IAFTC Newsletter. Volume 2. Issue 1. February 02, 2026.
Joshua Ott1
1Caselock, Inc., P.O. Box 285, Lebanon, GA 30146
This is an open-access article under the CC BY-NC-ND license.
Download PDF.
Abstract
This article examines the findings of a randomized clinical trial published in JAMA Psychiatry in 2023 evaluating the classification accuracy of Field Sobriety Tests (FSTs) with respect to cannabis exposure and driving impairment (as determined via a driving simulation). The study involved 184 adult cannabis users who were randomly assigned to placebo (0.02% THC), 5.9% THC, or 13.4% THC groups.
Certified Drug Recognition Expert (DRE) Instructors administered a battery of field sobriety tests following dosing to the participants. Driving simulations were also performed by the participants. The participants who received a placebo dose are the primary emphasis of this article. 49.2% of the placebo-dosed participants were classified as FST impaired, despite the pretreatment simulator performance showing no evidence of residual effects. The individual FSTs, which included the Walk and Turn, One Leg Stand, Modified Romberg Balance Test, Finger to Nose, and Lack of Convergence tests, demonstrated high false-positive rates.
These findings raise significant concerns regarding the accuracy of these FSTs in distinguishing impaired from unimpaired individuals. The results highlight the need for reassessment of current roadside testing practices and further research into more accurate and reliable evaluations.
Introduction
Field Sobriety Tests (FSTs) are widely used by law enforcement officers to assist in determining whether a driver should be arrested for DUI. These tests play a central role in arrest and charging decisions for driving under the influence (DUI), including cases involving suspected drug impairment. While multiple studies [2,3,4,5,6,7,8] have evaluated the Standardized Field Sobriety Tests (SFSTs), Horizontal Gaze Nystagmus, Walk and Turn, and One Leg Stand, in relation to Blood Alcohol Concentration (BAC), these studies validated classification accuracy relative to BAC thresholds rather than alcohol-related impairment as measured by independent performance outcomes. There is no known research validating these or any other FSTs for identifying drug and/or alcohol impairment. Despite this limitation, officers frequently rely on both standardized and non-standardized FSTs when evaluating suspected drug-impaired drivers.
A randomized clinical trial published in JAMA Psychiatry sought to examine the classification accuracy of FSTs with respect to cannabis exposure and driving impairment (as determined via a driving simulation). The study enrolled 184 adult participants between the ages of 21 and 55 who were active cannabis users and met the inclusion and exclusion criteria. Participants were required to abstain from cannabis use for at least two days prior to the experiment. A pretreatment driving simulation test demonstrated no residual effects of prior cannabis use.
Participants were randomly assigned to receive one of three different doses: a placebo dose (0.02% THC), a 5.9% THC dose, or a 13.4% THC dose. Following dosing, participants completed driving simulations and underwent a battery of FSTs administered by eleven certified DRE Instructors. The officers were blinded to the dosing condition and were asked to indicate whether they believed participants had received active THC or a placebo. Horizontal Gaze Nystagmus was not administered, as it is not expected to be present from cannabis use.
This article focuses on the results from the placebo-dosed group to examine the false-positive rates associated with commonly used FSTs. The findings raise important questions regarding the accuracy of the FSTs at correctly identifying sober subjects as not impaired and raise concerns for all DUI investigations, including cases involving only alcohol.
Overview of the Study
There were 184 participants ranging in age from 21 to 55 years old who were cannabis users.
Inclusion and Exclusion Criteria
Inclusion criteria required participants to:
-
Be 21–55 years of age
-
Have used cannabis four or more times in the past month
-
Hold a valid driver’s license
-
Have driven at least 1,000 miles in the previous year
Exclusion criteria included:
-
History of traumatic brain injury
-
Significant medical or psychiatric conditions
-
Positive pregnancy test result
-
Positive urine screen for nonprescription amphetamines, benzodiazepines, barbiturates, opiates, oxycodone, cocaine, methamphetamine, or phencyclidine
-
Past-year substance use disorder
-
Oral fluid THC concentration greater than 5 ng/mL on the testing day
Study Design
The participants were to abstain from cannabis use for at least 2 days prior to training and experiment days. On the experiment day, the participants were tested for drugs and alcohol. Prior to dosing, they performed a driving simulation.
