Does Dextroamphetamine Change Task-Specific Brain Blood Flow Without Raising Global Flow? An Eight-Person PET Experiment

Answer first: In a small, double-blind crossover experiment involving eight healthy adults, an oral dose of d-amphetamine (dextroamphetamine) altered selected task-related regional cerebral blood-flow signals without producing a statistically significant change in the study’s global cerebral blood-flow measurements. Accuracy on Raven’s Progressive Matrices increased from 78.5% to 87.4% (p = .035), whereas Wisconsin Card Sorting Test accuracy was not significantly different (93.9% versus 94.3%; p = .86). Two of 32 tested brain regions had nominally significant drug-by-task interactions; a third did not reach the usual threshold. The investigators chemically measured serum dextroamphetamine in only seven of the eight people. Neither the imaging result nor the serum measurement establishes a concentration in legally collected human venous whole blood, a validated threshold for driving impairment, or the cause of anyone’s conduct in a real incident.

These distinctions make this primary experiment useful as a lesson in what an imaging biomarker can and cannot support. It provides controlled evidence that a stimulant may change how particular neural systems participate in particular tasks. It does not support the much broader assertion that stimulant exposure uniformly improves all cognitive abilities, produces a characteristic global brain-perfusion pattern, or allows a forensic examiner to infer an individual’s functional state from one chemical value.

What original primary experiment is being evaluated?

Mattay, Berman, Ostrem, Esposito, Van Horn, Bigelow, and Weinberger published the original 1996 study in the Journal of Neuroscience, under the title Dextroamphetamine Enhances “Neural Network-Specific” Physiological Signals: A Positron-Emission Tomography rCBF Study. Its enduring question is not simply whether amphetamine is stimulating. The investigators asked whether a controlled oral stimulant challenge would alter cerebral blood-flow activity differently when the same individuals performed cognitive operations thought to recruit partly different networks.

The original report is available at the study’s primary DOI record. The meaningful evidence is contained in its actual Methods, Results, four numerical tables, figure explanations, and Discussion—not in a generic summary of dopamine, clinical stimulant prescribing, or functional brain imaging. This article uses the reported group comparisons and their limitations rather than extrapolating beyond those observations.

How many independent participants were studied?

Eight people participated: four men and four women, mean age approximately 25 years and range 22–32 years. They were medically screened healthy volunteers rather than patients undergoing forensic testing, suspected intoxicated drivers, people using methamphetamine, or individuals diagnosed with attention-deficit/hyperactivity disorder. Each participant was studied on two different days, usually one to two weeks apart. Each day involved several PET observations while performing cognitive and matched control tasks.

This repeated-measures design does not create 16 or 64 biologically independent participants. The same person’s placebo session and stimulant session are intentionally compared; four PET blood-flow observations per session reflect different task conditions, not four separately recruited people. Repeated time points and pixels inside a scan are measurements nested within that person’s data. Treating every PET image, region of interest, or response trial as a new human subject would artificially magnify the apparent sample.

What was administered, by which route, and when?

Investigators administered oral dextroamphetamine 0.25 mg/kg or an oral placebo approximately two hours before the brain-flow investigation, using a counterbalanced, double-blind crossover sequence. Drug and placebo order differed among subjects. The four task conditions were also arranged to reduce systematic order effects, while each person’s task order remained the same between sessions. The protocol limited recent caffeine and nicotine and gave volunteers task practice before scanning.

The chemical identity matters. This was d-amphetamine, not oral methamphetamine, not racemic amphetamine, not the multi-salt mixture sold as some ADHD products, and not the prodrug lisdexamfetamine. Dose is the administered amount per unit body mass. It is not a measured serum value, not an analytical concentration in a forensic blood tube, and not evidence of how much parent stimulant was present at some different time outside the laboratory.

Why were volunteers injected with radioactive water?

During each scan, the team administered an intravenous bolus of oxygen-15-labelled water, an imaging radiotracer, approximately one minute after a task began. The source specifies 37.5 millicuries for the PET tracer. That intravenous injection must not be misdescribed as an intravenous amphetamine dose. All experimental dextroamphetamine was administered orally. Labelled water was used as a physiologic marker of cerebral perfusion, not to assay the ingested drug.

A PET scanner detected the transient tracer signal across 15 brain slices. The researchers used an arterial input curve, including automated arterial blood withdrawal at approximately 3.8 mL/min, to help calculate regional cerebral blood flow (rCBF). The radiotracer time-activity calculation and quantitative chemical toxicology are different procedures: arterial radioactivity data tell the imaging model about tracer passage, not the molar or mass concentration of amphetamine. A person can have substantial tracer sampling information and still have no legally relevant quantitative whole-blood amphetamine result.

