Does Repeated Dextroamphetamine Change Reward Processing Without Faster Decisions? A 22-Man fMRI Study

Short answer: In a randomized, double-blind experiment involving 22 healthy men, four exposures to oral dextroamphetamine were associated with altered functional MRI signals during different stages of a laboratory gambling task. Compared with the placebo group, the dextroamphetamine group showed lower caudate BOLD responses during decisions, higher caudate responses while anticipating rewards, and an altered amygdala response to wins versus losses. Participants also reported stronger subjective amphetamine-like effects after repeated exposure. They did not show a corresponding drug-specific improvement in gambling-task reaction time. This was a small neuroimaging experiment, not a driving study or a blood-concentration/impairment validation.

The source is O’Daly and colleagues’ original 2014 PLOS ONE empirical article, Amphetamine Sensitization Alters Reward Processing in the Human Striatum and Amygdala, available at DOI 10.1371/journal.pone.0093955. The central forensic distinction is between a measured brain-imaging contrast, a subjective drug response, a reaction time on a specific computerized task, and a chemically measured drug concentration. They answer different questions.

What is the scientifically defensible conclusion?

Repeated low-dose exposure changed how particular brain regions responded during a tightly controlled reward task, but it did not uniformly change observable behavior. The authors reported significant group-by-session effects for caudate and amygdala fMRI signals, alongside selected self-reported mood and drug-response scales. However, neither the group-by-session reaction-time interaction nor the group-by-session-by-trial-type interaction reached significance. A separate main effect of session indicated that participants generally became faster when they repeated the task, which complicates a simple interpretation of improved performance as an amphetamine effect.

These results are compatible with a hypothesis that repeated stimulant exposure changes reward-related circuitry. They do not show that a particular person became impulsive, drove unsafely, gambled recklessly, had higher synaptic dopamine on a given trial, or committed any specific act. A group-level fMRI interaction is not a validated individual forensic diagnostic rule.

Who was studied, and how many independent people were there?

The investigators recruited 22 right-handed healthy male volunteers, allocated to a dextroamphetamine group or a placebo group in a randomized, double-blind, parallel-group design. The paper describes 11 men in each group. Each individual attended four administration sessions, and scans relevant to this particular report were obtained at the first and fourth visits. The same people returned; two scans or four drug administrations per person are repeated observations, not 44 or 88 independent subjects.

The participants were screened for relevant medical history, medication use, major family psychiatric and substance-abuse histories, and recreational drug use. They gave prior-use histories using a graded self-report scale. The researchers also collected urine for previsit recreational-drug screening. That screening should not be confused with a postadministration quantitative assay of the experimentally administered dextroamphetamine. Importantly, the paper says that no standardized psychiatric interview was performed during screening. All-male selection, the small sample, and prior low recreational exposure constrain generalization to women, adolescents, individuals with stimulant use disorder, patients receiving complex prescriptions, and people involved in legal incidents.

What was administered, and why is it not a methamphetamine experiment?

Men in the active group received 20 mg oral dexamphetamine on each of four occasions; the comparison group received placebo. The first three administrations were separated by approximately 48 hours, and the fourth occurred following a two-week interval. The regimen was designed to test a model of sensitization: whether reactions following repeated intermittent exposure differ from those seen with an initial dose.

The drug is dextroamphetamine, not dextromethamphetamine. They are distinct chemical substances even though both are stimulants and can influence catecholaminergic systems. Nor is the active treatment interchangeable with lisdexamfetamine, mixed amphetamine salts, a methamphetamine laboratory confirmation result, or any person’s unknown street-drug exposure. The specific dose was fixed at 20 mg rather than individually weight-adjusted, which the authors identified as a limitation. The controlled medication administration does not establish an equivalent plasma concentration for all participants.

What was the reward-processing task?

During an approximately 18-minute functional MRI run, volunteers performed a wheel-of-fortune-style computerized choice task. Gambling trials involved a possible £1 wager, while control trials required a color choice without a stake. Three displays suggested a 25%, 50%, or 75% chance of winning. Participants had a two-second response window; a missed gambling decision carried a prescribed loss. After the choice, a delay preceded feedback showing a win or loss and an updated running monetary balance.

Crucially, the outcomes were experimentally fixed trial by trial to produce roughly balanced gains and losses, rather than letting every displayed wheel generate its own natural payout sequence. The investigators analyzed separate phases: decision-making while the subject selected an option, anticipation while awaiting the result, and outcome processing when the result was delivered. This design can isolate an imaging contrast related to those task phases, but it is not a real-world gambling propensity measurement, a standardized roadside assessment, or a validated assessment of financial decision competence.

