Does Feeling Less Drunk Mean Better Vigilance After Alcohol? A 93-Adult Genotype Study

Answer: No. Feeling less intoxicated does not establish that a person is responding faster or making fewer attention lapses. In a 2026 controlled alcohol-administration study of 93 young adults, higher habitual-drinking scores were associated with lower self-rated intoxication, but not with better performance on a five-minute psychomotor vigilance task. After participants consumed 0.5 g of ethanol per kilogram of body weight, group-level vigilance measures worsened at 40, 70 and 100 minutes compared with the pre-drinking baseline. Breath alcohol concentration, subjective sensation and computerized response performance were measured separately; none of these variables is an interchangeable measure of a particular driver’s blood alcohol concentration or fitness to drive.

The original primary study is Motoi and colleagues, “Discrepancy between self-reported intoxication and psychomotor vigilance task performance after alcohol intake: associations with BrAC, alcohol-metabolizing genotypes, and drinking tendency,” Journal of Physiological Anthropology (published June 10, 2026), DOI: 10.1186/s40101-026-00434-w. The results support a distinction between reported sensation and one laboratory measure of alertness. They do not validate a clinical, roadside or courtroom test for deciding whether a specific individual was impaired, truthful or the cause of an accident.

Why can someone feel less drunk without performing better?

Self-assessment and performance answer different questions. A person saying “I do not feel very drunk” is reporting a subjective experience, which can reflect expectations, previous drinking, discomfort, social context and other influences. A psychomotor vigilance task, by contrast, records responses to visual cues and quantifies missed or slow responses under a particular laboratory protocol. Neither observation directly measures a blood ethanol concentration.

The Motoi study tested a particular explanation for this mismatch: differences in alcohol-consumption history and alcohol-metabolism genotypes might be associated with breath ethanol, intoxication ratings and vigilance in different ways. In its statistical model, greater AUDIT-C consumption scores were associated with lower intoxication ratings at 40 and 100 minutes, while the corresponding associations with vigilance lapse rate were not statistically significant. That is not proof that frequent drinking produces protective tolerance, nor proof that a person who feels sober is impaired. It shows that perceived intoxication cannot be substituted for measured performance.

Who participated, and how much alcohol was administered?

The researchers recruited healthy university-age adults and screened 104 people. After eligibility screening, 98 entered the experiment; five with the relatively uncommon ADH1B*1/*1 genotype were excluded from the primary comparison because their subgroup was too small for stable genotype-based inference. The main analysis therefore involved 93 individuals: 55 men and 38 women, mean age 22.3 ± 1.6 years. The experiment did not comprise 93 independent drinking episodes for every measurement or 104 fully analyzed participants.

Each analyzed participant was given an ethanol dose of 0.5 g/kg body weight using diluted vodka. The drink was divided into three servings approximately ten minutes apart, over a roughly 30-minute dosing period. The protocol included preparation, standardized pre-test conditions and repeat assessments, but it was not a randomized alcohol-versus-placebo driving trial. The investigators selected comparatively healthy, alcohol-tolerant volunteers and excluded several categories of illness, medication use and sleep disruption. These restrictions matter when generalizing to older drivers, patients, persons with sleep loss or people with different drinking histories.

The overall sample did not include individuals with ALDH2*2/*2: the investigators excluded participants with a positive alcohol patch test during screening. Saliva samples were used for ADH1B and ALDH2 genotyping; that assay identified inherited variants rather than measuring ethanol or reaction speed.

What was actually measured—breath ethanol, feelings or response behavior?

Three measurements had different evidentiary meanings. First, a ST-3000 breath alcohol detector assessed ethanol in exhaled breath; the mean of three consecutive readings was analyzed at each scheduled breath time. The reported matrix was breath alcohol concentration (BrAC), in mg/L. The experiment did not report paired venous whole-blood ethanol concentrations determined by forensic headspace gas chromatography. A hypothetical blood-equivalent conversion is not the same as an actual whole-blood specimen and assay.

Second, participants rated subjective intoxication on a visual analog scale from 0 (“not at all intoxicated”) to 100 (“very intoxicated”). A lower number means that the participant described feeling less intoxicated, not that the participant had passed a functional or legal safety threshold.

Third, participants completed a five-minute computerized psychomotor vigilance task (PVT) at baseline, 40, 70 and 100 minutes. They responded to visual stimuli appearing after unpredictable intervals. The study’s defined lapse rate was the proportion of valid responses requiring more than 355 milliseconds. A separate reciprocal-reaction-time score summarized reaction speed, with slower responses corresponding to lower reciprocal values. These are directly recorded task outcomes; they are not observed steering, actual road hazards, crash reconstructions or individualized legal impairment diagnoses.

For additional context on the differing chemical specimens, see breath alcohol measurement and headspace gas-chromatographic blood ethanol testing. They address analytical concepts distinct from a self-rating or visual-response test.

What happened to vigilance after drinking?

