How Much Did Alcohol Increase Lane Weaving in a 17-Driver Simulator Trial?

In a 17-person randomized crossover study, the higher alcohol condition increased average simulator lane weaving by 4.06 centimetres compared with placebo. The average alcohol value before driving was approximately 0.07%, reported from breath testing. The lower condition, averaging approximately 0.04%, did not produce a statistically significant overall lane-weaving difference from placebo. Neither result defines a universal impairment threshold for an individual driver.

Garrisson and colleagues’ 2022 study is useful in DUI evidence review because it separates a measured driving outcome from an intended alcohol dose, a breath-reported concentration and the driver’s own confidence. Those are related but different pieces of evidence.

The same drivers completed three conditions

Seventeen adults, eight men and nine women, completed placebo, lower-alcohol and higher-alcohol sessions in counterbalanced order. They were aged 20–30 years and held driving licences. Treatment order was randomized, and participants and the investigator were blinded. A nurse prepared the beverages separately.

The sessions targeted alcohol values of 0.05% and 0.08%, but the achieved pre-driving averages were lower: approximately 0.04% and 0.07%. Target concentrations are experimental goals, not observations that can replace the readings actually obtained.

After the absorption period, participants drove a simulated 105-kilometre highway route for about an hour. They were instructed to remain in the left lane, maintain 100 km/h and overtake slower vehicles when needed. The analysis excluded overtaking and certain other portions of the recorded drive.

What “lane weaving” measured

The main lateral-control measure was the standard deviation of lateral position, or SDLP. It describes variability around the car’s mean position in the lane. It is not the maximum distance the car crossed a lane line, a count of collisions or a direct probability of a crash.

Mean SDLP was 21.09 cm with placebo, 22.03 cm in the lower-alcohol condition and 25.15 cm in the higher-alcohol condition. The higher condition differed significantly from placebo (p = .016) and from the lower condition (p = .004) using the study’s adjusted pairwise significance criterion.

The higher-versus-placebo difference of 4.06 cm is a group mean within this experiment. It does not mean that every participant worsened by that amount. Nor does it allow a toxicologist to infer an unknown driver’s alcohol concentration from an observed amount of weaving.

The lower-dose null result deserves equal attention

The overall SDLP comparison between placebo and the lower-alcohol condition was not significant. That finding should be reported alongside the positive higher-dose result. It does not demonstrate that approximately 0.04% is safe, that no participant experienced an effect or that every other driving skill was preserved.

This small experiment had limited power to identify smaller differences. It also examined a particular sustained highway simulation rather than every aspect of driving. Its results cannot establish a lowest concentration at which any alcohol effect begins.

Speed variability illustrates another statistical distinction. Although the overall treatment test for speed variability was significant, the reported corrected pairwise comparisons were not. It would overstate the detailed results to claim that the higher condition produced a securely established pairwise increase in speed variability against placebo.

Time on task does not prove an alcohol concentration curve

The investigators also divided the drive into 20-minute segments. Both alcohol conditions showed an overall time effect for SDLP, but the corrected segment comparisons were narrower: the second segment exceeded the first in the higher condition. The lower condition’s segment comparisons did not reach the adjusted criterion.

Alcohol was not continuously measured throughout driving. The authors considered descending alcohol concentrations and declining vigilance as possible explanations for changes over time. Those explanations were not established by simultaneous concentration measurements. The study therefore cannot provide an individual elimination rate or justify retrograde BAC extrapolation.

Breath-reported alcohol was not a whole-blood assay

The study used a calibrated Lion Alcolmeter SD-400PA to measure breath alcohol in grams per 210 litres of breath. The paper expresses exposure using BAC percentages, but it did not obtain a whole-blood ethanol result for each reported value. Serum/plasma, urine and alcohol biomarkers were not interchangeable exposure measures in this experiment.

The pre-driving reading also preceded the post-driving cognitive test and confidence rating by about an hour. Assigning the same numerical concentration to all those observations would conceal an unmeasured interval during which alcohol could change.

Confidence and task performance did not move together

Self-rated confidence did not differ significantly across conditions (p = .819), despite the higher-condition lane-control difference. The four-choice reaction task also showed poorer higher-condition performance relative to baseline; data for that analysis were available for 16 participants. The paper reports differing exact p-values between its text and table, so the reliable conclusion is the reported significant higher-condition change rather than a single supposedly definitive exact p-value.

This is an objective-performance question, distinct from whether people accurately estimate their alcohol level or fitness to drive. Unchanged average confidence cannot show that every participant was unaware of an effect. Equally, confidence alone is not evidence that lane control remained intact.

What the experiment can establish in a DUI case

The controlled crossover design supports a group-level alcohol effect on this simulator’s lateral-control measure at the higher tested condition. Limits include the small selected sample, absence of a predefined statistical analysis plan, incomplete cognitive-task data and simulator conditions without real traffic consequences. Collision risk was not directly measured.

A case-specific opinion must still distinguish actual driving observations, the specimen tested, test timing and supported concentration estimates. The study does not convert lane weaving into BAC, establish a person’s legal status or prove that alcohol caused a particular crash.

Primary source

Garrisson H, Scholey A, Verster JC, Shiferaw B, Benson S. Effects of alcohol intoxication on driving performance, confidence in driving ability, and psychomotor function: a randomized, double-blind, placebo-controlled study. Psychopharmacology. 2022;239:3893–3902. https://doi.org/10.1007/s00213-022-06260-z.