Can a Hangover Disrupt Word Recognition and Tracking at Zero Breath Alcohol? A 25-Person Study

Short answer: Yes, in a controlled comparison of laboratory task performance, the same 25 social drinkers made more word-recognition errors and showed time-specific reductions in mouse-tracking accuracy on a morning after drinking, even though none had a breath-alcohol instrument reading above 0.00% on arrival. In the study, delayed recognition accuracy was lower during hangover, and the participants reported greater mental demand and effort. These are group-level experimental findings: they do not establish an individual’s blood alcohol concentration, driving impairment, eyewitness accuracy, or the cause of a particular event.

The original investigation was published by Ayre and colleagues in Psychopharmacology in 2022, DOI 10.1007/s00213-022-06150-4. It asked whether hangover changes the allocation of attention when a person must encode spoken words while keeping a mouse cursor on a moving visual target. The distinction between a breath result, an experimental cognitive outcome, and an individual forensic conclusion is essential.

What does a zero breath-alcohol result establish here?

The investigators used a regularly calibrated Lion Alcolmeter SD400PA when participants reached the laboratory between 8:30 a.m. and 12:30 p.m. on each test day. According to the primary report, no participant registered a breath reading greater than 0.00%. That is an observation about a particular breath instrument at a particular visit—not an independently measured venous whole-blood ethanol concentration by forensic gas chromatography and not a biological measurement of cognitive capacity.

Participants’ peak alcohol concentrations from the previous evening were estimated from self-reported drink counts, drinking duration, and total-body-water/elimination assumptions. The study’s demographic table reports a mean estimated prior-evening BAC of 0.135%. That number was not obtained from an actual prior-evening blood specimen, and it cannot be treated as a validated historical BAC for any specific participant. For more about measurement methods and interpretive boundaries, see breath alcohol measurement.

Who participated, and how were the visits organized?

Thirty-six people initially entered the study. Eight did not complete the required test visits, and three reported drinking on the evening before their supposedly abstinent control visit. That left 25 participants for the paired analysis, of whom 76% were women. Their mean age was 25.32 years. They were healthy social drinkers, recruited in Melbourne, with eligibility limits concerning smoking, medical and psychiatric conditions, medications, and other drug use.

Each eligible participant completed a screening appointment, a hangover morning after an ordinary planned drinking occasion, and a control morning after at least 24 hours without alcohol. The order of the two test conditions was counterbalanced. Test sessions were scheduled 5–14 days apart, and participants were asked to follow similar breakfast and caffeine restrictions. Researchers did not assign a fixed ethanol dose in the laboratory for the previous night: drink history was self-reported. This semi-naturalistic design helps study ordinary next-day experiences but limits control of sleep, beverage selection, activities, and other influences.

Participants reported a mean of 8.15 Australian standard drinks before their hangover session; the article defines an Australian standard drink as 10 g of pure ethanol. The observed range was 3–18.5 drinks. A drink count is a historical exposure report, not a simultaneous analytical specimen result. Pharmacokinetic concepts and estimates are discussed separately in forensic pharmacokinetics.

What exactly was the dual-attention experiment?

The dual-attention task combined visual mouse tracking with auditory word learning. On the screen, a red target moved across a white background. Participants attempted to keep a mouse cursor near it. At the same time, recorded, unrelated two- or three-syllable words were presented through headphones. Separate word lists were learned with and without concurrent tracking, with list order counterbalanced.

The tracking program sampled the separation between target and cursor every 100 milliseconds. Greater separation meant higher tracking cost—poorer alignment during the particular laboratory challenge. This was a task-specific pixel-distance measurement. It was not a gait measure, a field sobriety test, lane deviation on a public highway, or a validated clinical classifier of impairment.

After the encoding phase, participants completed an intervening 20-minute cognitive multitasking battery and questionnaires. They then saw a recognition test containing 40 words: 10 from each of the two original learning conditions plus 20 unfamiliar distractors. For each word, they answered whether it had appeared earlier. The design let researchers assess recognition accuracy, recognition speed, and different types of incorrect responses after a delay.

What happened to mouse-tracking accuracy?

