Answer: An experimental brain scan can demonstrate a group-level change in blood-oxygen-level-dependent (BOLD) responses after alcohol administration, but it cannot establish that a particular person was fearless, misread a threat, formed a criminal intent, or acted because of alcohol. In a randomized, single-blind study of 49 healthy social drinkers, Hur and colleagues found reduced fMRI reactivity to emotional faces in a small region of the left bed nucleus of the stria terminalis (BST) after alcohol compared with placebo. A predefined regional analysis also supported dampening across the BST and central nucleus of the amygdala (Ce). Those findings concern an indirect physiological signal during a laboratory image-classification task—not an individual diagnostic test of fear or legally relevant conduct.
The primary evidence is Hur et al., “Acute alcohol administration dampens central extended amygdala reactivity” (2018), Scientific Reports, DOI: 10.1038/s41598-018-34987-3. The study is informative for understanding alcohol-sensitive neural systems, provided its measurement and inference boundaries remain intact.
What did the researchers actually test?
The researchers compared a group assigned to consume alcohol with a group assigned to receive a beverage designed to seem alcoholic. They then acquired functional magnetic resonance imaging (fMRI) while participants distinguished pictures of emotional faces from pictures of buildings. The question was whether alcohol changed the difference in brain activity elicited by the image categories, with particular attention to the central extended amygdala (EAc).
The EAc comprises interconnected regions that include the BST and Ce. These structures are implicated in the processing of emotionally salient and potentially threatening cues. Their involvement in a laboratory fMRI contrast, however, is not interchangeable with a measurement of fearfulness, courage, judgment, aggression, or real-world threat recognition. The experiment was neither a forensic reconstruction nor a behavioral test of actions during a disputed incident.
Study design: 61 recruited, 49 analyzed
Investigators recruited 61 community participants aged 21–35. The final analysis comprised 49 participants, 46.9% women. Twelve were excluded for unusable anatomical images (three), a scanner problem (one), incidental neurological findings (two), inadequate task performance (three), or excessive head-motion artifact (three). Thus, the results reflect an analyzed subset rather than all initially recruited individuals.
Participants were healthy social drinkers who reported prior familiarity with the study’s upper alcohol dose. The exclusion criteria screened out several neurological, psychiatric, and alcohol-related histories. Assignment to alcohol or placebo was randomized with stratification by sex and race/ethnicity; the design was between subjects, not a crossover in which each participant served as their own alcohol-free comparison. It was single blind, leaving room for experimenter expectancy effects despite an elaborate placebo procedure.
Participants abstained from alcohol and other substances for 24 hours and from eating or drinking for three hours before the visit. Investigators checked initial sobriety using an Alcosensor IV breath instrument. Each participant consumed three beverage portions over 30 minutes before scanning. The alcohol beverage combined juice with 100-proof vodka; the nominal placebo replaced the vodka with water but also used a small amount of vodka and bitters on the surface and a trace of aerosolized vodka on the rim to support blinding. It should not be described as a chemically guaranteed zero-ethanol beverage.
Was a blood alcohol concentration measured?
No separate venous whole-blood ethanol assay is reported for these imaging comparisons. The investigators described a targeted “blood alcohol level” (BAL) around 0.09%, using a dosing formula based on body characteristics, with an intended range of approximately 0.06%–0.12%. They assessed alcohol levels before and after scanning in a study that identified an Alcosensor IV breath instrument as the assay device. Consequently, a reported study BAL should be identified as breath-assay-derived or breath-referenced, not silently recast as a forensic headspace-GC measurement of a venous whole-blood specimen.
There are at least four separate data types: ethanol administered in a drink, breath-instrument alcohol information, MRI BOLD contrast values, and picture-discrimination accuracy. None is a substitute for the others. The protocol did not establish a forensic blood-to-breath partition ratio for each participant, obtain evidence of an individual driver’s actual whole-blood ethanol at an incident, or validate an alcohol concentration cutoff for an individual’s emotional response.
The measured or estimated ethanol exposure describes conditions around the experiment. It does not transform an fMRI signal into an ethanol assay. For the broader distinction between measurement and inference, the same-site two-blood-sample alcohol-elimination research brief explains why a calculated value and a specimen result must remain separate.
