Can Amphetamine Improve Vigilance Without Improving Response Inhibition? A 24-Person Crossover Trial

Short answer: Yes—in a randomized, double-blind, placebo-controlled crossover experiment involving 24 healthy adults, single oral doses of 40 mg d-amphetamine and 100 mg lisdexamfetamine improved selected laboratory measures of vigilance and the accuracy of responses when no stopping cue was present. Neither medication significantly shortened the stop-signal reaction time, improved digit span, or changed performance across the five facial-emotion-recognition categories. Lisdexamfetamine improved digit-symbol processing speed; d-amphetamine did not significantly improve that test. A further complication is that, among participants receiving an active medication, higher modeled plasma amphetamine exposure at testing was associated with worse scores on certain tasks. The findings do not establish that either medication makes every person a safer driver, that a plasma result has a universal performance meaning, or that any measured behavioral change proves individual legal impairment.

The original primary experiment was reported by Dolder, Strajhar, Vizeli, Odermatt, and Liechti in Psychopharmacology (2018), DOI 10.1007/s00213-018-4849-0. Its measured endpoints, denominators, and biological matrix matter more than a broad label such as “amphetamine improves cognition.”

What is the central forensic question?

When someone tests positive for amphetamine or reports taking an amphetamine medication, it is tempting to move directly from drug exposure to either enhanced attention or impaired judgment. This trial illustrates why that shortcut is unsound. Researchers measured several distinct dimensions of performance, and the same treatment produced significant improvements on some dimensions, nonsignificant changes on others, and correlations in the opposite direction when individual plasma exposure was examined within an active-treatment condition.

A vigilance task that requires noticing rare events is not the same experiment as a task that requires inhibiting a response already being prepared. Neither task is direct observation of an individual operating a vehicle in ordinary traffic. The paper supports specific group contrasts under controlled experimental conditions, not a categorical rule about driving fitness, accident causation, professional competence, medication adherence, or a per-se amphetamine concentration.

Who took part, and how was the study controlled?

The investigators recruited 24 healthy adult volunteers aged 18–45, excluding many medical, psychiatric, medication, and substantial substance-use histories. Each person attended three experimental sessions and received oral d-amphetamine, oral lisdexamfetamine, and placebo in randomized order. The design was double-blind and crossover: treatment-condition observations arise from the same individuals, not from 72 independent people. Study visits began in the morning, with medication given at approximately 9 a.m.; investigators collected repeated blood samples and performed computer-based tasks several hours after administration.

Participants were not selected as patients with attention-deficit/hyperactivity disorder, sleep deprivation, a drug-impaired-driving arrest, a traumatic injury, or a substance-use disorder. The trial specifically evaluated responses in healthy non-sleep-deprived adults under laboratory conditions. It is therefore inappropriate to assume that the magnitude of its effects would transfer unchanged to a particular patient, a sleep-deprived driver, a tolerant long-term user, or someone who consumed alcohol or other drugs.

The same participants contributed to different tasks, but valid observations differed because of technical losses and instruction problems. There were 24 people for the digit-symbol task, 23 for digit span, 23 for the stop-signal task, and 21 for the Mackworth vigilance task. Reporting all outcomes as “24 per test” would overstate the precision of the smaller task-specific analyses.

What drugs and doses were actually administered?

The active conditions were a single oral 40 mg dose of d-amphetamine and a single oral 100 mg dose of lisdexamfetamine, compared with placebo. The authors describe the two active products as equimolar in terms of delivered d-amphetamine base, approximately 29.6 mg. These are administered product doses, not directly observed concentrations in any person’s brain or whole blood. The authors deliberately selected relatively high doses for their healthy-volunteer experiment; they note that 100 mg lisdexamfetamine exceeded the cited 70 mg maximum recommended daily dose.

Lisdexamfetamine is a prodrug converted into active d-amphetamine. Consequently, the named dosage in milligrams of lisdexamfetamine cannot be compared naively with the milligrams printed on a d-amphetamine product. Nor does matching their theoretical active moiety guarantee identical early time courses: activation, absorption, and concentration profiles can differ substantially while later pharmacodynamic maxima appear similar.

