top of page
  • jmedleyoffice

"I know I passed the roadside eye test, but I still got arrested for DUI!"

Updated: Apr 30

The beginning of the end of a Colorado DUI arrest is when the officer starts waving a pen or their finger in front of your eyes. When an officer asks you to, "Follow the object with your eyes and your eyes only- do not move your head." they are already on their way to placing you into handcuffs. The eye test is called HGN, and DUI officers think it is the source of their superpowers to detect impaired drivers. Very few people are released to go home after participating in this test. Read this article to see why.

Police officers in Colorado receive instruction during the basic police academy about how they are supposed to watch a driver's eyes go back and forth. These officers are told that scientific research has shown that this test can tell them whether a driver is over .08 BAC. Specifically, officers are told in their training that the test is 88% accurate in determining whether a driver is over the legal limit. This is false.

Is that information a lie!? Well, yes and no. Police officers are not scientists and they are never even shown the actual studies in their training. So when an officer swears to tell the truth in court and recites this misinformation, they typically do not know what they are saying is not true. The officers and scientists who publish the booklet used in the training that tells officers this false information ARE guilty of presenting fraud to the courts.

This false information is proven faulty through at least two different scientific lenses. 1. The lense of forensic science, and 2. the lense of statistical science.

Fraud in DUI Forensic Science

In any forensic science, there are two broad categories of testing. There are presumptive tests, also called screening tests. There is another level of testing called confimatory tests. These tests have different purposes in forensic science.

Confirmatory Tests

Confirmatory tests exist to create a high degree of certainty in the result of a test. A confirmatory test result is derived with the end goal of it being used as evidence in court to prove a person is guilty of a crime beyond a reasonable doubt. These results therefore are required to present a very, very small risk that the result can exist when an accused may be innocent. This quality is called specificity.

The desired quality of specificity usually requires the examiner to hold a high degree of training or education, and that the examiner use only scientific methods or devices that are capable of eliminating the possibility of innocence to a high degree of scientific certainty. As would be expected, these processes require time and often expensive equipment. It is not realistic to commit these expensive and time consuming resources to every single case in which DUI is only suspected.

Screening Tests

Screening tests, also called presumptive tests, are tests that can usually be conducted at roadside. Screening tests differ from confirmation tests in that they can be conducted by less intensively trained examiners, screening tests do not typically require a long time to administer, and they do not require expensive equipment.

Another desired quality of a screening test is that it be sensitive. Sensitivity is the quality of a test that allows the test to have a high degree of certainty that no guilty suspect will escape detection. In a DUI roadside investigation for example, officers do not want an impaired driver to be mistakenly released. Screening tests therefore must be very reliable at detecting indications of a driver even possibly being under the influence.

The weakness introduced by high sensitivity is lower specificity. In order to make sure no possibly guilty suspects are wrongfully released, a highly sensitive test is going to implicate more innocent suspects as being possibly guilty. This highlights the importance of confirmation testing as not only important to prove guilt, but also to clear the innocent people who were implicated by overly sensitive screening test.

The most common examples people in our normal lives see the difference between screening and confirmation tests include early pregnancy tests, and home COVID tests. Both of these tests are quick to perform, don't require scientific expertise, relatively cheap, and are very unlikely to be falsely negative, which means they are very sensitive. Although a positive test with either of these, if properly considered, is followed up by a visit to a physician for confirmation.

The police DUI eye test is merely a screening test. It is very sensitive to elevated alcohol levels in the human body. In data collected in a 1998 DUI study, 97% of all drivers who were over .08 showed a positive result on the eye test. That result demonstrates a highly sensitive test.

The 1998 study reveals another side of this data in terms of specificity. Of all of the drivers in the study who were under .08, 37% showed a positive result on the eye test. This specificity score of a mere 63% is unacceptable as a forensic test; over 1/3 innocent people are said to be guilty by the DUI eye test according to these numbers. The police training booklet does not tell officers anything about this low specificity in their training.

Another study completed by DUI police officers was published in 2007. In this study, 67% of test subjects who were under .08 were scored as positive for being over .08 on the eye test. This study, demonstrating a dismal 33% specificity with the eye test is not even mentioned in police DUI training.

Given the demonstrated sensitivity of the DUI eye test, it is a valuable tool for police officers in determining whether to arrest a driver or to let them keep driving. Kept in the context of only determining probable cause for arrest, the eye test is fair if taken in the context of other observations. It is in fact listed as a "pre-arrest screening" step in the police DUI booklet.

The eye test is not however accurate enough for use as evidence in court to present to a jury to be part of a possible conviction. The unacceptable lack of specificity should render the eye test inadmissible in a trial. Unfortunately, 40 years of bad science and case law that ignores science has allowed this false test in court and has led to false convictions.

So what are police officers told in their training booklet? That is where the statistical fraud takes place.

Fraud in DUI Statistics

Statistical results from research can be presented in terms of descriptive statistics, or inferential statistics. Let's compare the difference in these two categories.

Descriptive Statistics

These numbers simply describe the results of a specific study in terms of the distribution of results regarding the specific sample involved in that individual study. For example if a group of five Denver DUI police officers are surveyed about how many hours they worked in January 2023, and the results showed: the lowest number of hours was 120, the average number of hours was 170, and the highest number of hours worked by a DUI officer was 196. These results describe that sample of five DUI officers. It would not be scientific to make any generalizations about a larger group of Denver DUI officers, Denver police officers in general, and in no way could any statement be made about U.S. police officers.

Inferential Statistics

Some research studies can generate data that can be generalized, or inferred to a larger group or population. These types of studies require much more of a research design. Research design includes randomization, existence of control groups, a sample representative of the broader population, and a large enough sample to allow inferences to the larger group.

The 1998 study I mentioned above, involved only seven police officers, all of whom were very experienced and highly trained in DUI procedures. There were only about 290 drivers exposed to the eyes test, all in one city on the west coast of the United States in a few month period of one year. Without getting into too much detail, it is fair to say that there was no randomization, no control group, and seven exceptionally trained officers in one police department do not accurately represent officers who make DUI arrests across the country.

In spite of the absence of any methodologies justifying the characterization of the results of the 1998 study as "inferential," the police booklet ignores this flaw. The current police DUI training booklet that was published in 2018, tells officers, "... the test is 88% accurate."

You probably noticed the 88% number isn't the specificity or the sensitivity found in the study. It is simply the number of correct classifications made by those seven DUI officers using the eye test on all drivers in the study. This is exactly the fraud of presenting a descriptive statistic as if it were an inferential statistic.


Every day in Colorado drivers who are accused of DUI go to trial. During those trials, DUI officers come into court only armed with the deceptive information given to them in the booklet the State of Colorado gives them in their training. They present the eye test as a reliable test, confirmed to be 88% accurate. They will not even know that the very study that gave them that number, showed 37% of innocent people failing the test. This testimony is forensically false and statistically false, and if your DUI lawyer doesn't understand these falsities, you will not get effective representation.

6 views0 comments


bottom of page