Short answer: Not consistently. In a randomized, double-blind, four-condition crossover experiment involving 24 adults, alcohol increased errors across three laboratory attention tasks and their combined attention-error score. Adding an 80-mg-caffeine energy drink did not reliably cancel those errors. At 45 minutes, the alcohol-plus-energy-drink condition actually produced more visual-search errors than alcohol alone. A few working-memory outcomes improved with the combination at particular testing times, but those isolated improvements cannot establish that a person is unimpaired, fit to drive, or accurate as an eyewitness.
This distinction matters in litigation because a stimulant-containing drink can change a particular task outcome without reversing ethanol exposure or proving that attention, judgment, and practical performance have returned to normal. The source is the complete original study by Benson, Tiplady, and Scholey (2019), PLOS ONE, DOI 10.1371/journal.pone.0209239. Its reported findings concern a controlled laboratory study, not a validated roadside examination or a reconstruction of any actual collision.
What did the original experiment directly establish?
The investigators compared alcohol alone, a caffeinated energy drink alone, both together, and their corresponding placebos within the same 24 participants. The clearest pattern was an alcohol-associated increase in incorrect responses in several attention tasks, while the measured response speed on the available attention measures changed far less consistently. The study therefore illustrates why the apparent speed of someone’s response cannot safely be used as a substitute for checking whether that response was correct.
Its factorial design is unusually informative for this narrow question. Alcohol-plus-energy-drink effects could be compared both with alcohol alone and with energy drink alone; investigators also tested the same people at multiple times after beverage consumption. Nevertheless, one small controlled trial cannot provide a universal rule about all caffeine doses, brands, individuals, or real-world tasks.
Who participated, and how were the four conditions arranged?
Twenty-four healthy adults took part, including 12 women and 12 men. Mean age was 22.83 years; the reported range was 18–40. The participants regularly consumed caffeine and alcohol and were screened for specified medical and psychiatric exclusions. The sample does not represent every driver, witness, older adult, person with a substance-use disorder, or person using other sedating or stimulating medications.
Each person completed four sessions in a randomized sequence with at least 48 hours between them. The investigators used a double-blind, placebo-controlled crossover arrangement and matching sensory cues. At the start of every session, they obtained a baseline breath reading and conducted the tablet-based cognitive battery. Participants then consumed the assigned drink over ten minutes, and the cognitive tests were repeated at 45, 90, and 180 minutes.
The alcohol dose was 0.6 grams of ethanol per kilogram of body weight, delivered as vodka. The active energy-drink dose was a 250-mL serving containing 80 mg of caffeine. This is one alcohol dose and one energy-drink formulation—not a comparison of all possible alcohol–caffeine combinations. The design used a noncaffeinated, flavor-matched energy-drink placebo and sensory cues for the alcohol placebo.
| Study abbreviation | Administered treatment | Useful comparison; critical limitation |
|---|---|---|
| PLA | Water plus placebo energy drink | Baseline comparator for the experimental beverage conditions; not a no-expectancy real-world driving control. |
| ALC | 0.6 g/kg ethanol plus placebo energy drink | Alcohol-associated laboratory changes in the absence of the active energy drink. |
| ED | Water plus 250 mL energy drink containing 80 mg caffeine | Effects of the complete beverage, not an isolated pure-caffeine experiment. |
| AMED | 0.6 g/kg ethanol plus the same active energy drink | Tests whether that particular formulation modifies an alcohol effect on a specific outcome at a specific time. |
A counterbalanced crossover reduces between-person differences because each participant encounters all four conditions. It does not remove uncertainty from relatively small samples, particular laboratory tasks, incomplete blinding, repeated testing, or testing at different times of day.
Was ethanol measured in blood, breath, or inferred from the administered dose?
These are separate quantities. The investigators administered 0.6 g/kg ethanol and aimed for an approximate peak blood-alcohol-equivalent concentration of 0.05%. Those are a dose and an experimental target, not direct assays of a particular person’s venous whole blood.
The actual exposure observations were made with a Lion Alcolmeter SD400PA breath-testing instrument at baseline and 45 and 210 minutes. Although the authors use the shorthand “BAC” for the displayed or derived alcohol readings, the described protocol does not report separately drawing, preserving, and measuring venous whole blood using a forensic chromatographic method. Breath measurements must not silently be relabeled as laboratory-analyzed whole-blood ethanol.
