Does Alcohol Make Unknown-Odds Choices More Appealing? What a 100-Person Bar Study Found

Short answer: In an observational study of 100 adults recruited at a restaurant bar, people in higher breathalyzer-based alcohol reading groups chose a two-drink lottery with unknown odds more frequently than people in the lower group. The association was reported for the 54 men but not for the 46 women. The investigators did not randomly assign alcohol, collect forensic blood samples, test actual driving or determine anyone’s legal intent. These findings describe a choice made in one setting; they cannot establish that alcohol caused an individual to take a particular risk.

The primary evidence is Tyszka, Macko and Stańczak, Alcohol reduces aversion to ambiguity, Frontiers in Psychology (published January 15, 2015), DOI: 10.3389/fpsyg.2014.01578. This article distinguishes an observed preference for unknown probabilities from risk-taking with known probabilities, and it separates an exhaled-breath instrument reading from a directly measured whole-blood ethanol concentration.

What question did the researchers actually test?

The investigators asked whether people with higher alcohol readings were less averse to ambiguity: uncertainty about the probabilities of possible outcomes. In the classic distinction, a risky choice presents specified probabilities, whereas an ambiguous choice leaves the probabilities incompletely specified. Those are not interchangeable constructs. A preference for an unknown-odds lottery does not necessarily mean greater willingness to speed, fight, gamble substantial money, drive while intoxicated or engage in any other particular conduct.

Here the alternatives differed not only in how certain the probabilities were, but also in the size of the potential reward. The known-odds urn offered one drink; the unknown-odds urn offered two drinks. Choosing the second could reflect a greater attraction to the potential prize, a lower aversion to missing probability information, preferences concerning alcohol itself or a combination of such factors. The study’s comparison does not isolate one psychological mechanism.

How did the bar study recruit and evaluate participants?

The investigators approached customers in a restaurant located within a leisure center. The final study included 100 participants: 54 men and 46 women, ages 18–43 years (mean 26.3; standard deviation 5.35). The study was performed in an ordinary drinking environment rather than a laboratory in which subjects were randomly given an alcohol dose or placebo. It was therefore an observational, between-person comparison, not a controlled causal alcohol challenge.

Participants first estimated their own alcohol level by choosing among six stated ranges. They were then instructed to rinse their mouths with water and provide an exhaled-breath sample. The investigators required an interval of at least 20 minutes since the last drink or cigarette. Those safeguards address some immediate sampling concerns, but the publication does not present paired venous whole-blood gas-chromatography measurements, a forensic chain-of-custody assessment or a device-validation study for this exact application.

Finally, participants chose between two urns containing coupons for drinks available at the bar. The reward was sufficiently concrete to motivate participation, but it also means the choice might depend on a participant’s desire for additional drinks. No subsequent dangerous act, traffic offense, injury or driving performance was measured.

What were the two alternatives, and why were their odds different?

The unknown-odds option cannot be assigned an expected payoff without knowing how many of its 30 coupons were winners. The experiment did not provide that missing number to participants. Thus the results should not be recast as the fraction of people preferring an option with a known probability of success, nor as a demonstration that they made mathematically irrational choices. There was no separate measure of individual beliefs about how many winning coupons the ambiguous urn held.

What alcohol quantity did the investigators measure?

The device was identified as an Alkohit X100 breathalyzer. Although the primary article sometimes refers to the device’s output as a “blood alcohol level,” the described biological sample was exhaled breath, not drawn venous blood. A blood-style displayed result or a theoretical breath-to-blood conversion does not become a direct whole-blood laboratory assay merely because the paper calls it a blood level.

The authors placed the participants in three ascending device-reading categories: 32 in the low group, 39 in the middle group and 29 in the high group. The source displays inconsistent percent and per-mille symbols when describing these categories and its self-estimated ranges. Without an independently validated unit reconciliation from the original instrument record, it would be unsound to translate the printed group thresholds into forensic whole-blood legal-limit categories. The reliable comparison for this article is the study’s own ordered, breathalyzer-based group membership.

Subjective estimates and instrument readings showed a reported Spearman correlation of 0.48 among all 100 participants (p < 0.001). This moderate association is not a test of the absolute accuracy of self-estimates, and it says nothing about the agreement of that breath device with a separate forensic blood assay. Estimated intoxication categories themselves did not show the same association with urn choices that the measured device groups did.

For additional context about specimens and the distinction between an instrument reading and the tested material, see the site’s published breath alcohol measurement guide and headspace blood-alcohol testing guide. Those guides do not supply missing validation or blood results for this particular experiment.

