Answer: In two alcohol-challenge experiments, changing what participants were told about drinking did not reliably change how much beer they subsequently consumed or how strongly they reported craving alcohol. Yet in one experiment, the message participants received was associated with a different pattern of performance on a stop-signal task after alcohol. This distinction matters: a belief about alcohol, a subjective desire to drink, an observed computer response, and an individual’s ability to control conduct during a real incident are not interchangeable findings.
Knibb and colleagues tested these questions in two separate studies published in 2018. The investigators administered alcohol or a placebo in counterbalanced sessions, measured performance on a laboratory inhibitory-control task, and offered a subsequent beer taste test. Their findings support a narrow interpretation of how an experimental message may interact with measured task performance. They do not demonstrate that expectations alone cause intoxication, that any person lacked legally meaningful self-control, or that a computer task predicts a particular act. The original research is available at DOI: 10.1371/journal.pone.0201042.
What was the research question?
The researchers examined two related but distinct propositions. First, might a person’s beliefs about alcohol change the tendency to consume more alcohol after an initial drink, often described experimentally as an alcohol priming effect? Second, might a message about alcohol and self-control change performance on a test requiring a person to stop an already-prepared response?
These are not equivalent outcomes. Drinking beer during a timed laboratory taste test is an observed consumption measure. Craving is self-reported desire. Stop-signal reaction time is calculated from keyboard responses in a structured task. The study was not designed to determine criminal intent, intoxication-related legal defenses, eyewitness accuracy, field sobriety performance, or whether a defendant could have refrained from a particular act.
Who participated, and how was alcohol administered?
Study 1 enrolled 81 adults (44 men, 37 women; ages 18–49). Study 2 enrolled 82 different adults (29 men, 53 women; ages 18–48). Recruitment centered on the University of Liverpool. Participants had to report drinking at least 10 United Kingdom units of alcohol per average week, prefer beer, and satisfy specified health and eligibility criteria. These were selected social-drinking volunteers, not a sample of drivers, criminal defendants, patients undergoing forensic evaluation, or the general population.
Each person attended two sessions separated by at least 48 hours. The within-person comparison was a beverage containing 0.5 g of ethanol per kilogram of body weight versus a placebo beverage containing lemonade with vodka scent applied to the glass and surface. Alcohol/placebo order was counterbalanced, while the message condition varied between participants. The beverage manipulation was single-blinded rather than a guarantee that participants remained unaware of the actual drink.
The investigators used a breath alcohol instrument during sessions. They did not collect matched venous whole-blood samples for forensic headspace gas chromatography. The paper’s printed breath-concentration units and magnitudes have ambiguities and must not be silently converted into a claimed measured whole-blood BAC. The administered dose, device measurements, participant perceptions, and subsequent consumption are distinct categories of evidence.
How did Study 1 attempt to change beliefs about self-control?
Participants in Study 1 completed the stop-signal task and then received deliberately inaccurate feedback. One group was told its self-control score was high, with a displayed index of 92.6%; the comparison group was told its performance was average, with a displayed index of 51.2%. These percentages were fabricated experimental feedback, not validated accuracy scores or independently measured percentages of real-world self-control.
The feedback came after the stop-signal task in that session. Therefore, it could not explain the task performance already recorded on that occasion. Researchers then assessed perceived control, craving, subjective intoxication, and subsequent beer consumption. Their manipulation check did not show that the high-versus-average feedback successfully changed participants’ beliefs about their capacity to control drinking.
That failed manipulation check is central. A study that does not move its intended psychological variable cannot support a strong causal assertion that this variable does or does not affect a later outcome. The investigators did observe other patterns, but those results must not be presented as proof that the intended self-control belief treatment worked.
What did Study 1 actually find?
During the alcohol sessions, participants consumed more beer in the subsequent taste test than during placebo sessions, with a reported drink main effect p = 0.007. However, there was no significant drink-by-feedback interaction in consumption (p = 0.730), and no main effect of assigned high-versus-average feedback (p = 0.734). Craving was higher overall during alcohol than placebo sessions (p = 0.001), but the drink-by-time interaction was not statistically significant (p = 0.273).
The stop-signal reaction-time analysis did not establish a significant overall alcohol main effect (p = 0.105). For one separate outcome—stop-signal inhibition errors—the authors reported a drink-by-feedback interaction (p = 0.006). Their exploratory order analysis indicated that differences depended on which session involved alcohol and on prior feedback. Given the failed belief manipulation and session-order dependence, these results should not be collapsed into a simple statement that “thinking you have good self-control protects against alcohol.”
How was Study 2 different?
Study 2 moved the message before the beverage and the stop-signal task. The experimental script misleadingly suggested that consuming a small amount of alcohol reduces the urge to drink; the neutral comparison script described general research about alcohol and cognitive processes without making that specific claim. Participants then consumed alcohol or placebo, completed the cognitive task, reported their experiences, and entered a later beer taste test.