The alcohol-alone and alcohol-plus-energy-drink conditions did not show a statistically significant difference in the reported breath alcohol comparisons at 45 minutes (t(23) = 1.44, p = .163) or 210 minutes (t(23) = 1.24, p = .227). Failure to detect a difference is not proof that the two concentration distributions are identical, and the result is not evidence that an energy drink accelerates alcohol clearance. It only describes those comparison tests in this sample.
For any disputed forensic result, the analytical method, specimen, collection time, calibration, and uncertainty must be established independently. A general discussion of the limits of converting measurements across times appears in the same site’s analysis of alcohol elimination from two samples; nothing in this experiment supplies a case-specific elimination rate.
Which attention tests became less accurate after alcohol?
Three computerized tasks measured different components of attention. In Four-Choice Reaction Time, participants tapped a response corresponding to a lit circle. In Number Pairs, they judged whether selected digits matched. In Visual Search, they located an L-shaped target among differently oriented distractors. Errors, not just response time, mattered to each outcome.
Across testing, alcohol was associated with more incorrect answers on each of the three tasks: Four-Choice Reaction Time, F(1,23) = 7.92, p = .010; Number Pairs, F(1,23) = 7.02, p = .014; and Visual Search, F(1,23) = 8.52, p = .008. The aggregated attention-error score also showed an alcohol main effect, F(1,23) = 11.95, p = .002.
The study did not provide usable reaction-time data for the Four-Choice Reaction Time task because of a software failure. For the measured attention reaction-time composite, the authors did not report a significant treatment effect. This is not permission to conclude that all reaction times were measured perfectly or remained unchanged; the missing primary-task speed recording and individual Number Pairs interactions need to be acknowledged.
The strongest defensible interpretation is narrower than “alcohol makes everyone slower.” Participants in this experiment could continue to respond comparatively quickly while making more mistakes. That speed–accuracy separation is relevant when a disputed observation turns on whether someone noticed and correctly classified a stimulus rather than simply how promptly they acted.
Did the combination perform better or worse than alcohol alone?
Both outcomes appeared, depending on the task and time. At 45 minutes, the combined AMED condition generated more Visual Search errors than alcohol alone, t(23) = 3.11, p = .005, as well as more errors than energy drink alone, t(23) = 3.08, p = .005. This is an important negative finding for any blanket claim that a caffeinated drink counteracts alcohol-associated errors.
At 90 minutes, the Serial Sevens working-memory task gave a different result: participants made fewer errors in AMED than in ALC, reflected in the comparison t(23) = 2.48, p = .021. At 45 minutes, the authors also reported favorable AMED-versus-ALC comparisons for Digit–Symbol Matching errors (p = .022) and response time (p = .033). These are time-specific findings, not proof that cognitive performance generally normalized.
Study-wide conclusions must account for the uneven pattern: worsened Visual Search with AMED at one time, improvements on other tasks at selected times, and no consistent reversal across the battery. Even if an intervention improved a particular task for a particular participant, that would not establish competent performance on an unmeasured task such as hazard perception, accurate recall of an incident, or sustained lane control.
What does the study’s speed–accuracy pattern mean for interpreting behavior?
A person may make a decision quickly and still make the wrong decision. Conversely, a longer response need not signify failure to perceive the stimulus. Laboratory performance has at least two dimensions: speed and correctness. Treating one dimension as a proxy for the other removes important information and can lead to overconfident forensic inferences.
Here the authors characterized alcohol’s effects as a speed–accuracy trade-off: error rates on multiple attention outcomes increased without comparably broad slowing in the reaction-time measures that were actually available. One Number Pairs reaction-time interaction involved alcohol, energy drink, and time, underscoring that the pattern was not identical for every task.
Consider an illustrative claim that someone “responded immediately, so their attention was intact.” The experiment demonstrates why response latency alone cannot establish accuracy in laboratory tasks. It does not establish that any named person in a particular incident made an error, that a witness’s testimony is false, or that a driver’s moment-by-moment decision was caused by ethanol. Those are additional, individualized questions requiring evidence about the actual circumstances and behavior.
What did the working-memory tasks add?
