Does 20 mg Dextroamphetamine Reduce Late-Task Reaction-Time Instability? A 69-Person Conners CPT Trial

Short answer: In a blinded, randomized, single-session experiment involving 69 healthy adults, the 20 mg oral dextroamphetamine group showed less deterioration in the consistency of reaction times across a prolonged Conners Continuous Performance Test than the placebo group. The study’s clearest corrected result concerned change in hit-reaction-time error across the task (overall drug-dose effect, p = .003). An independently prespecified analysis also reported a dose-by-task-block interaction in signal-detection performance (p = .036). But neither dose significantly improved the aggregate Conners detectability score, overall misses, false alarms, gambling-task performance, or Wisconsin Card Sorting Task performance in the principal group comparisons. These narrow task findings do not show that a particular person taking amphetamine was safe to drive, free of impairment, or better at judgment under real-world conditions.

The primary evidence is MacQueen, Minassian, Henry, Geyer, Young, and Perry, Amphetamine Modestly Improves Conners’ Continuous Performance Test Performance in Healthy Adults, published in the Journal of the International Neuropsychological Society (2018), DOI: 10.1017/S135561771700090X. The analysis below addresses the original full methods, group data, statistical thresholds, figures, and limitations rather than treating the abstract’s general description of “improved vigilance” as a universal human-performance conclusion.

What exact scientific question did the experiment test?

Can a single oral dose of dextroamphetamine change the trajectory of a laboratory attention test in healthy adults? This question differs from whether the drug reduces an average reaction time or makes every participant more accurate. A person might respond with roughly the same overall average speed yet become less consistent as minutes pass. Conversely, average accuracy can look unchanged even if the last part of a repetitive task reveals a loss of sustained attention. The investigators deliberately looked for changes across the task in addition to standard summary scores.

This distinction matters when laboratory data appear in occupational, medical, fitness-for-duty, medication, or litigation discussions. The source experiment had neither a driving simulator nor an actual road test. Its principal outcome was computer-keyboard performance after random assignment to treatment—not detection of a concentration of amphetamine in an individual driver’s blood. The mechanistic explanation should remain bounded by those observations.

Who actually participated, and how were the groups allocated?

Sixty-nine distinct adults completed the experimental assessments. The paper reports 34 men and 35 women, ages 18 to 34 years, with median age 22. They were screened to exclude major psychiatric or recent substance-use conditions, relevant neurological or motor disorders, and recent drug findings on intake toxicology screening. This was a medically screened, healthy volunteer sample, not a cohort selected because of attention-deficit/hyperactivity disorder, sleep deprivation, driving incidents, or therapeutic treatment failure.

There were 29 participants assigned to placebo, 16 to 10 mg oral d-amphetamine, and 24 to 20 mg oral d-amphetamine. The design was parallel-group, single-session, randomized and double-blind: each person received one assigned condition. It was not a 69-person three-way crossover, and repeated test blocks did not create additional independent people. The investigators reported that the pharmacy later oversampled placebo and 20 mg to increase the comparisons anticipated to be most informative. That practical allocation choice should be disclosed rather than mistaken for equal-sized arms.

The administered compound was dextroamphetamine (d-amphetamine). It was not methamphetamine, not a racemic d,l-methamphetamine mixture, and not lisdexamfetamine. A nominal swallowed dose of 10 or 20 mg is also not a laboratory measurement of a participant’s plasma or whole-blood concentration. The investigators did not report a quantitative post-dose forensic amphetamine concentration-response series for these participants.

What was the test sequence after dosing?

After ingestion under the study procedures, participants began the Conners CPT-2 at approximately 80 minutes, the Iowa Gambling Task around 95 minutes, and the Wisconsin Card Sorting Task at approximately 120 minutes. These are approximate protocol times, not individualized peaks established by measuring each participant’s drug concentration. The investigators expected the time window to fall near the ascending distribution phase based on prior pharmacokinetic literature, but the present study did not validate that prediction by collecting a contemporaneous concentration series.

The CPT-2 consisted of 360 letter presentations: 18 sets of 20 trials, with interstimulus intervals of one, two, or four seconds. Participants were instructed to press for letters other than “X” and to withhold their response to “X.” About 10% of trials involved the non-target letter. The investigators organized the task into three successive 120-trial segments to examine whether vigilance changed as the test continued. Those segments are dependent observations on the same participant.

It is important not to describe this task as a roadside field sobriety test, a reaction-time measurement made behind the wheel, or a standardized prediction of an individual’s risk of causing a collision. It records responses to controlled visual stimuli using a keyboard. It does not recreate conflicting traffic demands, judgment of speed, unexpected hazards, distractions, medication interactions, fatigue from an overnight shift, or a collision reconstruction.

