Vicki E. Cody, Assistant Federal Defender, Sacramento, California, for the defendant-appellant.

    Mark J. McKeon, Assistant United States Attorney, Sacramento, California, for the plaintiff-appellee.

NOONAN, Circuit Judge:

    Kevin Vincent Brannon appeals his conviction of driving under the influence on a military base in violation of the Assimilative Crimes Act, 18 U.S.C. §13, which incorporates California Vehicle Code (CVC) §23152(a).

    Shortly after 1:00 a.m. on June 6, 1996, Brannon drove to McClellan Air Force Base in Sacramento County, California for the purpose of dropping a passenger off at a club on the base. Airman Sherri Wellman stopped Brannon at the gate and requested military I.D. No one in the car had the proper I.D. Brannon requested and received permission from Wellman to make a U-turn to leave the installation. However, he did not follow Wellman's instructions as to where to make the U-turn and was stopped by Staff Sergeant David Trudel.

    There was an open container of alcohol in the backseat from which one of Brannon's passengers was drinking. Trudel detected an odor of alcohol coming from Brannon's breath and asked Brannon to perform several field sobriety tests. In Trudel's opinion the effects of alcohol on Brannon were obvious. Brannon's performance of the tests was unsatisfactory, and he begged off completing them on the ground of an old knee injury. Staff Sergeant Eddy Sierra was called as back-up. He transported Brannon to a law enforcement facility on the base to determine the level of alcohol in his blood. Brannon was given the choice of a breath, blood or urine test. He elected the breath test, which was administered by Sierra using an Intoxilyzer 5000.

    The Intoxilyzer works by shining a beam of infrared light through a chamber full of uncontaminated air. The rate of absorption of the infrared light by the air is known. When new air is breathed into the chamber, the amount of light will be decreased if there is alcohol in the breath, because alcohol, apparently colorless, has a color just past red in the rainbow. The decrease in light due to alcoholic breath is measured and converted into an alcohol percentage. As the blowing continues, the percentage will rise if there is alcohol in the breath. The first breaths are mixed with air in the mouth, room air. Only as the blowing goes on does alveolar air, deep air at the bottom of the lungs in equilibrium with the blood, get into the air chamber. For that reason, if there is alcohol in the breath, the percentage registered will rise; it will never decrease. The usual length of a single test is seven seconds.

    Brannon was told to blow into the machine hard enough for it "to register," that is, give a steady whistle tone. He was to continue to blow until the whistle went away. Sierra inserted into the machine a card on which the test results would be printed.

    Brannon began to blow. After 3 seconds, the machine started registering. It registered .15 for 2 seconds. Brannon stopped blowing. The machine stopped registering and printed "Deficient Sample." After a few minutes the machine automatically purged itself and was ready for a second try. Brannon was told to blow again. He puffed out his cheeks but put no air in the machine. The machine spit out the card, printing "Subject Test Refused." Brannon started coughing and said that he had bronchitis. Sierra understood Brannon to be refusing to take the test. Sierra offered him no alternative test.

    Brannon was charged in a five-count information under the Assimilative Crimes Act, 18 U.S.C. §13, with driving while under the influence of alcohol, in violation of CVC § 23152(a); driving with a blood alcohol content of .08 percent or more, in violation of CVC § 23152(b); possessing an open container of alcohol while driving, in violation of CVC § 23222(a); making an illegal U-turn, in violation of CVC § 21461(a); and driving without a valid license, in violation of CVC §12500(a).

    Brannon consented to a trial before a federal magistrate judge. At the beginning of the trial, Brannon orally moved to suppress the evidence of the .15 digital display from the Intoxilyzer on the grounds that it was a partial test result without scientific reliability. He argued that only a complete test and printout could give a reliable reading of blood alcohol content. The magistrate judge deferred ruling on Brannon's motion until hearing from the government's expert witness Robert Lee Baker, a criminalist employed by the California Department of Justice for over 18 years and currently employed by Valley Toxicology. Baker described the workings of the Intoxilyzer 5000 as summarized above. He also testified that the Intoxilyzer 5000 in use at the base was checked every ten days and that, both before and after Brannon's use of it, it was in good working order. The magistrate judge denied the motion.

