Opinion

On its own motion this court transferred this case to itself for hearing and decision after the Appellate Department of the Orange County Superior Court reversed Bruce Robert McDonald’s convictions for driving under the influence of alcohol and driving with a blood-alcohol content of .10 percent or more. McDonald contends the trial court erred in its instructions to the jury on the use of the blood-to-breath partition ratio in determining the accuracy of a breath test. He also argues he was denied the effective assistance of counsel. We reverse McDonald’s convictions.

I

McDonald was tried by jury for the misdemeanor offenses of driving under the influence of alcohol (Veh. Code, § 23152, subd. (a)) and driving with a blood-alcohol content of .10 percent or more (Veh. Code, § 23152, subd. (b)). He was also charged with having suffered two similar prior convictions, although we do not know whether these allegations were later proven. At trial, California Highway Patrol Officer Pat Dillon testified he detained McDonald at about 1:30 a.m. on October 11, 1986, after observing him driving erratically. McDonald smelled strongly of alcohol, his eyes were red, watery, and bloodshot, he fumbled for his license in his wallet and his gait was unsteady. He admitted drinking five or six beers but felt he was “O.K. to drive.” After McDonald failed a field sobriety test Dillon arrested him for driving under the influence of alcohol (Veh. Code, § 23152, subd. (a)). McDonald chose to take a breath test, which registered .13 percent at 2:40 a.m.

Although Dillon had not mentioned a nystagmus test during direct examination, defense counsel asked him if he had performed any tests other than those mentioned during direct. Dillon answered negatively, but added that he had examined McDonald’s eyes. Thereafter, the parties entered into the following stipulation: “Deputy District Attorney David Kirkpatrick instructed California Highway Patrol Officer Pat Dillon not to mention the ‘nystagmus test’ or to make any mention of the word ‘nystagmus’ because it has been Deputy District Attorney Kirkpatrick’s experience before the judges of the North Orange County Municipal Court that the ‘nystagmus test’ or the term ‘nystagmus’ is not legally admissible and might even result in a mistrial unless first raised by the defense attorney.”

Dillon was recalled as a witness for the prosecution and described giving McDonald a nystagmus test before placing him under arrest. Although he made no mention of that test in his written police report Dillon recalled that McDonald showed signs of nystagmus. Dillon used that information in forming his opinion McDonald was under the influence of alcohol.

The prosecution’s expert witness, Russell Mumford, a criminalist at the county crime laboratory, opined that all persons are under the influence of alcohol for purposes of driving if they have a blood-alcohol level of .08 percent or higher. He explained the breath testing device relies on a blood-to-breath partition ratio of 2,100:1. According to Mumford, that figure is based upon the general public’s average partition ratio and is mandated by title 17 of the California Code of Regulations. Mumford first testified each person’s partition ratio remains constant, but then admitted the ratio could be affected by illness, medication, and environmental factors such as air pollution. He also acknowledged that although the average is 2,100:1, a given individual’s blood-to-breath partition ratio could be as high as 2,700:1 or as low as 1,550:1. Mumford denied knowing of any scientific studies which addressed the issue of whether a person’s partition ratio was constant. He testified that Dillon properly administered the nystagmus test to McDonald, and further observed that McDonald’s condition and driving pattern, as described by Dillon, were consistent with being under the influence of alcohol.

The defense called its own criminalist, Henry Greenberg. Greenberg testified that scientific studies indicate a person’s blood-to-breath partition ratio varies over time and does not remain constant. He agreed that the breath testing device relies on a ratio of 2,100:1 as mandated by the California Code of Regulations. Greenberg further testified that while an individual’s blood-to-breath partition ratio could be measured, it would be speculative to say the same ratio existed at an earlier time when the subject took a breath test to measure his blood-alcohol content. He was never asked to measure McDonald’s partition ratio, and opined it would be costly to conduct a series of tests to determine the variability of a person’s partition ratio. Greenberg agreed McDonald’s symptoms were consistent with the breath test result of .13 percent blood alcohol. He also admitted that, according to one 1980 study, the breath testing device McDonald used was accurate 85 percent of the time. He disagreed with Mumford on two additional points, however. Greenberg testified everyone is under the influence of alcohol for purposes of driving at a blood-alcohol content of .10 percent or more, not .08 percent. Moreover, he said Dillon did not administer the nystagmus test correctly.

