Answer: In a randomized laboratory experiment involving 142 university students who drank alcohol weekly, participants assigned 0.6 g/kg ethanol made more risk-taking responses on a simulated balloon task than those assigned a placebo beverage. The ethanol group also reported a larger increase in urge to drink. The observed balloon-task difference was modest: 33.37 versus 28.68 adjusted pumps (p = 0.019; Cohen’s d = 0.40). The authors’ statistical mediation analysis connected post-drink balloon decisions with change in drinking urge, but it did not establish that any named person’s real-world decision, impaired judgment, legal intent, or injury was caused by alcohol.
The original investigation is Rose, Jones, Clarke and Christiansen, Alcohol-induced risk taking on the BART mediates alcohol priming, Psychopharmacology 231:2273–2280 (2014), DOI: 10.1007/s00213-013-3377-1. It is a study of hypothetical monetary choices and self-reported desire for alcohol, not a driving test, formal forensic impairment examination or validation of an individual’s alcohol concentration.
What did the experiment actually test?
The researchers asked three related but distinguishable questions. First, does a controlled ethanol dose change the number of chances a participant takes on a computer-based risk task? Second, does post-drink risk-taking performance correlate with the participant’s self-reported customary drinking, above and beyond questionnaire measures of trait impulsivity and sensation seeking? Third, might a model using risk-task behavior statistically account for part of the alcohol-related increase in reported desire for another drink?
Those questions are not interchangeable. The first concerns a randomized group contrast on a defined behavioral outcome. The second is an association with recalled drinking history. The third is a mediation model involving a nonrandomized intermediate variable. A significant result in one category does not automatically validate conclusions in the others.
Who participated, and what were they given?
The investigators recruited 142 students, including 67 men, from the University of Liverpool. Participants were weekly drinkers, generally healthy, English-speaking, and excluded if they had current or historical alcohol or drug-use disorders or took medication that could interact with alcohol. Mean age was 20.33 years (standard deviation 3.74). Mean self-reported weekly consumption was 22.92 UK alcohol units, with one UK unit defined as 8 g of ethanol. These were not a representative sample of licensed drivers, patients, criminal defendants, eyewitnesses or persons with alcohol dependence.
Each person attended one experimental session and was randomized, stratified by sex, to ethanol or placebo. The ethanol beverage contained 0.6 g of ethanol per kilogram of body mass as vodka and lemonade, totaling 400 mL of beverage. Placebo consisted of 400 mL lemonade in glasses misted with vodka. The beverage was consumed in three portions over the structured drinking period. Participants then rested for 20 minutes before the post-drink questionnaires and computer task.
An administered 0.6 g/kg dose is not itself a measured blood or breath concentration. The paper’s baseline eligibility reading was 0.0 mg/L on a breath instrument, followed by readings around the task. No venous whole-blood specimen was separately assayed by forensic headspace gas chromatography in this experiment.
What does the Balloon Analogue Risk Task measure?
The Balloon Analogue Risk Task, or BART, is a game-like laboratory measure of repeated decisions under a reward-versus-loss contingency. Participants clicked to inflate each of 20 on-screen balloons. Every successful pump increased a temporary hypothetical reward by $0.05. The participant could bank the temporary amount at any point. If a balloon unexpectedly burst before banking, the temporary amount for that balloon was lost. The programmed bursting distribution had a mean explosion point of 64 pumps.
The article’s prespecified behavioral variable was adjusted average pumps: the average number of pumps made on balloons that did not burst. This is narrower than a generic count of all clicks, a direct measure of financial loss, a validated score for reckless driving, or a measure of willingness to injure another person. Choosing additional pumps can be strategically sensible in some environments and counterproductive in others. The task operationalizes a particular risk-reward tradeoff under the study’s rules; it does not directly observe a real-world transaction or harmful action.
Because the money was hypothetical and the observation was a single session, incentives, learning, experimenter context and participant expectations could affect the outcome. The authors themselves identified real-reward experiments and within-person designs as future research needs.
