Does Intravenous Methamphetamine Make People Jump to Conclusions? What a 17-Person Crossover Experiment Found

Short answer: In a randomized, blinded, three-visit experiment involving 17 analyzed healthy adults, intravenous methamphetamine induced mild psychotic-like experiences but did not produce a statistically significant change in how many beads participants inspected before deciding which jar a sequence came from. Adding oral amisulpride did not establish a drug-related reduction in information gathering either. The number of beads examined varied strongly with task difficulty and remained relatively consistent within individuals across visits. A small exploratory correlation connected baseline information sampling with symptoms in the methamphetamine-plus-amisulpride visit, but it was not reproduced as a significant association for methamphetamine alone. These are laboratory group results, not a validated measure of an individual person’s truthfulness, legally impaired driving, or blood methamphetamine concentration.

The original source is Ermakova, Ramachandra, Corlett, Fletcher and Murray’s 2014 peer-reviewed primary experiment, Effects of Methamphetamine Administration on Information Gathering during Probabilistic Reasoning in Healthy Humans, published in PLOS ONE. Its most instructive feature is a dissociation: an administered stimulant affected interview-rated mental experiences while the experimental measure of information sampling remained statistically unchanged. The distinction matters because stimulant exposure, altered subjective mental state, decision-task behavior and legal functional impairment are four different evidentiary questions.

What exactly did this experiment test?

The investigators asked whether a controlled catecholamine-releasing drug exposure would make healthy volunteers gather less evidence before making a probabilistic judgment. This response is conventionally called a “jumping-to-conclusions” tendency, but that label refers to a particular measurable task, not to every type of impulsivity or every conclusion made in ordinary life. The researchers also asked whether pretreatment with a dopamine-receptor-active antipsychotic, amisulpride, changed the cognitive response or the psychotic-like experiences associated with methamphetamine.

A third question involved individual stability: would someone who tended to sample few beads under placebo show the same pattern under different drug conditions? The investigators therefore compared drug conditions within the same people and calculated intraclass correlations. These questions are separable. A drug can change symptoms without changing task behavior; a stable task performance pattern can coexist with an acute response to a drug; and a correlation within seventeen people does not validate individual psychiatric or forensic prediction.

How many independent people participated?

Eighteen healthy volunteers were enrolled, including eleven men, with a reported enrollment mean age of 25.3 years and standard deviation of 4.9 years. One male participant was excluded because the investigators suspected a drug-administration error involving amisulpride on two visits. Accordingly, the principal analyzed sample was 17 independent human participants, not eighteen completed protocols and not fifty-one new people across three visits.

The design was within-participant crossover. Each person attended three study visits separated by at least a week. Each visit supplied repeated responses on the beads task, confidence ratings, psychiatric interview measures and digit-span assessments. The repeated trials improve information about within-person variability; they do not enlarge the biological denominator. The controlled experiment enrolled healthy volunteers rather than patients diagnosed with methamphetamine use disorder or schizophrenia. It therefore cannot be assumed to represent tolerant chronic users, acutely psychotic patients, older motorists or a litigation-specific population.

What substances, doses and routes were used?

One visit combined an intravenous methamphetamine infusion of 0.3 mg/kg over ten minutes with an oral placebo tablet. Another used the same intravenous methamphetamine exposure plus an oral 400 mg amisulpride tablet given approximately one hour before infusion. The control visit involved an intravenous saline infusion and oral placebo. The order was pseudorandomized and counterbalanced, and the relevant research staff and participants were blinded to the pharmacological condition.

The beads task was performed approximately three hours after the methamphetamine or placebo infusion. Thus a reported 0.3 mg/kg is an administered body-weight-adjusted intravenous dose, not a plasma, serum, or venous whole-blood drug concentration. The 400 mg amisulpride value is a tablet dose, not a measured receptor occupancy in these participants. The authors cited pharmacological occupancy findings from other literature to justify design choices; they did not chemically quantify each subject’s amisulpride or methamphetamine during this task.

What was the “beads task” and why did difficulty matter?

In the task, two hypothetical jars held different color proportions of beads. Participants encountered a sequence of actual beads and could inspect additional beads before declaring which jar was most likely responsible. One version made the evidence more informative, with a black-to-white ratio of approximately 85:15; the harder version used a ratio of 60:40. The main endpoint was “draws to decision,” abbreviated DTD: the number of sampled beads before deciding.

