Does Dextroamphetamine Shift Effort Choices Most When Reward Is Unlikely? A 17-Adult Trial

Short answer: In a controlled study of 17 healthy adults, oral d-amphetamine increased selection of a more effortful, potentially higher-paying button-press task. The increase was most apparent when the displayed probability of payment was low or intermediate. It did not significantly change how reward magnitude influenced choices, and the observed increase in button-pressing speed did not independently explain high-effort selection. These were repeated laboratory choices following nominal 10 mg and 20 mg doses of d-amphetamine—not observations of methamphetamine use, measured blood amphetamine concentrations, driving ability, or a person’s propensity to take dangerous risks.

The evidence comes from Wardle, Treadway, Mayo, Zald, and de Wit, whose original human crossover experiment appeared in the Journal of Neuroscience in 2011. The primary article is available under DOI 10.1523/JNEUROSCI.4387-11.2011. The critical distinction is between a change in willingness to perform a demanding laboratory action for uncertain money and a scientifically validated conclusion about any particular person or incident.

What did the investigators actually test?

They asked whether a controlled stimulant challenge would change a person’s tradeoff between effort, reward size, and probability of obtaining the reward. Every choice required weighing an easier, smaller potential payment against a harder, larger potential payment. A participant’s selection of the harder option was treated as an observable task choice, not as a direct measurement of personality, reckless driving, criminal intent, or chemical intoxication. Importantly, the experiment did not isolate risk preference independently from physical effort. The difficult task took more button presses and usually more time; the reward itself was uncertain for both the easy and hard options.

The hypothesis arose from work on dopamine and effort-related choices, but the human experiment administered a systemic drug rather than selectively manipulating a single brain circuit. D-amphetamine also changes signaling outside dopamine systems. The results are therefore best described as acute effects of the administered agent on this particular choice procedure, rather than definitive proof that one neurotransmitter or one brain region caused the behavior.

How many independent people participated?

The analyzed sample consisted of 17 independent adults, of whom six were men and eleven women. Participants were healthy volunteers, predominantly young adults; the table in the original paper reports a mean age of 22.7 years with a standard deviation of 3.5 years. The study recruited this subgroup from a larger program of drug and emotion research. This matters for inference: it was not a representative sample of licensed drivers, patients with substance-use disorder, people with ADHD, emergency-department patients, or persons being evaluated in a forensic investigation.

The same people returned for three sessions. Those visits are repeated observations within participants—not 51 independent adults—and choices within a session are repeated trials nested within people. Three individuals lacked task data from one session because of a computer malfunction. That incompleteness is relevant to analysis and precludes treating every participant as having a perfectly balanced three-visit record. The trialwise statistical method modeled within-person correlation, but it does not erase the uncertainty created by a small independent sample.

Which drug, dose, route, and timing were actually used?

Each volunteer attended three double-blind, counterbalanced conditions: an oral placebo, 10 mg oral d-amphetamine, and 20 mg oral d-amphetamine. Sessions were spaced by at least three days; the original report describes the total observed separation as ranging from three to 113 days. Capsules were administered at approximately 9:30 a.m., and the effort-choice task began at about 12:30 p.m., after other laboratory procedures. These facts describe nominal dosing and study scheduling; they are not measured pharmacokinetic concentrations in individual specimens.

D-amphetamine is not the same compound as d-methamphetamine, even though their names and some pharmacological actions are related. The Wardle study did not give methamphetamine, lisdexamfetamine, methylphenidate, or a mixed amphetamine-salt product as its experimental active conditions. Substituting one of those exposures would misidentify the study. There also is no demonstrated conversion from these oral milligram doses into any participant-specific forensic whole-blood amphetamine value.

How did the effort-and-reward task work?

The Effort Expenditure for Rewards Task, abbreviated EEfRT, repeatedly offered an easy versus hard button-press choice. To complete the easy task, the participant made 30 presses with the dominant index finger within seven seconds, with a potential payment of $1.00. For the harder option, the participant attempted 100 presses with the nondominant little finger within 21 seconds. The hard option offered a variable potential payment from $1.24 to $4.30. A choice was followed by actual task performance, then feedback about completion and whether a designated winning trial occurred.

On each trial, the display gave a probability cue of 12%, 50%, or 88%. Those probabilities applied to the possibility of winning after either choice. Successful task completion made the participant eligible for the specified payment; it did not guarantee payment. The experiment therefore mixed motor effort, opportunity for gain, uncertainty, and the effort cost of completing the choice. A person could choose the difficult option for several reasons. It would be incorrect to rename the percentage of difficult choices an individual “impairment score.”