What did participants see and do inside the scanner?
The MRI task alternated eight blocks of face images with nine blocks of building images. During face blocks, participants identified fearful versus neutral faces: 75% were fearful and 25% neutral. During building blocks, they identified houses versus skyscrapers. Each block lasted about 16.3 seconds and contained 16 brief images. Images were standardized, static, grayscale photographs; they were not actual people posing an immediate threat to the participant.
The response was a simple forced-choice button press on an MRI-compatible pad. The task did not ask how afraid a participant felt, whether they would flee, whether they believed a stranger intended harm, or whether they would take a risk. It had no aggressive-behavior trial, automobile, eyewitness-identification event, confrontation, or validated legal-capacity endpoint.
The scanner collected BOLD signals with a 3-tesla system and a multiband sequence. BOLD measures changes associated with local blood oxygenation and hemodynamics during neural activity; it is not direct electrical recording of neurons, not a direct assay of brain ethanol, and not a lie detector. The researchers modeled the difference between emotional-face and place blocks and compared alcohol and placebo groups, including analyses restricted to anatomically specified BST/Ce regions.
What were the actual results?
Participants generally performed the picture classifications accurately, but the alcohol group was less accurate. Mean discrimination accuracy was 83.2% (SD 8.2) in the alcohol group versus 91.1% (SD 4.9) in the placebo group, a descriptive difference of 7.9 percentage points (FW-S(1,47)=15.98; p<0.001). Accuracy was also higher for building pictures than face pictures across groups, while the group-by-image-condition interaction was not significant (F(1,47)=0.24; p=0.63). Task accuracy is an observed laboratory behavior; it does not specify a person’s driving ability or their reaction to a real threat.
For the voxelwise fMRI comparison, the alcohol group showed a reduction in left BST reactivity to faces relative to places: t=5.46, cluster volume 104 mm³, p<0.05 after familywise-error correction within the anatomically defined EAc region. At that corrected voxelwise threshold, the investigators did not find a significant amygdala cluster for the alcohol-versus-placebo interaction. Exploratory whole-brain testing found no additional significant clusters.
In a separate analysis extracting signals from independently defined BST and Ce regions, the group-level alcohol-dampening effect across regions was at the conventional significance boundary (FW-S(1,47)=3.93; p=0.05). The treatment-by-region interaction was not significant (p=0.88). This supports no detected difference in alcohol sensitivity between these two predefined regions; it does not prove identical biological effects in every brain or person.
Evidence table: observations versus defensible conclusions
| Study measure | Reported observation | Proper inferential limit |
|---|---|---|
| Final analyzable participants | 49 of 61 recruited; 12 excluded for imaging, performance, or related reasons | Selected healthy social-drinker sample; not a representative forensic population |
| Face/building discrimination | Mean 83.2% accuracy alcohol vs 91.1% placebo; p<0.001 | Difference on a simple laboratory button-press task, not a personal impairment threshold |
| Corrected voxelwise EAc analysis | Left BST reduction; t=5.46; 104 mm³; corrected p<0.05 | Localized group-level BOLD contrast, not a fear measurement or whole-brain shutdown |
| Predefined BST/Ce regional comparison | Across-region group effect p=0.05; treatment × region p=0.88 | Boundary-level pooled result; no detected differential sensitivity, not equivalence or an individual classifier |
| Faces versus buildings accuracy interaction | Group × condition p=0.63 | No reliable evidence that alcohol’s accuracy effect differed by these stimulus categories |
| Alcohol assay and brain measurement | Breath-device-referenced alcohol exposure plus MRI BOLD signals | Not direct venous whole-blood BAC, a brain ethanol concentration, or a case-specific causation test |
Why the BST finding and the Ce finding require different wording
Two statistical approaches appear in the same paper and should not be collapsed. The corrected voxelwise analysis identified the left BST, not a significant alcohol-related Ce cluster at the stated threshold. The a priori region-of-interest comparison extracted signals from anatomical BST and Ce regions, found an across-region group effect near its significance boundary, and found no statistically reliable treatment-by-region difference. These are related but nonidentical propositions.
