Answer: No. An unchanged Stroop interference score does not rule out an alcohol-related performance effect. In a controlled, 40-person crossover experiment, alcohol exposure was associated with an estimated 17-millisecond increase in reaction time across both easy and conflicting Counting Stroop trials, even though the additional time attributable specifically to conflicting stimuli did not show a detectable alcohol-related increase. Those are two different outcomes. The experiment provides useful evidence about how a narrow executive-function score can remain statistically stable while overall responding slows; it does not establish that an individual was unimpaired, fit to drive, or at a proven venous blood-alcohol concentration.
The original research is Riedel and colleagues’ 2021 study, Acute alcohol does not impair attentional inhibition as measured with Stroop interference scores but impairs Stroop performance, published in Psychopharmacology (DOI: 10.1007/s00213-021-05792-0). The distinction between a difference score and the underlying absolute responses matters when cognitive testing is offered in litigation as evidence for or against alcohol-related functional change.
What question did the alcohol-clamp experiment actually answer?
The investigators asked whether moderate, relatively stable alcohol exposure selectively increased attentional interference during a Counting Stroop task or instead caused a more general slowing of responses. Attentional interference is not identical to overall cognitive performance. It describes the extra difficulty created when a distracting stimulus feature conflicts with the feature a person has been instructed to use. A study may find that the conflict penalty is similar in alcohol and control sessions while still finding a main effect of alcohol on speed.
This is a measurement question, not a universal declaration about sobriety. The experiment did not administer a standardized field sobriety test, evaluate actual vehicles on roads, determine crash risk, or define an individual diagnostic cutoff. The reported statistical results belong to this task, this participant sample, and this controlled exposure protocol.
Who participated, and how was the crossover experiment organized?
Forty-six eligible volunteers consented; forty completed the necessary two Counting Stroop sessions and were included in the behavioral analysis. Participants were adults aged 25–55 years, with a mean age of 29.4 years. They had medium-to-high-risk drinking histories but no current or previous diagnosed alcohol use disorder under the study’s screening criteria. The investigators also specified right-handed, German-proficient participants with normal or corrected-to-normal vision, limiting generalization to other populations.
Each analyzed participant completed one intravenous ethanol session and one 0.9% saline control session in randomized order. Eighteen received alcohol first and twenty-two saline first. The alcohol-versus-saline assignment was described as single-blind for this component. The Counting Stroop assessment was a secondary outcome within a larger trial involving randomized naltrexone or placebo treatment; the main analysis was structured around the alcohol and trial-type factors, with additional medication-related analyses addressed in the source’s supplement. The larger-trial setting matters when assessing possible contextual influences and the scope of inference.
Was the target “80 mg% arterial BAC” directly measured in blood?
No direct arterial or venous blood sample established that concentration in the reported protocol. The investigators delivered ethanol intravenously under a computer-assisted infusion procedure intended to reach a theoretical arterial blood-alcohol concentration (aBAC) of 80 mg per 100 mL (often expressed as 80 mg/dL) in approximately 25 minutes and maintain it. The paper used repeated end-expiratory breath-alcohol measurements from a Dräger Alcotest 6810 med to guide the infusion. It reported a conventional 2,100:1 blood-to-breath conversion to approximate aBAC from breath measurements.
Thus three quantities must remain conceptually separate: administered intravenous ethanol, the instrument’s directly measured end-expiratory breath ethanol, and a blood-equivalent arterial estimate derived from the breath result and the model’s partition assumption. None is proof of a contemporaneous forensic venous whole-blood ethanol concentration determined by laboratory headspace gas chromatography. For more on that matrix distinction, see this site’s research resource on interpreting two blood samples and alcohol elimination.
How did the Counting Stroop task create conflicting and nonconflicting trials?
Each session used 160 computerized trials: eighty congruent and eighty incongruent. Volunteers saw a group of identical digits and had to respond to how many digits appeared, not to the numerical identity of the digit. When the two features matched, such as three instances of the digit 3, the trial was congruent. When they differed, such as two instances of the digit 3, the trial was incongruent. The mapping from quantities one through four to response keys was kept orderly to reduce extraneous mapping difficulty.
The investigators measured response latency in milliseconds, errors, and an inverse-efficiency score combining latency with accuracy. They then calculated an interference score by subtracting the congruent-trial measure from the corresponding incongruent-trial measure. Importantly, an individual’s interference score is not the same number as the individual’s overall response time. A participant who slows similarly in both trial types could have substantially changed absolute performance with little change in the subtraction-based interference score.