What did the scanner measure?

The study used a 1.5-tesla GE MRI system with blood-oxygen-level-dependent (BOLD) functional imaging. The authors analyzed time-varying magnetic-resonance signals associated with local hemodynamic responses to task events, using statistical models, anatomical regions of interest, and corrections for multiple voxel comparisons. The imaging volume was repeatedly acquired during the task; those repeated volumes are measurements of the same brain, not new people.

A BOLD parameter estimate is not a dopamine concentration. A caudate signal described in arbitrary BOLD units should not be converted into ng/mL of dopamine, ng/mL of amphetamine, a percentage of dopamine release, or a directly observed quantity of reward-seeking behavior. fMRI is an indirect physiological response influenced by neuronal and vascular processes. The paper interpreted the pattern through dopamine sensitization theory, but this study did not directly sample synaptic dopamine from the caudate or amygdala.

What biological specimen was actually used for drug exposure?

The researchers referred to blood plasma amphetamine concentrations when analyzing between-visit variability and stated that concentrations during scanning were not significantly different between the first and fourth amphetamine sessions. They statistically controlled for within-subject changes in plasma amphetamine when relating subjective sensitization to imaging contrasts. The article does not provide a fully stand-alone quantitative forensic assay validation, a detailed patient-level concentration series, or a numeric blood-plasma/whole-blood conversion that would permit forensic reconstruction.

Plasma is the liquid fraction separated from whole blood. Its concentrations cannot automatically be treated as concentrations measured in legal venous whole blood. The study’s separate prestudy urine drug checks concern screening for other recent drugs, not validated quantitation of postdose amphetamine in urine. No exhaled-breath, oral-fluid, hair, or vitreous amphetamine concentration was evaluated as an impairment predictor here. For a practical explanation of why physical matrices are not interchangeable, see the published biological matrices guide.

Did the participants become subjectively more responsive to amphetamine?

Yes, for selected self-report measures. Group-by-session interactions were reported for amphetamine-like experience (p = .015), drug-induced euphoria (p < .009), activity/vigor (p = .018), and a dreamy-to-attentive mood measure (p = .019). Participants supplied these ratings on validated questionnaire families and visual analogue mood scales during scheduled assessments over four hours. They are reports about how the participant felt, not chemical measurements or contemporaneous estimates of driving skill.

That distinction matters when interpreting the word sensitization. The original paper’s behavioral evidence was stronger for the subjective response than for the selected peripheral physiological and gambling-response measures. Pulse, blink rate, and blood pressure did not show the expected sensitization pattern. An individual may report a stronger drug effect without a corresponding measurable change in one task, and the reverse can occur in other studies. Neither pattern alone supplies a validated personal safety threshold.

Did gambling choices get faster with repeated amphetamine?

No drug-specific acceleration was demonstrated by the reported interaction tests. Reaction times differed by trial type: gambling choices were faster than no-stake control choices, with a main trial-type effect F(1,20) = 30.72, p < .001. They also differed with displayed probability of winning, F(2,40) = 10.23, p < .001. But both groups showed an important repeated-session effect, F(1,20) = 49.72, p < .001, consistent with a substantial practice effect despite pretraining.

Crucially, the treatment-group-by-session interaction was not significant (p = .569); neither was treatment-group-by-session-by-trial-type (p = .974). These are not proof that every conceivable executive function remained unaffected. They are specific null findings for the interaction terms the authors actually assessed in this task. The difference between an altered imaging signal and unchanged drug-specific reaction-time interaction is the essential result, not a contradiction to be edited away.

What changed in the caudate during decision-making?

The fMRI analysis detected group-by-session interactions in decision-related BOLD contrast within the left caudate (reported Z = 3.13, multiple-comparison-corrected p = .038) and right caudate (Z = 3.04, corrected p = .050). The plotted result showed lower decision-related signal after repeated amphetamine exposure relative to the first exposure and its placebo comparison. These locations belong to striatal circuits studied in reward and action selection.

The authors themselves were cautious: changes also occurred in the placebo condition, and the prominent practice-related reaction-time change could confound how participants engaged with the decision stage. They did not find a significant group-by-session-by-reward-probability interaction during decision-making. The result therefore cannot be paraphrased as a universal reduction in an individual’s striatal function or as proof that an exposed person could no longer make an informed choice.

What changed while a reward was anticipated?

