Can Attention Lapses Predict D-Amphetamine Liking? Why Two Impulsivity Tests Pointed in Opposite Directions in 165 Adults

Short answer: In a double-blind, four-session laboratory experiment involving 165 analyzable healthy adults, more attention lapses during a placebo session predicted less reported liking, feeling and desire for more d-amphetamine after 20 mg. Slower placebo-session stop-signal reaction time—indicating poorer response inhibition—predicted more positive subjective drug response. Those opposing associations mattered most at 20 mg, not at 5 mg. The study measured computerized performance and self-reported drug effects; it did not quantify amphetamine in forensic blood, establish an individual addiction outcome, or test driving impairment.

These results come from Weafer and de Wit, Inattention, impulsive action, and subjective response to d-amphetamine (2013), an original human experimental analysis. The study helps distinguish different components commonly bundled under the word “impulsivity.” It does not establish a universal prediction rule for an individual, nor does a higher pleasure rating prove later misuse. The primary publication is available at DOI 10.1016/j.drugalcdep.2013.05.021.

What exactly was the scientific question?

The researchers asked whether two placebo-session behavioral characteristics predicted how positively volunteers would describe d-amphetamine after receiving controlled oral doses. One characteristic was variability in a simple response task, interpreted as lapses in sustained attention. The other was the time required to inhibit a response when a stop signal unexpectedly appeared. These are not interchangeable constructs. Someone can occasionally lose attention yet inhibit an initiated response efficiently; another person may remain attentive but struggle to stop an already prepared action.

The distinction is crucial because earlier discussion of stimulant reward often treated impulsivity as one general personal trait. A single broad label hides mechanisms that may have different relationships with self-reported stimulation or liking. The trial offered a within-participant comparison against placebo while statistically accounting for age, sex and the order of the placebo visit. That design provided more disciplined evidence than an uncontrolled survey of whether “impulsive people like drugs.”

Who actually participated, and why does the denominator matter?

A total of 198 volunteers completed the broader experiment. Valid attention-lapse observations remained for 192 after missing observations and unusually extreme scores were excluded. Valid stop-task observations remained for 170 after missing measurements, outliers and failure to meet task-accuracy requirements. Only 165 people had the data needed for both predictors and the main analyses; the relevant independent sample is therefore 165 humans, comprising 89 men and 76 women, with a mean age of 23.5 years and standard deviation of 3.4 years.

Eligibility was deliberately restrictive: volunteers were 18–35 years old, met body-mass-index and education requirements, had no recent DSM-IV diagnosis, no lifetime history of substance dependence and no serious medical condition. The population was restricted to Caucasian participants because the experiment originated in a genetic-response research program. These criteria reduce some sources of variation, but they also limit generalization. Findings in healthy young adults should not be presented as though directly validated in people with stimulant-use disorder, ADHD, older drivers, hospitalized patients, or forensic defendants.

Each eligible person attended four sessions, but the biological sample size was not 660 distinct people. The four observations belonged to the same individual. Likewise, repeated questionnaire scores and computerized trials provide longitudinal within-person measurements, not extra independently recruited participants. This difference governs how effect estimates and claims about precision should be interpreted.

What drug was administered, and how were the sessions organized?

Each participant received placebo and three separate oral d-amphetamine doses: 5 mg, 10 mg and 20 mg, in a randomized, double-blind, within-person design. Sessions were spaced by at least 48 hours. Participants had been instructed to abstain from alcohol and other drugs for the preceding 24 hours, while maintaining ordinary caffeine and nicotine intake to avoid withdrawal effects. An opaque capsule was administered around 9:30 a.m.; repeated subjective and physiological measurements were obtained 30, 60, 90, 150 and 180 minutes later.

Computerized behavioral tests began approximately 90 minutes after the dose. The investigators treated the placebo-session test results as baseline indicators of inattention and impulsive action, although these tasks were not necessarily performed at a wholly drug-free, expectation-free orientation visit. That methodological distinction matters: a placebo-session score can serve as an experimental comparator without becoming a permanent diagnostic measure of someone’s personality or future drug-seeking behavior.