The paper reports a statistically significant hangover-by-time interaction in the cursor-tracking measure, F(29,696) = 1.63, p < .05. Average tracking costs were higher during the hangover visit at selected points in the word-presentation cycle, rather than uniformly at every sampled moment. Significant differences were identified approximately 700–900 milliseconds into word presentation and again 2,200–2,400 milliseconds into the 3-second cycle. The authors confirmed the selected pairwise findings using paired nonparametric tests.

A useful interpretation is that the hangover condition was associated with increased difficulty sustaining cursor alignment at certain periods of simultaneous verbal processing. The researchers did not record a separate target-tracking-only condition. Accordingly, their data cannot determine whether altered tracking reflects a specific attentional-allocation deficit, a more general psychomotor slowing, fatigue, or a combination. The original discussion explicitly identifies this limitation.

Did delayed word recognition change?

Yes, group-level recognition accuracy was lower in the hangover condition. The main effect of hangover was F(1,24) = 12.18, p < .01. The main effect of concurrent tracking load was also significant, F(1,24) = 9.13, p < .01. Participants generally remembered fewer words when the words had been learned while they were also tracking the moving cursor.

Importantly, the study did not find a statistically significant interaction between hangover and tracking load for recognition accuracy: F(1,24) = 0.036, p = .85. It would therefore be inaccurate to claim that hangover specifically magnified the recognition cost of divided attention. Both factors had detectable main effects, but the hypothesized additional synergistic effect was not demonstrated in these data.

The separation of error types matters. For the outcome the authors labeled accuracy of “no” responses, the reported group values were 62% ± 10.79 in hangover and 67% ± 9.67 in control. The investigators interpreted the difference as more false-negative recognition responses during hangover, t(24) = 3.13, p < .01. Because the terminology in the original report is easy to misread, the supported finding is the authors’ specific response-category comparison—not a reconstructed trial-by-trial confusion matrix. The paper’s corresponding false-positive comparison (72% versus 77%, p = .07) was not significant. Do not combine those two error categories into an unsupported blanket claim that every kind of recognition error necessarily increased.

Did people also respond more slowly?

Not in the straightforward way one might expect. Recognition response time had a significant main effect of attentional load, F(1,24) = 7.92, p < .05. But the response-time main effect of hangover was F(1,24) = 3.71, p = .07 and therefore did not meet the conventional .05 significance threshold. The hangover-by-load interaction in response time also was not significant (p = .42).

Forensic interpretation should separate accuracy from speed. A person can give an answer within an ordinary-looking interval while still committing a recognition error. Conversely, a statistically significant group difference on one task does not prove that a given individual’s apparently ordinary answers were wrong. These measures describe the test conditions and sample, not a universally validated witness-reliability instrument.

What did the experiment show about perceived effort?

Participants completed the NASA Task Load Index, rating mental demand, physical demand, temporal demand, their own performance, effort, and frustration. In the hangover condition, ratings were higher for mental demand (p < .01), effort (p < .01), and frustration (p < .001). Participants rated their own task performance as poorer (p < .02). Differences in physical and temporal demand did not satisfy conventional significance thresholds (p = .06 and .08, respectively).

Subjective awareness is useful context but is not direct measurement of visual tracking, memory, or ethanol in any biological matrix. Nor do these ratings show that a person who feels more effortful must be unable to compensate in every real-world situation. The study observed objective group performance differences and reports of elevated effort under its selected laboratory protocol; it did not quantify a universally effective or ineffective compensation strategy.

What do the primary results establish—and not establish?

Why do false-negative word responses matter in forensic interpretation?

A false-negative recognition response occurs when an individual fails to endorse something encountered before. The study’s difference between hangover and control conditions raises a research question about delayed recognition of previously encoded material. But a laboratory test involving random short words is not the same as recognition of a face, recollection of a crash, identification of a suspect, or reporting of an event under emotionally charged conditions.

Witness testimony involves many additional processes: the conditions of original perception, attention to relevant details, the interval before recall, intervening information, suggestion, stress, sleep, and the form of questioning. Ayre and colleagues did not study eyewitness lineups, false memories from suggestion, deception, or individual identification accuracy. It would be scientifically unsound to treat their 62% versus 67% group result as the probability that a particular witness is truthful or mistaken.

