Answer: Yes. A displayed zero on a breath alcohol instrument does not by itself establish that next-day response-time performance is free from potential alcohol-associated or sleep-related confounding. In a 2020 test–retest study involving 36 university students, the 18 students who chose a first testing morning after a party were slower on several cognitive tasks than 18 students first tested after abstinence and rest. When both groups returned for a later non-hangover session, the initial response-time pattern diminished. The study did not prove that residual ethanol caused the slowing, determine any participant’s forensic whole-blood alcohol concentration, or validate a diagnostic cutoff for individual impairment.
The question for a lawyer, forensic evaluator, or researcher is consequently narrower than “Was this person intoxicated?” It is whether an unmeasured or inadequately controlled next-day state could change the interpretation of a cognitive-test score. The evidence comes from the complete original research report of Murgia and colleagues, University Students’ Hangover May Affect Cognitive Research (2020).
What did the investigators actually test?
The research examined a potential measurement confound: psychological experiments may enroll students without knowing whether they consumed substantial alcohol the previous evening or slept poorly. Its central scientific outcome was response speed on laboratory tasks, not ethanol metabolism, driving competence, truthfulness, or the legal definition of intoxication. The project combined a 36-person two-visit laboratory study with separate campus interviews about previous-night partying.
The laboratory recruited 36 students, comprising 15 men and 21 women, ages 19–36 (mean 22.3 years; standard deviation 2.6). Eighteen were assigned analytically to a “hangover” group because they voluntarily attended their first session the morning after a party; the other 18 attended after a night of no alcohol and ordinary rest. Approximately one week later, all 36 returned under non-hangover conditions. The investigators did not randomly administer alcohol, standardize the prior drinking dose, or randomize the initial hangover state. The order was fixed: first testing (T0), then second testing (T1).
Why does a zero breath reading not settle the question?
At the end of testing, investigators used a G3 Ferrari Drinkontrol EVO 1YB00200 breath alcohol instrument. The paper says the estimated alcohol concentration was zero in all participants the morning of testing, even though the methods describe a breath-derived criterion below 0.02%. Those statements should be understood in the context of the reported device and its displayed values. There was no independently collected venous whole-blood sample analyzed by forensic headspace gas chromatography and no linked serum, plasma, oral-fluid, or urine ethanol measurement. Nor was there a validated clinical assessment establishing an individual’s alcohol clearance time.
A breath device measures ethanol in exhaled air and can report a device-specific result or blood-equivalent estimate. Such a reading is not the same physical observation as a GC-measured whole-blood ethanol concentration. A recorded zero at one breath-test time does not reveal an earlier peak, a previous-evening dose, or how a particular person performed while alcohol was present. Equally, slower responses at a later time do not demonstrate residual ethanol in blood. The paper’s result is a next-day task-performance association in participants with reported hangover, not a chemical finding of remaining intoxication.
How were the cognitive tasks performed?
The investigators used a computer and a five-button response box to collect reaction times and accuracy. Participants were told to respond as rapidly and accurately as possible, but the stimuli remained visible until they answered. Consequently, a participant could be slower without necessarily making more errors. Three tasks followed the same sequence for every participant and on both visits.
Task 1: odd-versus-even parity judgment
Forty experimental trials displayed single digits. The participant pressed one key for even numbers and another for odd numbers. Correct classification required seeing the stimulus, retaining a response rule, and initiating a key press. Even this seemingly simple task incorporates perceptual, decisional and motor components, so a slower response cannot be uniquely attributed to one psychological function.
Task 2: parity judgment with a color-dependent rule switch
The second task had 120 trials. On white-digit trials, the participant used the normal parity mapping. On black-digit trials—one-third of trials—the mapping reversed. The investigators analyzed the task overall and separately for ordinary-rule and reversed-rule trials. This design added rule retention and response selection demands. It did not establish a validated roadside executive-function or impairment test.
Task 3: arithmetic verification
The third task presented 48 calculations across three blocks and required judging whether the displayed answer was correct. It engaged mental calculation and short-term retention of intermediate values. Again, there was no response deadline and no actual field performance outcome. For reaction-time analysis, the investigators discarded incorrect trials and used each participant’s median correct-response time for each condition; they studied accuracy separately.