The jury was given the standard CALJIC instructions relating to the charged offenses and explaining how to view the results of the breath test. (CALJIC Nos. 16.830, 16.830.1, 16.831, 16.832, 12.61, 12.61.1.) In addition, according to the court file, the court instructed, on request of the prosecution, that “[t]he defendant is presumed to have a 2100 to 1 breath to blood partition ratio unless he presents evidence as to his personal ratio which raises a reasonable doubt that the 2100 to 1 ratio is not valid for him. General evidence of such a possibility of error in the partition ratio will not suffice to rebut this presumption.” This instruction was quoted differently in the settled statement on appeal: “The defendant is presumed to have a 2100:1 partition ratio unless he presents evidence as to his personal ratio which establishes that the 2100:1 ratio is not valid for him. That general evidence of such a possibility of error in the partition ratio will not suffice to rebut this presumption.”

The court rejected this related instruction requested by the defense: “In determining the accuracy of a reported blood alcohol level calculated from an analysis of the amount of alcohol in a sample of breath, you are instructed that the law of this state requires that the blood alcohol level be reported to you as 2100 times the breath alcohol level which was measured. [[[] The law does not require that you accept that reported blood alcohol level as correct. It is an expert opinion, and you may examine the facts it is based on before deciding whether to accept or reject it. [fl] In determining the facts the opinion is based upon, you are instructed that all of the evidence admitted during this trial was relevant evidence, that no evidence was admitted which you are not permitted to use, and that none of the evidence should or must be ignored by you.”

McDonald was convicted of both driving offenses and placed on probation on condition, inter alia, he serve 120 days in the county jail. The jail term has been stayed pending final resolution of this appeal.

Before the Orange County Superior Court appellate department, McDonald argued the court erred in its instruction on the partition ratio and in denying his motion to exclude certain statements because of a claimed Miranda violation. (Miranda v. Arizona (1966) 384 U.S. 436 [ 16 L.Ed.2d 694 , 86 S.Ct. 1602 , 10 A.L.R.3d 974 ].) He also contended he was denied the effective assistance of counsel as evidenced by trial counsel’s handling of the nystagmus evidence. In response to our request for further briefing McDonald raises only two issues: the claimed erroneous jury instruction and the effective assistance of counsel question, We conclude the instruction given was erroneous and reverse McDonald’s convictions without discussing the other issues raised on appeal.

II

The parties and the record are in dispute as to the exact wording of the instruction given. We conclude, however, either version was erroneous under the particular facts of this case. Although three published decisions from the Appellate Department of the Los Angeles Superior Court appear to conflict with this holding, closer analysis reveals an important factual distinction.

People v. Pritchard (1984) 162 Cal.App.3d Supp. 13 [ 209 Cal.Rptr. 314 ] explained that “[i]n order to convert the breath reading to that of a blood-alcohol reading for purposes of Vehicle Code section 23152, subdivision (b), a partition ratio must be used. A blood-breath ratio of 2,100 to 1 was used as mandated by title 17 of the California Administrative Code (Cal. Admin. Code, tit. 17, § 1220.4, subd. (f)). This is called the partition coefficient or partition ratio. However, there can be an error factor of 10 percent in the partition ratio of 2,100 to 1.” (Id., at pp. Supp. 15-16.) Distinguishing People v. Campos (1982) 138 Cal.App.3d Supp. 1 [ 188 Cal.Rptr. 366 ], the appellate department concluded the results of the breath test which measured .11 and .12 percent were sufficient evidence of guilt despite the 10 percent margin of error because the defendant did not present any evidence his personal partition ratio made the 2,100:1 ratio invalid as to him. The Pritchard court stated, “the individual partition ratio is an individual physical characteristic of the defendant, and presumably can only be obtained through means of physical tests upon him.” (People v. Pritchard, supra, 162 Cal.App.3d at p. Supp. 16.) Because of the defendant’s exclusive access to the information, the Pritchard court concluded he had the initial burden of producing evidence on the issue. And, because the defendant failed to present any evidence his partition ratio differed from the norm, the presumption his ratio was 2,100:1 was sufficient evidence to convict, despite a 10 percent margin of error.

Two weeks later the Los Angeles Superior Court Appellate Department filed its opinion in People v. Gineris (1984) 162 Cal.App.3d Supp. 18 [ 209 Cal.Rptr. 317 ]. There too the defendant challenged the sufficiency of evidence to convict him of violating Vehicle Code section 23152, subdivision (b). Gineris registered a breath test result of .11 percent. “The People’s expert testified that taking into account the possible variations in the human partition ratio, the time interval between the driving and the test, and the error margin of the machine reading, he could not say beyond a reasonable doubt that appellant had a 0.10 percent blood-alcohol level at the time of driving.” (Id., at p. Supp. 22.) The court recognized the constitutionality of the rebuttable presumption that the defendant had a blood-alcohol level of .10 percent or more, as prescribed in Vehicle Code section 23155, subdivision (a)(3). And, again adopting the so-called rule of convenience, the Gineris court held the evidence was sufficient to convict, despite the expert’s reasonable doubt, because the defendant had not presented any evidence his personal partition ratio varied from the norm. As in Pritchard, that conclusion followed the court’s statement that “[t]he defendant’s individual partition ratio is presumably obtained by means of an examination which he is best able to obtain by submitting himself privately for such scrutiny.” (Id., at p. Supp. 23.)

