Alcohol increased hostile interpretations of ambiguous facial expressions in one controlled experiment, but it did not significantly increase hostility ratings overall or demonstrate that anyone became violent. The distinction matters when scientific evidence is used to explain a disputed confrontation. Recognizing an emotion, interpreting hostility and acting aggressively are separate outcomes.
Eastwood and colleagues’ 2026 study examined judgments about briefly displayed faces and responses on a joystick task. It offers evidence about a possible perceptual pathway, rather than a test of whether alcohol caused an individual assault.
What the experiment administered and measured
Eighty-four social drinkers, half male, completed alcohol and placebo sessions at least seven days apart in a counterbalanced, double-blind design. Participants were aged 18–40, with a mean age of 22.5 years. The alcohol drink contained vodka and tonic, with a nominal ethanol dose of 0.4 g/kg. Dosing was capped at the amount calculated for 90 kg, so participants above that weight did not receive the full nominal weight-adjusted dose.
Participants drank within ten minutes, waited another ten minutes and then completed the hostility and approach/avoidance tasks in a fixed order. Together the tasks lasted approximately 30 minutes. The protocol intended to capture rising or near-peak exposure, but it did not continuously measure the alcohol curve throughout each task.
Post-session breath alcohol ranged from 8 to 32 micrograms per 100 mL of breath, with a mean of 16.62 and standard deviation of 4.81. These are breath concentrations, not grams per deciliter of whole blood. No paired venous whole-blood, serum or plasma series established an individual blood concentration or elimination rate.
The overall result was null; ambiguity changed the pattern
Participants classified briefly presented composite male faces as hostile or not hostile. Facial intensity ranged from ambiguous to clear emotional expression. Surprise-expression data were excluded because of a programming error, leaving angry, happy, sad, disgusted and fearful expressions in the analysis.
The primary overall alcohol effect was not statistically significant: p = 0.342. Descriptive global hostility ratings averaged 31.61% after alcohol and 30.51% after placebo. An account that simply says alcohol made all faces seem more hostile would misstate that result.
The exploratory drink-by-intensity interaction was significant, p = 0.002. Alcohol was associated with more hostile interpretations at the ambiguous end of the continuum, with the difference diminishing as expression became clearer. The model reported an approximately 3.48-percentage-point increase at its lowest-intensity reference. That estimate is a task-specific model result, not the probability that a witness misread a real person’s intentions.
For happy expressions, descriptive hostility ratings averaged 8.09% after alcohol versus 5.07% after placebo; the modeled drink effect was significant, p = 0.009. There were no significant drink main effects for the other four expressions. An additional exploratory interaction for angry expressions indicated greater hostility judgments at low intensity, rather than a uniform increase across all angry faces.
Faster joystick responses did not establish greater approach toward threat
The second task asked participants to push or pull a joystick according to the image border, rather than explicitly judging its emotional meaning. Three participants were excluded from this analysis for unusually high error rates.
Alcohol shortened response times across emotions by approximately 21–26 milliseconds. But the alcohol-by-approach/avoidance interactions were not significant for any emotion, all p values above 0.260. Faster responding in both directions is different from preferentially approaching a hostile face. State-anger ratings also did not show significant alcohol-related changes.
The task’s negative split-half reliability coefficient, r = −0.32, further limits interpretation of the approach/avoidance null result. It neither proves that motivation was unaffected in every person nor supports an increased tendency to attack.
Why this cannot reconstruct a confrontation
The laboratory presented isolated faces for 150 milliseconds, without the speech, relationship history, movement and setting of a real encounter. Participants were selected social drinkers, rather than a representative forensic population. Most identified the alcohol session, limiting the success of blinding: 83.3% believed they had alcohol in that condition versus 40.5% in the placebo condition. Pharmacological effects and expectations were therefore not cleanly separated.
The study was powered for an overall alcohol effect; interaction and subgroup findings require particular caution. The analysis also changed from the originally planned approach, and multiple emotions and intensities were examined. These qualifications belong alongside the positive findings.
A defensible use of the study identifies a possible change in interpreting ambiguous social cues while retaining the null global result and absence of observed violence. It does not establish what someone saw, whether a threat existed, intent, credibility or causation in a particular case. Our separate analysis of anger recognition and reported empathy addresses different social-cognition measures; those outcomes should not be substituted for hostility attribution.
Primary scientific source
Eastwood APR, Penton-Voak IS, Munafò MR, Attwood AS. Effects of acute alcohol consumption on hostile attribution bias towards emotional facial expressions and implicit approach/avoidance tendencies. Journal of Psychopharmacology. 2026;40(9):1392–1404. doi:10.1177/02698811251408743.
Leave a Reply