The experimental agent was d-amphetamine, not methamphetamine, methylphenidate, mixed amphetamine salts, or lisdexamfetamine. Those substances have different chemical structures or pharmaceutical activation pathways. Results of this specific oral d-amphetamine administration should not be transferred automatically to a different compound, route, dose, formulation, or real-world pattern of exposure.

How did the researchers measure attention lapses?

Participants completed a computerized simple reaction-time task drawn from the Automated Neuropsychological Assessment Metrics battery. The task required a rapid key press after a stimulus appeared at variable intervals. The selected index was a participant’s mean reaction time minus the modal reaction time, called deviation from the mode. The investigators interpreted an unusually long right tail of the response-time distribution as a sign of intermittent lapses in sustained attention.

This number is reported in milliseconds. It is not a count of actual missed roadway hazards, and it does not directly register distractibility during every moment of an ordinary day. Importantly, it was a distributional summary of a laboratory key-press task. Two people with the same average response time could have different lapse metrics if one makes occasional very slow responses. For the 165 analyzed volunteers, the placebo-session mean lapse score was 42.9 ms, with a standard deviation of 28.6 ms.

How did the stop-signal task measure a different kind of impulsivity?

The stop task required participants to respond promptly to ordinary “go” signals, but to withhold the action when an auditory stop signal occurred. The stop-signal delay was adjusted until successful inhibition occurred on roughly half of the stop trials. A stop reaction-time estimate was then constructed using mean go-response time and the final mean stop delay. Longer estimated stop reaction times indicate poorer inhibitory control under this task model.

Only participants meeting predefined performance criteria supplied valid scores. Stop-trial inhibition success needed to fall within 37.5%–63%, and target accuracy needed to reach at least 80%. The placebo-session mean estimated stop reaction time in the final sample was 192.9 ms, with a standard deviation of 45.8 ms. Greater placebo-session lapse scores were modestly associated with longer stop times (r = .24; p < .01), indicating some shared variance, not identity of the underlying constructs.

A stop reaction-time estimate is not a direct neural impulse measurement. It depends on task instructions, the independent-process modeling assumptions of stop-signal methodology, attention, motor responses and adequate task engagement. Neither a slow estimate nor a fast estimate, by itself, determines whether a person committed an impulsive act in an unrelated real event.

What did “liking the drug” actually mean?

Subjective drug response was recorded through the Drug Effects Questionnaire (DEQ): “Like Drug,” “Feel Drug,” and “Want More,” each scored on a 0–100 mm visual-analogue scale. The researchers also examined positive-mood dimensions from the Profile of Mood States (POMS), especially Elation, Vigor and Friendliness. Responses were collected over the post-administration period and summarized as area under the response-time curve for each condition.

For the statistical outcome, each active-dose area-under-the-curve score was compared with that person’s placebo score. That establishes a self-reported change under controlled laboratory administration. It is not the number of doses the participant later sought, a demonstrated likelihood of future addiction, a brain dopamine measurement, or a validated estimate of reward experienced by a person facing criminal or civil proceedings.

What were the strongest baseline predictors of drug liking?

The clearest pattern appeared after 20 mg. Greater baseline attention lapses predicted less DEQ liking, less feeling of a drug effect, and less wanting more. By contrast, longer baseline stop reaction times predicted stronger responses on all three DEQ scales. These associations were estimated simultaneously in regressions that first included age, gender and placebo session order, then added both behavioral metrics. Thus the opposite directions were not merely separate unadjusted correlations between unrelated groups.

The reported standardized beta coefficients convey the direction and relative strength of the modeled association, not individual drug concentrations or a calibrated probability of misuse. The table below reproduces a selected set of quantitative findings from the original regression analyses without turning standardized coefficients into an impairment scale.

