Does Dextroamphetamine Increase Memory Intrusions Without Improving Correct Recall? A 31-Person Study

Answer first: In a double-blind, within-person experiment involving 31 healthy young adults, oral dextroamphetamine given at memory retrieval increased erroneous free-recall intrusions for both pictures and words, without significantly increasing correct free recall. It also increased false recognition of words and reduced the ability to distinguish studied from unstudied words. The significant pattern was specific to the experiment’s staged laboratory tests: picture recognition was not significantly changed, and the investigators did not establish that every memory became inaccurate. Crucially, participants had learned the material two days earlier without the study drug. These findings neither quantify an individual’s blood amphetamine concentration nor determine the truthfulness of a real witness or the cause of an incident.

The original empirical study is Ballard, Gallo and de Wit, Amphetamine Increases Errors During Episodic Memory Retrieval, Journal of Clinical Psychopharmacology 34(1):85–92 (2014), doi:10.1097/JCP.0000000000000039. This article distinguishes what that study measured, what remained statistically nonsignificant, and which legal-toxicological questions it did not test.

What was the central experimental question?

Memory formation and memory retrieval are different processes. A substance given while a person first experiences information may affect encoding or consolidation; a substance taken when a person later tries to remember already learned information may alter access, response selection, confidence or willingness to report a candidate memory. The researchers deliberately separated those periods. They asked whether a stimulant administered only before the later test changed correct recall, erroneous recall, or recognition decisions concerning previously encountered emotional and neutral materials.

This design is especially important when someone cites a generalized proposition that a stimulant either improves cognition or makes recollection unreliable. A change in a laboratory intrusion count is not interchangeable with a change in stored factual knowledge. Nor does an overall effect on a memory task establish how a specific person would describe an unexpected real-world event. The proper starting point is the experimental stage, the task and its scoring rule.

Who participated, and how many independent people were measured?

Thirty-one distinct healthy volunteers, including 16 women, completed the study. Their average age was 22.6 years, with a standard deviation of 3.4 years. Researchers screened volunteers for significant medical and psychiatric exclusions; these were not patients undergoing treatment for attention-deficit/hyperactivity disorder, intoxicated motorists, forensic patients or people recounting a crime. Volunteers with most active serious psychiatric conditions or substance dependence, apart from permitted tobacco dependence, were excluded.

Each volunteer returned for several visits. Three dose conditions, multiple picture and word sets, and hundreds of stimulus judgments are repeated measurements within people, not additional independent participants. Three word-recognition records were lost to a computer malfunction, so the inferential word-recognition analyses used 28 people, not 31. That distinction matters for interpreting the degrees of freedom reported for false-word recognition and its accuracy result.

What precisely was administered?

The tested agent was oral dextroamphetamine, supplied as capsules containing either 10 mg or 20 mg, with an inactive placebo condition. The volunteers received each condition in a double-blind, counterbalanced, within-person arrangement. The capsules were swallowed with water. These are administered mass doses, not chemically measured blood concentrations. The experiment was not conducted using methamphetamine, methylphenidate, lisdexamfetamine, mixed amphetamine salts, or a standardized illicit stimulant product.

Drug identity, route and dosing stage are all essential. A 20-mg oral challenge before a memory test cannot be equated to a 20-ng/mL human whole-blood result or to the effects of an unknown street preparation. Nor did the article supply an individualized pharmacokinetic series connecting its capsule doses with quantitative parent-drug concentrations in any specified biological matrix.

When were items learned, and when was the drug taken?

For every experimental condition, an initial approximately one-hour encoding visit took place without the study medication. Participants viewed and evaluated a set of pictures and words. Two days later, they returned for a longer retrieval visit, ingested placebo, 10 mg or 20 mg of dextroamphetamine, and subsequently attempted the memory tasks. The investigators used different stimulus sets across conditions and counterbalanced their assignment.

Testing followed a fixed sequence after dosing: picture free recall at about 60 minutes, word free recall at about 75 minutes, picture recognition at about 90 minutes, and word recognition at about 120 minutes. Consequently, drug exposure at retrieval was experimentally separated from drug-free encoding. The design does not isolate every process contributing to retrieval, and task order was fixed, but it does not support describing these outcomes as an effect of taking amphetamine while first seeing the pictures.

What counts as a correct recollection versus an intrusion?

In free recall, participants wrote descriptions of remembered pictures or previously viewed personality-trait words. Four independent raters judged the picture descriptions, with a further rater resolving split decisions; reported interrater agreement was 97%. A correct item corresponded to material from that condition’s original study set. An intrusion was an attempted recollection that was not correctly attributable to that set. Intrusions included items remembered from earlier sessions, wholly extraneous descriptions and descriptions only weakly related to the actual study material.