What exactly was measured, modeled, and left untested?
| Component | Study method | Proper evidentiary interpretation |
|---|---|---|
| Alcohol exposure | Controlled intravenous ethanol versus saline visits | Known intervention and intended exposure, not a field drinking-history reconstruction |
| Alcohol measurement | End-expiratory breath readings from Alcotest 6810 med | Instrument measured breath ethanol; not a forensic venous whole-blood gas-chromatography result |
| Target “aBAC” | Computer-assisted target of 80 mg/dL using a 2,100:1 breath-to-blood approximation | Modeled arterial blood-equivalent target, not a directly assayed arterial specimen |
| Behavioral outcome | Button-press times, errors, inverse efficiency in a Counting Stroop paradigm | Observed laboratory-task performance in this sample |
| Interference measure | Incongruent minus congruent task performance | Within-task difference score—not overall speed or a universal executive-function assay |
| Unmeasured case outcomes | No actual driving, roadside tests, forensic venous specimen, collision reconstruction or individual classifier validation | Cannot infer particular-driver fitness, statutory BAC, or event causation directly |
What did the original results show?
The study’s main analysis compared the factors alcohol versus saline and incongruent versus congruent trial within the same forty participants. Alcohol produced a statistically detected overall reaction-time effect. Conflict also produced a substantial reaction-time effect. However, the alcohol-by-conflict interaction was not statistically detected. The estimated marginal differences are group-model summaries, not predicted reaction times for every individual or for every task.
| Prespecified or reported comparison | Original result | What it supports—and does not |
|---|---|---|
| Overall alcohol effect on reaction time | Estimated marginal mean +17 ms; F(1,39)=8.45; p=.006 | Group-level slowing on both trial types; not an individual impairment threshold |
| Incongruent versus congruent trials | Estimated marginal mean +49 ms; F(1,39)=208.73; p<.001 | Task conflict increased reaction time irrespective of alcohol condition |
| Alcohol × trial-type interaction | F(1,39)=0.94; p=.337; generalized η²<.001 | No detected additional conflict penalty from alcohol under this test; not proof of an exactly zero effect |
| Bayesian full versus reduced model | BF(full/reduced)=0.29, or approximately 3.5:1 favoring the reduced no-interaction model | Moderate relative model evidence for no extra interaction under specified models and priors |
| Overall error rate | Mean error proportion .055, SD .052 | Accuracy was generally high; does not establish perfect cognition or individual safety |
These figures come from the original study’s main Results section and its statistical-model comparison; they are neither a forensic test-validation dataset nor a set of real-world collision outcomes. The investigators also observed an alcohol-related change in the combined inverse-efficiency measure while the interference contrast remained comparatively stable.
How can responses slow while the Stroop interference score stays similar?
Suppose, for illustration only, a person responds to congruent trials in 500 milliseconds and incongruent trials in 550 milliseconds. The interference penalty is 50 milliseconds. If both types become 20 milliseconds slower, the respective response times become 520 and 570 milliseconds—yet the interference difference remains 50 milliseconds. This example is arithmetic, not a reproduction of participant-level study data. It demonstrates how the subtraction cancels a general shift shared by both types of trials.
Consequently, “interference unchanged” and “performance unchanged” are not interchangeable propositions. A test that isolates a cognitive construct by taking a difference may legitimately be insensitive to an effect that moves both constituent conditions together. A forensic interpretation must identify what the scoring method deliberately removes before claiming the score rules out an alcohol effect.
Does the nonsignificant interaction prove that attention was unimpaired?
No. A p value of .337 for the alcohol-by-trial-type interaction means the experiment did not detect that specified differential reaction-time effect using its analysis and sample; it does not measure the probability that the participant was unimpaired, prove exact equivalence, or rule out every smaller or differently expressed effect. The investigators calculated that their forty-person design could detect an effect on interference down to approximately Cohen’s d=.58 under the stated power assumptions. This does not exclude more modest effects with confidence, nor does it transfer to different executive tasks.
The Bayesian result supplies additional, but similarly constrained, evidence. Comparing a model that included the interaction with a reduced model containing the two main effects gave a Bayes factor of .29 in the full-versus-reduced direction. Reversing that comparison yielded about 3.5 in favor of the reduced model. That ratio compares the fit of specified competing models to the observed data; it is not an accuracy percentage for roadside decisions and should not be recast as proof that alcohol never affects inhibitory control.
Did the investigators test reaction-time tails or conflict adaptation?
Yes. To check whether an interaction was concealed within a subset of faster or slower responses, the researchers compared interference scores across reaction-time quantiles. Interference varied across quantiles, but the alcohol-by-quantile interaction was not statistically detected (F(3,117)=0.96; p=.413). They also assessed conflict adaptation—the tendency for the interference effect to change after a prior conflicting trial. Adaptation was present overall; the tested alcohol effect on adaptation was not detected (F(1,39)=0.10; p=.754).
Those supplemental findings strengthen the study’s account of this particular task. They do not transform it into a general test of safe braking, hazard recognition, divided attention, response inhibition during driving, or other behaviors that require different experimental measurements. Nor were the forty persons multiplied into thousands of independent subjects simply because many button presses were recorded.