What were the primary response-time results?
The authors reported statistically significant group-by-visit interactions for reaction time on all three tasks. An interaction tests whether the change from the first visit to the second differs between the students classified as initially hungover and those classified as initially non-hungover. Such an interaction is informative about possible state-dependent measurement variation; it does not, by itself, establish which component of that state caused the change.
| Task and denominator | Original reported interaction | What it can support | What it does not establish |
|---|---|---|---|
| Parity, 36 students | Group × Time: F(1,34) = 4.54; p < .05; partial η² = .12 | Test speed differed across visits according to initial group. | An individual’s diagnosis or a measured residual blood ethanol level. |
| Parity with rule switching, 36 students | Group × Time: F(1,34) = 7.07; p < .05; partial η² = .17 | Speed pattern also appeared in a more complex rule task. | A clinical classification of impaired executive function, or an observed driving effect. |
| Arithmetic verification, 36 students | Group × Time: F(1,34) = 5.42; p < .05; partial η² = .14 | Retest change differed even on a calculation task. | Proof that a particular medication, alcohol metabolite, or mechanism caused the slowing. |
The authors’ descriptive comparison shows the hangover group’s slowest performance at T0, followed by faster scores when that group was retested at T1 without reported hangover. By contrast, the overall group main effect in arithmetic verification was not significant. The published prose gives some individual contrast p-values that do not align cleanly with conventional two-sided calculations from its own t-statistics; the interaction results are therefore the more cautious summary of the paper’s comparative pattern. The article does not provide a reproducible individual cutoff or performance-based likelihood ratio for forensic identification.
Did accuracy decline alongside speed?
No statistically significant between-group change in accuracy was reported. At the first visit, error percentages across the three tasks ranged from 2.64% to 6.78% in the hangover group and from 3.34% to 6.71% in the other group. The ranges concern different tasks, not an individual person or a pooled clinical error rate. A participant’s accurate answer therefore did not guarantee an unchanged response time.
That distinction matters in evidence assessment. A witness, student, or examinee can perform a task correctly but slowly; conversely, a fast answer does not guarantee accuracy. The procedures permitted additional response time rather than requiring a fixed deadline. Therefore this study cannot say whether the same participants would make more mistakes on an urgent-driving, flight, weapon-handling, medical, or eyewitness task with different time constraints. It is inappropriate to convert this laboratory speed association into crash probability or a specific amount of roadside impairment.
How much drinking and sleep loss did the hangover group report?
Participants attending the first session after a party recalled consuming a mean of 7.42 alcohol units (SD 2.83) the previous night. The authors treated a unit as approximately 10 grams of ethanol and estimated a mean dose of 1.22 g/kg (SD 0.46) using body weight. These are retrospective self-reports and calculated dose estimates, not administered research doses or measured ethanol concentrations. The mean Alcohol Hangover Severity Scale score was 4.11 (SD 1.62).
The self-reported sleep difference is substantial and cannot be ignored: the hangover group reported 5.10 hours (SD 2.57) before T0, compared with 7.38 hours (SD 1.37) before T1. The non-hangover group’s respective reports were 6.97 and 7.08 hours. The study did not isolate sleep deprivation experimentally from the physiological consequences of drinking, differences in mood, hydration, diet, expectancy, or practice effects. Thus it is sound to identify potential alcohol-associated and sleep-associated confounding, but not to allocate the observed response-time difference quantitatively to any one cause.
What did the separate campus survey actually show?