Some ambiguous picture descriptions were rescored under a more liberal criterion; the authors reported no material change in their general findings and presented the stricter results. That sensitivity check does not make scoring infallible. It reinforces the need to distinguish a controlled, adjudicated memory-error definition from the more complicated question of whether a real person’s account contains a mistaken but sincerely held detail.

What did placebo performance show before comparing drug doses?

Even without active dextroamphetamine, recall was incomplete and intrusions occurred. Under placebo, participants recalled on average about 25% of studied pictures while producing approximately 2.9 erroneous picture descriptions. They recalled about 14% of studied words while generating approximately 7.7 word intrusions. Those baseline observations demonstrate that an error does not, by itself, establish stimulant exposure: errors arose in the inactive-drug condition too.

The word and picture tasks differed markedly in difficulty and scoring. Personality-trait words could invite semantically related but unstudied responses; pictures supplied distinctive visual scenes. Rates should not be compared as if the same number of items, chance level or error type applied. Repeated measures and task-specific denominators matter more than a generic statement that one task produced more mistakes.

Evidence table 1: which retrieval outcomes were significant or genuinely null?

The table supports a narrow conclusion: more erroneous recall attempts were observed, not a reliable gain in recovery of studied items. Neither the test statistics nor an error count converts to the probability that one particular recollection is false. Confidence analyses were subdivisions of a larger outcome family and should not be presented without the broader pattern and its nonsignificant results.

Did dextroamphetamine improve correct picture or word recall?

No statistically significant improvement was established. The original report describes no drug effect on the overall amount of correctly recalled studied pictures or words, or on confidence concerning correct recall, regardless of their predefined normative emotional categories. In both tasks the striking difference lay in intrusions: more attempted recollections were incorrect while accurate recall did not show a corresponding demonstrated increase.

This is not equivalent to saying that the drug erased actual memories. A nonsignificant correct-recall comparison means the experiment did not demonstrate the specified effect under the reported test. It is also not proof that every possible correct-recall effect is exactly zero. Estimating an individual’s true-memory capacity would require more granular performance data, measurement precision and validation beyond these group analyses.

Did intrusions increase only when participants lacked confidence?

No. The original picture-intrusion analysis found a statistically significant drug-related increase within both moderate-confidence and high-confidence categories, although the high-confidence follow-up reached significance for the higher 20-mg dose. The low-confidence picture-intrusion result, p = .067, did not meet the stated .05 criterion. That null must remain visible when describing the pattern of confidence-qualified errors.

For word intrusions, the broader count increased significantly, while the reported confidence-bin analysis reached significance for the lowest-confidence category at the higher dose. The specific word-confidence pattern was not identical to the picture-confidence pattern. It would therefore be inaccurate to state that dextroamphetamine uniformly produced confidently asserted false recollections at every dose and for every type of material.

Was recognition of pictures impaired?

No significant change in picture recognition was demonstrated. Under placebo, participants already correctly identified approximately 96% of studied photographs while erroneously endorsing approximately 4% of unseen pictures. The article reported no drug-related change in overall picture recognition accuracy, hits or false alarms; comparisons were nonsignificant, with reported p values above .06. Near-ceiling hits and a low false-alarm rate restrict how much movement the task can reveal.

This creates an important contrast inside the same experiment. Participants could generate more false descriptions during unrestricted picture recall yet not show a detectable change when recognizing old pictures among distractors. Free generation and recognition are not interchangeable measures. The discrepancy cautions against treating a single category of memory error as evidence that all memory processes were globally compromised.

What happened in the word-recognition task?

The word-recognition endpoint yielded a different result from picture recognition. The placebo lure false-alarm rate was about 33%. It rose to approximately 38% after 10 mg and 40% after 20 mg, with a statistically significant dose effect, F(1,27) = 6.7, p = .015. By comparison, the hit rate for genuinely studied words was approximately 86% under placebo, 86% at 10 mg and 85% at 20 mg; an overall change in hits was not established, p = .38.

Because participants accepted more novel words as previously seen without a corresponding improvement in recognition of actual targets, their discrimination accuracy declined significantly, F(1,27) = 8.8, p = .006. The sample for this endpoint was 28 individuals because three participants’ word-recognition task data were lost through a computer malfunction. Treating those 28 analyzable participants as 31 or treating their hundreds of responses as independent volunteers would misstate the strength of the evidence.

Evidence table 2: the raw recognition distinction across three doses

The original Table 2 provides finer positive, neutral and negative item-category values. The authors’ summary reported no significant overall picture-recognition effect. The word false-alarm percentages are rounded group means, not numerical amphetamine levels, diagnostic cutoffs or estimates of the likelihood of deception. The absence of active-dose aggregate picture values from this reorganized table must not be filled by interpolation.