Attention-lapse coefficients were negative, whereas stop-time coefficients were positive. The difference is not evidence that one group “liked amphetamines” categorically and another did not. Both were continuous predictors with overlapping individual outcomes. The study does not supply a clinical threshold for classifying a named person as high- or low-risk based on either score. The POMS analysis also shows why a broad assertion that all measures moved in opposite directions would be incorrect: attention lapses did not significantly predict the three POMS scales at 20 mg.

Was the same relationship present at 5 mg and 10 mg?

No. At 5 mg, adding the two placebo-session behavioral predictors did not yield significant associations with the DEQ measures. At 10 mg, the combined behavioral predictors improved regression models for the DEQ Like Drug and Feel Drug outcomes, with the same general negative attention-lapse and positive stop-time directions. The stronger and broader result occurred at 20 mg. This dose pattern is part of the scientific conclusion and should not be omitted to manufacture a dose-independent prediction rule.

For the POMS scales, adding the behavioral components improved the explained variance for Elation and Vigor at 20 mg; the overall Friendliness model increment was only a trend (p = .06), even though the individual stop-time coefficient for Friendliness was reported significant. At 5 and 10 mg, adding these behavioral components did not significantly improve any POMS model. The appropriate distinction is between a particular predictor coefficient and the statistical test of how much an entire set of predictors improved a model.

Did d-amphetamine itself improve either computerized task?

Yes, but that result answers a different question. When investigators compared participants’ task scores under drug and placebo, repeated-measures analyses indicated improvement in both the attention-lapse and stop-time measures overall. Attention-lapse improvement was statistically significant for 10 and 20 mg, but not for 5 mg (p = .18). Estimated stop reaction time improved significantly at all three active doses. These are within-person average task effects after administered drug, distinct from the predictive association between placebo-session task scores and the volunteers’ later subjective ratings.

Positive values in this table represent improved performance relative to the participant’s placebo condition, as coded by the authors. They do not equal reaction times after drug, and the variation across participants was considerable. In particular, the standard deviations for stop-time change exceeded the mean improvement at every dose. It would therefore be misleading to describe the figures as a guaranteed improvement for every individual or to treat a 15-ms group change as a forensic driving-safety margin.

Did stronger performance improvement mean stronger subjective pleasure?

Not in a simple one-to-one way. The researchers also correlated each person’s within-person drug-related performance change with their subjective drug-response change. At 20 mg, greater improvement in attention lapses was associated with smaller increases in DEQ Feel Drug and Like Drug (reported correlations −.28 and −.22, both p < .01), while the analogous Want More association did not pass the conventional significance threshold (p = .051). That latter p-value is not a demonstrated positive finding.

By contrast, more pronounced drug-related improvement in stop reaction time showed positive associations with several subjective mood responses, especially Elation and Vigor at the higher dose (correlations .28 and .26, respectively; both p < .01). These observations are compatible with differing behavioral and subjective processes. They do not demonstrate that a performance improvement mechanically produces euphoria or that euphoria causes better inhibition. Both outcomes could reflect other individual differences not experimentally isolated here.

Why can baseline associations and drug-induced changes point in different directions?

A baseline predictor refers to a measured characteristic before the active drug condition, whereas a drug-induced change refers to how that participant’s performance differed after dosing. These are different estimands. The fact that a highly inattentive participant later improves on a reaction-time task does not require that the same person report strong pleasure from the drug. Similarly, a participant with poorer baseline response inhibition might show improvement while also reporting a larger subjective effect. The analysis deliberately compared these patterns rather than treating all forms of impulsive behavior as a single quantity.

The results invite mechanistic hypotheses involving reward, attention and inhibitory-control systems, but the study did not directly measure dopamine receptor availability, synaptic dopamine concentrations, neural firing, or the pharmacodynamic pathway connecting an individual’s score to the drug’s subjective effect. Explanations involving D2 receptors mentioned in the discussion should be recognized as background hypotheses supported by other research, not directly observed measurements in these 165 volunteers.