To assess whether initial drinking histories might vary by testing day, researchers also approached students present on campus between 10 a.m. and 1 p.m. on a Thursday and a Tuesday of the same week. On each day, they asked 500 students whether they had partied and drunk more alcohol than usual the preceding night. The survey was separate from the 36-person experiment and did not administer the cognitive tasks or independently test blood alcohol in these 1,000 respondents.
| Campus interview day | Answered yes to partying and drinking more than usual | Denominator | Measured construct |
|---|---|---|---|
| Thursday morning | 66 students (13.2%) | 500 students approached that day | Self-reported previous-night behavior, not diagnosed hangover or current chemical intoxication. |
| Tuesday morning | 19 students (3.8%) | 500 students approached that day | The same self-report question on a different weekday. |
The authors reported a between-day χ² = 28.40, p < .001, and a similar directional result the following week (χ² = 16.95, p < .001). They did not supply the second week’s respondent breakdown in the reported passage, so it should not be back-filled. These counts are not a general-population rate of clinically established hangover, a prevalence estimate for every university, or an evidentiary probability applicable to a named litigant. They show how the pool of volunteers accessible on a given weekday could vary in a specific setting.
What is the forensic significance of an uncontrolled pretest state?
A cognitive score is an observation made under a particular set of conditions. If the purpose of an assessment is to compare sessions, study a treatment, or evaluate a person’s claimed functional change, then the conditions preceding each test can be relevant to the strength of the inference. The 2020 experiment gives a concrete example: an observable speed difference arose in a group with different preceding-night drinking and sleep histories even though the breath instrument showed zero during the testing visit.
This does not justify dismissing a poor score as “just a hangover” or assuming an examinee consumed alcohol. Nor can an expert determine that alcohol, rather than inadequate sleep, caused an individual score without person-specific evidence. It instead supports a targeted records-and-methods inquiry: Was the person asked about recent drinking? Was sleep documented? Did the testing protocol record response latency and accuracy separately? Was there a second session, matched practice exposure, and a prespecified interpretation rule? Was a breath reading incorrectly represented as a measured venous whole-blood concentration?
For a scientific expert, the relevant concept is construct validity: did the task actually measure the construct claimed, under conditions sufficiently controlled to justify the interpretation? For trial counsel, the corresponding question is whether the proffered inference exceeds the population, specimen, instrument, and task that generated the evidence. These are issues of study design and evidentiary weight, not a new legal exclusion rule invented by this research.
Can an apparently normal accuracy score exclude a reaction-time effect?
Not from this study. The investigators found interaction patterns for correct-response speed without a corresponding statistically significant difference in error rate. That does not imply speed effects are universally more sensitive than accuracy measures. The test design allowed participants to take as long as needed, which can favor accuracy at the expense of speed. A time-limited operational task might produce a different combination of errors and delays, and the article did not test that counterfactual.
The careful inference is also asymmetric: knowing that some students were slower while still accurate does not allow the inverse conclusion that every slow individual has a hangover. Slow performance has many potential causes; the original study did not provide diagnostic sensitivity, specificity, a false-positive rate, or a classifier trained on case-level records. It would be scientifically unsound to use its group averages to attribute an individual test result to alcohol, sleep loss, dishonesty, or neurological disorder.
What are the study’s strongest and weakest design features?
The within-person retest is useful because it compared each initially hungover participant with their own performance at a later visit. The non-hangover comparison group gives a simultaneous reference for general repetition and visit effects. Both features strengthen the description relative to a simple one-time anecdotal comparison.
However, the design remains observational as to alcohol exposure. Students could choose when to appear; participants were not blinded to their symptoms; the initial visit was always before the retest; the tasks repeated in the same fixed sequence; and the groups were small. Sleep differed markedly between the hangover group’s visits. No pre-drinking baseline, controlled ethanol dosing, blood pharmacokinetic curve, objective sleep monitoring, or case-based performance threshold was collected. Multiple outcomes and contrasts were assessed. The paper’s plotted reaction times do not provide a numerical case-specific threshold merely because the group interaction was significant.
External validity is similarly limited. The volunteers were university students present for campus activities, not a representative sample of criminal defendants, injured plaintiffs, first responders, drivers, witnesses, or older clinical populations. The separate campus questionnaire was a self-report sample taken at specific times in one university context. Extrapolating its proportions to another locale or its cognitive findings to an operational safety task would require independent validation.
How should specimen and effect language be kept separate?
