were attempting to prove a wilful, deliberate, premeditated killing to convince the jury Dellinger had committed first degree murder. The prior assault, occurring in close proximity to the time of death, was relevant to prove Dellinger’s mental state. It would assist the jury in determining the appropriate degree, if they found him guilty of murder.
The testimony regarding the bruise and Dellinger’s admission had substantial probative value and was not cumulative of other evidence. Because of the similarity of the conduct, the admitted agency and the proximity in time, the trial court properly concluded the probative value of the past conduct outweighed its prejudicial impact. Most importantly, Dellinger’s mental state was “actually in dispute” and his prior assault was relevant to prove the willfulness and deliberation necessary to a finding of first degree murder.
D. Prejudicial Impact
“The admission of any evidence that involves crimes other than those for which a defendant is being tried has a ‘highly inflammatory and prejudicial effect’ on the trier of fact.” {People v. Thompson, supra, 27 Cal.3d at p. 314, fn. omitted.) “[Substantial prejudicial effect [is] inherent in evidence of prior offenses. [Citation.]” {People v. Sam (1969) 71 Cal.2d 194, 206 [77 Cal.Rptr. 804, 454 P.2d 700], italics added.)
The Attorney General contends the admission of the other crimes evidence was not prejudicial because a limiting instruction was read to the jury and the case was not a close one. The error requires reversal if it is “reasonably probable that the jury would have reached a more favorable result without it. {People v. Watson (1956) 46 Cal.2d 818, 836 [299 P.2d 243].)” {People v. Tassell, supra, 36 Cal.3d at p. 89.) Because of the in[300]*300herently prejudicial nature of other crimes evidence, the courts have recognized that limiting instructions are frequently inadequate to protect the accused. (See, e.g., People v. Gibson, supra, 56 Cal.App.3d at p. 130.) “No limiting instruction, however, thoughtfully phrased or often repeated, could erase from the jurors’ minds the picture of . . .” a “cokehead” brutally twisting the leg of an innocent toddler until it snapped. (People v. Guerrero, supra, 16 Cal.3d at p. 730.)
Moreover, the case was indeed very close. The evidence was entirely circumstantial. The People’s case rested on the questionable validity of a biomechanical engineer’s unverified tests and conclusions, the inference that a person who had used cocaine in the past would administer it to a child or leave it accessible to her, and the inference a child with a prior spiral fracture must have been murdered by her stepfather, the defendant. Any potential for prejudice was exacerbated by the prosecutor’s repeated references to the prior cocaine use and the spiral fracture during closing argument. Under these circumstances it is reasonably probable the jury would have returned a more favorable verdict if the evidence had been excluded. Consequently, the error necessitates reversal.
Ill
Was the trial court’s failure to instruct the jury sua sponte to unanimously agree on which act constituted murder (CAUIC 17.01) reversible error?
The record reflects considerable confusion surrounding the causes of death and the alleged criminal acts by Dellinger. The cause of death was described as swelling of the brain due to blunt force trauma to the head. However, ingestion of cocaine was found to be a contributing cause. The jury was instructed they could find the defendant guilty of first degree murder if they found a willful, deliberate, premeditated killing or if they found murder by poison; but they were not instructed they had to unanimously agree on which act constituted murder. Dellinger claims some jurors may have decided he gave Jaclyn cocaine while others may have decided he inflicted a death blow to her head. Dellinger further contends this constitutionally impermissible possibility triggered the trial court’s sua sponte obligation to instruct the jury on the unanimity requirement. He is correct.
“A trial court must instruct sua sponte on those general principles of law which are ‘. . . closely and openly connected with the facts before the court, and which are necessary for a jury’s understanding of the case. ’ [Citation.]” (People v. Crawford (1982) 131 Cal.App.3d 591, 596 [182 Cal.Rptr. 536].) A court’s sua sponte obligation extends to instructing the jury on certain principles of law which control how they approach their [301]*301task. (People v. Madden (1981) 116 Cal.App.3d 212, 215 [171 Cal.Rptr. 897].) Specifically, juries must be informed that their verdict must be unanimous. (See Cal. Const., art. I, § 16; People v. Wheeler (1978) 22 Cal.3d 258, 265 [148 Cal.Rptr. 890, 583 P.2d 748].) To protect this constitutional guarantee of juror unanimity the court must instruct the jury, in language similar to CALJIC No. 17.01,4 to agree on the act or acts constituting the offense. (People v. Diedrich (1982) 31 Cal.3d 263, 280-282 [182 Cal.Rptr. 354, 643 P.2d 971]; see also People v. Madden, supra, 116 Cal.App.3d at pp. 216-217 for decisional history on the necessity for a unanimity instruction.)