What was directly measured, and what was only reported or modeled?
| Evidence category | Actual study variable | What it cannot establish by itself |
|---|---|---|
| Administered exposure | Assigned oral ethanol at 0.6 g/kg, or an alcohol-cued placebo | A measured serum or forensic whole-blood BAC; the concentration at a later unrelated event |
| Direct physiological sample | Exhaled breath alcohol readings near the task, in mg/L | Independently assayed venous whole-blood ethanol; urine, oral-fluid, or vitreous ethanol |
| Observed task behavior | Adjusted average pumps across 20 hypothetical BART balloons | Actual road behavior, violence, carelessness, consent, truthfulness or legal intent |
| Self-reported current state | Alcohol Urge Questionnaire score before and after drinking | An observed second drink or proof a participant would later consume alcohol |
| Self-reported background | Recent drinking recalled on a two-week Timeline Follow Back; questionnaire trait scores | Forensically confirmed intake, a validated individual history, or an alcohol-use diagnosis |
| Statistical model | Regression and mediation estimates using task and questionnaire results | A demonstrated personal causal chain from ethanol to a particular act or injury |
Separating these categories matters because a behavioral effect and a biological measurement answer different questions. A BART score may differ after a controlled drinking assignment without any evidence about whole-blood ethanol, driving competence, or event-specific causation.
How much alcohol was measured on the breath instrument?
The authors reported mean breath-alcohol readings in the ethanol condition of 0.36 mg/L (SD 0.10) before the task and 0.36 mg/L (SD 0.08) after the task. These values are measurements of exhaled breath reported in mass per volume of breath. They should not be silently reprinted as a measured blood concentration of 0.36 mg/mL, 0.36 g/L, or 0.36%.
A blood-alcohol-equivalent number, if created with a specified breath-to-blood partition convention, would be a conversion subject to an assumption, not a newly drawn specimen. The paper did not characterize the individual breath-to-blood partition ratios or quantify analytical uncertainty for a later forensic test. In litigation, the actual instrument printout, instrument calibration and contemporaneous collection records remain separate evidence from this experimental report.
Did ethanol change observable risk-task choices?
Yes, in this sample and under this task definition. The ethanol group averaged 33.37 adjusted pumps (SD 11.66), compared with 28.68 (SD 11.49) for the placebo group. The reported between-group contrast was t(140) = 2.38, p = 0.019, d = 0.40. The difference is about 4.69 pumps on the adjusted measure, not 4.69 extra actual injuries, losses, crimes or seconds of impaired response.
Although this randomized contrast supports a group-level experimental effect on the BART outcome under these conditions, it does not quantify the probability that an individual student was affected. The investigators found neither a statistically significant main effect of sex nor a sex-by-treatment interaction in their additional task analysis (p > 0.1). A nonsignificant interaction does not prove identical pharmacological or behavioral responses for every man and woman.
Did participants report wanting to drink more?
Participants rated current desire or urge to drink before and after their assigned beverage. Both groups showed an increase, but the ethanol group increased more. The reported group-by-time interaction was F(1,140) = 23.38, p < 0.001; the between-group comparison of change scores was t(140) = 4.83, p < 0.001, d = 0.81. At baseline the groups’ urge scores did not differ significantly (p = 0.68), whereas their post-drink scores did (p = 0.001).
The finding matters for the paper’s specific hypothesis about alcohol priming—increased self-reported motivation to drink following an initial beverage. However, no subsequent open-bar consumption trial demonstrated that these participants actually chose to drink additional alcohol as a consequence. The measured outcome was a questionnaire score. It is not a validated individual prediction of drinking or a direct demonstration of real-world compulsion.
What did the regression and mediation models add?
The researchers examined whether BART performance was associated with participants’ retrospectively reported usual alcohol intake, while controlling for Barratt impulsivity and sensation-seeking scores. In the ethanol group, adjusted pumps contributed to a model of weekly alcohol consumption (F-change(1,67) = 5.01, p = 0.03; standardized β = 0.27) and to a model of reported binge frequency (F-change(1,66) = 4.97, p = 0.032; β = 0.27). Comparable added predictive associations were not significant in the placebo group.