Participants were randomly assigned to one of three groups:
-
Placebo group (0.02% THC)
-
Low-dose THC group (5.9% THC)
-
High-dose THC group (13.4% THC)
(63) participants received the placebo dose. For purposes of this paper, the primary focus is on the placebo-dosed group.
After the dosing, the participants performed driving simulations followed by Field Sobriety Tests (FSTs) four times during the remainder of the day. (The results provided were from the first driving simulation and FST administration.)
A question that must be asked is, were the participants in the placebo group under the influence or experiencing residual effects from marijuana usage prior to the experiment? The authors addressed this concern stating, “we found no differences in use intensity or time since use and no evidence of residual effects on pretreatment simulator performance.”
Field Sobriety Test Administration
Eleven certified DRE Instructors administered the Field Sobriety Tests (FSTs). This is the highest level of certification that a law enforcement officer can obtain for DUI enforcement. The officers were asked, “Which treatment do you think the participant received?” Answers were given using a 5-point scale (from “strongly believed…real marijuana” [1] to “strongly believed…placebo” [5]).” The officers did not observe the driving simulation.
Tests Administered
The following tests were used:
Horizontal Gaze Nystagmus (HGN) was not administered because it is not expected to be present from cannabis use.
FST Results
Overall, officers classified 49.2% of participants in the placebo-dosed group as “FST impaired,” using the authors’ terminology. The authors noted, “of participants classified as FST impaired, officers strongly or somewhat believed that 99.2% of participants had received THC, suggesting they suspected all poorly performing participants to be under the influence.”
Field Sobriety Test outcomes in this study function as a binary classification, in which participants are categorized as either “FST impaired” or “unimpaired.” In a binary classification framework where all subjects are known to be unimpaired, an ideal test would approach a false-positive rate near zero; a rate approaching 50% indicates the test performs no better than random assignment. Among placebo-dosed participants, 49.2% were classified as “FST impaired,” a rate consistent with chance-level classification rather than meaningful discriminatory accuracy.
These numbers have important implications. Put into a real-world context, officers usually suspect that a driver may be under the influence prior to administering the FSTs based on the driving and initial interaction with the driver. If the FSTs then identify the driver as “FST impaired,” it is highly likely that the officer will charge them with DUI.
The results for each of the tests individually are provided in the table below.
Comparison to Prior Research
It is important to note that the San Diego Study(5), which is currently relied upon by law enforcement to cite the accuracy of the SFSTs (HGN, Walk and Turn, and One Leg Stand), has similar false-positive rates. The Walk and Turn false-positive rate was 52%, and the One Leg Stand was 41%. In that study, the false positives had alcohol in their systems but were below the legal limit. It has been argued that these individuals may still have been impaired. The present study counters that argument by demonstrating comparable false-positive rates among placebo-dosed participants.
There are no known studies that have been conducted to validate the Lack of Convergence, Modified Romberg Balance, or Finger to Nose tests.
Driving Simulation Results
The authors reported that “FST impairment had a sensitivity of 80.9% and specificity of 35.7% relative to driving simulator impairment.” When driving simulator impairment is used as the reference standard, the reported specificity corresponds to a false-positive rate of 64.3% (1 − specificity) for the FSTs.
The Study did not document the results for each of the groups, but it provided a table showing the overall results. The study reported overall simulator results showing 112 participants classified as “not impaired” and 68 classified as “impaired,” totaling 180 participants. The reason four participants were not included in this summary was not explained.
The reported simulator data allows for the possibility that if all 63 placebo participants were classified as “not impaired” on the driving simulator, this would suggest that 49 THC-dosed participants (40.4%) were also classified as “not impaired” during simulated driving. If any of the placebo-dosed participants were classified as “impaired” on the driving simulator, then the percentage of THC-dosed participants classified as “not impaired” would increase. These findings suggest that recent cannabis use does not necessarily correspond to driving impairment.
