Do Amphetamine and THC Change Emotional Memories More Than Neutral Memories? A 48-Person Recognition Experiment

Answer first: A controlled experiment found improved delayed picture-recognition accuracy after oral dextroamphetamine and reduced accuracy after oral dronabinol (THC), in separate volunteer groups. But the most consequential limitation is statistical: the investigators did not detect a significant dose-by-picture-valence interaction in either drug group. Findings for positive pictures were statistically significant in category-specific tests, whereas the corresponding negative-picture comparisons were not, and neutral-picture accuracy did not show the same pattern. Neither drug produced the hypothesized positive-versus-negative memory bias. These laboratory observations cannot establish what a particular eyewitness genuinely saw, whether a specific remembered account is accurate, or a person’s drug concentration or legal impairment.

The original empirical source is Ballard, Gallo, and de Wit, Pre-encoding administration of amphetamine or THC preferentially modulates emotional memory in humans, Psychopharmacology 226:515–529 (2013), doi:10.1007/s00213-012-2924-5. This explanation follows the original methods, participant exclusions, numerical results and stated limitations rather than relying on its headline alone.

What specific scientific question did the study ask?

The researchers asked whether administering two pharmacologically different psychoactive drugs before people viewed emotionally positive, emotionally negative and neutral pictures would affect recognition of those pictures two days later. Dextroamphetamine was expected to strengthen memory formation, while delta-9-tetrahydrocannabinol was expected to impair it. Their additional hypothesis was that effects would be more prominent for emotionally charged pictures than for neutral ones. They also considered whether a subject’s positive mood during exposure would bias recognition in favor of pleasant material.

That is a narrower question than whether a stimulant creates accurate eyewitness testimony or whether cannabis exposure causes an inaccurate allegation. The operational outcome was discrimination of studied photographs from distractor photographs under a laboratory schedule. The protocol did not examine unexpected real-life incidents, police interviews, cross-examination, traumatic narrative recall or witness identification procedures.

Who participated, and which sample size belongs to which result?

The investigators recruited 25 adults into the dextroamphetamine experiment and a different 25 adults into the THC experiment. The amphetamine group included 12 women, and the THC group 11 women. The participants were healthy adult volunteers with current or past occasional recreational cannabis experience, generally in their twenties. Each drug experiment used three repeated conditions on its own participants: placebo, lower active dose, and higher active dose.

The key delayed-memory analyses included 24 independent adults in each experiment, or 48 distinct people total for those endpoints. One amphetamine-group participant viewed the wrong image set on an encoding visit, and one THC-group participant did not comply with the high-dose memory instructions. Their memory-task data were excluded. Other reported demographics, mood and physiological summaries could refer to the 25-person groups. A sound account must not silently substitute the 50 recruited individuals for the 48 people whose delayed-recognition data were analyzed.

Three drug sessions on 24 people do not create 72 independent experimental people. Multiple photographed scenes, valence classifications, cardiovascular observations and delayed recognition responses are repeated measurements nested within volunteers. A participant-level conclusion requires appropriate attention to that structure.

Were amphetamine and THC given together?

No. The drugs were studied in separate volunteer groups. Dextroamphetamine recipients did not also receive a combined THC-plus-amphetamine condition, and the THC group was not subjected to a within-person amphetamine-versus-THC challenge. The experiments can be discussed side by side because they applied a similar recognition procedure, but their drug effects cannot be interpreted as a randomized direct comparison between the two chemicals within the same subjects.

Consequently, the study cannot quantify whether THC counteracts amphetamine, whether taking both compounds changes memory, or whether one drug reverses the other’s pharmacology. An interaction between two drugs requires direct coadministration or a suitable factorial comparison; neither occurred here. The distinction is particularly important where toxicology reports contain more than one detected analyte and an advocate attempts to infer joint effects from independent single-drug studies.

What were the actual substances, doses, and administration routes?

The active stimulant was oral dextroamphetamine supplied in capsules at 10 mg or 20 mg. The other experiment used oral synthetic delta-9-THC (dronabinol) at 7.5 mg or 15 mg. Placebo capsules contained the same inactive filler, and investigators counterbalanced the three administration conditions within each group under double-blind procedures. Capsules were swallowed with water, not inhaled, smoked, injected or administered through a patch.

These are nominal administered milligram amounts, not concentrations in an individual’s blood, plasma, serum or urine. The research also should not be relabeled as a study of illicit methamphetamine, racemic amphetamine salts or a marijuana flower formulation. Molecular identity, isomer composition, delivery route and rate of absorption are material to whether an outside forensic comparison is defensible.

When were pictures viewed and when was memory tested?

About 90 minutes after capsule ingestion, each participant viewed a study set containing 60 photographs: 20 normatively positive, 20 normatively neutral and 20 normatively negative scenes. The pictures came from standardized affective image sets designed to differ in emotional content. On each viewing trial, the participant also evaluated how positive, negative or arousing the image felt. The visual exposures were structured, expected and timed; they were not spontaneous events witnessed during everyday life.