There were four trials at each difficulty level. Participants did not earn monetary rewards for rapid decisions, and the task imposed no explicit information-sampling cost. In the easier 85:15 condition, each bead often supplies stronger evidence than in the 60:40 condition; collecting fewer beads can therefore reflect rational sensitivity to evidence strength, not necessarily pathological haste. A low DTD score must be considered in relation to the actual likelihoods, the subject’s task instructions and baseline decision strategy.

The investigators separately measured confidence estimates while displaying bead sequences, including one sequence that shifted its favored jar. Those confidence ratings are not automatically equivalent to deciding to stop sampling, and neither should be confused with memory span or a clinical psychiatric diagnosis.

First evidence table: what happened to draws to decision?

Across the three visits, the authors found no significant drug-condition main effect on DTD; their text gives p = .630. The interaction between drug and task difficulty likewise was not significant (p = .472). In contrast, the effect of the jar-ratio condition was significant: more beads were drawn for the harder 60:40 task (reported F = 83.97, p printed as .000, which should be understood as a value below the publication’s display precision rather than literal zero). This is evidence that participants differentiated easier from harder inference tasks while the study failed to demonstrate a drug-related shift in stopping decisions.

Does a nonsignificant drug comparison prove there was no possible effect?

No. The study demonstrates a failure to detect a treatment effect with its sample, dose, task conditions and analytical design. Statistical nonsignificance is not a formal equivalence test and does not prove that every possible cognitive effect is exactly zero. With only seventeen analyzed people, moderate or context-dependent changes might evade detection. Other probabilistic tasks penalize additional sampling or add financial rewards; this version did not. Different task incentives might produce different results.

Nevertheless, it would be scientifically incorrect to describe this experiment as showing that intravenous methamphetamine made participants gather significantly less information. The published group means did not follow a consistent dose-independent pattern of fewer draws, the source reported a null treatment result, and the interaction test did not demonstrate differential drug effects as the evidence became less informative. Both positive and null findings need to be reported together rather than selectively invoking the general reputation of stimulants.

How stable was information sampling across the three visits?

The authors used intraclass correlation coefficients to assess whether different participants’ tendencies persisted across sessions. The easier 85:15 condition yielded an ICC of 0.857; the harder 60:40 condition yielded an ICC of 0.917. Within this experiment, those are relatively strong group-level measures of ranking consistency: volunteers who tended to stop after fewer or more draws often remained comparatively low or high samplers when the visit condition changed.

High ICC does not mean a task can detect whether a witness is telling the truth, whether a suspect was intoxicated, or whether a patient will develop psychosis. ICC quantifies relative stability within a particular sample and assessment procedure. It is not a sensitivity/specificity estimate, a legal reliability coefficient for individual testimony, or an independent replication in an external population. The authors interpreted the finding as supporting a relatively trait-like aspect of information sampling, subject to the boundaries of their experiment.

Did methamphetamine produce psychotic-like symptoms?

Yes, on the study’s structured psychiatric interview scale. Experienced psychiatrists administered selected components of the Comprehensive Assessment of At-Risk Mental States (CAARMS), including unusual thought content, non-bizarre ideas and perceptual abnormalities. These measurements concerned mild experimentally induced psychotic-like experiences, not verified schizophrenia diagnoses or an assessment of legal competency. The reported CAARMS mean was 0.18 under placebo, 1.94 under methamphetamine alone, and 3.12 under methamphetamine with amisulpride.

The authors reported a significant overall treatment-condition effect (p = .002). Comparisons with placebo were positive for methamphetamine alone (p = .032) and for the amisulpride-plus-methamphetamine condition (p = .008). However, the difference between methamphetamine alone and the combination was not statistically significant (p = .396). The combined condition’s numerically higher symptom mean should not be described as proof that amisulpride significantly worsened methamphetamine-induced symptoms. Nor did this particular 400 mg amisulpride pretreatment demonstrably abolish the reported mild experiences.

Second evidence table: significant effects versus genuine nulls

Did the researchers also test subjective confidence?

Yes. Participants rated confidence as successive beads appeared, including sequences that shifted the apparent most likely jar. The source reported no significant drug effect on average confidence for either sequence (p = .188 and p = .925). Its separate confidence comparisons at beads 1, 10 and 20 also were nonsignificant (p = .564, .395 and .894, respectively). These nulls matter because confidence and stopping time can move independently. The study cannot justify an assertion that methamphetamine significantly created overconfidence in this particular probabilistic-inference procedure.