These are within-sample associations with self-reported history, not a prospective demonstration that pumping a balloon forecasts someone’s future alcohol consumption or a validated clinical screening rule. The model relationship does not establish an individual risk threshold or causation, and retrospective self-report is not independent verification of actual alcohol intake.
The mediation analysis treated randomized drink assignment as the exposure, post-drink BART responses as a possible intermediary, and change in questionnaire urge as the outcome. The paper reported statistically significant component paths (assignment-to-BART p = 0.019; BART-to-urge change p = 0.023) and a model-based indirect effect. Because the intermediary was not randomized, a mediation calculation cannot exclude all confounding between the BART behavior and urge response. It does not show that making riskier decisions caused later craving, much less that craving caused another purchase or an offense.
Which primary results belong on an evidence chart?
| Study question | Primary reported result | Defensible interpretation |
|---|---|---|
| Exhaled breath near BART, alcohol arm | 0.36 ± 0.10 mg/L before; 0.36 ± 0.08 mg/L after (means ± SD) | Direct breath readings, not independently drawn whole blood |
| Risk taking, adjusted nonburst balloon pumps | Alcohol 33.37 ± 11.66; placebo 28.68 ± 11.49; t(140) = 2.38, p = 0.019; d = 0.40 | Group-level laboratory behavior contrast; hypothetical reward |
| Change in self-rated urge to drink | Drink × time interaction F(1,140) = 23.38, p < 0.001 | Different questionnaire changes, not directly observed repeat drinking |
| Weekly consumption association, alcohol subgroup | F-change(1,67) = 5.01, p = 0.03; β = 0.27 | Regression against recalled prior drinking, not prospective prediction |
| Binge-frequency association, alcohol subgroup | F-change(1,66) = 4.97, p = 0.032; β = 0.27 | Group association with retrospective binge metric, not a diagnosis |
| Proposed statistical indirect pathway | Assignment-to-BART p = 0.019; BART-to-urge change p = 0.023 | Mediation model supports a hypothesis but cannot prove an individual’s causal chain |
Every row has its own denominator, variable type and purpose. Neither a behavioral test statistic nor an urge-questionnaire effect should be mistaken for an assay result or an individual legal threshold.
Why might the study be useful in alcohol-related litigation?
The study helps test a categorical claim such as “alcohol cannot affect willingness to accept reward-related uncertainty without obvious physical signs.” It shows an experimentally observed group change on one objectively scored computerized task after a controlled ethanol dose. That is relevant background science for evaluating how attorneys, investigators or expert witnesses frame alcohol-related decision research.
The same study also limits opposite categorical claims. The BART is not a roadside field sobriety test, driving simulator, validated violence-prediction instrument, capacity test, or reconstruction of a disputed contract or financial transaction. A participant’s propensity to take an extra hypothetical balloon pump is not an observed willingness to cause physical harm. Whether a real person understood consequences, consented, acted intentionally, was negligent or was impaired at a specific historical time must be assessed from evidence appropriate to that issue.
For example, in dram-shop, personal-injury or criminal proceedings, proof that ethanol may change laboratory risk-taking behavior would not identify where a person purchased a beverage, whether intoxication was observable to a particular server, what that person actually consumed at the relevant time, or whether alcohol caused a subsequent event. Those facts require contemporaneous records and jurisdiction-specific legal analysis outside this experiment.
What confounders and design limits matter most?
The study enrolled university students who drank at least weekly; mean self-reported AUDIT score was 15.17. It did not sample a broad general population, a forensic case series, or a clinically dependent cohort. It used one randomized assigned-beverage session per participant and a structured laboratory setting. Hypothetical rewards, expectations generated by the vodka-misted placebo, timing during the ascending exposure period, and sample selection all affect generalizability.
Adjusted balloon pumps were measured only after drink assignment, not as a repeated within-person sober-to-drunk change in that same participant. The investigators’ separate statistical tests of drinking history relied on self-reported two-week intake. The urge outcome was an immediate self-report, not the number of later drinks actually consumed. The mediation calculation is therefore best read as an exploratory mechanism model, not a demonstrated sequence of events within an individual.