How do misses, false alarms, reaction time, and response-time error differ?

A miss is failure to press when an ordinary target letter calls for a response. A false alarm is pressing for the non-target “X” when the response should be withheld. Hit reaction time is a latency for correct target responses. Hit-reaction-time error captures inconsistency or dispersion in those latencies rather than their average. A person can retain similar mean speed while becoming more erratic in response timing, and faster responses are not automatically more accurate.

The paper also used a detectability or signal-discrimination statistic, labeled D-prime (D′), and a response-style measure (beta). D′ is intended to capture discrimination between target and non-target events, rather than directly measure time-to-brake. The “ADHD Confidence Index” is a composite function constructed from task and demographic variables; it is not a clinician’s new ADHD diagnosis and not a forensic impairment probability. Terms must be reported as the source authors used and defined them, not converted into legal thresholds.

What exactly improved under dextroamphetamine?

The clearest corrected outcome was the change in hit-reaction-time error as the CPT progressed. The placebo arm’s source Table 2 mean change across blocks was +0.053 (standard deviation 0.068); the 10 mg arm was +0.033 (SD 0.077); and the 20 mg arm was −0.024 (SD 0.056). A positive value indicates increasing inconsistency across the analyzed task blocks under the paper’s metric. The overall dose comparison had F(2,62) = 6.50 and p = .003, satisfying the prespecified corrected significance threshold of p < .005 for the set of principal CPT measures. The 20 mg arm differed from placebo in post hoc testing reported as p < .042.

The 10 mg mean was numerically intermediate, but the source did not establish a statistically supported pairwise advantage of 10 mg over placebo on that outcome. Nor did the experiment prove that 20 mg is an optimal dose for all adults: there was no wider individualized dose-response optimization, no repeated dosing course, and no person-level rule that could be translated into driver classification.

Separate analyses across three 120-trial blocks identified a dose-by-block interaction for D′: F(4,124) = 2.7, p = .036. The original protocol allowed the conventional .05 threshold for this prespecified time-course interaction. Within the placebo arm, detectability changed across blocks, p = .010; the corresponding block comparisons in the 10 mg and 20 mg groups were p = .14 and .11. The absence of a significant within-arm trend in the active conditions is not proof that attention was preserved for every treated individual. The statistically relevant statement is the interaction across groups and task segments, interpreted with the smaller active-arm denominators and variability.

Two different statistical questions should not be confused

Looking only at the overall CPT detectability mean gives a very different impression from examining change over time. Source Table 2 reported overall D′ values of 0.81 (SD 0.47) for placebo, 0.78 (SD 0.40) for 10 mg, and 0.81 (SD 0.42) for 20 mg, with an overall dose p = .982. That null result is real and should not be erased by the separate dose-by-block interaction. The task may be sensitive to changes in vigilance trajectory while the overall detectability averages remain similar.

Similarly, the source Table 2 aggregate mean hit reaction time was 392.58 ms with placebo, 357.22 ms with 10 mg, and 365.09 ms with 20 mg, but the reported overall drug-effect p = .098. A numerical difference in those group means does not authorize a claim of statistically established faster reaction time at the study’s corrected threshold. The 20 mg benefit concerned time-course inconsistency, not a demonstrated global improvement on every summary metric.

Evidence table 1: the study design and what each task measured

Evidence table 2: positive, trend-level, and null outcomes

Why the study’s multiplicity rules matter

With many task outputs, a conventional .05 threshold applied blindly to every comparison makes it easier to encounter a nominally significant result by chance. The investigators addressed this for the main outcomes by dividing .05 by the number of measures in each task, using a stricter p < .005 for the main CPT comparisons and p < .006 for the gambling and card-sorting measures. The hit-reaction-time error time-course comparison survived that rule. The ADHD Confidence Index drug effect of p = .026 did not, even though the 20 mg versus placebo post hoc contrast was described as .009. Labeling the composite a proven diagnostic improvement would both ignore the adjustment and misunderstand what the index represents.

The separate prespecified dose-by-block analysis used α = .05 for its time-course hypothesis. That explains why its p = .036 was discussed as an interaction without implying that an overall dose comparison with p = .982 became significant. It is not scientifically coherent to switch between these endpoints without identifying the statistical family and what was being tested.

For legal review, the exact endpoint, hypothesis, sampling unit, effect-size estimate, multiple-comparison procedure and denominator should appear alongside any quoted p value. A small p value indicates evidence about a study-level null hypothesis under the tested model; it cannot assign an individual motorist or patient to a state of competence.

Did amphetamine reduce impulsive responding?