    The magistrate judge found Brannon guilty of driving under the influence and sentenced him to pay a fine of $600 and a two-year term of limited supervised probation which would terminate upon payment of the fine and obtaining a valid driver's license. (Upon a probation revocation proceeding during the pendency of his appeal, Brannon's probation was modified to 90 hours of community service.) The charge of driving with a blood alcohol level of .08 or more was dismissed by the government prior to trial. Brannon was acquitted on all of the other counts.

    Brannon appealed his conviction to the district court, which upheld the admissibility of the Intoxilyzer test and held that there was sufficient evidence to support Brannon's conviction.

Brannon appeals.

    The case turns on whether a reading on an uncompleted breathalyzer test is admissible evidence. The defense stipulated to Baker's "qualifications as an expert with the one exception of the contested evidence, the partial breathalyzer results." The defense followed up its in limine objection to admission of the breathalyzer evidence with the argument that Baker's testimony did not show that an uncompleted test had reached "the level of general acceptance within the scientific community," did not cite any studies to the effect that a partial test was reliable, and did not offer any proof that the machine was accurate when the test was not completed. On appeal, the defense argues that Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993) was not complied with.

    The case is close. Twenty-five years ago breathalyzers were certified as accurate by the National Highway Traffic Safety Administration of the Department of Transportation. 38 Fed. Reg. 30459 (1973). Their methodology is well-known and unchallenged. See California v. Trombetta, 467 U.S. 479, 481, 489 (1984). Baker is stipulated to be an expert on complete breathalyzer tests.

    The defense treats an uncompleted test as something different from what is the acceptable norm. A single uncompleted test is not what the scientific literature recommends. See, e.g., Kurt M. Dubowski, Quality Assurance in BreathAlcohol Analysis, 18 J. Analytical Toxicology 306 (1994); 2 Richard E. Erwin, Defense of Drunk Driving Cases § 18.03 (3d ed. 1995). But as the method by which a breathalyzer works is unchallenged, the objection to an incomplete test goes not to the scientific technique employed but to the reliability of the result reported. Daubert does not forbid admission of the report. Baker's testimony "pertain[s] to `scientific knowledge.' " Daubert, 509 U.S. at 590.

    An analogy may be helpful. Suppose a child's temperature is taken in the old-fashioned way by thermometer. The doctor says to keep the thermometer under the tongue for three minutes. The child spits out the thermometer after 60 seconds. The reading is 101 degrees. More likely than not, most people would agree, the child has a fever. The scientific basis for temperature-taking is well-established. The methodology of the thermometer is well-known. The incompleteness of the test would not invalidate the knowledge or the methodology.

    Every scientific method known to humanity is subject to error, both mechanical and human. The best way of running a breathalyzer is to have a period of observation of the subject fifteen minutes before the test, to have blank tests preceding each of two actual tests, and to run a control test following the actual test. Dubowski, supra, at 308. Omission of these safeguards lowers the reliability of the tests, it does not eliminate them as scientific aids to a factfinder investigating drunkenness.

    As the defense pointed out at trial, Sierra, the operator of the machine, could have been mistaken in saying that he saw a reading of .15. The possibility was appropriately explored by cross-examination. It is not a reason to exclude Sierra's testimony.

    California, like several other states, requires that there be two complete tests for breathalyzer evidence to be admitted. Cal. Code Regs. tit. 17, §1221.4(a)(1). We are not bound by this procedural requirement. United States v. DeWater, 846 F.2d 528, 529-30 (9th Cir. 1988). Prudent though the California rule may be, we do not believe that it is within our power to add it to the requirements of the Federal Rules of Evidence.

    No other federal appellate court appears to have considered the precise question we address here. Several state supreme courts have held partial breathalyzer results to be admissible. E.g., Temple v. State, 679 So.2d 611 (Miss. 1996); People v. DeMarasse, 647 N.E.2d 1353 (N.Y. 1995); cf. State v. Kennedy, 657 A.2d 773 (Me. 1995). We join in their reasoning that the deficient sample did not make the test invalid and that any error was likely to favor the defendant. We add that in the present case the breathalyzer result does not stand alone but coheres with the observations of the officers and with the belated excuses Brannon made to avoid completing the field sobriety tests and the breath test.