What biological specimens did this experiment actually analyze?

Participants underwent preliminary urine and breath screens to check compliance with pre-session abstinence instructions. The original report did not present a validated quantitative d-amphetamine assay in collected venous forensic whole blood, nor a quantitative series in separated plasma or serum, oral fluid, hair, sweat, vitreous humor, or exhaled breath. No chemical drug concentration, specimen collection tube, chromatographic method, limit of detection or quantitative calibration curve was reported for the endpoints studied. A reported oral dose in milligrams is therefore not a measured concentration in ng/mL.

This point is not semantic nitpicking. Whole blood includes cells and plasma; serum and plasma are different specimen fractions; urine or oral-fluid screening serves different purposes from quantitative drug measurement. Even if a person has an accurately recorded administered dose, individual absorption, metabolism, timing and specimen partitioning cannot be reconstructed from this paper’s performance tables. For background on how specimen type changes inference, consult the site’s guide to biological matrices.

How should a forensic report separate chemical exposure from subjective response?

A well-defined forensic opinion identifies the evidence that supports each step. A laboratory result may identify or quantify a compound in a specific collected specimen. A drug-administration record establishes an observed experimental event. A questionnaire records an individual’s reported experience. A computerized key-press task quantifies response behavior under the task conditions. The Weafer and de Wit experiment involved observed administration, task metrics and subjective ratings, but did not supply quantitative blood amphetamine values. None of these evidence types is a substitute for the others.

For example, a participant’s DEQ Want More score is not a chemical indication of recent use and is not a clinical diagnosis of dependence. A relatively long stop reaction time is not itself a measure of intoxication. A mean decline in group reaction-time lapses does not show that a specific witness perceived a particular event accurately. A causal account of a collision would require accident reconstruction and person-specific facts far beyond these laboratory measures. General information about evidentiary analytical specificity appears in the site’s confirmatory drug-testing guide.

Can these results diagnose addiction vulnerability?

No validated prospective diagnostic prediction was established. The study assessed how healthy participants reacted during supervised laboratory sessions. It did not follow participants over years to determine which individuals later developed a stimulant-use disorder, used drugs nonmedically, escalated dose, drove while impaired, or sought treatment. The two baseline scores and self-reported liking may inform hypotheses about vulnerability, but a concurrent statistical association is not a clinically established individual-risk classifier.

The distinction is especially important because a negative association between inattention and liking could be interpreted in opposite ways: weaker reward might protect against recreational use, while some theories suggest people who feel less effect could take more drug to reach a desired subjective state. This experiment was not designed to adjudicate those long-term possibilities. It would be improper to select one story as though it were directly proven by the placebo and dose-response sessions.

Why do sample selection and repeated testing constrain interpretation?

The final analysis excluded 33 of the 198 study completers because valid scores on both predictor tasks were not simultaneously available. Many exclusions reflected task criteria rather than known drug-related adverse outcomes. Selection can improve measurement quality but may also alter the distribution of baseline impulsivity compared with a more representative population. The study does not establish that the modeled relationships would be unchanged among participants who cannot complete the stopping task at the required accuracy.

The placebo baseline measure was obtained during one of four randomized sessions, which means practice or order effects could remain despite including placebo-session order in the model. The laboratory environment, compensation, blinding and short observation period also differ from a person choosing and using medication outside a trial. Self-report scales are susceptible to expectancy and reporting biases even under blinding. Numerous combinations of six outcomes and three doses were examined; the reader should attend to the most robust patterns rather than treating every borderline comparison as definitive.

What results were genuinely nonsignificant or unmeasured?

Several negative findings narrow the claim. At 5 mg, baseline behavioral predictors did not produce significant DEQ relationships. Attention-lapse improvement under 5 mg was nonsignificant (p = .18). At 20 mg, attention lapses were not significant predictors of the three POMS positive-mood outcomes (reported p > .18). The incremental POMS Friendliness regression at 20 mg had p = .06. Within-person improvement in lapses versus the 20-mg DEQ Want More change had p = .051. These observations should not be converted to “trends proved” or quietly omitted.