Five ideas should remain distinct. Exposure refers to the preceding-night self-reported drinking history and calculated grams per kilogram. Specimen measurement refers only to exhaled-breath alcohol testing at the study visit; no forensic whole-blood, serum, urine, oral-fluid, or biomarker such as EtG or PEth was assayed here. Subjective state refers to symptoms captured by the hangover questionnaire. Observed performance refers to actual timed responses and accuracy in three computerized laboratory tests. Individual impairment and causation would require additional validation and case-specific data; neither was established.
Those distinctions prevent a common reasoning error. A person may report feeling hungover with no positive displayed breath reading; that is not an analytical contradiction. But one must also avoid asserting that a negative or zero displayed breath reading proves absence of all prior alcohol-related effects, or that observed slowing identifies a specific causal pathway. An evaluator considering measured blood ethanol or historical drinking estimates should separately inspect the actual assay, collection time, analytical uncertainty, and scientific support for any time-based extrapolation. A related methodological overview discusses why two blood-sample measurements need careful interpretation; that page is background, not evidence that this study assayed blood.
What would a defensible case-specific analysis require?
One should first state the asserted proposition precisely: is the dispute about current ethanol exposure, prior-night consumption, score reproducibility, observed performance, diagnosis, or causal responsibility? The study cannot be stretched to answer all these questions. A case-specific analysis would need contemporaneous records of the actual testing instrument and specimen, time since reported consumption, sleep, medication and illness context where documented, prior and subsequent test administrations, task timing, scoring instructions, and any independently validated interpretive cutoff. Competing explanations and uncertainties should be considered before arriving at an opinion.
For research reading rather than case advice, the Attorney Scientists research library organizes source-first discussions, while the site’s editorial and source policy explains attribution and review limits. This article is an AI-assisted synthesis of a named primary publication; it is not represented as an original experiment, an adopted personal opinion by a named attorney, an individual diagnosis, or a substitute for qualified examination of the full case record.
Frequently asked questions
Does a 0.00 breath result mean there cannot be a next-day cognitive effect?
No. A breath reading characterizes the device’s measurement at that time; this small study documented slower laboratory response times in students reporting hangover despite zero displayed breath ethanol. The finding does not identify residual blood ethanol or a universal next-day effect.
Did the researchers perform an evidential whole-blood BAC test?
No. Their measurement was an exhaled-breath instrument; the paper did not report an independently assayed forensic venous whole-blood concentration, blood biomarker, or individual ethanol elimination profile.
How many participants underwent both cognitive testing sessions?
Thirty-six, in two groups of 18: one group had initial hangover symptoms and a later non-hangover retest; the other was non-hungover at both sessions. The separate weekday interviews involved 500 respondents per day, not another 1,000 performance-test subjects.
Did accuracy decline on the three tasks?
The paper did not find statistically significant accuracy differences. Its principal comparative signal involved response time on correct trials, in tasks without fixed response deadlines.
Were alcohol dose and sleep deprivation experimentally separated?
No. Previous-night consumption was recalled rather than administered by protocol, and sleep was shorter before the hangover visit. The design cannot assign a precise fraction of slower responses to ethanol effects as distinct from sleep loss or other accompanying changes.
Can this article establish a particular driver’s or witness’s impairment?
No. It tested students performing parity and arithmetic tasks, not on-road driving, witness identification, crash causation, intoxication thresholds, or an individual diagnostic classifier. Its group results cannot decide a named person’s condition.
Does a higher Thursday survey percentage prove campus-wide hangover prevalence?
No. The 13.2% Thursday and 3.8% Tuesday figures reflected a self-report question about partying and drinking more than usual the preceding night in two samples of 500 campus attendees. They are neither objective current-alcohol measurements nor verified clinical hangover diagnoses for a general population.
Primary scientific source
Murgia M, Mingolo S, Prpic V, Sors F, Santoro I, Bilotta E, Agostini T. University Students’ Hangover May Affect Cognitive Research. Frontiers in Psychology. Published September 29, 2020; 11:573291. doi:10.3389/fpsyg.2020.573291. Interpret all results within the actual two-visit observational design and the limitations identified above.