It is therefore imprecise to write that the study proved every amygdala component “shut down” or that the central amygdala independently passed the same corrected voxelwise test. The authors’ interpretation of similar dampening across the two divisions rests on the regional model, and an interaction p=0.88 is not proof of mathematical equality. A null interaction can also reflect the available precision and power for detecting differences.
The researchers conducted a control analysis using 15 performance-matched subjects per group and obtained similar neural findings. That reduces concern that different task engagement alone explains the BST signal, but it does not eliminate every possible physiological, scanner, or selection explanation. The observation remains strongest as a group contrast in a specific fMRI protocol.
Does lower fMRI reactivity mean a participant experienced less fear?
Not directly. The protocol did not independently collect a validated, time-locked fear rating or defensive-response measure for each image. It employed fearful-looking faces to engage neurocircuitry associated with salience and threat processing. The paper itself notes that static fearful faces do not necessarily elicit robust manifestations of fear or anxiety, such as startle potentiation. A reduced hemodynamic response to these stimuli therefore cannot establish a subjective emotional experience for any one participant.
Even when a neural response and an emotion covary in other experiments, reverse inference remains problematic: observing altered activity in a brain region does not license the conclusion that a specific psychological state must have occurred or disappeared. The EAc contributes to several interacting functions, and fMRI voxel or regional contrasts do not uniquely identify a person’s motive or degree of caution.
To ask whether a person was frightened, perceived danger, or disregarded a warning during a particular episode, an evaluator would need contemporaneous behavioral, testimonial, environmental, medical, and toxicological evidence. This paper supplied none of those individual incident records.
Could the results explain a fight, threat misinterpretation, or an accident?
Only as limited background science, not as an event-specific proof. Random assignment to a beverage and measurement of a later fMRI response support an experimental inference about the study protocol: alcohol administration influenced group-average responses to standardized emotional stimuli. There was no test of actual fights, assaults, defensive force, driving, hazard detection on the road, eyewitness reliability, legal intent, or consent.
Neither a group-average brain effect nor a lower face-classification score permits an expert to calculate the probability that a named defendant misperceived a threat. The study provides no validated personal BOLD cutoff, no classifier sensitivity or specificity, no calibration across forensic clinical populations, and no demonstrated pathway from its 104-mm³ cluster to a specific disputed act.
In litigation, keep at least four questions apart: Was ethanol present, and in what specimen and amount? What did the person actually perceive or do? What external circumstances could have contributed? Is there evidence establishing that ethanol, rather than another factor, caused the act or outcome? Answering the first question does not automatically establish the later ones.
Why expectancy, selection, and imaging choices matter
The study’s single-blind arrangement did not blind investigators fully to all aspects of treatment administration, leaving potential expectancy bias. The placebo was deliberately credible and contained small alcohol-associated sensory cues; participants in that group estimated that they had consumed about two drinks. The design is useful for separating some pharmacological effects from expectations, but not for assigning a precise subjective ethanol effect to every participant.
The final sample excluded 12 recruits, including three whose task accuracy fell more than two standard deviations below the mean and three with excessive head motion. Such exclusions can be reasonable for imaging quality, yet they constrain extrapolation to intoxicated people who cannot complete tasks well or remain sufficiently still in a scanner. The participants were younger, screened, familiar with comparable drinking, and tested in a controlled setting rather than during a medical crisis or criminal encounter.
Hemodynamic BOLD signals additionally depend on data preprocessing, registration, filtering, smoothing, regions of interest, and statistical thresholds. The authors implemented several safeguards, including correction for multiple comparisons in the EAc region and predefined anatomical comparisons. Those safeguards improve internal analysis; they do not convert the research protocol into a standardized individual forensic diagnostic instrument.