Any comparison with another paper should match outcome definition and timing before combining findings. A balloon-pumping task, delay-discounting problem, stop-signal task, known-probability lottery, and naturalistic driving trial all operationalize different constructs. They are not interchangeable “alcohol makes people reckless” instruments. Likewise, a nonsignificant relationship in the placebo arm is not direct proof that no such relationship exists in sober people generally.
How should a forensic reader test an expert’s interpretation?
First establish the proposed claim’s endpoint. Is the expert discussing a directly measured alcohol concentration, a modeled earlier concentration, a physiological effect, a questionnaire response, a computerized action, or a case-specific legal conclusion? Then identify the independent human denominator, actual comparison, sampling time, statistical adjustment, effect magnitude and relevant outcome measure.
For this Rose et al. study, a source-faithful opinion should use exhaled breath mg/L for the instrument results, describe the observed BART pump difference as a group-level hypothetical-task finding, identify the urge measure as self-report, and characterize mediation as a statistical analysis rather than an unconfounded causal demonstration. An individual reconstruction would require separately authenticated records of actual alcohol administration or drinking, specimen testing, behavior, and the disputed event.
Related research-reading resources include the Attorney Scientists research library, its editorial and source policy, and an explanation of why two blood samples do not automatically settle an earlier BAC. These are context links, not additional experimental evidence for the Rose results.
Frequently asked questions
Does this paper prove alcohol makes every person take more risks?
No. It reports a statistically significant average difference in adjusted balloon pumps between randomized groups of 142 students. Variation within groups remained substantial, and a mean group difference is not an individual classification rule.
Was the 0.36 mg/L result a blood alcohol concentration?
No. The source explicitly reports a breath alcohol concentration in mg of ethanol per liter of exhaled breath. It is not an independently analyzed forensic venous whole-blood specimen and should not be rewritten using blood units without identifying the conversion assumptions.
Did the study observe people purchasing or drinking a second alcoholic beverage?
No. The immediate priming outcome was change in a questionnaire about desire to drink. It is not an observation of later alcohol purchases or consumption.
What does adjusted average pumps mean on the BART?
It is the average number of pumps on balloons that did not explode, across 20 trials with hypothetical rewards. It is not a count of roadway risks, legal violations, crashes, monetary losses or physical injuries.
Does the mediation analysis establish that risk taking caused craving?
No. The mediation model found an indirect statistical association consistent with that hypothesis. Task behavior was not independently randomized, and model assumptions cannot prove an unconfounded individual causal mechanism.
Can a BART score establish that a witness, driver or defendant was impaired?
No validated case-specific classification or threshold was established. The article measured hypothetical task choices and reported urge in university students; it did not assess a named person’s testimony, actual driving performance, criminal state of mind or injury causation.
Does a significant group effect tell an expert what BAC was present during a later event?
No. The observed experimental breath readings arose at specified session times after a standardized dose. They do not measure a different person’s whole-blood BAC, establish a retrospective elimination rate, or validate a historical impairment estimate.
Bottom line: separate what the experiment showed from what a case requires
Rose and colleagues offer original evidence that a controlled 0.6 g/kg ethanol condition produced more hypothetical balloon pumping and a larger increase in self-reported drinking urge than an alcohol-cued placebo in a 142-student experiment. Their associations and mediation model are useful for generating and critically evaluating scientific explanations of alcohol-related motivation. They cannot substitute for direct forensic specimen evidence or prove a particular individual’s risky act, state of mind, clinical impairment, negligent service, or accident causation.
Primary source: Rose AK, Jones A, Clarke N, Christiansen P. Alcohol-induced risk taking on the BART mediates alcohol priming. Psychopharmacology. 2014;231:2273–2280. https://doi.org/10.1007/s00213-013-3377-1. This is an AI-assisted research appraisal of a published primary study, not a statement that the named attorney personally evaluated these results or formed a case-specific expert opinion.