The study did not randomize multiple dose levels of each drug. An observed association between higher concentration and poorer task results within the active conditions is not itself evidence from a separately randomized low-dose-versus-high-dose experiment.

What specimen was analyzed for amphetamine?

Researchers obtained timed venous blood samples in lithium-heparin tubes and measured amphetamine in plasma using an ultra-high-pressure liquid chromatography–tandem mass spectrometry procedure described in the original investigation and its companion pharmacokinetic report. Plasma is the liquid component after separation; it is not equivalent to untreated forensic venous whole blood, serum, urine, oral fluid, or a hair result. These different matrices have different analytical and physiological meanings.

A further distinction is especially important here. To align amphetamine exposure with the exact timing of each task, the authors used compartmental modeling to estimate individuals’ plasma concentrations from their sampled time courses. Thus, some reported exposure–response correlations use modeled plasma estimates at the test times, not a directly assayed blood sample drawn precisely during every task. Both the physical specimen and the estimation procedure are part of the evidence.

No legal whole-blood driving threshold was evaluated. The paper did not measure blood-to-plasma conversion factors for the participants or validate an individual’s impairment probability from a single concentration. See the published biological matrices guide for why a result should not migrate across specimens without specific evidence.

Which measures actually improved?

Table 1 presents a compact selection of the authors’ original Table 2 findings. Values are reported as the study’s means with standard errors of the mean (SEM), not population reference ranges. The final column identifies the reported main drug-effect test; a particular medication’s comparison with placebo is distinguished in the adjacent interpretation. “Not significant” means that this study did not demonstrate that tested contrast under its analysis—not that a change of exactly zero was established.

These are meaningful changes on a ten-minute vigilance task and certain response measures. They are not the same as improved braking decisions, detection of actual roadway hazards, error-free complex work, or absence of unwanted pharmacologic effects. The two active drugs also increased subjective ratings of stimulation and alertness, making the contrast between feeling alert and measured task-specific performance especially important.

Why is go-trial accuracy not evidence of improved stopping?

The stop-signal paradigm contains two competing processes. On ordinary “go” trials, the subject must select and execute the appropriate response. On the subset of “stop” trials, a late signal instructs the person to cancel that response. Go-trial accuracy is an indicator of correct execution when responding is required; stop-signal reaction time estimates the latency of the inhibitory process under the task’s model.

Both active medications improved go-trial accuracy from a placebo mean of 96% to 98%. However, the estimated stop-signal reaction times of 216, 211, and 205 milliseconds did not differ significantly in the reported analysis. Accuracy on stop trials likewise showed no statistically significant main drug effect. Calling this experiment proof that amphetamine “improves inhibition” would confuse a significant accuracy result from one trial type with an unsupported result from another.

This distinction can matter when evaluating a claim about failure to stop, an allegedly impulsive movement, or divided attention. One laboratory metric cannot adjudicate the real-world event without evidence about the person’s specific task demands, exposure history, baseline function, and contemporaneous behavior.

What did the vigilance test show—and whom did it include?

The Mackworth Clock Test required subjects to detect rare deviations in a regularly progressing visual sequence. The trial used a shortened ten-minute administration, placed late in the testing battery. Three participants’ data were unusable because one did not complete the test and two misunderstood instructions, leaving 21 independent people for this outcome. The test included correct detections, misses, and false alarms, so its apparent benefit was not defined by a single response speed.

Both active medications increased accuracy, reduced missed critical signals, and reduced false alarms compared with placebo. However, because the trial was conducted during a limited laboratory period, the result does not demonstrate sustained benefit across an overnight shift, a long-distance drive, a multihour trial, or a complex task requiring coordination with other people. Fatigue and practice history can interact with drug effects in ways the study was not structured to isolate.

A particularly revealing feature emerges when those between-condition improvements are compared with the participants’ variation in modeled plasma exposure within treatment. Better average performance with active medication does not necessarily mean that a higher plasma value within an active condition predicts better performance.