Did emotional content determine the drug effect?

The study classified words and pictures as normatively positive, neutral or negative, then separately asked participants how they themselves perceived those items. These are different variables. During the memory tests, the investigators reported no dose-by-normative-valence interaction for the memory measures, with p values above .31. Thus the primary drug-related changes in intrusions and word false alarms were not shown to depend on the predefined emotional category in the tested analyses.

A separate exploratory analysis found increased generation of subjectively positive picture descriptions and words during free recall, particularly with the 20-mg dose. Reported dose-by-subjective-valence statistics included F(1,30) = 4.4, p = .044 for picture descriptions and F(1,30) = 11.6, p = .002 for words. That is an observed positivity in the items people generated, not a demonstrated enhancement in their accuracy. The analysis counted correct and incorrect answers together, a distinction that matters when reporting a memory bias.

Why is the distinction between a hit and a false alarm indispensable?

A recognition hit is a previously encountered item correctly accepted as old. A false alarm is a new item incorrectly accepted as old. A person who says “yes, I saw it” more liberally can show both more hits and more false alarms without having a better memory. For that reason, the investigators used hit rate minus false-alarm rate as a discrimination-accuracy measure.

In this experiment, the word hit rate remained approximately unchanged while false endorsements of unstudied words increased. Discrimination therefore declined. This is much more specific than the assertion that the drug simply enhanced recollection, and it differs from a false identification of a person under actual lineup conditions. The laboratory outcome tells us how the studied yes/no decision task changed, not why a particular courtroom witness answered a question in a specific way.

Were subjective stimulation and blood pressure measured instead of drug concentrations?

Researchers assessed subjective drug effects, mood, heart rate and blood pressure. They observed the expected stimulant-related changes: active doses increased pulse and pressure and various measures of positive mood, while participants generally preferred the active drug over placebo. At 20 mg, more volunteers correctly identified the capsule condition as a stimulant and reported greater motivation to complete the memory tasks. These contextual measures show pharmacodynamic responsiveness, not the mass of dextroamphetamine circulating in whole blood.

Physiological and subjective observations can be informative without being specific chemical assays. Heart rate can vary with exercise, anxiety, caffeine, sleep and many other factors. Mood scales measure reported feeling states. None supplies an enantiomer-specific or parent-versus-metabolite concentration, or validates a numerical rule converting recognition errors into a forensic drug level.

What did the urine screening and breath test actually establish?

Before laboratory visits, participants were checked with a qualitative urine drug screen and a breath-alcohol device for protocol eligibility. The presence of screening does not mean the investigators quantified the administered dextroamphetamine in postdose human venous whole blood, separated plasma or serum, oral fluid, hair, sweat, vitreous or another specimen. Nor did they report confirmatory quantitative parent-amphetamine pharmacokinetics with a validated limit of quantitation for the retrieval visits.

Forensic interpretation requires keeping each physical matrix attached to the question it answers. The site’s independently verified guide to biological matrices explains why a urine screen, a breath-alcohol eligibility test and a quantitative venous whole-blood assay are not interchangeable. The companion confirmatory testing guide distinguishes an assay-specific identification from a behavioral suspicion. Neither guide converts these group memory-task statistics into an individual chemical measurement.

Can the experiment establish the reliability of a specific eyewitness?

No. Participants knowingly studied standardized stimuli, returned after a scheduled delay and were instructed to write picture descriptions and recognize previously presented items. The study did not examine a spontaneous crime scene, a traumatic assault, an unanticipated collision, lineup identification, social suggestion, repeated police interviewing, courtroom cross-examination or intentional deception. Most importantly, there is no demonstrated person-specific classification of true and false testimony from the observed group statistics.

The experimental results are relevant to evaluating overbroad claims such as “stimulants invariably strengthen accurate memory” or “every stimulant-exposed account is unreliable.” Neither statement follows. Assessing one witness would require independent facts about the event, the form and timing of any actual substance exposure, contemporaneous observations, confidence development, outside information, corroboration and the witness’s opportunity to perceive what occurred. This experiment supplies neither a truth detector nor a legal credibility cutoff.

Does the research establish individual intoxication or driving impairment?

No driving simulator, road test, crash investigation or medically validated impairment classification was performed. Writing down words and recognizing pictures are not validated substitutes for lane control, divided-attention driving, hazard response or operational safety. Even if an observer had consumed dextroamphetamine, the study does not provide a chemical whole-blood level, individual concentration-effect calibration, last-dose clock or probability of dangerous driving for that observer.