Not on the central CPT indicators in this experiment. False alarms, a response made when the letter “X” required withholding, averaged approximately eleven or twelve across groups and had a dose comparison of p = .982. Perseverative or very rapid responses did not show a supported drug effect, with p = .815 in Table 2. The response-style beta comparison was likewise nonsignificant at p = .191. The investigators discussed the possibility that a low baseline frequency of impulsive responses limited room to observe improvement.

This is not the same thing as proving that amphetamine has no effects on response inhibition at other doses, in sleep-deprived drivers, or in an ADHD population. A null group comparison may reflect a true small effect, measurement precision, task sensitivity, sample size, restricted baseline performance, or a mixture. The proper claim is tied to the particular test, dose, study population and time of observation.

Did judgment improve on the Iowa Gambling Task?

The Iowa Gambling Task presents four decks with differing patterns of modest rewards and larger occasional penalties. Participants choose virtual cards, receive feedback, and may learn which strategies lead to better long-term theoretical payoffs. It is relevant to studying reward-related choice, but it does not recreate real investment behavior, prosecutorial decision-making, medical decision-making, or judgment at an intersection.

The investigators reported no significant dose effect on the total score (p = .795) or the other analyzed gambling measures. Numerical total scores were 9.31 for placebo, 12.88 for 10 mg, and 14.25 for 20 mg, with broad standard deviations. Selecting the largest mean without its variance and null test would exaggerate the result. The appropriate conclusion is that a demonstrated time-course change in keyboard vigilance did not translate into a statistically established group difference in these card-choice measures.

What did the Wisconsin Card Sorting Task add?

The Wisconsin task requires adaptation to changing classification rules by using feedback about card sorting. It is often described as a measure of set shifting and aspects of executive function. In this experiment the total-correct mean was 48.21 for placebo, 51.63 for 10 mg, and 48.67 for 20 mg, with an overall dose p = .988. Total errors, perseverative errors, categories completed, and failures to maintain a set also failed to show supported differences under the study’s analytic criteria.

Neither gambling nor card sorting is a perfect proxy for a particular individual’s real-world judgment. The fact that two additional laboratory tasks were included improves the interpretive specificity of the primary result: the evidence points toward a selective difference in sustained attention metrics, rather than broad improvement across decision-making or cognitive flexibility. That is a more defensible reading than calling dextroamphetamine a general-purpose cognitive enhancer.

What blood, plasma, urine, and oral-fluid chemistry can this paper establish?

The protocol used an intake urine toxicology screen to help identify recent substance use that might disqualify participants. It did not report a paired postdose quantitative concentration series in venous whole blood, separated plasma or serum, urine, hair, oral fluid, or exhaled breath. The experiment’s primary exposures were assigned oral dose groups. A screening assay for eligibility is not a quantitative pharmacokinetic measurement and should not be described as a blood concentration.

Forensic interpretation must therefore resist apparently useful but unsupported conversions: 20 mg swallowed is not a specified ng/mL in blood; a drug’s expected peak time from previous studies is not a measured individual peak; and a laboratory performance difference cannot create a blood concentration threshold for impairment. These distinctions parallel the general principles explained in the site’s biological-matrices guide, confirmatory-testing guide, and therapeutic drug-monitoring guide.

To interpret an actual toxicology finding in a legal record, the analyst would need the specimen type and collection time, the target drug and any metabolites, identification and quantitation procedures, calibration and uncertainty, administration history, and independent contemporaneous behavior. None of those case-specific elements can be filled in by assuming that a concentration lies on the 10 mg or 20 mg arm of this study.

Would a better laboratory attention score prove safe driving?

No. Safe driving is a composite functional and legal question that includes attention, hazard identification, visuomotor coordination, divided attention, decision-making, response inhibition, and actual environmental conditions. The MacQueen experiment did not measure a vehicle’s lane position, stopping distance, braking during an unexpected hazard, speed selection, collision frequency on a course, or any real crash. A narrower keyboard reaction-time consistency result cannot be substituted for an unmeasured driving outcome.

In the same way, a null change in a gambling task cannot prove judgment is unaffected in every individual. The experiment was conducted in generally healthy, screened volunteers who were not selected for adverse medication reactions, sleep debt, comorbid conditions, or frequent polysubstance use. Generalizing the group result to a different population requires separate evidence. A forensic expert could cite the paper as evidence that stimulant exposure may alter selected attention metrics while leaving other task metrics unchanged, but not as a direct measurement of impairment in the litigant.

Which design limits matter when comparing individual cases?