Did the medication improve working memory or facial-emotion recognition?

No statistically significant improvement was detected for digit-span performance: the mean total scores were 15.39 with placebo, 15.83 with lisdexamfetamine, and 15.26 with d-amphetamine. Digit span and digit-symbol coding test different constructs, even though both are colloquially called cognitive tests. Faster symbol coding does not demonstrate a larger capacity to retain or reproduce the sequence used in digit span.

The Facial Emotion Recognition Task also showed no significant treatment effect for recognition of neutral, happy, sad, angry, or fearful facial expressions. For example, the original table lists fear-recognition mean percentages of 52.4, 50.8, and 54.9 across placebo, lisdexamfetamine, and d-amphetamine, respectively, with the drug main effect not significant. This is not a validated demonstration that every participant’s emotion perception was unchanged; it is a group-level null for that particular task and test power.

For an eyewitness, clinician, or attorney, facial-affect labeling in a computer test is not interchangeable with reading threats, remembering voices, deciding whether to intervene, or correctly reconstructing the intentions of a person during a specific episode.

What happened to empathy and subjective experience?

Selected social-cognition ratings changed while many others did not. d-Amphetamine increased direct emotional-empathy ratings for positively valenced stimuli, but the overall direct-empathy measure and responses to negative stimuli were not significant. Neither active condition changed the cognitive-empathy accuracy measure significantly. Some pleasantness and attractiveness ratings in the picture-rating task increased, including for neutral stimuli; the authors did not demonstrate a content-specific change that would explain actual sexual or interpersonal conduct.

Meanwhile, both active medications produced much larger subjective reports of stimulation, alertness, and positive drug effect. Table 2 juxtaposes selected subjective and social outcomes. It is important not to confuse a self-rating on a 0–100 visual analog scale with physiological plasma exposure or an objective performance test.

Not every significant image-rating contrast was an effect on sexuality, empathy, or socially consequential choices. The study measured responses to standardized images under a particular instrument. It did not observe relationships, consent decisions, violent behavior, criminal intent, or reliable person-specific truthfulness markers.

Did lisdexamfetamine and d-amphetamine reach plasma at the same time?

The paper shows a timing difference consistent with the prodrug mechanism. Peak plasma amphetamine levels occurred at approximately 3.3 hours after d-amphetamine and 4.6 hours after lisdexamfetamine. The companion pharmacokinetic analysis, referenced by the authors, estimated a mean 0.6-hour longer lag and a mean 1.1-hour later peak after lisdexamfetamine, with considerable between-person variation. These figures describe the studied products and schedule, not a universal onset time for any particular individual.

During the earlier emotion-recognition and empathy tests at approximately three to three-and-a-quarter hours, modeled plasma amphetamine levels were higher after d-amphetamine than after lisdexamfetamine; later, during several cognitive tests, the contrast was small or absent. The timing of an observation therefore matters. A laboratory finding collected three hours after one product cannot automatically be compared with a different product measured later without considering absorption, prodrug conversion, and testing conditions.

The source physically collected timed blood specimens and then modeled test-time plasma exposure. It was not designed to establish a validated forensic “hours since dose” estimator, does not provide case-specific metabolic phenotyping, and did not measure active amphetamine in every subject’s legal forensic whole-blood specimen at an alleged incident time.

How could higher plasma amphetamine correlate with poorer performance?

This was one of the study’s most consequential qualifications. During the digit-symbol task, higher modeled plasma amphetamine exposure correlated with a smaller placebo-adjusted increase in correct answers: reported correlation coefficients were approximately −0.47 for lisdexamfetamine (p<.05, n=24) and −0.53 for d-amphetamine (p<.01, n=23). During the vigilance task, higher exposure also related to less improvement in some measures. For example, in the active-treatment conditions the authors reported positive associations of plasma concentration with false alarms (r≈.50, n=21, and r≈.49, n=20; both p<.05), meaning more false alarms at higher exposure within those comparisons.