What did PET directly quantify—and what remained modeled?

The PET method yielded estimated cerebral blood-flow rates, conventionally reported as milliliters per minute per 100 grams of brain tissue. For regional analyses the investigators normalized each regional value to the subject’s global mean. They drew regions of interest using corresponding magnetic-resonance images and compared the change caused by a reasoning task with the change seen during its related sensorimotor control task. Their central interaction concerned whether the effect of drug versus placebo differed between two tasks in a given region.

This is a functional neuroimaging response, not a direct assay of dopamine, norepinephrine, neuronal firing, neurotransmitter release, brain drug concentration, or cell survival. The source discusses possible monoaminergic mechanisms, but PET perfusion alone cannot separate those mechanisms definitively. Nor is an imaging region an independent subject: the reported analysis aggregated repeated measurements from the same eight people.

Which cognitive tests were paired with the images?

The Wisconsin Card Sorting Test (WCST) asks the participant to infer and then update changing rules for sorting displayed stimuli using feedback. Its visual-motor control condition uses a simpler matching response with much less abstract rule discovery. Raven’s Progressive Matrices (RPM) presents visual patterns with a missing component and asks participants to select the appropriate completion. RPM likewise had a simpler matched visual-control task.

The tasks were chosen because they have overlapping but different cognitive demands and brain-network associations. WCST requires feedback-guided set shifting and rule maintenance; RPM involves analogical pattern reasoning and extended visuospatial processing. Under the protocol, both task performance and tracer-based rCBF were observed, permitting a within-person comparison of behavioral accuracy and relative brain activation. A difference between the two tasks does not mean one is a universal assay of executive function or that the laboratory tasks reproduce the perceptual and motor demands of traffic.

What did the behavioral results actually show?

Raven’s accuracy was higher during the dextroamphetamine session than during placebo: the study reported 78.5% versus 87.4%, with a paired p value of .035. However, the number of Raven items completed did not reach statistical significance. WCST accuracy, conceptual-level responding, categories completed, and perseverative errors also did not show statistically significant drug–placebo differences. The authors pointed to extensive WCST practice and performance near a ceiling as possible reasons why a benefit, if present, might have been hard to observe.

The exact distinction is important: one tested measure of RPM correctness improved in this particular eight-person experiment. That finding does not establish that all executive operations improved, that every volunteer benefited, or that a stimulant improves safety-sensitive performance. Conversely, a nonsignificant WCST comparison is not proof that the two sessions are equivalent: with only eight individuals, the trial has limited precision for small behavioral differences.

The standard deviations in this table describe between-observation spread in reported group summaries; they must not be silently substituted for standard errors or precision bounds on paired treatment differences. No person-level performance threshold was validated against a field impairment finding.

Did overall cerebral blood flow change significantly?

No statistically significant global effect was demonstrated. The original global-flow analysis found no main effect of drug (p = .94), no main effect of task (p = .43), and no drug-by-task interaction (p = .35). The more granular matched comparisons likewise failed to establish a global difference in the WCST, its matching control, RPM, or its matching control. For example, global rCBF on WCST was 42.8 ± 1.6 mL/min/100 g after placebo and 41.0 ± 1.6 after dextroamphetamine, with p = .18.

Here the published ± values are standard errors of the mean, unlike the standard deviations for the behavioral table. That statistical label is not interchangeable. A global null is compatible with differential regional task patterns because an average across many anatomical areas can be stable while selected regions vary relative to other areas. But a significant regional comparison does not prove that the whole brain receives more blood or that stimulant treatment increases total cerebral perfusion.

Which regional PET effects were observed?

The investigators examined 32 brain regions of interest for drug-by-task interactions. The reported pattern was distinctive in two regions. In the superior portion of the left inferior frontal gyrus, dextroamphetamine increased the task-specific rCBF contrast during WCST but reduced it during RPM, relative to placebo. In the right hippocampal region, the reverse task preference appeared: the WCST contrast declined while the RPM contrast increased. These are interactions between task and drug, not measurements of parent amphetamine in either brain region.

The reported tests were F(1,7) = 16.7, p < .0046 for left inferior frontal and F(1,7) = 18.71, p < .0035 for right hippocampus. A right superior frontal interaction gave p = .053; calling that third observation statistically established would misstate the study. The 32-region search raises a multiple-comparison question: the paper’s nominal ROI p values should not automatically be characterized as a demonstrated family-wise-error-controlled result. They show the reported regional signal pattern in a small exploratory physiological experiment.