WALLACH, Judge, dissenting:

    I dissent. An incomplete breathalyzer test is not admissible under the circumstances here present. [1] In Daubert v. Merrell Dow Pharmaceuticals, Inc., 113 S.Ct. 2786 (1993), the Supreme Court stated that "under the [Federal ] Rules [of Evidence] the trial judge must ensure that any and all scientific testimony or evidence admitted is not only relevant, but reliable." Id. at 2795. Here, the majority upholds the admission of this evidence based on a hurried observation [2] of a flashing LED readout. [3] Experts recommend, and some states have adopted, admission of such breath alcohol evidence when there are two completed tests with printouts of the results. Under the Daubert standard, the government failed to show that the results of the partial breathalyzer test were reliable.

    As recognized by the majority, experts in the industry recommend against reliance on a single uncompleted test because such results may be unreliable. See Opinion at 7793 (citing Kurt M. Dubowski, Quality Assurance in BreathAlcohol Analysis, 18 J. Analytical Toxicology 306 (1994); 2 Richard E. Erwin, Defense of Drunk Driving CasesS 18.03 (3d ed. 1995)). Even a completed test is unreliable if, inter alia, the test subject has used breath-fresheners or mouthwash or has regurgitated gastric content. See Dubowski, supra, at 309.

    Indeed, some states have implemented requirements regarding the administration of breath alcohol tests to assure their reliability, such as California's requirement set forth by the majority. See Opinion at 7794. These recommendations and requirements illustrate the unreliability of incomplete breath alcohol tests and provide a sound basis for rejecting results from such a test.

    The two state cases cited by the majority are clearly distinguishable. In Temple v. State, 679 So.2d 611 (Miss. 1996), the Mississippi Supreme Court upheld admission of an uncompleted digital intoxylizer reading of alcoholic blood level where the defendant failed to blow long enough. Id. at 611. The court stated it was "dealing with a machine and people trained to calibrate it and administer it whose credentials to verify its accuracy at some interim period may be questioned", id. at 613, and noted it was "problematic to allow incomplete results where the machine's testing mechanism has not run its full course." Id. However, the court found that on the record before it, "with no real attack on the credentials of the witnesses, or their experience in these circumstances, and no objection to their opinion testimony, the decision to admit the evidence will not be reversed. " Id.

    Here, defense counsel repeatedly objected to the admissibility of the incomplete test results. On this record it appears the Supreme Court of Mississippi would not have upheld the admission of the test.

    In People v. DeMarasse, 647 N.E.2d 1353 (N.Y. 1995), the Court of Appeals for New York found the results of an Intoxilyzer test that printed a reading of .217 and the words "deficient sample -- value printed was highest obtained" was admissible because the People had established a sufficient foundation. See id. at 1354. The foundation consisted of testimony that the machine was in proper working condition, the test was properly administered, and testimony that "in Intoxilyzer terminology a `deficient' sample is not an invalid sample, but merely one in which the subject did not breathe for a long enough period of time to reach sufficient deep lung air to give the most accurate reading obtainable." Id. Here, no printout registers any reading.

    Because the evidence of Intoxilyzer results lacked an adequate foundation and its reliability was highly questionable, it was inadmissible under Daubert. The issue is not one of weight; it is about the essence of admissibility of scientific evidence. Since the trial judge was not convinced of the defendant's guilt based on the remaining evidence, I would reverse.

[*] The honorable Evan J. Wallach, Judge, United States Court of International Trade, sitting by designation.

[1] The majority's analogy to the use of an "old-fashioned" thermometer is illustrative of the central problem with its opinion. See Opinion at 7793. If one of the new LED thermometers is, according to instructions, supposed to be placed against a child's forehead for fifteen seconds, a doctor is hardly justified in assuming that a two second measurement gives an "adequate" reading and that the temperature could only go higher. To support a scientific basis for admissibility one must presume that the manufacturer knows the limits of its own testing devices. Out of such assumptions do malpractice cases arise.

[2] The housing patrol officer who observed the number testified that it appeared for "maybe two seconds". Appellant's Excerpt of Record 93.

[3] The machine normally produces a printout of numerical results. The one produced in evidence stated on its face "DEFICIENT SAMPLE". Appellant's Excerpt of Record 194.