Breathalyzers: 40% Margin of Error is Acceptable Accuracy?
Posted by Lawrence Taylor on June 20th, 2006With more than a little federal coercion, all states have now passed laws making it a criminal offense to drive with .08% alcohol in your blood. And most people suspected of violating that law are given breath tests to determine their blood-alcohol concentration (BAC).
The breathalyzer will take a small sample of the suspect’s breath and estimate how much alcohol is in it — and, then, estimate how much may be in the blood. And what that machine says is pretty much the end of it. There will be no second tests. There will be no cross-examination of the machine.
Are these machines so reliable and accurate that we have permitted them to become judge and jury and to determine guilt beyond a reasonable doubt?
Ignoring the many flaws of the machines for the moment (see “How Breathalyzers Work — and Why They Don’t”), scientists universally recognize an inherent error in breath analysis, generally of plus or minus .01%. That means that if everything is working perfectly (an unlikely scenario), a .13% breathalyzer test result can be anywhere from .12% to .14%. This has been acknowledged by courts across the country (see, for example, People v. Campos, 138 Cal.Rptr. 366 (California); Haynes v. Department of Public Safety, 865 P.2d 753 (Alaska); State v. Boehmer, 613 P.2d 916 (Hawaii), recognizing an even larger .0165% inherent error).
What does that tell us about the accuracy of these breathalyzers? Well, let’s take a common test result of .10%. Taking inherent error into consideration — and assuming the machine was working perfectly, the officer administers the test correctly, and the suspect’s physiology is normal and perfectly average — the true BAC could be anywhere from .09% to .11%. In other words, the true BAC can be 10% in either direction — or, put another way, anywhere within a 20% margin of error.
These machines have a 20% margin of error?
That’s right. A person accused of driving with over .08% BAC can be convicted by a machine whith a built-in 20% margin of error. Would you be comfortable with an airline pilot who worked with a 20% range of error? With a bank teller who gave you $90 cash for a $100 check? How about the sole evidence in a criminal case where guilt must be proven beyond a reasonable doubt?
But it gets worse. Most states have laws which establish standards for breath alcohol analysis. These set forth minimum levels of accuracy in a given test, usually determined by the requirement that two separate tests produce results within a given range. California’s requirements, for example, are fairly typical: to be admissible in court there must be two test results that are within .02% of one another.
What does that mean? Well, let’s again assume that you breathe into the machine and it produces a .10% reading. You are now required to breathe into the machine a second time. What test result would be necessary for the evidence to be considered acceptably accurate and legally admissible? .08? .09? 10? .11? .12? Actually, any of these readings would satisfy the .02% requirement: anything from .08% to .12% would render the test results “accurate”.
In other words, a 40% range of error in that second test is deemed sufficiently accurate to sustain the prosecution’s burden of proof “beyond a reasonable doubt”.
Close enough for government work…in a DUI case.
However, that is an atypical circumstance, and those error ranges are designed to flag the wayward instrument. If you were to look at a large population of replicate results, a very high majority of those would be well within the 0.02 g/dL range. I would be interested in seeing that data, but I doubt it would serve your purpose much.
In addition, your example improperly refers to a 40% range of error in the second test. A more accurate depiction would be ± 0.02 g/dL error, and the difference is important. In your example, the replicate readings can be 0.10 g/dL and any one of the following: 0.08 g/dL (20% less), 0.09 g/dL (10% less), 0.10 g/dL (no difference), 0.11 g/dL (10% more), and 0.12 g/dL (20% more). Those numbers are dependent upon the fact that you chose 0.10 g/dL as your benchmark. A higher (or lower) BAC will change those percentages, because the ± 0.02 g/dL requirement is fixed, regardless of the BAC.
Comment by Tox Lab Guy — June 21, 2006 @ 11:03 am
Dear Mr. Taylor:
I practice in Pennsylvania and the district attorneys rely on Commonwealth v. Rick, 366 A.2d 302 (1976) to circumvent the hearsay rule. In the above-referenced DUI case, a lab report was admitted over a hearsay objection because the police testified that he had spoken to the chemist who related he would be on vacation during the preliminary hearing but would be available to testify at trial.
While I’m aware that this has been standing law for a very long time, the problem I see again and again is that the police do not contact the crime lab to verify that they will be available to testify at trial, but testify that the chemist will be available to testify when in fact they have not contacted anyone from the crime lab.
While I have argued to many magistrate’s without success that Rick does not support the facts described above, is there a case to support my position. i.e. do police need to obtain veritication each time or can they just rely on the assumption that someone will appear to testify as a chemist?
If you would be so kind, I have another question. If someone is validly parked in a restaurant parking lot and a police officer allegedly made arrests there within the past 2 years, is there a Pennsylvania or U.S. Supreme Court case that shows that when a fully uniformed police officer, driving a marked police car, shines his spotlight onto defendant’s said car (without blocking the defendant’s car from leaving), that the police officer made a show of force that necessaitates proof of reasonable suspicion for the stop to be valid?
I am aware that the other arrests must be close in time to make this parking area suspect, but I cannot find a case that only has the facts stated above. There are many other cases related but they all seem to have additional facts that change the situation my client faces.
Thanking you in advance for your help. I love to read your website.
tommyshaffer@hotmail.com
Comment by Thomas W. Shaffer, Esq. — June 22, 2006 @ 2:40 pm
ToxLab Guy:
I’m not sure what you mean by an “atypical circumstance”. I agree that a high majority of tests will reflect duplicate test results within the + / – .02% range. Of course, my point is not that law enforcement is not getting results within this range, but rather that this range is entirely too broad: a permissible second-test variation of 40% is no indication of acceptable scientific accuracy — nor of legal proof beyond a reasonable doubt.
As to my “improperly referring” to a 40% range of error, I use this illustrative figure because the second test is designed as an accuracy check on the first. The question is what the permissible range is for that second test for there to be legal “accuracy”. If the first test is .10%, the permissible range is .08%-.12%, which represents a .04% range. This means that the accuracy check on a .10% reading is 20% in either direction, or a total allowable range of 40%
Of course, the 40% depends upon the numerical value of the first test; the numerical range varies with the vlaue of the initial test. I could, for example, have chosen .08% as an example, in which case the range of error would be 50%. I chose .10% as an example because (1) mathematically it is easier to illustrate the point, and (2) in my involvement with thousands of DUI cases over 30 years, it has been my experience that the majority of DUI test results fall within the .08% to .12% range.
It may well be that we are engaged in semantics. In any event, I appreciate your well-reasoned comments and continuing contributions to this blog.
Comment by Lawrence Taylor — June 23, 2006 @ 8:31 am