Other endpoints were not measured at all: no quantitative postdose drug levels, no diagnostic conversion to stimulant-use disorder, no independent objective measure of experienced euphoria, no real driving or crash data, and no blinded witness accuracy assessment. An unmeasured endpoint is not a negative statistical result. It is a gap outside the scope of the experiment. Keeping those categories separate is central to faithful forensic use of behavioral science.

What would a careful expert compare before applying this evidence to a case?

The most useful approach is to specify the case-specific proposition first. Is the question whether a person was exposed to d-amphetamine; whether a specimen contained the parent drug; whether cognitive performance was altered; whether they deliberately chose to ingest it; whether a particular act was impaired; or whether the exposure caused an incident? These questions require different datasets. The present experiment supplies statistical evidence about laboratory baseline task scores and reported feelings after known oral doses, not validation for all of those downstream propositions.

An analyst should examine whether the compound, route and actual specimen match; whether the task operationally resembles the behavior alleged; whether the relevant participant population is comparable; whether the direction and size of an effect matter at the individual level; and whether null outcomes constrain the proposed inference. When timed patient concentrations actually are available, the site’s therapeutic monitoring overview explains why timing and specimen still govern interpretation. Here, there were no quantitative concentrations to integrate with the behavioral scores.

Seven frequently asked questions

1. Did people with more attention lapses like d-amphetamine more?

No. At the 20-mg dose, the adjusted model associated more baseline attention lapses with less subjective Like Drug, Feel Drug and Want More response. The result concerns continuous group-level regression relationships and is not an individual classification tool.

2. Did poorer stop-signal inhibition predict stronger positive drug feelings?

Yes, in the opposite direction: longer placebo-session estimated stop reaction time predicted greater DEQ responses at 20 mg and selected positive-mood ratings. That result does not demonstrate that inhibitory-control failure caused the subjective response.

3. Were there 165 people or 660 people because each attended four sessions?

There were 165 independently analyzed volunteers, not 660 different humans. Each person participated in placebo, 5-mg, 10-mg and 20-mg sessions; those repeated observations are paired within a person.

4. Was the 5-mg attention benefit statistically significant?

No. The mean improvement in attention-lapse score at 5 mg was 3.6 ms with SD 36.3 ms, but the original report gives p = .18. By contrast, the 10-mg and 20-mg attention-lapse comparisons reached p < .001.

5. Did researchers measure amphetamine concentration in venous whole blood?

No quantitative postdose venous whole-blood d-amphetamine values are reported. Recorded oral dose, preliminary abstinence screens, subjective ratings and computerized performance do not substitute for a validated quantitative blood assay.

6. Does a positive Like Drug rating prove someone later became addicted?

No. The study did not prospectively determine addiction outcomes and excluded people with a lifetime history of substance dependence. A reported temporary subjective effect cannot establish a later diagnosis, misuse or a specific person’s future behavior.

7. Can the result establish whether a driver was impaired or an eyewitness was accurate?

No. There was no on-road driving, validated simulator impairment threshold, observed traffic incident, or eyewitness accuracy experiment. Inferring an actual person’s conduct, impairment, credibility or event causation would require independent case-specific evidence.

Bottom line

The central evidence is not that “impulsive people like stimulants.” It is that two different behavioral characteristics related to opposite directions of self-reported response under a controlled 20-mg oral d-amphetamine condition: more baseline attention lapses tracked weaker subjective response, while poorer baseline stop-signal inhibition tracked stronger response. Significant findings at 20 mg coexisted with genuine nulls at lower doses, in selected mood models and in several within-person correlations. The original study is valuable for refining experimental questions, but it provides neither a quantitative forensic blood-drug standard nor a validated person-specific addiction, driving or incident-causation conclusion. The complete primary publication should be consulted at Weafer and de Wit’s original DOI.