Why are the 12%, 50%, and 88% cues central?

Probability was not an afterthought. The investigators specifically asked whether drug effects differed when a reward was relatively unlikely. In the lowest-probability condition, a participant could choose a demanding task while having only a 12% chance of being paid on a successful trial. At 88%, winning was relatively likely. Because a larger payment can still have low expected value at low probability, the design could probe a form of cost tolerance that a simple maximum-speed button test cannot capture.

The task’s expected value was computed from the potential hard-task reward multiplied by the displayed win probability. This is a mathematical description of the stated choice, not a pharmacokinetic prediction, a blood concentration, or a general measure of financial prudence. Both the magnitude and probability cues independently influenced task selection, and their combined expected-value effect entered the authors’ trial-level models.

What was the overall d-amphetamine effect on hard-task choice?

The investigators’ main generalized estimating equation analysis associated increasing drug condition with a higher likelihood of selecting the hard, higher-potential-reward option: the model-2 drug coefficient was 0.138, standard error 0.03, and reported p = .004. The study also reported a clear contrast between the 20 mg and placebo conditions (p < .001) and between the 20 mg and 10 mg conditions (p = .012). In contrast, the 10 mg versus placebo pairwise analysis did not cross a conventional .05 threshold (p = .085).

These numbers support a dose-associated shift in the observed choice procedure; they do not prove that every subject changed, that all people would respond similarly, or that the same effect applies at untested doses. The regression coefficients belong to a model of correlated binary task choices, not to milligrams per milliliter, percentages of blood alcohol, or an individual’s absolute probability of harmful behavior.

Primary evidence table: what changed, and what did not?

Values here are taken from the original trial-level results and table, not reconstructed from a plot or a later review. Several coefficients have different adjustment structures because they come from separate models. They must not be added together or treated as mutually independent effect estimates.

Was the drug effect concentrated in low-probability choices?

Yes, within the limits of this task. The drug-by-probability interaction was significant (p = .011). Follow-up analyses reported a drug-related increase in difficult-task selection at 50% probability (b = 0.178, p = .001) and at 12% probability (b = 0.262, p < .001). The increase was not statistically significant in the 88% probability condition. This result is more specific than saying that the stimulant simply made participants press faster or value every large reward more.

A low probability here is a computer-generated chance of earning the stated payment, not the probability of a real collision, a medical complication, or an arrest. The researchers also noted that the nonsignificant high-probability comparison was unlikely to reflect complete response saturation, because difficult-task selection under the 88% cue was not at 100%. No direct on-road risk or accident outcome was tested.

Did the result track expected value rather than payment size alone?

A separate model indicated that the drug-related choice effect differed by expected value: drug × expected-value coefficient −0.134, p < .001. A median split used in follow-up analyses showed a significant drug-associated increase on lower expected-value trials (b = 0.179, p = .001). On higher expected-value trials, the estimate was near zero (b = 0.007, p = .844). This is compatible with increased willingness to pursue a difficult and uncertain option when the computed expected return was less attractive.

Nevertheless, these probability, expected-value, and reward-magnitude variables are mathematically and behaviorally linked. The analysis does not isolate a single pure mechanism called “risk seeking.” A person may choose to work harder for an unlikely payment for reasons that would not translate to traffic decisions, interactions with police, or courtroom credibility.

Why did reward magnitude fail to explain the stimulant effect?

Reward size certainly influenced choices during the task: larger potential hard-task payments predicted more hard-task selections in the baseline behavioral model. But the drug-by-reward-magnitude interaction was not statistically significant (b = −0.056, p = .158). In practical terms, reward magnitude still mattered while the drug did not demonstrably change the slope relating reward size to choices. A finding about the interaction does not mean money was irrelevant or that all reward-related brain processes were unchanged.

This separation is important when interpreting testimony that a stimulant made someone more susceptible to money, less concerned about consequences, or necessarily more motivated by a specific payment. The experiment gives a constrained statement about choices between two defined button-press requirements. It does not provide a validated monetary valuation scale suitable for attributing motive in an individual legal matter.

Could faster finger movements alone explain the choices?