What did the 100 participants’ urn choices show?

The authors reported more frequent selection of the unknown-odds, two-drink urn at higher device-reading categories when all participants were considered together. Their three-level categorical test was χ²(2, N = 100) = 6.77, p = 0.03. In other words, the distribution of choices varied across the study’s ordered alcohol-reading groups more than would be expected under that particular null model. The statistic does not show how much any person’s preference changed after drinking, because the same participant was not observed once sober and again intoxicated.

When the investigators analyzed men and women separately, the association was detectable in the male subset but not in the female subset. These are subgroup association tests; the paper did not establish a sex-by-alcohol interaction with an explicit direct comparison of the two subgroup effects. A significant test in one subgroup and a nonsignificant test in the other is not, by itself, proof that the subgroup effects differ statistically.

These tests were conducted on a convenience sample in one setting. They did not measure the odds that a randomly selected intoxicated driver would choose an ambiguous action, nor calculate a likelihood ratio suitable for identifying a person’s decision-making state in litigation. A statistical association can be scientifically relevant while remaining too indirect for a case-specific opinion.

Does the apparent male–female difference establish a biological sex effect?

No. The results are consistent with a sample-specific association among the men and no clear association among the women. But that is narrower than proving that alcohol biologically affects men’s and women’s ambiguity preferences differently. The participants were not randomly assigned alcohol doses, cultural roles or drinking histories; the study did not directly test a between-sex difference in the slopes or strengths of those associations.

The investigators discussed a risk-as-value explanation involving gendered social expectations. That is a theoretical interpretation, not a mechanism independently manipulated or measured in this study. No neural or endocrine measurements were collected. Alternative explanations may include drinking patterns, motivations to obtain free drinks, selection into the bar sample, preexisting attitudes toward uncertain rewards and contextual cues. The article provides no basis for assigning any of these mechanisms to an individual participant.

The absence of a significant female-subgroup association (p = 0.77) cannot establish equivalence or that alcohol has no effect in women. Nor can this study justify stereotyping men’s or women’s judgment in a particular criminal, civil or employment matter.

Why can this observation not prove alcohol caused the choice?

The key identification problem is that participants chose how much to drink before study recruitment. Higher measured readings might correlate with how often people drink, their social setting, preference for alcohol, or the type of decision-maker who attends the venue. Those attributes could also influence selection of an urn awarding additional alcoholic drinks. Without alcohol randomization, a sober control condition for the same people, or adequate confounder adjustment, the study cannot isolate an alcohol-caused within-person change.

The experiment measured one decision at one point in time. It did not follow each participant through ascending and descending breath-alcohol curves, determine when an individual’s alcohol peaked, establish dose, or assess consequences of the urn selection. The authors described the study as naturalistic, which is useful for context but does not remove observational bias.

A laboratory dose-controlled crossover could better isolate a pharmacological effect by comparing choices within individuals under alcohol and placebo, using the same incentives. Even that would require attention to expectancy and the task’s ecological validity before the result could be applied to a real person making a different decision.

Can the breathalyzer-based groups be treated as forensic BAC ranges?

They should not be. In a forensic record, the measurement must remain attached to the matrix and method that produced it. This study describes an exhaled-breath device, not a venous whole-blood specimen collected into a documented container, accessioned into an accredited laboratory and measured by a validated chromatographic method. It reports no parallel reference blood specimens, instrument traceability package, replicate agreement, analytical uncertainty budget or participant-specific blood–breath relationship.

The apparent percent/per-mille inconsistency in the publication adds a distinct reporting problem. One cannot repair ambiguous units by silently substituting a statutory conversion factor or a familiar legal cutoff. A forensic toxicologist would first need the original device settings, the author’s actual unit convention, calibration and instrument records, and details of sampling conditions before making a concentration comparison at all.

Even a verified breath result near the time of this experimental choice would not determine ethanol concentration at an earlier or later disputed event without additional facts. The site’s published pharmacokinetics overview explains why absorption, distribution and elimination models require their own assumptions and should not be mistaken for measured concentrations.

What is the appropriate interpretation for a litigation question?

A careful report could state that, in this study, stronger breathalyzer-based alcohol reading categories were associated with greater selection of an unknown-probability two-drink lottery across 100 bar patrons, with the detectable subgroup association confined to the male analysis. That is a bounded description of the authors’ observation. It would be an overstatement to conclude from this result that a named person was unable to assess danger, knowingly accepted a particular risk, intended an outcome, had impaired consent capacity, or would have made a different decision while sober.

