Can Low Alcohol and a Memory Task Worsen Real-Highway Lane Keeping in Novice Drivers?

Quick answer: A small randomized crossover experiment conducted in actual highway traffic found that alcohol conditions targeting 0.2 and 0.5 g/L blood-alcohol equivalents increased lane-position variability compared with placebo. Adding an auditory memory task increased weaving, and the combination was particularly consequential for novice drivers in the lower-dose condition. But the study measured driving performance in 18 supervised volunteers using breath-derived alcohol estimates. It did not directly analyze their venous blood, measure crash rates, or establish that a particular driver at a particular reported BAC caused an incident.

What does this real-highway alcohol study actually show?

Forensic discussions often treat a chemical alcohol result and a driver’s observable ability as though they are the same fact. They are not. A laboratory concentration describes an analyte in a specified specimen. Lane-keeping performance describes behavior under specific driving conditions. Jongen and colleagues examined the latter question in a controlled experiment involving real highway driving, rather than a computer simulator. Their 2018 open-access primary study is “Single- and dual-task performance during on-the-road driving at a low and moderate dose of alcohol” (DOI: 10.1002/hup.2661).

The central finding was a dose-related increase in the standard deviation of lateral position (SDLP), a continuous measure of lane-position variability, under breath-alcohol conditions selected to correspond to nominal BACs of 0.2 and 0.5 g/L. A separate task requiring auditory word learning increased the demands on drivers’ attention. The experiment offers evidence that low-dose alcohol can affect a measured driving skill under supervised conditions; it cannot convert an isolated toxicology result into a reconstructed driving performance score.

Who drove, where, and under what experimental conditions?

The investigators recruited 18 healthy drivers aged 18–25 years, divided into nine newly licensed novices and nine more experienced young drivers. Each group had four men and five women. Novices had held a license for no more than one month and reported less than 1,000 km of independent driving. Experienced participants had held a license for at least one year and accumulated at least 7,500 km. Thus, the comparison was principally about licensing and practice within a narrow young-adult sample, not a comparison of adolescents with middle-aged drivers.

Each participant attended three treatment sessions in a randomized, double-blind, placebo-controlled crossover design, with at least seven days between sessions. The conditions were alcohol placebo, a dose targeted to 0.2 g/L, and a dose targeted to 0.5 g/L. Alcohol mixed with orange juice was administered in five portions spaced ten minutes apart; the individual dose was adjusted when necessary to approach the target. Placebo orange juice contained a very small masking drop of ethanol. A standardized meal preceded dosing.

Participants then completed a roughly 100-km highway circuit in an instrumented vehicle, accompanied by a licensed instructor able to use dual controls. The requested speed was approximately 95 km/h. The approximately one-hour test took place in actual traffic, but within a highly standardized, supervised protocol. The instrument recorded vehicle speed and position in the lane; segments containing overtaking or traffic disturbances were removed from the analytic road-tracking record. The study did not involve unsupervised intoxicated driving.

Were the alcohol levels directly measured in blood?

No. The authors administered alcohol to target nominal blood-alcohol concentrations, but obtained alcohol readings from breath using a Lion Laboratories SD-400 Alcoholmeter. The instrument’s breath reading was converted to a BAC estimate; the paper subsequently called that estimate “BAC.” This terminology matters when the study is offered in a case involving a hospital serum result, a venous whole-blood headspace gas-chromatography result, or a statutory breath result.

At the start of the road test, the reported mean estimated concentrations for novice drivers were 0.24 g/L in the low condition and 0.52 g/L in the moderate condition; corresponding means in experienced drivers were 0.26 and 0.51 g/L. The source reports these as 0.24, 0.52, 0.26 and 0.51 mg/mL, numerically equivalent to g/L. These are condition means from a breath-derived calculation, not 18 individually measured venous blood concentrations or a validated conversion for any later forensic specimen. For background on what a breath measurement actually represents, see breath-alcohol measurement and its analytical limits.

What is SDLP, and why is it useful without being a crash measurement?

SDLP is the standard deviation of a vehicle’s lateral position within a lane, commonly expressed in centimeters. A greater value indicates more variation in lateral positioning under the test conditions. Here, lateral position and speed were sampled at four measurements per second; road-tracking outcomes were assembled from retained five-kilometer segments. This is a quantitative behavioral outcome, not a police officer’s subjective impression and not a laboratory ethanol concentration.