The primary vigilance analysis included 92 people, because one participant had missing performance data. Across the post-drinking assessment times, there was a significant overall time effect on both the proportion of attention lapses and reciprocal reaction time. The lapse-rate time effect was F(2.68, 235.78) = 9.26, p < .001; for reciprocal reaction time it was F(2.58, 227.00) = 15.36, p < .001. The degrees of freedom reflect the study’s repeated-measures modeling, rather than a new number of independent participants.

Compared with each participant’s pre-drinking reference, lapse rates were significantly higher at 40 minutes (p < .001), 70 minutes (p < .001) and 100 minutes (p = .010), after the study’s stated multiple-comparison procedure. Reciprocal reaction time was also lower at all three time points (p < .001, p < .001 and p = .020, respectively), indicating slower responses by that metric.

These results describe change in this defined five-minute task; they do not show that every participant was uniformly slower on every individual response. Nor do they establish whether residual fatigue, repeated testing or other unmeasured influences contributed to the 100-minute result. No actual driving outcome was tested.

Evidence table: which findings came from which measurement?

Did ALDH2 or ADH1B genotype predict who performed better?

The study compared variation in two alcohol-metabolizing genes, but the results were not a simple gene-to-impairment classification. Participants with the ALDH2*1/*1 genotype had lower group BrAC values than ALDH2*1/*2 carriers at selected times: 40, 50, 60 and 90 minutes, with adjusted pairwise p values of .040, .012, .030 and .037. Comparisons at 70, 80 and 100 minutes were not statistically significant. Subjective intoxication likewise differed by ALDH2 group at 40, 50 and 60 minutes but not later time points.

By contrast, the main vigilance analysis did not identify significant main effects or genotype-by-time interactions involving either ADH1B or ALDH2. The correct interpretation is not that genotype has no biological importance, or that all genotypes perform identically. It is that the specified study, sample, dose, outcomes and analysis did not demonstrate genotype-related differences in the PVT measures. Detecting a breath or subjective-rating difference does not establish a comparable behavioral difference.

This was a narrow genotype sample with exclusions, a modest single alcohol dose, a relatively brief observation window and a single principal vigilance task. Those choices limit the ability to infer different risks for unrelated genetic backgrounds or for complex roadside performance.

What did the AUDIT-C drinking-history score actually show?

AUDIT-C is a three-question questionnaire about usual alcohol consumption; its score reflects reported drinking tendency, not ethanol at a particular time. The researchers included this score in two observed-variable structural equation models, separately fitted at 40 minutes (near the measured breath peak) and 100 minutes (later in the observed period). A model path is a conditional statistical association within the specified model; it is not independently observed causation or a calibrated individual prediction equation.

At 40 minutes, the standardized path from AUDIT-C to the participant’s subjective intoxication was β = −0.24, p = .013. The path from AUDIT-C to the vigilance lapse outcome was β = −0.14, p = .11, not statistically significant. At 100 minutes, AUDIT-C again had a negative association with reported intoxication (β = −0.28, p = .002) but no significant association with lapse rate (β = −0.08, p = .53).

This distinction is the main forensic lesson. A higher score indicating a greater tendency to drink was accompanied by a tendency to feel less intoxicated in the fitted models, without evidence from these paths of proportionately better vigilance. It does not show that a specific person with a high AUDIT-C score actually performed poorly, nor does it prove deliberate deception, chronic tolerance, or safe performance in any incident.

Did breath alcohol predict both feelings and attention lapses?

At 40 minutes, higher measured BrAC was associated with higher self-rated intoxication (standardized β = .28, p = .002) and more vigilance lapses (β = .27, p = .001) in the same model. At 100 minutes, the BrAC-to-subjective-intoxication path remained significant (β = .33, p < .001), while its path to lapse rate was not (β = .08, p = .52).

Do not interpret the 100-minute null path as proof that alcohol no longer affected vigilance. The separate repeated-measures comparison still found worse task scores at 100 minutes than at the pre-drinking baseline. These tests ask different statistical questions: within-person change over time versus cross-person associations in a model at a particular time. A significant result in one analysis and a nonsignificant result in another are not automatically contradictory.

Similarly, a regression coefficient such as β = .27 is not a 27% chance of impairment, a 27% increase in measured BAC or an error rate for an individual legal determination. It is a standardized association conditional on the model’s other measured variables.

Evidence table: why the 40-minute and 100-minute models differ

These numbers come from the original observed-variable structural equation models; they do not compare independent experiments or assign a probability of truth, competence or driving impairment. The study’s model used sex, genotype, AUDIT-C, BrAC, subjective ratings and lapse-rate variables. Nonrandom drinking history and limited unmeasured confounder control prevent interpreting an arrow as a proven mechanism.

Can the reported breath decline rate reconstruct a driver’s prior blood alcohol?

