Operator error in breath tests?

The Intoxilyzers didn’t fail. They faithfully detected the presence of alcohol when 1,147 drivers blew into them.

But the mistakes made by humans were so basic – did anyone notice, for example, that 0.009 is a bigger number than 0.005 – that they raise questions about training and oversight in the Philadelphia Police Department.

And at least one defense attorney, a DUI expert in Harrisburg with scientific training, argues that Pennsylvania’s regulations are too weak to ensure that basic errors are caught and fixed statewide – an assertion that some in law enforcement dismiss.

City officials announced last month that police had been using four breath-testing instruments whose accuracy had not been proved in mandatory checks. The evidence in drunken-driving cases between September 2009 and November 2010 would be reviewed, District Attorney Seth Williams said.

Most convictions for driving under the influence rely on far more than a breath-alcohol test, which normally is administered after an arrest to get a measurable reading that will seal the case. Although the review will be a costly headache for overworked prosecutors and judges, most cases are unlikely to be thrown out. Letters were sent several weeks ago to lawyers for 416 defendants whose cases involved one instrument that was identified in February. None of them has requested a new trial, the District Attorney’s Office said Friday.

In fact, it is not clear that the instruments in question registered an over-the-limit blood-alcohol concentration when they shouldn’t have, or even that they were all calibrated incorrectly. The errors were in certifying that they were correct, so now there is no way to know that a 0.08 – the legal definition of drunk – was, in fact, a 0.08.

Thomas E. Workman, who teaches scientific evidence at the University of Massachusetts Dartmouth School of Law, used this analogy: “If you don’t have an inspection sticker, it doesn’t mean that the car is bad. It just means that you can’t verify that the car is working correctly.”

The Intoxilyzer 8000, a distant descendant of the original Breathalyzer (and the earlier drunkometer), is one of the newest and most common breath-testing instruments used for evidence by police around the country. It is the size of a toaster oven, weighs 17 pounds and typically is plugged in at the station. CMI Inc. of Owensboro, Ky., manufactures and sells the device for around $6,000 to $8,000 depending on configuration, features, and accessories.

The Philadelphia Police Department, which files 8,000 to 10,000 DUI cases a year, owns about eight of them.

The Intoxilyzer relies on the technology of infrared spectrometers, instruments that are widely used in chemistry laboratories to identify compounds by detecting how molecules absorb infrared light.

Ethanol, for example, absorbs light at specific frequencies. When you blow into the Intoxilyzer, it detects how much light has been absorbed by your breath at known wavelengths and converts that into a measure of blood-alcohol concentration.

Just like a violin is designed to play fine music but must be tuned to produce a “C,” these instruments are built to detect alcohol but must be trained to recognize a concentration of, say, 0.10. The device is then considered to be calibrated.

And just like a violin’s tuning would be checked before every concert, and adjusted if necessary, a breath-tester is regularly checked to ensure that its calibration is correct.

“Every instrument is going to have drift in it. The calibration of those instruments is not going to stay dead on in the course of the year,” said Ted Vosk, a defense attorney and DUI consultant near Seattle.

The calibration of a device like the Intoxilyzer is checked by running through it a fluid that is known to simulate a particular breath-alcohol concentration. Pennsylvania law requires several levels of checks. The most rigorous, done annually or before an instrument is returned to service, requires simulations at three concentrations, typically 0.05, 0.10, and 0.15.

Perfection is a scientific impossibility; going out several decimal places will always produce a reading above or below. A key issue is how much it varies.

To certify the calibration of an instrument in Pennsylvania, each simulation must be run five times. The amount of each deviation is recorded: a reading of 0.143, for example, would be 0.007 deviation from a perfect 0.15. The five deviations are then added up and averaged. An average that exceeds 0.005 is one finding that requires the instrument to be taken out of service.

In Philadelphia, two instruments were incorrectly certified as 0.005 – the average deviations were actually 0.0054 and 0.0058 – by leaving off the fourth decimal place, said Lynn Nichols, an assistant district attorney who is leading the review of cases. A third was off by 0.009 but was certified anyway. A fourth was done correctly but not all the calculations were shown in the paperwork, Nichols said, so that evidence is being considered invalid as well.

Some defense attorneys argue that scenarios such as this are appearing more frequently because lawyers and judges, intimidated by the complexities of science, historically allowed expert witnesses to make their case without the documentation to prove it.

A U.S. Supreme Court ruling in 1993 directed federal trial judges to consider certain evidence behind experts’ conclusions. A 2009 National Academy of Sciences report that lamented a lack of standards in forensic science cited as an example the appropriate reporting of ranges of deviations for instruments such as the Intoxilyzer.

It also pointed to wide disparities in standards and training from state to state.

Challenges to breath-testing equipment similar to Philadelphia’s have recently compromised thousands of cases in Florida; Alaska; Washington, D.C.; and Washington state. In some, bad simulator solution was used to calibrate the instrument. In others, technicians were said to have signed off on work that they did not examine.

“The majority of the time when these kinds of things happen, it is a lack of training,” said C. Stephen Erni, executive director of the Pennsylvania DUI Association, which is affiliated with law enforcement. Erni sees Philadelphia as an anomaly.

Justin McShane, a Harrisburg-based defense lawyer with training in spectroscopy, said he believed that the problem was systemic.

“Pennsylvania is the least-regulated [state] in terms of the verification of data in the United States,” he said.

Some states also centralize the purchase of breath-testing equipment, McShane said, whereas here every county or police department buys its own, so long as the device is on a state-approved list. Some states post all certification data online and also employ technical supervisors to double-check local results, he said, comparing those moves to the “double redundancy” norm of the pharmaceutical industry.

George C. Geisler Jr., a director of law enforcement services for the DUI association, dismissed McShane’s criticism as “absolutely untrue.” Philadelphia’s situation “was essentially caused by the carelessness of one individual,” Geisler said.

Frank Baranyai, a breath-test instructor for Pennsylvania’s Institute for Law Enforcement Education, also disagreed. Training, calibration, and accuracy in Pennsylvania are all “pretty much the same as in other states,” he said.

But procedures may vary.

Baranyai is on contract to certify the calibrations of all 11 breath-testers in Crawford County. The paperwork is available but there is no one checking on him.

“Once I complete the form, it goes into a logbook with that particular unit. No one else checks,” said Baranyai, a retired police officer.

But they tend to concur on one key point: It is very difficult to intentionally fool the machine.

Involuntary impacts

To whatever extent a measure of blood-alcohol concentration can be thrown off by an individual, it is generally involuntary – and it usually raises the reading.

Because the instruments detect alcohol, defense lawyers say, they can react to someone who works around ethanol (as in aircraft maintenance); to a diabetic in a dangerous state of ketoacidosis (the body may produce isopropanol); or to a person with acid reflux (alcohol in the stomach is far more concentrated than the blood level that is measured on the breath).

Prosecutors maintain that filters in the instrument flag such interfering substances.

Intentional manipulation

Pennies under the tongue and many other claims are urban myths.

One trick that could lower the reading – at least in theory – is hyperventilating, which gets rid of some of the alcohol in your lungs, before you blow into the machine. (Holding your breath could raise it.)

But police are supposed to observe suspects for 20 minutes before the test, and then have them blow twice. Plus, intoxication makes it harder to pull off. “I could tell a client, ‘If you are arrested, do this,’ but if they are drunk, they are not going to think to do these things,” said Thomas E. Workman, a DUI lawyer in Massachusetts.

– Don Sapatkin