All regional values were derived from perfusion imaging and anatomical normalization. The different degrees of freedom in the source’s regional and global analyses should not be casually replaced with an assumed common complete-case count. Individual regional effect sizes with validated forensic decision rules were not supplied.

How was the stimulant chemically measured?

In a distinctly separate procedure, the investigators obtained serum samples and used a gas-chromatographic assay to quantify parent dextroamphetamine. The paper reports an analytical sensitivity of 5 ng/mL. This is how the authors described the method; the original paper does not provide enough information to reclassify that number as a modern separately validated detection limit, quantitation limit, reporting cutoff, or forensic decision threshold.

Because of technical difficulties, serum results were available in seven of the eight volunteers, not eight. At roughly two hours after oral dosing the reported serum range was 24–51 ng/mL and the mean 36.14 ng/mL. The five serial samples at approximately twelve-minute intervals were not significantly different across the scanning period, F(4,30) = 0.26, p = .90. That supports the limited observation of little detected within-session change over that sampling window; it does not establish constant concentrations before or after the protocol or identical absorption and elimination in all people.

Why does the serum result not equal a legal whole-blood result?

Serum is the liquid phase obtained after blood clots and is separated from cells. Plasma is the liquid anticoagulated fraction; venous whole blood contains cells plus liquid and is a distinct analytical matrix. Even when the same parent molecule is present in all three, concentrations cannot responsibly be translated using an assumed universal one-to-one ratio. The source did not present contemporaneous legally collected venous whole-blood dextroamphetamine concentrations, a validated serum-to-whole-blood conversion, measured amphetamine enantiomers other than the administered identity, or a retrospective conversion for an unrelated individual.

In particular, the arterial blood withdrawn for PET radiotracer input is not a hidden venous whole-blood toxicology study. Likewise, a quantified parent drug in laboratory serum is not an assay of urine, saliva/oral fluid, sweat, hair, vitreous humor, breath, or tissue. The specimen type and the method must be tied to the actual result. For a general explanation of why specimen choice changes interpretation, see the site’s published biological matrices guide.

Does the measured serum concentration prove an effect in a specific person?

No individual concentration-to-performance calibration was validated. Seven people had serum concentrations within the reported range, while behavioral outcomes and regional rCBF were evaluated using paired laboratory conditions. The existence of a modest mean association between oral dosing and selected outcomes is not a validated equation linking one person’s serum value to a predicted percentage of Raven questions correct, a WCST error rate, a PET image, or a safe or unsafe driving threshold.

The crossover design improves control of some stable person-to-person differences because the same participants experienced both conditions. Yet it cannot remove every source of practice, expectation, order, mood, learning, scan noise, and time-of-day variability. The small group, incomplete serum sampling, and narrow cognitive task set prevent precise individual prediction. The study also did not administer methamphetamine or study a mixture of ethanol and dextroamphetamine, so it cannot support additive-drug predictions outside the experiment.

What did the study report about blood pressure, pulse, and mood?

Mean systolic pressure on the dextroamphetamine visit increased from about 103.4 to 114.4 mmHg at two hours, but the paired comparison was nonsignificant (p = .09). Pulse moved from about 66 to 68 beats per minute, also nonsignificant (p = .60). The investigators reported no significant drug or task effects on arterial carbon dioxide pressure. These results should not be reinterpreted as a clinically meaningful normal vital-sign guarantee or as proof of a measured individual autonomic response.

The separate mood questionnaires had a mixture of positive and null tests. The published anxiety scale had a higher score after the drug condition, with a nominal p = .02; several POMS domains were not significant. Mood scales do not measure a forensic amphetamine concentration, a legal standard of impairment, or real-world decision-making reliability. Because the amphetamine mood-rating table contains change scores with unconventional direction and reporting, those entries require their source scoring conventions before a categorical claim about subjective happiness can be made.

Why is a region-specific effect not the same as a universal cognitive benefit?

Brain systems are recruited differently by visual analogy problems and feedback-based rule sorting. A change in the contrast for one anatomical region during one task could occur alongside an opposite change during another task and alongside no statistically established difference in the whole-brain average. That is the central empirical pattern here. The observed RPM accuracy difference gives a behavioral connection, but the much more extensive set of nonsignificant WCST endpoints constrains any headline claiming that all reasoning or all executive function improved.

One tempting error would be to turn the two nominal regional interactions into a general biological stimulant signature. Another would be to treat the nonchanging global flow as evidence that nothing physiological happened. Neither follows. The correct interpretation is conditional: under this exact oral dose and imaging protocol, these volunteers showed selected task-dependent regional perfusion differences, one tested RPM accuracy improvement, and multiple tested nulls. Whether the pattern generalizes to another dose, clinical population, repeated administration, fatigue state, or safety-critical setting remains an unanswered question.