Discussion
The results of this study show high false-positive rates for individual FSTs and when officers used all the tests combined to classify someone as “FST impaired.” The implications are significant, particularly given that the study was conducted in a controlled environment free from roadside factors such as weather, uneven surfaces, traffic, and distractions that could affect a person’s performance. Additionally, officers in this study relied solely on FST performance, without behavioral or driving indicators, which differs from real-world conditions. Even so, placebo-dosed participants were frequently incorrectly classified as “FST impaired.”
The authors stated: “Officers also knew that many participants would receive placebo, and it is surprising that THC exposure was assumed in almost all individuals who performed poorly on the FSTs. Officers may encounter situations in which they suspect recent cannabis use (eg, noticing cannabis paraphernalia or odors, drivers stating that they use cannabis); such information could potentially influence the belief that poor FST performance may be causally related to cannabis use.”
Law enforcement training places substantial emphasis on FST performance in arrest decisions. Given this reliance, the high false-positive rates observed in this study warrant serious scrutiny into the current tests used by law enforcement.
“Cognitive, or confirmation, bias refers to seeking or in interpreting evidence that supports existing beliefs or hypotheses often outside of awareness. Law enforcement in state legal jurisdictions emphasizes that THC-related impairment, and not just exposure, is the question of interest. Confirmation bias, common in the general population, can remain despite advanced training (including in law enforcement and forensic sciences and it may have been a factor in this study.”
Conclusions
The findings of this study raise substantial concerns regarding the accuracy and reliability of FSTs. False-positive rates approaching or exceeding 50% suggest that FST outcomes alone may offer little discriminatory value between impaired and unimpaired individuals. In some cases, decisions may be no more accurate than chance.
There is a clear need to either improve the specificity of existing tests or develop new methods capable of accurately and reliably distinguishing impairment from non-impairment. Until such measures are implemented, both law enforcement and the legal community should be informed of the limitations and error rates of these tests.
Future research should utilize sober participants to examine the effects of environmental conditions, footwear, age, and physical injuries on FST performance.
Acknowledgements
The author used ChatGPT to assist in drafting the abstract and introduction and to improve clarity and conciseness of the manuscript text, including the conclusion. The AI tool was used solely for language refinement and organizational assistance. All substantive content, data interpretation, conclusions, and final editorial decisions were made by the author, who takes full responsibility for the accuracy and integrity of the work.
Conflict of Interest Disclosures
The author is a consultant and expert witness for DUI and BUI cases, but has received no funding or compensation for the preparation of this article.
References
[1]Marcotte TD, Umlauf A, Grelotti DJ, Sones EG, Mastropietro KF, Suhandynata RT, et al. Evaluation of Field Sobriety Tests for Identifying Drivers Under the Influence of Cannabis: A Randomized Clinical Trial. JAMA Psychiatry; 2023.
[2] Burns M, Herbert M. Psychophysical Tests for DWI Arrest. U.S. Department of Transportation National Highway Traffic Safety Administration; 1977.
[3]Tharp V, Burns M, Moskowitz H. Development and Field Test of Psychophysical Tests for DWI Arrest. Southern California Research Institute; 1981.
[4]Anderson T, Schweitz R, Snyder M. Field Evaluation of a Behavioral Test Battery for DWI. U.S. Department of Transportation National Highway Traffic Safety Administration; 1983.
[5]Stuster J, Burns M. Validation of the Standardized Field Sobriety Test Battery at BACs Below 0.10 Percent. United States. National Highway Traffic Safety Administration; 1998.
[6]Burns M, Dioquino T. A Florida Validation Study of the Standardized Field Sobriety Test (S.F.S.T.) Battery. United States. National Highway Traffic Safety Administration; 1997.
[7]Burns M, Anderson E. A Colorado Validation Study of the Standardized Field Sobriety Test (SFST) Battery. U.S. Department of Transportation National Highway Traffic Safety Administration; 1995.
[8]Burns M. The Robustness of the Horizontal Gaze Nystagmus Test. Southern California Research Institute; 2007.