Measurement hierarchy for a case involving alcohol and emotional behavior
| Evidence category | What it can establish if properly documented | What it does not automatically establish |
|---|---|---|
| Venous whole-blood ethanol assay | An ethanol concentration in that specimen at its verified collection time, within method uncertainty | A person’s earlier concentration, emotions, actions, or causation without additional evidence |
| Exhaled-breath ethanol reading | A device-reported breath result subject to device method and sampling conditions | An independently assayed venous whole-blood value or directly observed neural effect |
| Administered drink dose or projected BAL | A protocol quantity or model-based expectation under specified assumptions | A chemically measured individual concentration |
| fMRI BOLD BST/Ce contrast | A hemodynamic task-related signal with defined analysis and uncertainty | Direct neuronal firing, ethanol content of the brain, subjective fear, or criminal intent |
| Image-classification response | Accuracy on the experimental faces/places task | Road performance, truthful testimony, threat perception in a real confrontation, or a personal diagnosis |
| Actual case behavior and circumstances | Potentially relevant observations if contemporaneously documented and evaluated | By itself, the unique physiological cause of a disputed act |
For methodological background and source review, use the same-site forensic research library for lawyers and research source and editorial policy. Neither contextual page substitutes for the original study or for admissibility and reliability analysis in a specific jurisdiction.
What records would an expert need before drawing a case conclusion?
A case-specific assessment would begin with the contemporaneous record, not a hypothetical scan. Relevant materials can include verified blood or breath collection times, specimen matrix, analytical method, quality assurance, measured result and uncertainty; witnessed drinking and medication history; surveillance or other direct observations; incident chronology; interviews; contemporaneous statements; and plausible alternative explanations.
If an actual fMRI study of a person is being offered, request the complete acquisition and processing protocol, raw and quality-controlled data, motion exclusions, individualized statistics, pre-registration or analysis plan, reference population, repeatability, neurovascular confound assessment, and validated intended-use claims. An image with highlighted activation is not, by itself, a validated proposition about fear or intent. Even the published experiment did not supply a personal classification threshold.
Frequently asked questions about alcohol, the amygdala, and forensic fMRI
Did the study prove alcohol switches off the amygdala?
No. The corrected voxelwise result identified a reduction in a left BST cluster. The predefined regional model suggested broadly similar average dampening across BST and Ce, without demonstrating a complete shutdown of either structure.
Did researchers directly measure blood ethanol?
The protocol identified an Alcosensor IV breath instrument for its alcohol checks and used a dosing model targeting a BAL. It did not report a paired forensic venous whole-blood assay as the comparator for the imaging results.
Was the behavioral task a measure of real fear?
No. Participants categorized static faces and buildings by button press. They did not encounter real threats, and there was no validated individual fear-rating or defensive-action endpoint for the image presentations.
Does a statistically significant fMRI result diagnose impairment in one person?
No. Group-level statistical significance is not a validated individual threshold. The study established neither sensitivity and specificity for a personal impairment diagnosis nor a rule linking an fMRI cluster to unlawful conduct.
Can these data show alcohol caused violence or a driving collision?
No. Violence, actual driving, collision occurrence, and forensic causation were not measured. Moving from a controlled neural contrast to an incident explanation requires independent facts and a separately defensible causal analysis.
Why is the alcohol-by-region interaction important?
The predefined regional analysis found no statistically reliable difference between the treatment effects in BST and Ce (p=0.88). That is consistent with similar average patterns, but it is not a proof that their true effects or every person’s reactions were identical.
What is the most accurate one-sentence description for a legal filing?
“In a 49-person randomized, single-blind laboratory study, alcohol administration reduced group-average BOLD reactivity to emotional faces in a corrected left BST analysis, without validating any individual’s fear, impairment, intent, or event causation.”
Bottom line
Hur and colleagues offer credible experimental evidence that acute alcohol can modify a specific human brain network’s response to emotional-image cues. The paper’s meaningful findings include a corrected left-BST cluster, an across-region effect at p=0.05, and lower accuracy on the pictures task. The equally important boundaries are the breath-versus-blood distinction, the indirect nature of fMRI BOLD, the difference between classifying pictures and experiencing fear, and the absence of any validated individual impairment, legal-intent, or case-causation test.
Primary source: Hur J, Kaplan CM, Smith JF, et al. Acute alcohol administration dampens central extended amygdala reactivity. Scientific Reports. 2018;8:16702. doi:10.1038/s41598-018-34987-3. This educational evidence appraisal is not a case-specific medical or legal opinion and does not imply personal expert adoption of the findings.