The active drug conditions sped button pressing, so the investigators directly considered a motor-speed explanation: perhaps the hard task felt physically easier when fingers moved faster. However, button-press speed was not a significant predictor of choosing the hard option in the reported control analysis (coefficient 0.000237, p = .348). The drug effect on completion rate was also nonsignificant (b = −0.036, p = .327). These null findings narrow, but do not completely eliminate, motor-performance explanations.

A null coefficient within a small experimental sample is not proof that psychomotor speed and effort evaluation are biologically unrelated. The authors’ narrower conclusion was that measured tapping speed did not appear to be the main statistical explanation for the change in choice under the tested conditions. No driving simulator, hazard-perception test, braking-response test, or divided-attention road task was administered as part of this report.

Second evidence table: probability, expected value, and motor controls

This table describes study models, not independently verified clinical decision thresholds. The high-probability row faithfully states that the analysis was nonsignificant without assigning an invented p-value. Repeated task trials and incomplete visits remain nested within just 17 people.

What specimen was measured: dose, screening, or blood drug level?

The exposure described in the paper is the oral d-amphetamine dose administered. Entry and visit safety procedures included urine drug screening and a breath-alcohol test, but the published choice analysis did not report serial quantitative amphetamine measurements in venous whole blood, physically separated plasma or serum, oral fluid, hair, urine drug concentrations, or exhaled breath amphetamine. A 10 mg capsule does not establish a particular blood value. Neither a drug-screen result nor the number of elapsed minutes after dosing substitutes for chemical confirmation of a specific analyte in an identified specimen.

For an actual forensic toxicology comparison, the analyst would need to know the sampled matrix, collection chronology, preservative and tube, analyte identity, validated method, concentrations and units, calibration range, measurement uncertainty, and relevant biological variability. None of those individual event-specific pharmacokinetic facts can be inferred from this study’s nominal milligram treatment arms or regression results.

How should investigators distinguish d-amphetamine from methamphetamine?

A forensic report should explicitly identify which chemical was administered or measured. The active medication here was d-amphetamine. It was not administered as methamphetamine, and the experiment did not analyze a methamphetamine-positive patient or a forensic blood specimen. Drug names may share “amphetamine” while describing different molecular structures, formulation pathways, pharmacokinetics, and source interpretations. The task effect therefore cannot be relabeled as a measured methamphetamine effect without independent evidence.

Similarly, lisdexamfetamine is a prodrug converted to d-amphetamine, but this experiment did not administer lisdexamfetamine. A legitimate comparison with stimulant medication, overdose, workplace exposure, or toxicology testing must use the source trial’s actual molecule, route, nominal dose, and endpoints. It must also separate drug use from measurable drug exposure and measurable exposure from functional impairment.

What does the generalized estimating equation model contribute?

The task created many binary choices per person. Generalized estimating equations, or GEEs, allowed the researchers to analyze choices while accounting for correlation among repeated observations within participants. The models included potential reward magnitude, displayed win probability, expected value, and trial number. Interactions asked whether drug-related changes differed according to these task variables. A reported interaction p-value is not itself a diagnostic classification rate for whether a person was intoxicated.

There were five reported GEE model configurations, and the interpretation depends on which variables and interactions were included. The 0.138 drug main-effect estimate is not the same model term as the −0.134 interaction with expected value. The observational unit for the statistical calculation may be a trial, but the number of biologically independent participants remained 17. Thus, narrow inferential confidence is more defensible than claims of broad generalizability to all people or doses.

Which limitations matter most in an expert evaluation?

First, the study was small, with 17 independent healthy young volunteers and some missing visit-level task data. Second, the trial only tested two active oral doses and one laboratory reward task. Third, the hard and easy options differed in muscle effort, finger choice, response duration, and money, making it difficult to localize the drug effect to one underlying psychological feature. Fourth, expectancy was managed with blinding, but participants could experience subjective stimulant effects. Fifth, the experiment used systemic pharmacology, not a selective measurement of synaptic dopamine in a named human brain region.

Finally, the study had no quantitative amphetamine pharmacokinetics, no directly measured functional driving performance, no clinically injured subjects, and no follow-up linking difficult-task selections to later acts. These are genuinely unmeasured outcomes, not nonsignificant outcomes with p-values. A factfinder should not be presented with a fabricated negative test result or a claim that these variables were ruled out experimentally.

How might the finding be used responsibly in forensic testimony?