Most of the reported cases involving multiple criminal acts by a defendant also involve potential multiple offenses. (.People v. Diedrich, supra, 31 Cal.3d at p. 280; People v. Moore (1983) 143 Cal.App.3d 1059, 1064 [192 Cal.Rptr. 374]; People v. Crawford, supra, 131 Cal.App.3d at pp. 595-600; People v. Madden, supra, 116 Cal.App.3d at pp. 214-215; People v. McNeill (1980) 112 Cal.App;3d 330, 335 [169 Cal.Rptr. 313].) Characteristically, a defendant could have been charged with several different counts but was charged only once. (See, e.g., People v. McNeill, supra, 112 Cal.App.3d at p. 335.) Here there was only one offense and one victim but there were several hypotheses as to which act or acts caused Jaclyn’s death. As long as there are multiple acts presented to the jury which could constitute the charged offense, a defendant is entitled to an instruction on unanimity.
The Attorney General’s attempt to prevent reversal by disclaiming reliance on the cocaine theory is unconvincing. It is fundamentally unfair to develop two alternate theories at trial, secure a general verdict, and then abandon one of them on appeal. It is impossible to determine whether the jury unanimously agreed on one of the acts postulated by the prosecution when a CALJIC No. 17.01 instruction is not given.
There is no need to add to the literature on which standard is appropriate for determining whether the error is harmless. (See People v. Deleito (1983) 147 Cal.App.3d 458, 470-472 [195 Cal.Rptr. 233].) Here the instructional error was prejudicial under both the Chapman and Watson standards. (See Chapman v. California (1967) 386 U.S. 18, 24 [17 L.Ed.2d 705, 710, 87 S.Ct. 824, 24 A.L.R.3d 1065]; People v. Watson (1956) 46 Cal.2d 818, 836 [299 P.2d 243].) The evidence suggesting Dellinger was responsible [302]*302for Jaclyn’s ingestion of cocaine was entirely different from the evidence suggesting he had physically assaulted her. It is very plausible the jurors could have distinguished between the two acts by way of argument and evidence. (Cf. People v. Deletto, supra, 147 Cal.App.3d at p. 473.) Some jurors may have believed Dellinger administered a death blow to the child; others may have been convinced he poisoned her. “It is this unacceptable possibility which taints the verdict in this case.” (People v. Crawford, supra, 131 Cal.App.3d 591, 596.) “While it is of course possible that the jurors agreed unanimously . . ., such agreement would necessarily be fortuitous in the absence of a proper instruction.” (People v. Madden, supra, 116 Cal.App.3d at p. 219.) As the court concluded in Madden, “ ‘Since we cannot say that the jurors agreed unanimously upon the act constituting the offense charged ... we have no assurance that a miscarriage of justice did not occur.’ ” {Ibid.)
IV
The California Supreme Court recently held that a finding of assaultive felony child abuse cannot support a conviction of second degree felony murder. {People v. Smith (1984) 35 Cal.3d 798 [201 Cal.Rptr. 311, 678 P.2d 886].) Dellinger argues that Smith is dispositive and requires reversal. However, it is impossible to determine whether assaultive child abuse was the basis of either the jury verdict or the trial judge’s reduction. We do not need to consider this issue because any new trial will be governed by Smith.
V
Did the trial court err by refusing to admit defendant’s polygraph results ?
The trial court followed the direction of the California Supreme Court by disallowing evidence of defendant’s polygraph examination without the People’s stipulation. {People v. Adams (1975) 53 Cal.App.3d 109, 119 [125 Cal.Rptr. 818].) Dellinger argues that Witherspoon v. Superior Court (1982) 133 Cal.App.3d 24 [183 Cal.Rptr. 615] is now controlling, even though it was decided after his trial. The Legislature, however, has accepted the court’s invitation in Witherspoon to totally exclude polygraph evidence. (Evid. Code, § 351.1.) The decisional law at the time of trial is now codified. In either case, the results of defendant’s polygraph examination were inadmissible.
VI
Did the jury instructions erroneously omit subjective awareness of the risk to life as a necessary element of implied malice ?
Dellinger argues the jury was not advised that implied malice required a finding he was subjectively aware of the risk to life created by his actions. [303]*303He contends the jury was instructed on the awareness of a duty imposed by law rather than the awareness of a risk to life. Specifically, he claims the two jury instructions explaining implied malice, CALJIC Nos. 8.11 and 8.31, were erroneous as a matter of law.
Although not technically inaccurate, the standard jury instructions had injected some confusing and unfortunate language on awareness of a duty imposed by law. This additional verbiage was mere surplusage. However, both CALJIC Nos. 8.11 and 8.31 have been revised and now exclude any unnecessary and confusing language on societal duties. Hence, on retrial, the jury instructions will more clearly define the meaning of implied malice.
The admission of the evidence of the biomedical engineer’s expert opinion, Dellinger’s prior use of cocaine, and Jaclyn’s leg injury, and the failure to advise the jury to unanimously agree on which act constituted murder, require reversal of Dellinger’s conviction.
The judgment is reversed.
Trotter, P. J., concurred.