The key distinction is between two different statistical questions. The first compares average outcomes after active treatment versus placebo. The second asks whether people with a higher estimated concentration within an active condition had larger or smaller placebo-adjusted changes. Significant improvement in the first analysis can coexist with an adverse concentration–performance slope in the second. Collapsing both into a slogan—“amphetamine boosts performance” or “higher concentration always impairs”—misstates the evidence.

The authors discuss an inverted-U possibility: moderate stimulation may benefit some functions, while greater exposure may offer less benefit. But because each product was tested at one active dose rather than a randomized series of doses within a person, the study cannot define an individualized optimal dose or concentration. Between-participant pharmacokinetic and baseline differences can contribute to exposure–effect correlations. Correlation does not establish the mechanism or predict a new individual’s score with forensic precision.

Does an improved laboratory score mean someone was not impaired?

No. “Impairment” is not a single universal variable. A person can be more vigilant on one simple laboratory task yet show no improvement on inhibiting a response, on working memory, or on a divided-attention decision. The present paper did not use a driving simulator, naturalistic road observation, police sobriety assessment, accident reconstruction, or a validated binary classification of individually impaired versus unimpaired drivers.

Similarly, a normal-looking digit-span result is not proof of no meaningful pharmacologic effect. The participants reported marked subjective stimulation, and performance depended on task design and analysis. Test conditions offered controlled instructions, repeated task exposure, and laboratory monitoring that differ from actual emergencies, traffic, interpersonal encounters, and complex professional decisions.

A forensic report should specify exactly what an assessment measured, when it was administered, what kind of person was examined, and which outcome was significant. It should not use general stimulant pharmacology to fill gaps in a contemporaneous case record.

Can a positive drug result identify the dose or timing in a case?

Not from this trial alone. The researchers knew which products and doses were administered and collected serial samples under supervision. Real-world testing frequently begins with one sample taken after an uncertain interval, sometimes after medical treatment or additional substances. An identified amphetamine molecule does not reveal whether it originated from lisdexamfetamine, d-amphetamine, a different amphetamine preparation, or another relevant source without a supported analytical and history-based explanation.

One concentration also cannot reveal how many tablets were taken, the precise clock time, the subject’s baseline tolerance, or a performance score. Sampling matrix, specimen timing, analytical specificity, metabolism, dose formulation, and concurrent drugs all affect the interpretation. A mass-spectrometric result can reliably answer an identification or quantitation question within its validation; it is not automatically an individual event reconstruction. The published confirmatory drug testing guide explains the analytical portion of that distinction.

What would a stronger case-specific assessment require?

First, identify the exact medication formulation and administration history, while distinguishing a documented dispensed dose from one reported by the subject or inferred from testing. Second, inspect the original toxicology report, collection time, specimen type, units, compound identity, quantification range, and quality controls. If plasma was measured, do not silently relabel it as whole blood; if a concentration was interpolated from a model, describe it as estimated rather than directly sampled at that instant.

Third, independently evaluate the behavior alleged to matter: failure to stop, lane position, divided attention, sustained vigilance, impulsive response, or memory. There should be a valid connection between the behavioral claim and the task actually tested. Fourth, examine credible non-drug explanations such as fatigue, visual conditions, acute illness, distraction, training, speed, or task complexity. Finally, identify which case-specific evidence supports an inference about causation and what remains unknown. The study’s group statistics cannot replace those steps.

Where results may inform prescribed medication management rather than legal responsibility, the published therapeutic drug monitoring guide sets out why clinical dose decisions and forensic questions should not be conflated.

What limits the strength and generalizability of this experiment?

The biological sample comprised only 24 selected, generally healthy adults, with as few as 21 valid observations for some tasks. The crossover design strengthens within-person treatment comparisons but does not turn repeated sessions into independent participants. One relatively high dose of each active product was studied, not a titration series. The order and spacing of tests can influence performance, and several tasks have learning effects or ceiling constraints.