The study supports the limited general proposition that an acute oral d-amphetamine challenge can shift the allocation of effort among computer-mediated monetary options, particularly under relatively low displayed reward probability. It does not establish a person’s drug concentration, exact ingestion time, legal impairment, crash responsibility, dangerousness, truthfulness, or ability to form intent. Nor does increased willingness to select a hard button task mean decreased safety judgment on a road or increased aggression in an interpersonal encounter.

Expert testimony should distinguish the protocol and reported statistical comparisons from any case-specific inference. A proper foundation would identify actual drug and formulation, the person’s biological test and specimen matrix (if available), the time relationship between dosing and alleged conduct, the behavior actually at issue, and competing explanations. Without such additional evidence, treating the population-level EEfRT choice result as a validated predictor of an individual act would overstate its scope.

What would a careful cross-examination establish?

An appropriate inquiry would establish the precise active compound and oral dose; that only 17 adults were independent participants; that three sessions were not three separate cohorts; that the choice outcome was a hard-versus-easy button task with money and computer-displayed probabilities; and that the important positive interaction involved reward probability, not reward magnitude. It should also establish that some comparisons were explicitly nonsignificant, including 10 mg versus placebo, the reward-size interaction, high expected-value follow-up, and motor-speed predictor.

The same inquiry should establish that the paper collected eligibility-screening information without reporting postdose quantitative amphetamine whole-blood concentrations, and did not test licensed driving, actual crash risk, or individual legal impairment. These are distinctions about what the researchers observed and did not observe; they are not a claim that all psychostimulant effects are harmless.

Frequently asked questions about d-amphetamine and effortful reward choices

Did 10 mg of d-amphetamine significantly increase hard-task choice relative to placebo?

The reported 10 mg versus placebo comparison had p = .085, which did not meet the conventional .05 threshold. The 20 mg condition showed significant contrasts with placebo and 10 mg. A trend in one small sample should not be described as a proven 10 mg effect for every individual.

Was the effect greater when winning money was unlikely?

Within this EEfRT design, yes. The drug-by-probability interaction had p = .011, with reported drug-associated selection effects in the 12% and 50% probability conditions. The 88% comparison was not significant. These values describe computer-generated payment probabilities, not accident probabilities.

Does the study prove that d-amphetamine makes a person take more dangerous real-world risks?

No. The outcome was choosing to perform a demanding finger-press task for potential money. No road behavior, violence, criminal event, actual risky investment, or individual harm was directly measured. The task mixes work effort, payment probability, and reward size.

Did larger potential payments become more influential after the stimulant?

The drug-by-reward-magnitude interaction was nonsignificant, b = −0.056, p = .158. Reward magnitude influenced choices generally, but the investigators did not demonstrate a stimulant-specific increase in the effect of larger payment amounts.

Could the effect simply be faster button pressing?

D-amphetamine increased motor speed, but speed was not a significant statistical predictor of hard-option choice in the reported analysis, p = .348, and task-completion differences were also nonsignificant, p = .327. That limits a simple motor explanation without proving zero biological contribution.

Did the researchers measure blood amphetamine or methamphetamine at task time?

No quantitative postdose whole-blood, plasma, or serum amphetamine concentrations were reported in this experiment. The intervention was oral d-amphetamine, not methamphetamine. Eligibility drug screening and alcohol breath screening are different from measuring postdose blood analyte concentrations.

Can an expert use this paper to calculate a person’s intoxication, impairment, or crash cause?

No validated calculation follows from the reported task model. For a case-specific assessment, an expert would need independently established exposure, specimen matrix and concentration where relevant, collection and event times, validated functional evidence, and alternative explanations. The paper provides a constrained laboratory observation, not an individual forensic diagnostic rule.

Bottom line for the scientific record

A 17-person, counterbalanced crossover experiment found that acute oral d-amphetamine shifted willingness to select a difficult monetary task when the probability and expected value of reward were low. At the same time, an interaction with reward magnitude was nonsignificant, button-pressing speed did not account for the choice effect, and neither actual postdose blood drug concentrations nor meaningful driving outcomes were measured. That combination—positive effects and bounded null results—is the useful forensic lesson. Treat the result as evidence about one experimental choice process, not as proof of what any particular person thought or did.

Primary study: Wardle MC, Treadway MT, Mayo LM, Zald DH, de Wit H. Amping Up Effort: Effects of d-Amphetamine on Human Effort-Based Decision-Making. Journal of Neuroscience. 2011;31(46):16597–16602. DOI 10.1523/JNEUROSCI.4387-11.2011.