Working memory is not a single test result. The researchers used Memory Scanning, Serial Sevens, Digit–Symbol Matching, and Visuospatial Working Memory. Conditions containing the energy drink were associated with fewer Memory Scanning errors on an overall main-effect test (F(1,23) = 5.42, p = .029), but the follow-up analysis found no significant pairwise differences at individual times. Digit–Symbol Matching showed some energy-drink-related accuracy and speed effects. The Visuospatial Working Memory reaction-time interactions were not accompanied by significant between-treatment differences at an individual measured time.
For the composite working-memory reaction-time outcome, an energy-drink-by-time interaction was reported, F(2,46) = 4.27, p = .020. At 90 minutes ED was faster than placebo in a nominal follow-up comparison (p = .040), but the investigators cautioned that some composite conclusions did not persist after accounting for multiple comparisons.
It would be unsound to compress this heterogeneous collection into “caffeine restored working memory.” Some measures, at some times, moved favorably; others showed no significant effect, and one attention measure moved unfavorably when alcohol and energy drink were combined. An isolated positive test result cannot substitute for a validated battery or case-specific functional evidence.
What are the quantitatively supported takeaways?
The table summarizes selected original, explicitly reported tests. These p-values belong to distinct analyses in a relatively small repeated-measures experiment; they are not individual diagnostic probabilities, classification accuracies, absolute error counts, or estimates of collision risk. “More errors” describes the study-level direction, not a constant effect in every participant.
| Outcome / contrast | Reported statistical result | What it supports—and does not |
|---|---|---|
| Four-Choice Reaction Time, errors | Alcohol main effect: F(1,23) = 7.92; p = .010 | More errors with alcohol across conditions; reaction-time data for this task were lost to a software problem. |
| Number Pairs, errors | Alcohol main effect: F(1,23) = 7.02; p = .014 | Task-specific association with incorrect judgments; not an eyewitness-veracity test. |
| Visual Search, errors | Alcohol main effect: F(1,23) = 8.52; p = .008 | Group-level visual-search accuracy reduction, not a roadside eye or vision diagnosis. |
| Combined attention-error measure | Alcohol main effect: F(1,23) = 11.95; p = .002 | Convergent laboratory error finding; no driving or accident outcome was measured. |
| Visual Search, AMED versus ALC at 45 min | t(23) = 3.11; p = .005 | More errors after the combination than alcohol alone at this time; not a general rule for every task. |
| Serial Sevens, AMED versus ALC at 90 min | t(23) = 2.48; p = .021 | Fewer errors in the combination condition on one working-memory task at this time; not proof of restored fitness. |
| Breath exposure, ALC versus AMED | 45 min: p = .163; 210 min: p = .227 | No detected difference between these treatment readings, not directly assayed whole-blood equivalence or a verified elimination effect. |
How should statistical multiplicity change confidence in isolated differences?
The primary investigators intentionally did not correct individual follow-up comparisons for multiple testing, explaining that they wished to detect potentially harmful interactions. That choice increases sensitivity to possible signals but also increases the risk that one or more nominal p < .05 findings occur by chance when numerous tests are conducted.
The authors themselves distinguished findings that would remain after adjustment from more fragile follow-up effects. They described the main alcohol effects on the individual attention outcomes and attention composite as comparatively robust, along with the 45-minute AMED-versus-ALC Visual Search difference. By contrast, not every working-memory or composite time-specific finding survived such scrutiny. The public conclusion here therefore emphasizes convergent attention-error evidence rather than marketing a favorable isolated result.
Statistical significance also does not report the probability that a particular person was impaired; a study test cannot be converted into a criminal standard of proof or a numeric collision-causation estimate without additional evidence and a suitable validated inference procedure.
Could the reported energy-drink effect be attributed to caffeine alone?
No. The active drink contained 80 mg caffeine but also represented a real formulated product. The authors specifically noted that the placebo lacked other active ingredients, including taurine and B vitamins, and that glucose in the placebo could affect cognitive comparisons. Consequently, ED-versus-PLA is a comparison of formulations, not a controlled pure-caffeine isolation. The protocol did not measure blood or saliva caffeine concentrations to connect a person’s systemic caffeine concentration to their performance.
It also matters that many participants were regular caffeine consumers who abstained beforehand; relief of caffeine withdrawal might partly influence some outcomes. The experiment tested one fixed caffeine dose, while ethanol was administered by body weight. A universal alcohol-to-caffeine antidote relationship cannot be estimated from those two specific regimens.
What are the main limitations before applying the evidence to litigation?