Many distinct outcomes were examined across subjective effects, emotional recognition, empathy, image ratings, processing speed, memory, stopping, and vigilance. Individual nominal significance tests should be interpreted in light of this broad testing battery and the original analysis rather than treated as 100% certain, independent discoveries. Null findings may also reflect inadequate power to detect small changes rather than biological equivalence.

Plasma amphetamine was objectively assayed in timed samples, but task-time values were fitted by a pharmacokinetic model as described by the authors. The paper’s measured outcomes were laboratory tasks, not observed road behavior or case-specific injury mechanisms. No participant-level diagnostic cutoff for impairment or lawful driving was validated. The timing and size of effects after single doses in non-sleep-deprived people cannot automatically be extrapolated to long-term therapy, combined drug exposure, clinical intoxication, or a later event.

What conclusion can this study legitimately support?

It supports a differentiated result: in a 24-person randomized crossover, single 40 mg oral d-amphetamine and 100 mg oral lisdexamfetamine improved selected measures of vigilance and go-trial accuracy, while neither improved measured stop-signal inhibition or digit span significantly. Both produced pronounced self-rated alertness and stimulation; facial emotion recognition was not significantly altered. Lisdexamfetamine improved digit-symbol correct responses, and higher modeled plasma exposure within active conditions sometimes correlated with less cognitive benefit.

It does not support a universal drug-concentration-to-impairment equation, an inference that every exposed person performed better or worse, or a conclusion about the cause of an individual accident. The highest-value forensic lesson is to keep the administered medication, physically measured plasma, model-estimated exposure, named task, group-level statistical result, and individual legal question in separate evidentiary categories.

Frequently asked questions

1. Did the study show that amphetamine always improves attention?

No. Both active products improved a short Mackworth vigilance task and go-trial accuracy in a selected healthy sample, but other measured domains showed no significant change and higher modeled exposure was associated with less improvement on some tasks. One favorable test does not characterize all attention functions.

2. Did either drug significantly improve stopping a response?

No. Stop-signal reaction time was not significantly affected, although go-trial response accuracy improved. These are separate variables within the task and should not be labeled interchangeably as response inhibition.

3. How many independent participants were actually tested?

Twenty-four people completed the crossover experiment, with task-specific valid counts of 24 for digit-symbol substitution, 23 for digit span and stop-signal measurements, and 21 for the vigilance task. Three sessions per person do not create 72 independent volunteers.

4. Does a plasma amphetamine measurement equal whole-blood amphetamine?

No. This study physically analyzed plasma and used modeled plasma concentrations to align exposure with certain task times. It did not validate direct conversion to forensic whole-blood concentrations or an individualized behavioral threshold.

5. Why did higher estimated plasma levels sometimes correspond to worse performance?

The authors observed inverse within-active-condition associations for some task measures, consistent with the possibility of diminishing benefit at higher exposure. Because the trial used one active dose per drug, these correlations do not prove a causal optimum or predict a person’s individual capacity from a single plasma result.

6. Did the two amphetamine products affect all domains equally?

No. Both improved vigilance and go-trial accuracy. Lisdexamfetamine, but not d-amphetamine, significantly increased digit-symbol correct responses compared with placebo. Neither demonstrated a significant drug effect on digit span, stop-signal reaction time, or recognition of the five tested facial expressions.

7. Can this experiment determine whether a particular driver was impaired or responsible for a crash?

No. It included no real driving, collision reconstruction, case-specific witness assessment, or validated individual impairment classifier. Any case analysis requires timed, validated specimen results and independent behavioral and event evidence, with alternative explanations considered.

Original primary publication

Dolder PC, Strajhar P, Vizeli P, Odermatt A, Liechti ME. Acute effects of lisdexamfetamine and D-amphetamine on social cognition and cognitive performance in a placebo-controlled study in healthy subjects. Psychopharmacology. 2018;235:1389–1402. Original full primary Methods, Results, Tables 1–2, Figures 1–3, and Discussion: doi:10.1007/s00213-018-4849-0. Results of studies cited by these authors are not presented here as if observed in this experiment.