The appellate department considered a related question in People v. Herst (1987) 197 Cal.App.3d Supp. 1 [ 243 Cal.Rptr. 83 ]. There, a defense expert “testified that partition ratios varied and that during the absorption of alcohol the partition ratio was considerably lower than 2,100 to 1. Defendant, however, presented no evidence as to her partition ratio.” (Id., at pp. Supp. 3-4.) The trial court struck the expert’s “general evidence,” a ruling the appellate department upheld, citing Pritchard. More importantly, the court found no error in instructing the jury in language nearly identical to that of the instruction given here, stating it satisfied Pritchard and Gineris. (Id., at p. Supp. 4, fn. 1.) The Herst court rejected the claim the partition ratio of 2,100:1 “impermissibly relieved the prosecution of ‘proving beyond a reasonable doubt each and every fact upon which the proof of defendant’s blood alcohol level depended.’. . . The instruction did not shift the burden of proof for an element of the offense, as in [People v. Roder (1983) 33 Cal.3d 491 ( 189 Cal.Rptr. 501 , 658 P.2d 1302 )]. Rather, the instruction concerned the equivalent of an affirmative defense that defendant’s partition ratio deviated from the norm.” (People v. Herst, supra, 197 Cal.App.3d at p. Supp. 4.)

Of the three appellate department decisions, only Herst addresses the question raised here, the propriety of the instruction given. The language of the instruction, however, appears to come directly from Pritchard. (People v. Pritchard, supra, 162 Cal.App.3d at p. Supp. 17.) But in none of the three cases was there admissible evidence that a person’s partition ratio is not constant and may vary over time. Similar evidence was disallowed in Herst , but apparently no objection was made to the evidence presented by the defense expert here. Indeed, Greenberg disputed the notion that a defendant could measure his or her partition ratio because it would only be speculative that the ratio had been the same at the time of the prosecution’s breath test. Although the prosecution’s expert claimed one’s partition ratio was constant, he contradicted himself by admitting the ratio could be affected by external factors which vary over time. The significance of Greenberg’s testimony cannot be ignored because it directly contradicts facts Pritchard, Gineris and Herst presumed: that one’s partition ratio is constant and can be measured by the defendant. The rule of convenience should not apply unless the “exonerating fact is peculiarly within the defendant’s knowledge . . . .” (People v. Pritchard, supra, 162 Cal.App.3d at p. Supp. 16; see siso People v. Montalvo (1971) 4 Cal.3d 328, 334 [ 93 Cal.Rptr. 581 , 482 P.2d 205 , 49 A.L.R.3d 518 ].) As in Montalvo , the defense here does not “have any substantially greater ability to establish [the exonerating fact] than does the prosecution.” (People v. Montalvo, supra, 4 Cal.3d at pp. 334-335.)

We therefore hold it was error to instruct the jury that it should presume McDonald had a 2,100:1 partition ratio unless he presented evidence as to his personal ratio which either proved the contrary or raised a reasonable doubt. The effect of the instruction was to nullify Greenberg’s testimony even though it was admitted without objection. The fact presumed by the instruction—constancy of one’s partition ratio—was in doubt. The resolution of that doubt was a factual question for the jury. Yet the court’s instruction effectively took that question from the jury and cast in stone a fact not proven.

Both the prosecution and amici curiae on its behalf urge we consider scientific evidence on these questions by incorporating into the record exhibits not offered at trial. They also suggest we appoint our own expert to resolve the factual question. An appellate court does not ordinarily decide factual issues. More importantly, we must decide the legal question presented—the propriety of the court’s instruction—in light of the facts before the jury. If Greenberg’s testimony is incorrect, or not generally accepted in the scientific community (a conclusion we doubt in light of the testimony of the prosecution’s own expert witness) we presume the next prosecutor will object accordingly in the trial court.

The erroneous instruction is not harmless and it infected both counts charged. Although only the subdivision (b) count of Vehicle Code section 23152 relates directly to the blood-alcohol content, the breath test results were offered in support of the subdivision (a) count as well because the jury was instructed it could presume McDonald was under the influence if his blood-alcohol measurement was .10 percent or more. (CALJIC No. 12.61.) Thus, the reliability of the breath test results was critical on both counts and the error contaminated each. In light of these conclusions we need not address appellant’s other contentions.

Judgment reversed.

Scoville, P. J., and Sonenshine, J., concurred.

Respondent’s petition for review by the Supreme Court was denied March 30, 1989.