Evidential Breath Alcohol Testing in Michigan OWI Cases: The Science, the Statutes, and the Limits of the Reported Number

For decades, Michigan's primary evidential breath alcohol analyzer has been the BAC Datamaster and then the Datamaster DMT, both manufactured by National Patent Analytical Systems of Mansfield, Ohio. Michigan has now transitioned to the Intoxilyzer 9000, manufactured by CMI, Inc., of Owensboro, Kentucky. I have been court-qualified as an expert witness on the Datamaster in Michigan, and I now consult and testify on the Intoxilyzer 9000 as well. Both instruments operate on the same fundamental principle—infrared absorption of light by the ethanol molecule—but they differ in their optics, filter configurations, software, and communication messages, and each presents its own categories of foreseeable failure.

Infrared Spectroscopy and the Beer-Lambert Principle

Both instruments work by directing infrared radiation through a sample chamber filled with the subject's breath and measuring how much of the infrared light source is absorbed at specific wavelengths. Different molecules absorb infrared light at characteristic wavelengths corresponding to the vibrations of their chemical bonds, so the absorption pattern can be used to identify molecules and estimate their concentrations. The quantification is grounded in the Beer-Lambert Law, which holds that the quantity of light absorbed is proportional to the concentration of the absorbing molecule in the sample. The relationship is powerful under controlled laboratory conditions, but it depends on assumptions imperfectly satisfied at a police station at three in the morning—chiefly, that the absorption being measured is attributable to ethanol alone, and that the sample in the chamber is a representative aliquot of deep-lung alveolar air rather than a mixture contaminated by mouth alcohol or some other volatile organic compound.

The wavelength region matters because ethanol is not the only volatile organic compound found in human breath that absorbs infrared light in the relevant ranges. The BAC Datamaster measures the carbon-hydrogen bond vibrations in the 3-micron region. The Intoxilyzer 9000 takes a different approach: it measures the carbon-oxygen vibration in the 9-micron region using a pulsed infrared LED source and four single-wavelength filters integrated into a stationary detector. As Mark Thiessen explains, the change in wavelength region is significant because other molecules with carbon-oxygen bonds also absorb infrared radiation in the 9-micron region, and a number of common substances—dimethyl sulfoxide, diethyl ether, and certain esters and ethers used in flavorings and industrial solvents—exhibit absorption profiles that overlap with that of ethanol. Mark Thiessen, The Intoxilyzer 9000, Counterpoint, vol. 1, no. 1 (Fall 2015); see also Dominick Labianca, Breath-Alcohol Analysis: A Commentary on Ethanol Specificity in the 3 Micron and 9 Micron Regions of the IR Spectrum, 24 Journal of Forensic Toxicology 92 (2006).

The Intoxilyzer 9000 also differs from its predecessor in a way that matters for specificity: where the older Intoxilyzer 5000 used a five-filter chopper wheel that included a blank or zero filter, the 9000 has no such true-zero filter and is programmed to read zero if its four detectors do not register a response other than alcohol. As Thiessen observes, an assumed zero is not the same as a measured zero, and an instrument never challenged with a particular interferent cannot meaningfully claim to have ruled it out.

Henry's Law and the 2100:1 Partition Ratio

Even a perfectly specific, perfectly calibrated infrared instrument would not measure blood alcohol. It would measure breath alcohol—a different physical quantity, related to blood alcohol by Henry's Law. Henry's Law holds that, at equilibrium, the concentration of a gas dissolved in a liquid is proportional to its concentration in the gas phase above the liquid. By statutory and regulatory convention in Michigan, that relationship is fixed at 2100:1: one milliliter of blood is presumed to contain the same amount of alcohol as 2,100 milliliters of alveolar breath. The Datamaster and the Intoxilyzer 9000 both use the 2100:1 ratio to convert measured breath alcohol concentration into a reported blood-equivalent number, and both assume that the breath sample is drawn from deep alveolar air and that the subject's actual partition coefficient matches the population convention.

The peer-reviewed literature has documented for many years that the actual breath-to-blood ratio varies widely. As Jan Semenoff catalogues in Reflections on DUI Defenses (Counterpoint-Journal, 2021), measured ratios reported in the scientific literature range broadly across the population, with values published as low as 900:1 and as high as 4200:1, and the ratio also varies as a function of where in the absorptive curve a subject falls at the moment of testing. A subject whose actual ratio is below 2100:1—or who is still in the absorptive phase, when the arterial-venous gradient inflates the breath reading relative to the venous blood—will produce a breath measurement that overstates the true blood alcohol concentration. The 2100:1 convention treats a population approximation as if it were a physical constant, without any individualized assessment of the subject in front of the machine.

The Observation Period and Mouth Alcohol

A breath alcohol instrument is designed to measure ethanol that has crossed from pulmonary capillary blood into alveolar air. It is not designed to measure ethanol residing in the oral cavity, esophagus, or stomach as a result of a recent drink, recent regurgitation, dental work, oral piercings, or alcohol-containing products such as mouthwash or breath spray. Mouth alcohol contamination produces a sample whose ethanol concentration in the chamber is much higher than the alveolar concentration, and a falsely elevated reading is the result.

For that reason, every modern evidential breath testing protocol requires a deprivation or observation period during which the operator must continuously observe the subject and confirm that nothing is placed in the mouth, that there is no regurgitation, and that no alcohol is reintroduced into the oral cavity. A deprivation period is satisfied only by continuous, attentive observation; an operator who steps away to write a report, who watches the subject only intermittently, or who observes the subject's hands rather than the subject's mouth has not satisfied the requirement, and any reading produced after such an interrupted observation is, at best, of unknown reliability.