Its strength is sensitivity to a specific operational driving skill under repeatable conditions. Its boundary is equally important: SDLP does not by itself measure hazard perception, emergency braking, judgment at an intersection, real-world collision frequency, or the cause of a particular crash. A controlled shift in group-average lane variability does not establish that a given person’s momentary lane movement was attributable to alcohol rather than roadway geometry, distraction, fatigue, equipment, weather, or another influence.

How much did lane-position variability change with alcohol?

Across the study’s conditions, SDLP rose relative to placebo by a mean of 1.12 cm at the nominal 0.2 g/L condition (95% confidence interval 0.06–2.18 cm) and by 3.33 cm at the nominal 0.5 g/L condition (95% confidence interval 1.82–4.83 cm). The alcohol effect on SDLP was statistically significant overall; each alcohol condition differed significantly from placebo in the authors’ simple contrasts. The result supports a dose-associated effect on the measured task within this experimental population, not a universal centimeter-per-BAC rule.

The authors also reported an overall effect of driving experience on SDLP, but the experience-by-alcohol interaction for the primary SDLP analysis was not significant (p = .81). It would therefore misstate the evidence to say that the paper proved a different overall alcohol dose-response slope for novices versus experienced drivers. The more specific novice-versus-experienced result concerned the interaction among task demand, driving experience, and alcohol condition.

How was divided attention tested?

During a marked portion of the highway drive, participants listened to an auditory list of 30 monosyllabic nouns presented through an in-car speaker, then immediately recalled as many as possible. This learning-and-recall procedure was repeated over three trials, and the total correctly recalled words was used as the secondary-task measure. There was no visual phone display, texting task, hand-held device manipulation, or sudden emergency event. The study can inform divided-attention reasoning without being recast as a direct test of texting while intoxicated.

Compared with driving without that additional task, dual-task driving increased SDLP by an average of 1.13 cm (95% confidence interval 0.56–1.70 cm). A significant three-way interaction among alcohol condition, task, and driving experience was reported (p = .04). The investigators found greater novice SDLP during the auditory task than during simple road tracking at the lower 0.2 g/L condition, whereas the analogous within-condition task increase for experienced drivers appeared at 0.5 g/L. However, these are subgroup patterns from very small groups; they are not proof that every novice has measurable lane-keeping deterioration at a given level or that an experienced driver is safe below 0.5 g/L.

Did alcohol also impair the memory task?

For auditory word recall, the nominal 0.2 g/L condition showed a mean difference of −2.78 words compared with placebo (95% confidence interval −7.53 to +1.97), which was not statistically significant. At 0.5 g/L, the mean difference was −7.22 words (95% confidence interval −12.79 to −1.65), with p = .01. A negative change means fewer correctly recalled words under the study procedure. This is evidence about task performance under those conditions, not a clinical diagnosis of memory dysfunction or an accident-specific impairment determination.

A finding that one specific secondary test did not reach statistical significance at a lower condition does not mean that alcohol caused no cognitive effects. The outcome, timing, sample size, testing procedure, and statistical uncertainty determine what the data support. Conversely, a significant decrement at the higher condition does not permit an expert to estimate how many words a particular driver could recall during a roadside encounter.

Does this prove the legal BAC limit should be the same for every novice driver?

No single 18-person experiment can set a universal legal threshold. The authors interpreted their findings as supporting a lower legal limit for inexperienced drivers, given the changes observed at the lower target condition. That is a policy interpretation based on a controlled performance experiment. Legal limits vary by jurisdiction and regulated driver category; the scientific data must not be mistaken for the law applicable to a particular case.

The findings do challenge the scientifically simplistic proposition that a low estimated alcohol concentration necessarily implies normal driving performance. But they do not establish a converse proposition that every measurable ethanol level proves the same degree of impairment. Concentration, tolerance, task demand, driving experience, timing, and observable conduct remain separate evidentiary considerations.

What are the major limitations of this on-road trial?

Small, selected sample: There were only nine volunteers in each experience category, all young adults screened for health and participation criteria. The authors’ own power discussion notes reduced ability to detect some between-group differences. Subgroup contrasts and null results therefore require care.