Several features limit generalization. First, each participant completed one treatment condition, so there was no within-person baseline comparing that participant’s own untreated and treated task responses. Individual differences were handled by randomization, covariates and group-level analysis rather than personal paired trajectories. Second, group allocation was unequal, with only 16 adults in the 10 mg arm; nonsignificant differences in smaller groups do not automatically show equivalence.

Third, the tests occurred over a relatively narrow period after a single dose. The paper does not address multiple days of medication, cumulative sleep restriction, stimulant tolerance, withdrawal, co-use of alcohol, or post-peak performance. Fourth, participants were deliberately screened to exclude several clinical and substance-use conditions, limiting inference in populations often encountered in workplace or courtroom contexts. Fifth, outcome analyses included multiple test families; correct interpretation depends on knowing which comparison survived which multiplicity adjustment.

Sixth, the key improvement is a study-specific measure derived from sequential test performance, not an everyday number such as milliseconds required to brake. Seventh, no one-to-one chemistry and performance relationship was tested. Finally, the authors’ discussion of possible norepinephrine and dopamine pathways offers a plausible mechanistic context, not direct measured transporter occupancy or proof of a particular individual’s neurochemical state.

How should counsel or an expert witness use this evidence?

The most defensible use is to distinguish a drug’s capability to change particular controlled task outcomes from a blanket assertion about either improvement or impairment. A cross-examination that reports “amphetamine improves attention” without describing 69 participants, the one-session between-group design, the corrected p = .003 for across-block reaction-time error, the aggregate D′ null, and the Iowa/Wisconsin null results would be materially incomplete. So would a categorical statement that amphetamine had no cognitive effects because most main task means were nonsignificant.

A scientifically bounded opinion can identify which findings were genuinely measured and which questions remain open. It should preserve the original dose names, separate administered dose from assayed concentration, explain the independent n, and identify the difference between a group comparison and a case-specific diagnosis. It should also disclose that a concentration, if later found in a particular case, would require its own validated method and temporal interpretation. Neither adverse event responsibility nor criminal culpability follows mechanically from this laboratory report.

Frequently asked questions

1. Was the study about methamphetamine or dextroamphetamine?

It tested oral dextroamphetamine at 10 and 20 mg against placebo. Methamphetamine and other amphetamine-related compounds are chemically and pharmacologically distinguishable. This paper did not administer methamphetamine, and its findings should not be relabeled as a methamphetamine driving trial.

2. How many independent people were tested?

Sixty-nine adults: 29 placebo, 16 at 10 mg, and 24 at 20 mg. The three successive blocks of a 360-trial CPT were repeated observations within a person, not separate participants. The design was single-session parallel-group randomization, not crossover.

3. Did the 20 mg group actually react faster overall?

Mean aggregate response time was numerically shorter in the active groups, but the source overall dose comparison for hit reaction time had p = .098, not significant under the corrected main-outcome criterion. The established result concerned how reaction-time error changed across blocks, not a validated general acceleration in milliseconds.

4. Does the significant time-course result contradict a nonsignificant overall D-prime?

No. The dose-by-block D′ interaction (p = .036) addresses trajectory over sequential task segments. The overall D′ comparison (p = .982) addresses group averages aggregated across the task. Both can be true; neither permits a claim that every participant improved.

5. Did amphetamine improve impulsivity or the ability to make gambling decisions?

Those effects were not statistically established here. The CPT false-alarm comparison was p = .982, the Iowa Gambling total-score comparison was p = .795, and the Wisconsin total-correct comparison was p = .988. Nonsignificance is not proof that all possible clinically important effects are absent.

6. Can a 10 mg or 20 mg dose be converted into a blood concentration or DUI cutoff?

No. No paired postdose quantitative forensic whole-blood or plasma concentration-response dataset was reported for these participants. Administered dose, blood concentration, response on the CPT, and statutory impairment standards are different evidentiary variables. The paper validates no individual impaired-driving concentration threshold.

7. Does this study prove that stimulant medication makes a particular person a safer driver?

No. There were no directly measured driving outcomes, no real crashes, no case-specific medication histories, and no individualized competence classifier. The most it supports is a selective group-level change in a controlled measure of sustained attention at the tested dose and time. A person-specific opinion requires evidence about that person’s behavior and relevant clinical and toxicological circumstances.

Primary study and evidence boundary

The original research is MacQueen DA, Minassian A, Henry BL, Geyer MA, Young JW, and Perry W, Amphetamine Modestly Improves Conners’ Continuous Performance Test Performance in Healthy Adults, Journal of the International Neuropsychological Society 24(3):283–293 (2018), DOI 10.1017/S135561771700090X. This is a critical interpretation of a specific experimental report, not a case-specific toxicology result, personalized medication recommendation, validated driving-safety determination, or proof of cause in an individual event.