No individual reconstruction was validated. The authors estimated a mean BrAC decline slope of −0.00152 mg/L per minute (95% confidence interval −0.00173 to −0.00135 mg/L per minute) using a mixed-effects model over their observations. This is a model-based average slope for exhaled breath in a selected controlled-dose group, not a person-specific measured hepatic elimination constant. Converting it to a statutory blood ethanol decline would require separate evidence on specimen relationships, sampling conditions, absorption/elimination phase and the case at issue.

Even during a declining group-average breath period, individuals can have different concentration-time trajectories. The study did not collect an independent series of venous whole-blood GC samples for deriving blood-to-breath factors or validating defendant-specific extrapolation. See the site’s discussion of pharmacokinetics and toxicology modeling for why the measurement and model distinction matters.

Can a self-reported low intoxication rating prove sobriety or truthfulness?

It cannot. A 0–100 subjective scale records how intoxicated a participant says they feel, not their true drinking chronology or a legally verified capacity. A person could accurately report feeling little effect while exhibiting measurable slowing on one task, or could have an elevated breath reading without a measured deficit on a different task. The paper did not include independently verified statements about recent drinking in a criminal investigation, adverse-event findings, eyewitness-identification tasks or a test of deception.

The appropriate scientific inference is narrower: **subjective experience and laboratory performance have distinct sources of variability and cannot be treated as substitutable diagnostic instruments**. Expert testimony must not turn this group association into an opinion about a particular witness’s credibility, driver culpability or physiological cause of a collision without additional case-specific evidence.

What are the important limitations before using this study in litigation?

The study was confined largely to healthy, young adults of East Asian ancestry with selected drinking ability and genotype profiles. It did not randomly assign long-term drinking tendency, compare every relevant genotype, or include an alcohol-free parallel control condition. It involved one fixed weight-adjusted oral ethanol dose, a 100-minute observation horizon and a single primary five-minute vigilance task. Some outcomes had missing values, so the BrAC, subjective and PVT denominators differed.

Investigators could not fully control sleepiness, fatigue, repeated-testing practice effects or other influences on PVT results. Structural equation model paths depend on assumed model structure and measured covariates. Neither good fit indices nor small p values validate an individual impairment classifier or confirm biological mediation. The research did not observe actual driving, retrograde blood ethanol results, field sobriety testing or crash causation.

The study remains useful when an expert must explain why “I felt fine” is not equivalent to “my attention was objectively normal.” It is not a shortcut around specimen integrity, analytical specificity, the particular event timeline or the need to evaluate alternative explanations.

Frequently asked questions about feeling sober and measured vigilance

1. Can someone feel sober while alcohol still affects reaction time?

Yes, self-rated feelings and reaction-time outcomes may diverge. In this study, measured vigilance worsened after drinking while usual-consumption scores related to subjective ratings but not significantly to the PVT lapse outcome. This does not determine any particular individual’s performance.

2. Is breath alcohol the same as a measured blood alcohol concentration?

No. The investigators directly sampled exhaled breath using an ST-3000 instrument and reported mg/L BrAC; they did not independently measure paired venous whole-blood ethanol by forensic gas chromatography. Assay, matrix and conversion must be distinguished.

3. Did people with higher AUDIT-C scores perform better on the vigilance task?

The fitted AUDIT-C-to-PVT lapse associations were not statistically significant at either 40 or 100 minutes. There is no basis in these results for saying that greater habitual consumption protected task performance.

4. Did ALDH2 genotype determine whether a participant was impaired?

No. Selected group breath values and intoxication ratings differed with ALDH2, but genotype effects on PVT measures were not significant. No diagnostic genotype cutoff for impairment was established.

5. What counted as an attention lapse in this experiment?

The primary measure was a response taking more than 355 milliseconds during a five-minute computerized visual-stimulus task. This is a study-defined laboratory threshold, not an independent crash or legal-impairment criterion.

6. Does a nonsignificant breath-to-lapse path at 100 minutes mean full recovery?

No. The 100-minute model tested differences among participants, while the separate time analysis still found significantly worse vigilance than the pre-drinking baseline. The study followed participants for only 100 minutes and did not prove universal recovery.

7. Can this paper establish what a driver or witness could do during a disputed event?

No. It offers group-level biological and behavioral context, not a validated individual forensic classification. A case-specific opinion requires actual specimen evidence, chronology, analytical uncertainty, behavioral facts and competing explanations.

Primary scientific source and take-home conclusion

Motoi M, Shin N, Hama M, and colleagues. Discrepancy between self-reported intoxication and psychomotor vigilance task performance after alcohol intake: associations with BrAC, alcohol-metabolizing genotypes, and drinking tendency. Journal of Physiological Anthropology. 2026;45:19. https://doi.org/10.1186/s40101-026-00434-w.

Bottom line: Following a controlled ethanol dose, breath ethanol, how drunk a participant felt and how quickly they responded on a vigilance task did not behave as one interchangeable signal. Greater usual-drinking scores were associated with feeling less intoxicated, not with demonstrated protection against lapses. The paper provides useful caution against equating subjective confidence with functional safety, while imposing clear limits on translating laboratory group results into a particular person’s BAC, impairment or legal responsibility.