The pattern differed during the delay before win/loss feedback. The left caudate showed a reported Z = 3.15 with corrected p = .043; the right showed Z = 3.14 with corrected p = .038 for the group-by-session comparison. During this phase, caudate BOLD estimates rose in the dextroamphetamine group after repeated administration rather than falling as observed for the decision phase.

Among the actively treated participants, change in the right caudate anticipation response correlated with the change in subjective amphetamine-like experience after statistical adjustment for between-session plasma concentration differences: r = .623, reported one-tailed p = .027. A correlation in a selected ROI is evidence for an association in this small sample, not independent proof of a causal biological pathway, drug-induced aggression, addiction, excessive risk-taking, or a specific motive in court. The source did not show a significant interaction with displayed win probability for this phase either.

What happened in the amygdala after wins and losses?

The outcome phase supplied a different observation: a group-by-session interaction for the right amygdala when contrasting wins with losses (reported Z = 3.53, corrected p = .006). The authors characterized the relative win-versus-loss response as blunted following repeated amphetamine exposure. Change in outcome-related amygdala BOLD response also correlated with change in subjective amphetamine-like experience, with a reported r = .636, p = .048 after adjustment for between-session plasma amphetamine differences.

The amygdala finding does not demonstrate that losses became more psychologically painful, that a particular person became emotionally numb, or that real-world reward learning was irreversibly altered. A win-minus-loss BOLD contrast can change because responses to wins change, responses to losses change, or both do. The paper discussed competing explanations and the limited interpretability of the BOLD contrast. The experiment recorded neither actual harmful gambling decisions nor downstream legal behavior.

Primary evidence table: brain-imaging and exposure associations

These values are quoted at the level of the original statistical contrast. The reported spatially corrected p-values belong to specified imaging analyses; the two correlations involve a distinct statistical test. They should not be pooled as though they were repeated measures of one endpoint.

Primary evidence table: what changed and what did not

Both tables separate distinct model terms, so an apparent disagreement between a positive mood measure, an imaging contrast, and a null behavioral interaction does not justify discarding one result. The absence of a statistically significant interaction is also not evidence that two treatments are precisely equivalent.

Does a stronger subjective effect mean a greater quantified drug exposure?

Not necessarily. Repeated exposure can alter subjective response even when average plasma concentration during scanning does not differ materially between experimental visits. The source reported no significant first-versus-last scan plasma concentration difference, while noting within-subject variation and adjusting imaging–subjective correlations for that variation. A qualitative statement about the group’s plasma pattern does not provide the individual concentration, collection time, dose absorbed, protein binding, erythrocyte partition, or fully validated whole-blood analytical result necessary for a person’s forensic interpretation.

Different questions require different evidence. A clinical or legal specimen interpretation must specify exactly what was collected and measured; the confirmatory drug testing guide explains why an analyte-specific validated method matters. Even a quantified blood measurement cannot, standing alone, yield the exact brain BOLD contrast or subjective state found during this distinct experimental paradigm.

What does this research show about actual decision-making?

It shows task-phase-specific differences, not a broad observed deficit. Subjects had to select a color in a constrained two-second window, with prescribed probability cues and predetermined outcomes. Their recorded reaction time depends on instructions, practice, expected probability, reward context, and perhaps drug exposure. The lack of a differential session effect in this test does not make amphetamine behaviorally neutral in all settings, just as altered caudate activity does not establish poor decision quality in all settings.

Nor does the measured contrast tell whether a person consciously experienced desire for a reward or would choose to gamble outside the scanner. The paper did not collect a direct index of drug liking versus wanting; its discussion identifies that omission. The imaging outcome cannot retrospectively supply that missing measure. The appropriate inference is narrower: selected hemodynamic brain responses to standardized reward cues differed between treatment histories during this specific controlled task.

Why do source design and statistical uncertainty matter?

Eleven people per randomized arm makes estimates vulnerable to sampling variability and individual response heterogeneity. The first-versus-fourth-visit comparison cannot isolate every contributing factor—expectation, scan habituation, practice, fixed rather than weight-titrated medication, and contextual cues may all matter. The investigators attempted to standardize administration settings and scanning time, but participants did not spend the full postdose interval inside the scanner, and body weight was not remeasured on every visit. The relatively long four-second repetition time could reduce sensitivity to some imaging effects.