Unlike Study 1, the manipulation check in Study 2 supported an effect of the script on stated beliefs about control over drinking. The researchers reported a drink-by-script interaction p = 0.021 for the manipulation check. This does not establish that the script altered the pharmacokinetic concentration of ethanol or that every participant genuinely believed the script in the same way. It describes a group-level questionnaire response under the protocol.
Study 2 found a stop-signal interaction—but not greater beer consumption
The central stop-signal outcome was stop-signal reaction time (SSRT). A longer calculated SSRT is generally interpreted as slower inhibition of a prepared response within that paradigm. In Study 2, the investigators reported a significant drink-by-script interaction, F(1,72) = 6.00, p = 0.017. Within the experimental-message group, the mean SSRT was 246.96 milliseconds after alcohol versus 217.31 milliseconds after placebo (within-group comparison p = 0.012). Within the neutral-script group, the means were 202.18 versus 208.71 milliseconds, respectively (p = 0.480).
That pattern is an interaction, not a validated rule that alcohol impairs response inhibition only when a person expects a particular effect. The task was administered under controlled research conditions; the script itself suggested alcohol would reduce further urge to drink, not that every individual would acquire a fixed degree of motor or behavioral control. The analysis showed a statistically detectable difference between group response patterns, not individual classification accuracy.
The same study found more stop-signal inhibition errors after alcohol than placebo across message conditions (F(1,72) = 8.22, p = 0.005). It did not identify a significant script effect or alcohol-by-script interaction on subsequent beer consumption: the reported drink main effect for the taste test was p = 0.798, the drink-by-script interaction p = 0.373, and the script main effect p = 0.963. Nor did the authors find a significant script main effect on overall craving (p = 0.769).
Evidence table: separate the measured endpoints
| Study and actual outcome | Reported comparison | What the result supports |
|---|---|---|
| Study 1; beer consumed in taste test, n = 81 enrolled | Alcohol-versus-placebo main effect p = 0.007; drink × assigned feedback p = 0.730 | More beer was consumed overall after alcohol sessions; no detected moderation by assigned feedback. |
| Study 1; self-control feedback manipulation | High versus average fabricated feedback failed the investigators’ belief-change check | The intended belief mechanism was not convincingly altered; no causal conclusion about successful belief change. |
| Study 2; integrated SSRT, 74 datasets after exclusions | Drink × script F(1,72) = 6.00, p = 0.017 | Mean laboratory stopping latency varied by beverage and message combination. |
| Study 2; stop inhibition errors, 74 datasets | Drink main effect F(1,72) = 8.22, p = 0.005 | More stopping failures under alcohol in the sampled task; not an individual forensic diagnostic test. |
| Study 2; beer consumed in taste test, 82 enrolled | Drink p = 0.798; drink × script p = 0.373; script p = 0.963 | No significant effect detected on this later consumption outcome. |
Statistical tests used outcome-specific analyzable samples. Study 1 excluded five stop-signal datasets because of technical loss or analysis criteria; Study 2 excluded eight. It would be incorrect to treat all 163 recruits as participants in a single pooled stopping-latency comparison.
Why a computerized stop-signal task is not a legal test of self-control
On ordinary go trials, participants pressed a key matching an arrow. On a minority of trials, an auditory tone signaled that they should suppress the prepared response. The task used a tracking algorithm for the interval between the arrow and stop tone. SSRT is calculated from responses under that algorithm; it is not a blood measurement, a reading of intoxication, a direct observation of criminal behavior, or a clinical diagnosis.
Everyday decisions and legally disputed events can involve perception, memory, motivation, social context, choices made over many seconds or minutes, and evidence from multiple witnesses. A laboratory inhibition task isolates a constrained part of response control. Even a statistically significant average slowing does not determine what one participant would do in another setting or whether alcohol caused a disputed act.
Likewise, the taste test observed amounts consumed after a research drink while participants were instructed to rate samples of beer. It did not measure sustained heavy drinking in ordinary life, an inability to refuse alcohol, an alcohol-use-disorder diagnosis, or any later driving event. Craving scores capture reported feeling, not the number of drinks necessarily consumed.
Measurement table: what was direct, reported, or inferred?
| Variable | Evidence category | Not established by that variable |
|---|---|---|
| Oral ethanol dose of 0.5 g/kg | Administered amount based on body weight | Individual absorbed dose, contemporaneous whole-blood BAC, or legal impairment |
| Portable breathalyzer session reading | Instrument breath-related measurement/display | Direct venous blood measurement or automatically interchangeable BAC units |
| Belief manipulation questionnaire | Participant self-report and manipulation check | Actual biological ethanol effect, validated resistance to coercion or conduct |
| SSRT in milliseconds | Task-derived estimate of response inhibition | Measured blood ethanol, crash risk, intent, or individualized legal causation |
| Number of inhibition errors | Observed errors during specified computer trials | A general percentage chance of an accident, assault, or inability to choose |
| Beer volume in a taste test | Observed laboratory intake under task instructions | Real-world dependency, criminal conduct, or future drinking certainty |
Why the reported breath readings require special caution
The paper reports breath-device readings but prints some values with a concentration format that should not be treated as a validated blood concentration. Importantly, the authors acknowledge that participants were not given water following the priming dose, so residual alcohol in the mouth could have elevated some breath readings. The taste test also introduced additional ethanol later in the session, further changing what a later breath reading represents.