Controlled rather than ordinary traffic behavior: Driving occurred on a specified highway circuit in a dual-control car with an instructor, target speed, standardized task, and removal of disturbed segments. That strengthens experimental control while limiting extrapolation to emergencies, difficult intersections, aggressive maneuvers, and crashes.

Breath-based exposure measure: The reported “BAC” is an instrument-derived breath estimate. The paper does not establish a person-specific breath-to-blood relationship or validate a retrospective estimate from a later collected legal blood specimen. Readers comparing test matrices can consult the difference between breath estimates and directly analyzed blood-alcohol specimens.

Single added cognitive task: Auditory word recall does not reproduce all distractions. An experimentally defined dual task differs from conversation, navigation, emotional stress, visual texting, or an unpredictable road hazard.

Selected performance endpoints: SDLP and recall were the reported measures. Neither is a direct measurement of collision probability, legal fault, or the cause of an event. The numerical results are group contrasts, not an individual correction or threshold for an expert to apply mechanically.

Timing and kinetics: The administered doses were aimed at targeted condition levels and breath estimates were obtained around the road test. The paper does not establish absorption and elimination curves capable of reconstructing an incident-time BAC in a different person; for that separate question, see forensic pharmacokinetics and its individual uncertainties.

How should a forensic expert use the study in a DUI or collision case?

A scientifically bounded use is to explain that a controlled study found changes in lane keeping at relatively low alcohol exposure levels and that concurrent cognitive load may matter. An expert can describe the study design, source of the alcohol estimates, population, uncertainty intervals, and limits on applying group averages to one person. This may be relevant to evaluating claims that lower alcohol levels cannot affect performance or that experience necessarily protects a driver from divided-attention effects.

An inappropriate use would be to take a defendant’s single breath, plasma, serum, or blood value and assert a particular centimeter increase in lane weaving; to equate a standard deviation with evidence of crossing a line at a specific scene; to infer that a crash was caused by alcohol without reconstruction and other evidence; or to present the study’s policy discussion as binding criminal law. In a defensible evaluation, analytic validity, timing, observations, roadway evidence, alternative causes, and applicable legal elements must be assessed independently.

Frequently asked questions

Did the researchers let intoxicated people drive on a public road?

Participants completed a supervised, standardized highway test in an instrumented dual-control car with a licensed instructor. This was an approved research protocol, not an observational study of unsupervised impaired motorists.

Is 0.2 g/L the same number as 0.02 g/dL?

Yes, numerically 0.2 g per liter equals 0.02 g per deciliter. The caveat is analytical: the study’s reported BAC-equivalent values came from breath measurements converted to estimates, not direct venous blood testing.

Did alcohol worsen lane keeping at the lower target level?

Yes, the mean SDLP difference from placebo was +1.12 cm for the nominal 0.2 g/L condition, with a 95% confidence interval of +0.06 to +2.18 cm. This is a group mean under standardized conditions, not an individual diagnostic cutoff.

Did the study prove that all novice drivers are more impaired than experienced drivers?

No. An overall experience-by-alcohol interaction on SDLP was not significant. A narrower task-by-experience-by-alcohol interaction was significant, showing a different pattern when the auditory task was added. With nine people per group, the difference must not be generalized as a rule for all drivers.

Was distraction tested by texting or mobile-phone use?

No. Researchers used an auditory word-learning and immediate recall test. That added mental demand, but it was not a visual or manual phone-distraction experiment.

Can the study convert a blood test result into a weaving score?

No. It supplies mean behavioral contrasts from controlled exposure conditions. It does not validate a formula that predicts the lane keeping of a specific person from an isolated specimen concentration.

Can the paper prove alcohol caused a particular collision?

No. The investigators measured lane variability and memory performance, not collision causation. Case-specific reconstruction requires independent evidence about the person, vehicle, roadway, event sequence, and plausible competing explanations.

Primary scientific source

Jongen S, van der Sluiszen NNJJM, Brown D, Vuurman EFPM. Single- and dual-task performance during on-the-road driving at a low and moderate dose of alcohol: A comparison between young novice and more experienced drivers. Human Psychopharmacology: Clinical and Experimental. 2018;33(3):e2661. https://doi.org/10.1002/hup.2661. The study’s original Methods, Results, confidence intervals, and discussion are the basis for the numerical statements on this page.