That is not a universal claim that methamphetamine never changes confidence. The measured outcome was confidence in an artificial bead sequence under these task instructions, after one controlled exposure, with a small healthy volunteer cohort. A forensic opinion about recollection, perceptual error or suggestibility in a particular witness would require data on those actual functions and the person’s contemporaneous circumstances—not an indirect analogy from computerized or physical task behavior.

Did working memory explain the unchanged decision style?

To investigate whether cognitive capacity might mask drug effects on sampling, the researchers administered forward and backward digit span. A forward span reflects immediate ordered retention; backward span adds a manipulation component. Neither measure differed significantly between drug visits (forward p = .312; backward p = .247). The result constrains the specific hypothesis tested: the experiment did not identify a drug-related improvement in these span scores large enough to establish that a memory gain counterbalanced an opposing information-sampling effect.

Digit span is only one family of memory tasks. It is not a complete test of executive function, procedural learning, episodic memory, real-world judgment or legal decision capacity. A nonsignificant span contrast also cannot prove that no person experienced any change. The correct reading is that the authors measured these particular scores and did not report significant differences under the three crossover conditions.

What was the exploratory baseline-thinking association?

Participants who sampled fewer beads in the easy 85:15 task during placebo tended to report more psychotic-like symptoms during the combined methamphetamine-plus-amisulpride visit. The Spearman correlation was r = −0.517, p = .028. For unusual thought content specifically, the authors reported r = −0.547, p = .019. These are negative associations: higher baseline sampling aligned with lower later symptom scores in that condition.

Crucially, the analogous relationship for the methamphetamine-only visit was not significant, r = −0.127, p = .616. The harder 60:40 baseline task also did not significantly correlate with combination-visit symptoms (r = −0.216, p = .390), or methamphetamine-only symptoms (r = −0.077, p = .760). This pattern is narrower than claiming that a general “jumping to conclusions” trait predicts stimulant psychosis.

The exploratory analysis involved many related comparisons within seventeen people and was not externally validated. The participants’ mild CAARMS scores were measured after controlled drug administrations, not through a clinical longitudinal diagnosis. The data therefore cannot yield a tested classification threshold, an individual probability of psychosis, or a finding that a particular defendant, patient or witness was more likely to become psychotic.

What physical biological specimens and drug concentrations were measured?

This paper did not present a calibrated postdose chemical analysis of methamphetamine or amisulpride in any participant’s legal venous whole blood, separated plasma, serum, urine, oral fluid, sweat, hair, exhaled breath, vitreous humor, milk or tissue. Its recorded principal endpoints were an administered infusion amount, an oral tablet amount, choices on bead tasks, confidence ratings, digit span and psychiatric scale scores. Those are different measurement classes.

Because blood was not chemically analyzed for the administered drug in this dataset, there is no individual ng/mL value, method detection threshold, lower limit of quantitation, internal-standard recovery, whole-blood/plasma partition ratio or validated concentration–response curve. References in the Methods to amisulpride plasma peak timing or receptor occupancy are citations to prior pharmacological work and timing assumptions, not measured concentrations in these seventeen volunteers.

Likewise, the presence of intravenous administration does not imply a forensic venous sample was drawn and tested. The route of putting a drug into a participant and the matrix used to measure it are conceptually and physically separate. An infused mg/kg dose cannot be converted into an observed legal whole-blood concentration without a separate validated pharmacokinetic and sampling record.

How do the reported statistics limit the claimed findings?

Multiple hypotheses and endpoints were investigated: DTD by difficulty, drug-order effects, confidence at various sequence points, digit spans, CAARMS comparisons and several exploratory correlations. Different tests therefore have different inferential meaning and may be sensitive to multiple testing. The positive drug effect on mild symptoms is supported by source comparisons against placebo; the null DTD result cannot be erased by that positive result. Conversely, the measured symptom response does not prove that the experiment was too insensitive to detect any cognitive changes at all.

The original statistical presentation also contains some unusual formatting in the degrees of freedom and test-statistic reporting. Such issues warrant checking the source, statistical methods and any correction materials before testimony that depends on exact reconstructed degrees of freedom. They do not authorize replacing an observed nonsignificant outcome with an imagined significant result or reporting an unmeasured effect size. The two evidence tables above preserve the outcome values and p-level distinctions the published source actually communicates without claiming to independently reanalyze unavailable individual participant data.

What limitations prevent a human impairment inference?

The trial was small, involved young healthy adults, used a single methamphetamine infusion dose and a single amisulpride tablet dose, and measured task behavior at one approximate postinfusion interval. The investigators examined brief acute drug effects rather than chronic dosing, sleep deprivation, naturalistic polysubstance exposure, psychiatric crisis, tolerant repeated users, or an intoxicated driver in a roadway environment. The bead task supplied no reward or explicit penalty for sampling more information. It is not a direct substitute for judgment about traffic hazard detection, medical consent, occupational safety, criminal intent or testimonial veracity.