Why do Stroop interference, response inhibition, and driving impairment need separate labels?
Interference control concerns filtering irrelevant stimulus information while selecting the correct task response. Response inhibition, in paradigms such as stop-signal or go/no-go tasks, concerns withholding or canceling a response. Driving involves a larger set of interacting functions: perception, sustained attention, judgment, divided attention, movement coordination and environmental demands. Laboratory outcomes can illuminate one component without measuring the others directly.
It is therefore unsound to argue that a normal conflict-difference score disproves alcohol-related psychomotor change generally. Conversely, a significant average slowing on a desktop task does not by itself establish how a particular motorist steered, whether a specific collision would have happened otherwise, or whether that person crossed a jurisdiction’s legal impairment standard. These questions require independent evidence and a reasoned bridge from research setting to case facts. The site’s forensic toxicology research library organizes additional specimen and performance topics separately.
What are the main sources of uncertainty for case-specific use?
Exposure measurement: the target relied on breath measurements and a standardized partition conversion, not a certified case sample. Route of administration: intravenous clamping reduces one kind of exposure variability but does not reproduce oral absorption, changing post-drink levels, or every real-world consumption pattern. Participant selection: forty selected medium-to-high-risk drinkers without diagnosed alcohol use disorder are not a representative cross-section of all drivers, ages or medical conditions. Behavioral specificity: a Counting Stroop difference score is not synonymous with safe vehicle control or a validated individual forensic classifier.
Study design and precision: the task was a secondary endpoint of a broader medication trial, the sequence could allow practice, and the model’s sensitivity is finite. Naltrexone/placebo assignment in the parent study and source supplementary analyses should be accounted for rather than ignored. Individual results, alternate tests, exposure uncertainty and case timing could change an expert’s opinion; a group-average result is not a personalized reconstruction.
When deciding what a cognitive result means, attorneys and experts should first specify the construct, actual measured variable, sampling matrix, analytical or behavioral instrument, comparison condition, participant denominator, model assumptions and intended legal proposition. The scientific editorial principles described in the site’s research source policy are relevant to maintaining that separation.
Frequently asked forensic questions
Does a normal Stroop interference score mean alcohol caused no slowing?
No. Riedel and colleagues detected an overall estimated 17-ms reaction-time increase under their controlled alcohol condition despite finding no detectable change in the extra incongruent-minus-congruent penalty. A difference score can remain stable while both underlying response times change.
Was the alcohol concentration a directly assayed forensic whole-blood BAC?
No. End-expiratory breath ethanol was measured and converted using a 2,100:1 partition assumption to a theoretical arterial blood-equivalent concentration during an intravenous clamp. Neither a venous whole-blood gas-chromatography result nor a driving-time BAC was obtained for a forensic case.
Did the experiment test actual driving or collisions?
No. Participants pressed keys during a computerized Counting Stroop task. No on-road driving, collision investigation, standardized roadside sobriety test or crash-causation outcome was included in the reported experiment.
Does p=.337 establish that alcohol’s effect on interference was exactly zero?
No. It means that this analysis did not detect the specified interaction at the study’s sample size and precision. A nonsignificant effect is not an individual guarantee, a universal null result, or an equivalence test.
What does the Bayes factor of approximately 3.5 mean?
The reported model comparison favored a reduced model without an alcohol-by-trial-type interaction by approximately 3.5 to one against the stated full model. That is relative evidence about model descriptions of the study data—not an estimate of test sensitivity, specificity or individual fitness to drive.
Could a different cognitive task show a different alcohol effect?
Yes. Different paradigms isolate different processes and may be susceptible to exposure, training, task complexity and participant characteristics. The original paper explicitly distinguished attentional-interference scoring from response-inhibition tasks; results from one should not be silently transplanted onto the other.
What can a forensic expert reasonably cite from this experiment?
An expert can cite the controlled group’s slower average reaction times, the preserved tested interference contrast, the methods for calculating each outcome, and the uncertainty that limits extrapolation. A case conclusion about a particular person’s BAC, functional impairment, or causation requires separate qualified case evidence and cannot be read directly from these forty volunteers.
Scientific conclusion and original primary source
The defensible conclusion is narrow: in this controlled forty-person Counting Stroop crossover, alcohol was associated with slower overall responding while the specifically defined interference contrast showed no detected increase. The study is a warning against treating an unchanged derived score as proof of unchanged absolute function. It is equally a warning against translating an average laboratory response into an unvalidated individual BAC or crash opinion.
Primary paper: Riedel P, Wolff M, Spreer M, et al. Acute alcohol does not impair attentional inhibition as measured with Stroop interference scores but impairs Stroop performance. Psychopharmacology. 2021;238:1593–1607. https://doi.org/10.1007/s00213-021-05792-0.