The participant group was small, comparatively young, and selected for healthy regular users. Laboratory tablet tests, performed while participants remained at a research center, are not the same as controlling a moving vehicle, monitoring a fast-changing roadway, remembering a crime, or making a legally significant statement. The experiment did not report a validated person-level sensitivity or specificity for intoxication or impairment; it did not test collision causation, witness credibility, or legal decision-making.
Additional limitations include restricted doses and test times; a software failure removing Four-Choice Reaction Time latency data; possible knowledge of whether alcohol had been consumed despite blinding; morning versus afternoon testing; caffeine abstinence; unmeasured caffeine pharmacokinetics; and multiple comparisons. The source provides group statistics, not the medical histories or objective incident records needed to decide what happened to a specific person.
Importantly, no separate venous whole-blood assay, no serial forensic blood draw, and no validated breath-to-blood individual conversion were documented in this protocol. Nor did the study directly measure alcohol in the brain, receptor occupancy, or a person’s subjective belief that they were safe. A reported experimental dose and a breath-testing display should not be presented as mutually interchangeable chemical evidence.
How can an attorney or scientific reviewer use this study responsibly?
The defensible litigation use is to test overbroad generalizations. If someone argues that quick replies demonstrate accurate attention, ask whether correctness was independently measured. If a party claims that consuming an energy drink “cured” alcohol-related impairment, ask which endpoint, dose, time, population, and specimen actually support that claim. If a report extrapolates an administered dose to a precise historical blood concentration, ask for the separate specimen or validated pharmacokinetic model supporting that step.
In evaluating actual evidence, separate four layers: the administered beverage; the chemical matrix actually sampled; the laboratory task result; and any asserted individual real-world behavior. Each transfer between layers introduces assumptions. An analytical ethanol result does not automatically quantify attention, and a group-level computerized error difference does not identify a cause of a particular accident.
Readers seeking general research resources or the site’s methodology may consult the scientific resources collection and editorial standards. These contextual links do not substitute for the original Benson–Tiplady–Scholey article, which should be read with its methods, figures, and limitations intact.
Frequently asked questions
Does caffeine in an energy drink sober a person up?
This study did not establish that. Adding the energy drink did not produce a statistically detected difference from alcohol alone in the study’s measured breath comparisons at 45 or 210 minutes. An energy drink is not proof of ethanol elimination, accurate attention, or restored fitness.
Can someone respond quickly while becoming less accurate after alcohol?
Yes, that distinction appears at the group level in these computerized tasks. Alcohol increased errors in three attention outcomes, while a corresponding broad reaction-time slowing was not demonstrated in the recorded attention composite. One task’s reaction-time data were unavailable due to software failure.
Did alcohol mixed with energy drink always perform worse than alcohol alone?
No. The combination produced more Visual Search errors at 45 minutes but fewer Serial Sevens errors at 90 minutes, along with favorable isolated Digit–Symbol outcomes. The direction depended on the test and time, and no consistent overall reversal or exacerbation was established.
Was the reported alcohol concentration measured directly in whole blood?
No independent venous whole-blood laboratory ethanol assay is described. Investigators measured exhaled breath using an Alcolmeter and used blood-alcohol-style terminology. The oral 0.6-g/kg dose and intended 0.05% target are distinct from an actual forensic whole-blood measurement.
Can this trial establish whether a particular driver was impaired?
No. It did not measure actual on-road driving or supply validated person-level impairment thresholds. The experimental group findings may inform questions for an expert but do not resolve a named individual’s functional ability or incident causation.
Does a nominal p value below .05 prove that an energy drink reversed impairment?
No. The trial tested many tasks and time points without routinely correcting each follow-up comparison for multiplicity. A single comparison may be hypothesis-generating; broader claims require convergent evidence and suitable clinical or forensic validation.
Could energy-drink ingredients other than caffeine have contributed?
Yes. This was a comparison of complete beverage formulations, not isolated caffeine. The authors discuss other active ingredients, glucose-containing placebo, caffeine abstinence, and unmeasured caffeine concentrations as limitations on assigning a specific mechanism.
Primary scientific reference
Benson S, Tiplady B, Scholey A. Attentional and working memory performance following alcohol and energy drink: A randomised, double-blind, placebo-controlled, factorial design laboratory study. PLOS ONE. 2019;14(1):e0209239. doi:10.1371/journal.pone.0209239. Study interpretation above is limited to the methods, reported statistical results, and constraints of this original primary publication.