There was no validated on-road driving endpoint, no simulator crash endpoint, no laboratory whole-blood concentration, and no case-specific toxicological chronology. CAARMS findings concern mild induced experiences, not a forensic determination of functional incompetence or mental state during an alleged act. The crossover offers better control of stable between-person characteristics than an uncontrolled observational comparison, but it does not transform the findings into individualized legal evidence or broad population cutoffs.

How should forensic experts describe this paper in testimony?

A precise expert description would say that an acute controlled intravenous methamphetamine exposure in a 17-person analyzed crossover produced a statistically supported increase in selected mild psychiatric interview ratings while information sampling on the classical beads task showed no significant drug main effect or drug-by-difficulty interaction. The same participants were more consistent across sessions than a casual description of drug-induced decision changes might suggest, but exploratory associations were condition-specific and not independently validated.

Do not say that methamphetamine “caused jumping to conclusions” in this study; it did not establish that result. Do not say it proved normal cognitive functioning, since it did not comprehensively test all domains. Do not call the psychotic-like experiences a proven psychotic disorder, and do not convert 0.3 mg/kg infused dose to a fictitious measured blood level. For a particular litigation matter, the laboratory’s case specimen, exposure history, confirmed analytical concentrations, timing, sleep and health conditions, observed behavior and applicable scientific validation require separate examination.

Frequently asked scientific and forensic questions

1. Did everyone receive three independent doses of methamphetamine?

No. Each analyzed volunteer completed three crossover visits: placebo/saline, intravenous methamphetamine plus placebo tablet, and intravenous methamphetamine after oral amisulpride. This is seventeen independent people with repeated treatments, not fifty-one unrelated subjects or three methamphetamine infusions per person.

2. Was methamphetamine given orally, intranasally, or intravenously?

The study used a controlled ten-minute intravenous methamphetamine infusion of 0.3 mg/kg. The 400 mg amisulpride was given orally about one hour before infusion. The study cannot directly establish effects of different street-product exposures, routes or doses.

3. Did methamphetamine significantly increase “jumping to conclusions”?

No significant change in draws to decision was detected across drug conditions (reported p = .630), and the drug-by-task-difficulty interaction was nonsignificant (p = .472). The more difficult jar ratio did substantially increase draws. A statistical null does not guarantee no effect under other doses or paradigms.

4. Did amisulpride prevent the reported mental-state effects?

Not demonstrably in this protocol. CAARMS ratings increased versus placebo both under methamphetamine alone and under the combination. The comparison between those two active conditions was nonsignificant (p = .396). A nonsignificant difference does not prove amisulpride was protective or harmful.

5. Did the researchers measure methamphetamine in whole blood?

No. Their principal quantitative outcomes were behavioral and interview ratings. They did not report participant-specific forensic venous whole-blood methamphetamine, separated plasma or serum concentrations, drug confirmation, calibration, or a pharmacokinetic concentration–effect relationship.

6. Does baseline bead sampling reliably predict a person’s future psychosis?

No clinically validated prediction rule was developed. One exploratory placebo-bead correlation with symptoms in the combination condition was significant (r = −0.517, p = .028), but the methamphetamine-only correlation was not (p = .616), and no external validation or individual threshold was provided.

7. Can this paper prove impaired driving, unreliable testimony, or incident causation?

No. The volunteers did not undergo driving, crash, witness-recall or legally relevant incident testing. A small laboratory task and mild symptom scale cannot establish impairment, truthfulness or cause in a named person’s case without independent evidence and validation for that precise inference.

Source and final evidence boundary

Primary experiment: Ermakova AO, Ramachandra P, Corlett PR, Fletcher PC and Murray GK (2014), Effects of Methamphetamine Administration on Information Gathering during Probabilistic Reasoning in Healthy Humans, PLOS ONE 9(7):e102683, DOI 10.1371/journal.pone.0102683. For an independent explanation of why nominal administered drug dose does not substitute for a chemically tested specimen, see the published same-site guides to biological matrices, confirmatory drug testing, and therapeutic drug monitoring. These general guides do not provide missing concentration measurements for this trial. The scientifically defensible result is a selective acute psychiatric symptom effect alongside a null laboratory information-sampling drug comparison, without any validated individual DUI, legal impairment or case-causation finding.