A reader should preserve the original instrument record, units, sampling time, time of each drink, observation period, calibration documents, and matrix before drawing a forensic inference. In this research article there was no independently confirmed matched venous whole-blood ethanol series to resolve person-specific conversion. A single reported research breath concentration does not permit a reliable claim that the person had a particular forensic BAC or that a particular mental or behavioral state was caused by ethanol.
Limitations that matter in litigation
The samples were selected adult social drinkers recruited around one university. The beverage experiment was single-blinded, and participants sometimes inferred that placebo drinks contained alcohol. Study 1 did not successfully change its intended belief construct. Study 2 changed a questionnaire measure of expected control but did not alter observed beer drinking or craving by script. Stop-signal results were sensitive to the experimental message and, in Study 1, to session order.
The sample of usable stop-signal recordings was smaller than the number enrolled. The tasks were administered in a semi-naturalistic bar laboratory, not during an accident, arrest, courtroom examination, or actual choice with legal consequences. The authors did not record forensic blood specimens, assess divided attention during driving, establish a legally meaningful impairment threshold, or follow participants to determine case-specific behavioral outcomes.
Null or non-significant tests do not establish that two groups are identical; nor does one significant interaction establish a general mechanism in every individual. Consider the study’s hypotheses, number and nature of tested endpoints, exclusions, and the difference between a pre-specified comparison and an exploratory order analysis before characterizing its evidentiary weight.
Questions to ask when this paper is offered as forensic evidence
A careful scientific review begins by asking which result is actually being invoked. Is the assertion about increased consumption, self-reported craving, inferred stopping latency, count of stopping errors, or a numerical breath reading? Those endpoints cannot be substituted for one another.
Next, determine whether the underlying experimental manipulation worked and whether the sample and testing sequence represent the case under review. If an expert invokes the SSRT interaction, ask whether they acknowledge that Study 2 did not show corresponding increased beer consumption and that Study 1’s feedback manipulation was unsuccessful. If a breath number is invoked, require identification of the actual specimen, units, device, timing, and confirmation protocol. If causation is asserted, identify the independent case evidence bridging a laboratory average to the particular event.
Frequently asked questions
Did the researchers show that beliefs alone cause alcohol intoxication?
No. Alcohol was administered as a separate intervention, beliefs were manipulated through feedback or scripts, and the researchers assessed questionnaires and tasks. The study did not measure a physiological intoxication state produced by belief in the absence of ethanol.
Did the messaging make participants drink more beer?
Not in the reported comparisons between assigned message conditions. Study 1 found greater beer consumption after alcohol than placebo overall, but not a significant drink-by-feedback effect. Study 2 found no significant drink, script, or interaction effect on taste-test drinking.
Did alcohol impair stopping performance in every condition?
No. Study 2 found a drink-by-script interaction in derived SSRT: the alcohol/placebo difference was significant in the experimental-message group but not in the neutral-message group. The same study found an overall increase in inhibition errors after alcohol. These are different measures, not mutually exclusive claims.
What does SSRT measure?
Stop-signal reaction time is a model-derived timing estimate based on responses and stop cues in a computer task. It does not directly measure blood alcohol or a person’s capacity to control a particular real-world action.
Are the reported breath alcohol values equivalent to laboratory whole-blood BAC?
No. Breath measurement, any instrument’s reporting format, and measured venous whole blood are different. The paper also identifies a possible residual-mouth-alcohol limitation. It supplies no contemporaneous forensic whole-blood series validating a person-specific conversion.
Can the study establish that a defendant lost control because of drinking?
No. It documents group-level experimental responses. Whether alcohol contributed to a specific person’s conduct requires case evidence, appropriate timing, validated measurements, and analysis of alternative explanations.
Does a nonsignificant finding prove that alcohol never affects inhibition or craving?
No. A nonsignificant result means that a specified comparison did not meet the study’s evidentiary threshold. It does not prove equivalence, establish a universal absence of effects, or resolve what happened in another person.
Primary evidence and related research resources
Original study: Knibb G, Roberts CA, Robinson E, Rose A, Christiansen P. The effect of beliefs about alcohol’s acute effects on alcohol priming and alcohol-induced impairments of inhibitory control. PLOS ONE. 2018;13(7):e0201042. Read the complete original peer-reviewed study.
For a question-led overview of laboratory versus legal inference, consult this site’s forensic toxicology research library. Its source and editorial policy describes how scientific literature is distinguished from individual case opinions. An example of why calculated estimates require separate scrutiny appears in the two-blood-samples alcohol-elimination brief. These are navigation resources, not additional empirical sources for this appraisal.
Evidence boundary: This educational interpretation concerns the identified 2018 research, not an independent examination of any person’s blood specimen, alcohol-test record, driving, conduct, or litigation file. It is not a case-specific expert opinion.