The group comparisons incorporated multiple-comparison correction for selected anatomical regions, which is stronger than simply labeling any uncorrected bright voxel significant. Yet region-of-interest effects in a small study remain hypothesis-generating with respect to individual legal questions. Some correlations used a one-tailed p-value; one should preserve that specification rather than re-describing it as two-tailed. The authors also did not establish a diagnosis or a validated clinical sensitization threshold. Statistical significance in a scan or a mood inventory is not a calibration standard for legal impairment.

Forensic interpretation: what is measured, modeled, and absent?

Directly administered: 20 mg oral dextroamphetamine at specified visits, or placebo. Recorded psychological outcomes: questionnaire ratings of subjective stimulation and mood. Observed task outcomes: reaction time on a particular rewarded wheel-choice task. Imaging measurements: scanner BOLD signals modeled by anatomical location and task phase. Referenced chemical exposure: plasma amphetamine concentrations used in adjustment, with no complete individual-level quantitative forensic dataset published here.

Not directly established: a person’s legal venous whole-blood amphetamine concentration, an oral-fluid or hair cutoff, a method for estimating last drug use, a specific validated impairment threshold, a measured driving maneuver, a real collision, or why an identified witness acted as they did. BOLD activity is a hemodynamic proxy rather than a direct neurochemical assay. Reported drug experience is subjective rather than a forensic blood level. The participants’ group response is not an individualized case reconstruction. See also the therapeutic drug monitoring guide for interpretation of timed plasma and serum samples without treating a concentration as a behavioral verdict.

How should the paper be used in expert analysis?

The strongest use is as an example of dissociation: repeated stimulant exposure can change selected subjective and brain-imaging measurements without a matching drug-specific change in one straightforward behavioral performance metric. In evaluating another claim, ask whether it invokes dextroamphetamine or methamphetamine, whether exposure was controlled, how many independent people were studied, what physical matrix was tested, whether the analyte and its units were specified, what task was actually performed, and whether the statistical result was significant after the appropriate correction.

If an expert invokes this paper to assert that elevated caudate activation proves intoxication, or that the absence of a reaction-time interaction proves an individual was unimpaired, neither conclusion follows. The empirical record supports limited study-specific associations. Broader interpretations require independent pharmacokinetic, behavioral, clinical, and case evidence. Any extrapolation beyond this task should be labeled as an inference, not reported as an observation.

Frequently asked questions

1. Did repeated dextroamphetamine improve gambling-task reaction time?

No treatment-specific improvement was demonstrated by the reported group-by-session interaction (p = .569). A strong overall practice/session effect occurred, so becoming quicker on a repeated computerized task should not be equated with a drug effect.

2. Did the researchers administer methamphetamine?

No. The active drug was oral dexamphetamine (dextroamphetamine), 20 mg per administration, not methamphetamine. The paper is not an assay-validation study for identifying either compound in an unknown forensic specimen.

3. Were 88 participants tested because there were four visits?

No. There were 22 independent men—11 in each treatment group—each with repeated visits. Repeated sessions and scanner volumes increase observations per participant, not the number of people recruited.

4. Did the fMRI signal measure dopamine release directly?

No. It measured blood-oxygen-level-dependent signals, interpreted through models of reward processing and dopamine-related circuitry. Neither dopamine nor amphetamine concentration can be numerically reconstructed from those arbitrary BOLD units.

5. Did the study measure actual blood amphetamine?

It referenced separated plasma amphetamine concentrations and adjusted selected correlations for between-visit differences. It did not publish a validated legal whole-blood amphetamine cutoff or a complete individual forensic assay dataset. Plasma and whole blood must not be silently interchanged.

6. Can the results establish whether one driver was impaired?

No. There was no individual road-driving assessment, driving simulator, documented collision, or validated person-level decision threshold. Controlled laboratory brain and mood group measurements cannot replace case-specific behavioral and toxicological evidence.

7. Were all physiological and behavioral endpoints significantly altered?

No. Selected subjective inventories and imaging contrasts were significant, but drug-specific reaction-time interactions were null, and pulse, blink-rate, and blood-pressure sensitization were not demonstrated. A scientifically faithful account includes both categories.

Original primary research

O’Daly OG, Joyce D, Tracy DK, Azim A, Stephan KE, Murray RM, Shergill SS. (2014). Amphetamine Sensitization Alters Reward Processing in the Human Striatum and Amygdala. PLOS ONE 9(4):e93955. Read the original full empirical paper and associated journal record: https://doi.org/10.1371/journal.pone.0093955. A journal correction in 2019 added omitted funding acknowledgment; the primary numerical results discussed here were not revised by that funding correction.