A person’s ability to understand risk, form intent, make a voluntary choice or act reasonably depends on many forms of evidence not supplied by this paper: contemporaneous conduct and statements, the actual decision alternatives and incentives, an appropriately interpreted toxicological timeline, relevant medications or medical conditions, and case-specific witness and documentary evidence. A generic two-urn choice cannot replace that evaluation.

The result also should not be treated as an impairment threshold. The participants were not subjected to standardized field sobriety testing, on-road or simulated driving tasks, psychomotor performance measurements, accident reconstruction or validated assessments of decision-making capacity. Nor did the investigators attempt to diagnose an individual with intoxication or estimate a probability of violence or recklessness from the breath result.

How is this study different from an alcohol-and-risk-taking experiment?

In many risk-taking tasks, the subject knows the probabilities or can learn them through repeated trials. Here, one urn offered a known 50% probability and one drink, whereas the other offered two drinks at an unspecified probability. The study therefore addresses tolerance of missing probability information in a reward choice, not a calibrated comparison of driving hazards, reaction-time impairments, loss aversion or delayed gratification.

It is also different from tests where participants estimate their own intoxication. In this investigation self-estimated categories correlated moderately with device readings, but the self-estimates were not associated with the urn preference in the way device categories were. The measure of subjective intoxication and the choice of urn must not be collapsed into a single construct.

Finally, an association between naturally varying breath readings and the chosen urn is not equivalent to a randomized alcohol-versus-placebo estimate. Both questions may be worth studying, but one cannot infer the answer to the causal question from the observational question alone.

What source materials would strengthen a case-specific evaluation?

If this general research is offered to support or challenge an opinion in litigation, examine the complete paper and the actual contested event separately. For the study itself, the key review points include the breathalyzer model and units, eligibility and recruitment, timing after drinking, the meaning of the unknown-odds prize, subgroup denominators, omitted raw choice frequencies, and whether any direct male–female interaction was tested.

For the disputed event, inspect original testing records and the actual specimen matrix; the timing and method of collection; device calibration or laboratory controls; relevant physiological and pharmacokinetic circumstances; and independent evidence of what the person understood and did. Neither the published group χ² results nor the observed ρ = 0.48 substitutes for that evidence.

Frequently asked questions

Did the study prove that alcohol makes people reckless?

No. It observed associations between breathalyzer reading categories and one unknown-odds, two-drink lottery choice. Recklessness is a different construct requiring independent factual and legal analysis.

Was a participant’s whole-blood BAC directly measured?

No. The Methods identify an Alkohit X100 breathalyzer. The source does not report paired chromatographic measurements of drawn venous whole blood.

Did everyone choose between lotteries with the same prize?

No. The known-odds urn awarded one drink with 15 winners among 30 coupons; the ambiguous urn offered two drinks, but its proportion of winning coupons was not specified.

Does a male-subgroup p-value of 0.02 prove a different effect in men?

No. A significant association among men and a nonsignificant association among women do not establish that the two effects differ. A direct statistical interaction test would be needed for that inference.

Does p = 0.77 mean alcohol has no effect on women?

No. It means this sample did not provide statistically detectable evidence of the tested association among the 46 women. The study does not establish zero effect or equivalence.

Can the reported low, medium and high readings be converted into legal BAC ranges?

Not reliably from this article alone. It relied on a breath device and contains inconsistent percent/per-mille notation. Valid forensic unit interpretation requires original device settings and measurement records.

Can this study establish a defendant’s intent or an eyewitness’s credibility?

No. The study did not measure individual legal intent, memory reliability, witness truthfulness, driving impairment or collision causation. Any such opinion requires independently supported case-specific evidence.

Primary scientific source and editorial scope

Primary study: Tyszka T, Macko A, Stańczak M. Alcohol reduces aversion to ambiguity. Frontiers in Psychology. 2015;5:1578. https://doi.org/10.3389/fpsyg.2014.01578. The design, denominators, urn conditions and statistical values above are drawn from the complete original Methods, Results and Discussion, not an abstract-only summary.

This research brief is an educational, AI-assisted interpretation of a published primary study; it is not a signed expert report, a statement that the site’s named professional personally reviewed or adopted the article, or a legal opinion about a specific person. Consistent with the site’s published research-brief approach, case-specific conclusions require review of actual records, competing evidence and the applicable legal standards.