Manufacturers attempt to backstop the observation period with so-called slope or residual-alcohol detectors that look for the rapid rise-and-fall absorption profile characteristic of mouth alcohol. The peer-reviewed literature, as summarized in Reflections on DUI Defenses, indicates those detectors fail to identify mouth alcohol contamination at rates reported between approximately 37 percent (Harding 1992) and 48 percent (Simpson 2004), with some failures occurring after more than fifteen minutes of deprivation. The slope detector is a useful feature, but it is not a substitute for a properly observed subject.

While conducting the Intoxilyzer 9000 Validation Study for Michigan adaptation, Mark Fondren, Breath Alcohol Technical Leader of the Michigan State Police Forensic Science Division, found that the Intoxilyzer did not properly flag mouth alcohol in numerous cases. Contaminated readings ranged from 37% to 60% above estimated true alveolar BAC. At one point, he blew a 0.138 while actually under the legal limit, and in several other instances, he provided samples in the 0.10 range, which can easily result in a conviction for drunk driving. Even after compliance with the 15-minute observation period, he blew a 0.086 even though his bodily alcohol content was less than the legal limit. Despite these failings, he testified under oath that this proved the machine worked because some of the breath samples had been flagged by slope detection and rejected as mouth alcohol.

The Minimum Sample and the Inability to Provide

Every evidential breath instrument requires the subject to deliver a sample meeting minimum criteria for flow rate, exhalation time, volume, and slope. Subjects who cannot meet those criteria—often older subjects, smokers, subjects of small stature, and subjects with respiratory conditions—are at risk of being charged with refusing the test even when they have made every genuine effort that physiology permits.

The most recent peer-reviewed treatment of this issue is Aaron Olson, Small Samples, Big Problems—The Inability to Provide a Sample in Breath Alcohol Testing: Case Reports, 10 Forensic Science International: Synergy 100584 (2025), discussed in Aaron Olson and Charles Ramsay, Errors in Toxicology Testing and the Need for Full Discovery, 11 Forensic Science International: Synergy 100629 (2025). Olson and Ramsay catalogue recurring problems bearing on the reliability of the volume-and-slope criteria, including manufacturer-acknowledged software errors that caused valid samples meeting the minimum volume to be rejected, the use of single-point calibrations falling outside the range of subject readings, and the refusal of major breath alcohol manufacturers to sell their devices to scientists independent of law enforcement. The discovery problem is structural: breath samples cannot be retained for re-analysis, so the digital expirogram—the time-resolved record of flow rate, ethanol concentration, and total volume during the exhalation—is the only objective evidence of what actually happened during the test. Where the expirogram is preserved and produced, careful analysis can establish whether the subject reached a meaningful plateau, whether the reading reflects a rising or falling slope rather than alveolar air, and whether the volume was sufficient.

Measurement Uncertainty, Calibration Tolerance, and the 2020 Michigan Matter

Every analytical measurement carries an uncertainty, and a forensic measurement reported without an uncertainty estimate is incomplete. Calibration verifications on the Datamaster and the Intoxilyzer 9000 are typically required to fall within plus or minus 0.01 grams per 210 liters of a target value. As Semenoff notes, that single tolerance window itself reflects the inherent uncertainty of the measurement, and broader analyses of total uncertainty have placed the figure substantially higher when biological, instrumental, and procedural variables are combined. A reported result must be understood with that uncertainty in mind, with additional sources of variability—partition-coefficient variation, body temperature, breathing pattern, and the absorptive state at the moment of the test—not captured in the calibration window.

The reliability of breath testing depends not only on the science of the instrument but on the integrity of the people who maintain it. In 2020, the Michigan State Police suspended use of the Datamaster DMT statewide after evidence emerged that contractors hired to perform calibrations had forged records to make it appear that quality assurance checks had been performed when they had not. Olson and Ramsay describe the matter at section 9.3 of their 2025 paper. The episode underscored what defense counsel had long argued: digital data exists on these instruments, can be preserved, and is essential to evaluating any particular result. It also underscored why careful discovery practice—calibration logs, maintenance records, operator certification records, and where available the underlying digital expirogram—is not a luxury in OWI litigation but a basic predicate of fair adjudication.

What Expert Review Can Offer, and a Measured Conclusion

When I evaluate a breath testing case, the work begins with a step-by-step review of the entire test record: the printout, the breath testing log book and instrument maintenance records, the operator's training and certification records, the in-station video covering the full observation period and test sequence, the implied-consent advice, any preliminary breath test record, and any contemporaneous medical or physiological information bearing on the subject's ability to provide a sample. Where digital expirogram data has been preserved, I review it; where it has not, that absence is itself a finding. The result is an analysis identifying every departure from the applicable protocol, every issue with calibration or instrument logs, and every category of measurement uncertainty the prosecution will not address in its case-in-chief. None of that work guarantees a particular outcome. What careful review can promise is that the number on the printout will not pass through the case unexamined.

Evidential breath testing is a useful forensic tool. It is also an indirect, assumption-laden estimate that must be evaluated with the same scientific discipline any other measurement deserves. In Michigan, MCL 257.625 places enormous weight on the reported number, and that weight is justified only when the protocol has been followed, the instrument has been properly maintained, and the subject's individual physiology has not made the population-average assumptions of the test inapplicable. When any of those conditions has failed, the case calls for careful, case-by-case review.