Legal conclusions must also distinguish exposure from effect and effect from case causation. This small controlled retrieval experiment tested group memory-task performance after known doses, not the conduct or responsibility of any named person. Its results can help frame scientifically testable questions, but the evidence required for a driving or litigated incident conclusion is independent of these data.

Which limitations most constrain generalization?

The experimental participants were young, selected and medically screened. They were not a random sample of forensic examinees, older adults, people using multiple substances, or patients prescribed long-term stimulant medication. Both tasks involved standardized, repeated laboratory materials rather than autobiographical events. Repeated tests reduce between-person variability but still allow learning, order, fatigue and other laboratory influences that are not present in every outside setting. Memory categories, confidence subdivisions and multiple follow-up contrasts also create a multiplicity issue; the article does not establish a universal familywise-adjusted discovery rule for every listed comparison.

Three participants lacked usable word-recognition records because of computer failure. Near-ceiling picture recognition and relatively high word false-alarm baselines could limit sensitivity. Subjective positive-valence classification differed from standardized valence. Finally, the authors reported no validated chemical blood-versus-performance model and did not test individual witness accuracy, a real-world crash or specific impairment. These limits should accompany the interesting positive findings rather than erase or exaggerate them.

What would stronger forensic evidence require?

A concentration-effect study would need timed validated analyte-specific measurements in an expressly identified specimen, with laboratory quality controls, detection and quantitation limits, and a prespecified model of variability. An eyewitness-reliability study would need event-relevant encoding and retrieval, documentation of actual exposure and intervening influences, and validated person-specific prediction rather than only a laboratory mean difference. A study claiming drug-specific emotional effects would need appropriate interaction tests and adjustment for multiple comparisons.

Before translating a group effect into a case, an analyst should ask whether the event happened during drug exposure or after it, which drug and formulation were actually used, what was measured chemically rather than reported subjectively, and how alternate causes of an apparent memory discrepancy were assessed. The Ballard experiment answers none of those case-specific questions by itself. Clear recognition of that boundary protects the scientific value of what it did find.

Seven frequently asked forensic questions

1. Did amphetamine increase false memories without improving correct recall?

The 31-person crossover experiment found significantly more erroneous free-recall intrusions for pictures and words, with no significant improvement in correctly recalling studied items. “False memory” here means a specifically scored laboratory intrusion, not an adjudication that an actual eyewitness account is knowingly false.

2. Were all memory tests impaired?

No. Picture-recognition performance was not significantly changed. Word-recognition false alarms increased and discrimination declined, whereas correctly recognized studied words did not show a significant overall change. Free recall and recognition must be described separately.

3. Were 31 people analyzed for every endpoint?

No. Thirty-one independent adults completed the experiment, but three participants’ word-recognition records were lost during a computer malfunction. Word-recognition analyses used 28 adults. Dose visits and many stimulus responses do not create additional independent people.

4. Were participants already on dextroamphetamine when they first studied the material?

No. The researchers presented the source pictures and words under drug-free conditions; the assigned oral capsule was taken about two days later before recall and recognition. The study addresses drug exposure at retrieval, not drug exposure during initial encoding.

5. Did the drug make positive emotional memories more accurate?

No such general conclusion was demonstrated. More subjectively positive descriptions and words were generated, but that exploratory measure included incorrect as well as correct responses. Normative emotional-category dose interactions for the memory measures were not significant.

6. Did the researchers quantify human whole-blood amphetamine?

No quantitative postdose parent dextroamphetamine concentration in human legal venous whole blood, separated plasma or serum, or another drug matrix was reported. The experimental values were capsule doses, screening tests, physiological measurements and behavioral memory outcomes.

7. Can the results decide witness credibility, driving impairment or case causation?

No. The researchers did not test the truth of an actual eyewitness, driver performance, a particular incident, or an individual chemical concentration-effect threshold. Real-world inferences require independent case-specific evidence that this controlled laboratory source does not supply.

Bottom line

Oral dextroamphetamine at retrieval increased certain erroneous recollections and false-word recognition in a 31-person controlled experiment, while correct free recall and picture recognition did not show corresponding significant improvement or impairment. The findings describe distinct laboratory memory processes and remain bounded by the sample, task, dose, timing and missing chemical blood measurements. They are relevant to scientific claims about stimulant-associated memory error, not a license to classify an individual’s testimony, diagnose intoxication or assign legal responsibility.

Original primary study: Ballard ME, Gallo DA, de Wit H. Amphetamine Increases Errors During Episodic Memory Retrieval. Journal of Clinical Psychopharmacology. 2014;34(1):85–92. https://doi.org/10.1097/JCP.0000000000000039.