What are the principal statistical limitations?

The study was small, with eight people and only seven measured serum profiles. Regional analyses searched 32 anatomical ROIs, a setting in which multiplicity affects confidence in isolated nominally small p values. Some tasks were highly practiced, which can restrict measurable variation. The number of completed task items and the actual accuracy values answer different questions. The published PET global analysis reports denominator degrees of freedom different from the regional analyses, so the reader should use the particular sample and model behind each estimate rather than silently assuming uniform complete-case information.

Reported p values describe evidence against specific null models; they are neither probabilities that the competing forensic claim is true nor estimates of how many drivers would be impaired. One significant cognitive test among several cannot be promoted to a pre-established universal biomarker. Equally, nonsignificant comparisons with eight volunteers should be identified as unproven differences, not affirmative proof of equivalence or absence. The report contains no prospectively validated classification threshold, sensitivity, specificity, predictive value, or error rate for deciding whether a later person is impaired.

How far can the mechanism be inferred from PET?

Dextroamphetamine affects monoaminergic neurotransmission, and the authors offered plausible neurochemical explanations for the task-specific network modulation. But the PET experiment did not measure synaptic dopamine concentration, serotonin release, norepinephrine turnover, transporter occupancy, or receptor binding during the challenge. The radiotracer was labelled water used to estimate perfusion. A regional blood-flow interaction is downstream physiological information; competing neural and vascular explanations remain possible within this design.

The authors’ interpretation that stimulant exposure accentuated differences between task-engaged networks is best read as an interpretation of the observed interaction, not a quantitative neurotransmitter assay. No imaging output can be converted into a numerically precise drug effect at a specific nerve terminal without a suitable independently validated model. Anatomical location, regional rCBF, blood parent amphetamine, behavioral test score, and clinical or legal impairment are separate constructs requiring separate evidence.

What would a defensible forensic comparison actually require?

A forensic review should first establish the legal or clinical proposition at issue. Is the question compound identity, concentration at collection, timing relative to an event, observed psychomotor performance, a particular cognitive error, or causal contribution to a collision? The Mattay experiment directly addresses none of those individualized case outcomes. Its gas-chromatographic serum measurement documents exposure under the experimental protocol, while PET and laboratory tasks document group-level physiological and behavioral outcomes under the same protocol.

For a real specimen, the expert would separately need collection records, specimen type, chain of custody, analytical validation and calibration, identification criteria, quality controls, reporting range, timing, other exposures, relevant clinical context, and contemporaneous observations. The site’s published confirmatory drug-testing overview discusses why confirmed chemical identity and a defensible interpretation of effect are different undertakings. These documents cannot be replaced by a PET study involving unrelated volunteers.

What can be said about driving, witnesses, accidents, and causation?

The researchers did not conduct on-road driving, controlled simulated driving, field sobriety testing, witness-recall validation, workplace safety outcomes, or injury/collision reconstruction. They collected no incident files. Neither Raven accuracy nor WCST responses were linked to crash responsibility or a person’s ability to safely operate a vehicle. Consequently, an assertion that a tested concentration proves impairment, nonimpairment, intent, credibility, timing of ingestion, or causation would go beyond both the direct chemical analysis and the imaging experiment.

It is equally invalid to infer that an individual who received dextroamphetamine must have improved reasoning or could not be impaired in other domains. A higher mean score on one visual-reasoning test does not exclude slower reaction time, altered judgment, sleep deprivation, unusual dosing, competing drugs, or unrelated pathology in a different setting. Those possible factors are not findings of this original study; they illustrate why individual case questions cannot be answered from a single narrowly designed experiment.

Seven evidence-based frequently asked questions

What is the scientifically supportable conclusion?

The most defensible interpretation is narrow and informative: oral dextroamphetamine modulated the regional perfusion contrast between two distinct reasoning tasks in an eight-person laboratory crossover; the Raven accuracy endpoint improved while WCST endpoints and whole-brain blood-flow tests did not show significant change. Seven participants had measured serum parent dextroamphetamine, but that evidence belongs to serum and to the experiment’s time window. It is not a legally collected venous whole-blood case measurement, a universal stimulant-performance curve, or evidence that any identified person was or was not impaired.

Primary empirical citation: Mattay VS, Berman KF, Ostrem JL, Esposito G, Van Horn JD, Bigelow LB, Weinberger DR. Dextroamphetamine Enhances “Neural Network-Specific” Physiological Signals: A Positron-Emission Tomography rCBF Study. Journal of Neuroscience. 1996;16(15):4816–4822. Read the original empirical study by DOI.