Nor can this controlled word-list comparison determine whether an individual was intoxicated at an earlier time. The breath measurements were collected on the morning of the test, not at a prior driving or witnessing event, and the reconstructed evening eBAC was model-dependent. See the site’s forensic toxicology evidence library for the broader distinction between analytical testing and behavioral inference.

Could the memory effect reflect sleep, fatigue, or task design?

Potentially. The study was semi-naturalistic: investigators counterbalanced the order of morning visits but did not experimentally control each participant’s amount of alcohol, overnight sleep, food, social activity, or subjective fatigue. The intermediate 20-minute multitasking battery could itself increase cognitive strain before recognition testing. The paper notes that recognition accuracy was comparatively low even in the non-hangover, single-attention condition, raising the possibility of test-specific floor effects or fatigue.

The tracking task did not independently evaluate tracking alone, and the study involved only 25 complete paired observations, with women comprising 76% of the sample. Some data handling included substitution of scattered missing cursor measurements and winsorization of outliers; the authors report sensitivity analyses using nonparametric checks. Those choices do not nullify the observed associations, but they underscore that the result is a bounded experiment rather than a population-wide forensic prediction rule.

Several hangover severity correlations with subjective mental demand, effort and frustration were statistically significant. In contrast, most severity-to-objective-task correlations were not. This is a further reason not to translate a self-reported hangover score into a person’s precise impairment level, event-specific behavior, or witness reliability.

How should a toxicologist describe this study in a case review?

A scientifically calibrated description is: In a 25-person counterbalanced comparison, a hangover morning was associated with reduced delayed word-recognition accuracy and greater tracking cost at selected periods during an auditory-word/visual-tracking task, despite morning breath-alcohol displays of 0.00%; an interaction between hangover and concurrent tracking load on recognition accuracy was not demonstrated. That summarizes the study’s operational endpoints without substituting an inference it did not test.

Case assessment must then ask what specimen was actually collected, when the testing occurred, whether ethanol was quantified in whole blood or another matrix, what competing causes of poor performance exist, whether a particular person performed the relevant task, and whether any valid individual-level diagnostic or causal inference is available. Neither the absence of measured ethanol in a breath display nor a population-average cognitive result can answer those case-specific questions alone.

Frequently asked questions

Does a breath reading of 0.00% prove someone is cognitively unimpaired?

No. In this study, all 25 participants had morning breath instrument readings no higher than 0.00%, yet selected group differences in laboratory recognition accuracy and tracking were observed. The result is not a calibrated impairment test for any particular person.

Did Ayre and colleagues measure blood ethanol by forensic gas chromatography?

No paired venous whole-blood forensic ethanol analysis is reported. The study used morning breath screening and computed a prior-evening estimated BAC from reported drinking and model assumptions.

How many people were included in the completed crossover analysis?

Twenty-five social drinkers. Thirty-six initially participated, but eight did not complete test visits and three drank before the intended alcohol-free control condition.

Did the study establish that hangover and divided attention interact to worsen word recognition?

No. Both hangover and tracking load had significant main effects on delayed recognition accuracy, but their interaction was not significant (p = .85).

Does a false-negative recognition response mean the participant lied?

No. It is an incorrect response within an experimental word-memory paradigm. The study did not evaluate deception, actual testimony, suggestibility, or truthfulness.

Was the average prior-night BAC of 0.135% measured directly?

No. The value was calculated from self-reported drinking duration and quantity combined with estimated body-water and elimination parameters. It must not be attributed to a blood test.

Can the study identify which driver or witness was impaired the next morning?

No. It provides small-sample group comparisons on particular laboratory tasks, with no validated individual classifier, no actual driving measurements, and no case-specific causal analysis.

Primary scientific reference

Ayre E, Benson S, Garrisson H, Cox KHM, Verster JC, Scholey A. Effects of alcohol hangover on attentional resources during a verbal memory/psychomotor tracking dual attention task. Psychopharmacology. 2022;239:2695–2704. doi:10.1007/s00213-022-06150-4. This article interprets the full original investigation; its task outcomes and analytical limits should not be mistaken for a validated forensic BAC-to-impairment conversion.