Two days after each study session, the participant returned for a recognition test. That test combined the 60 previously studied pictures with 60 previously unseen distractors. Participants decided whether each image was old or new. Each person underwent distinct picture sets across the three dose conditions, which reduces direct stimulus reuse but introduces the usual need to control for image-set and order effects.

Because the test was delayed and the participants were not given another capsule at retrieval, the protocol addressed lasting recognition following pre-encoding exposure rather than only momentary response speed while acutely experiencing a drug. Nevertheless, those results cannot isolate initial attention from memory consolidation without additional mechanistic intervention.

How were recognition accuracy, hits and false alarms defined?

A hit meant that a participant correctly recognized a photograph shown during the earlier session. A false alarm meant calling a genuinely new photograph old. The main recognition-accuracy variable was hit rate minus false-alarm rate. A higher hit rate does not automatically establish better discrimination if it is accompanied by more false alarms; subtracting the latter helps address general willingness to respond “old.”

This outcome is a fraction based on repeated laboratory picture decisions, not an assigned probability that a single recollection is truthful. Likewise, an improved recognition average does not show that autobiographical details, interpersonal statements, event sequences or peripheral crime-scene information would all improve. False alarms were already low and recognition hits generally high; such near-ceiling behavior limits the size and interpretation of observable changes.

Evidence table 1: did recognition hits move similarly across picture categories?

The differences in the final column are simply higher-dose minus placebo descriptive group means computed from the original source data. They are not independently tested differences, and the table does not add a new statistical significance declaration. The two experiments should not be combined into a fabricated amphetamine-versus-THC treatment-effect estimate. Differences in hit rates must also be considered alongside false alarms and the investigators’ inferential analyses.

What was significant in the dextroamphetamine experiment?

The authors reported improved overall recognition accuracy with increasing dextroamphetamine dose: F(1,23) = 6.27, p = .02. Both active doses improved accuracy compared with placebo in the reported follow-up comparisons, with stated p values no greater than .01. Additional analysis attributed much of the improvement to increased correct recognition of studied pictures, with the overall hit-rate effect F(1,23) = 7.27, p = .013.

In category-specific analyses, positive-picture recognition accuracy improved significantly, F(1,23) = 5.69, p = .026. The negative-picture accuracy comparison approached but did not meet the conventional .05 criterion, p = .052. Neutral-picture accuracy was not significantly altered. Positive- and negative-picture hit rates separately showed significant effects, but a change in hits is not equivalent to a complete improvement in hit-minus-false-alarm recognition accuracy.

False alarms also changed in an analysis testing a nonlinear dose pattern, F(1,23) = 5.25, p = .031. Because the false-alarm baseline was low, the observation should not be transformed into a general theory that amphetamine produces unfailing or perfectly accurate recall. It is a group result from a particular 48-hour photo-recognition paradigm.

What was significant in the separate THC experiment?

THC exposure was associated with reduced overall recognition accuracy: F(1,23) = 6.37, p = .019. Both active oral doses yielded statistically poorer recognition accuracy than placebo in the reported follow-up comparisons. The overall studied-picture hit-rate effect was also significant, F(1,23) = 8.61, p = .007.

For positive pictures, accuracy fell significantly, F(1,23) = 5.81, p = .024. The negative-picture accuracy comparison, p = .081, did not reach conventional statistical significance, and neutral accuracy showed no significant drug effect. The significant THC positive-picture hit-rate result, F(1,23) = 12.38, p = .002, must be distinguished from the negative-picture hit-rate p = .077, which remained nonsignificant. Unlike the amphetamine experiment, the investigators did not observe a THC effect on false-alarm rates.

These results support observed impairments in selected laboratory recognition outcomes after oral dronabinol. They do not prove that every pleasant or unpleasant real-world memory would be inaccurate, nor can the magnitude be imported into an individual with an unspecified cannabis product, route, tolerance or time since exposure.

Why does the nonsignificant dose-by-valence interaction matter?

The study’s strongest interpretive qualification is that the formal dose-by-picture-valence interaction was not statistically significant in either the amphetamine or THC accuracy analysis. A main effect of a drug across picture categories and significant tests within selected categories are not the same as proof that the drug’s effect truly differs across categories. Researchers explored preplanned positive, neutral and negative contrasts, but those exploratory patterns cannot replace the missing statistical interaction.

That is why a forensic writer should not announce that the experiment definitively proved a selective mechanism targeting emotional but not neutral memory. The source supports overall drug effects plus informative category-specific observations. Its emotional-specificity hypothesis remains more qualified than a short headline might suggest. An apparent positive-picture effect and a nonsignificant neutral-picture effect are not by themselves sufficient to establish that the two effects differed significantly from one another.

This distinction is not a technicality. If a case theory turns on the claim that a particular drug uniquely strengthened fear memories, pleasant details or negative recollections, the claim requires an interaction or another directly appropriate test. The present study did not establish that category-specific differential effect to the standard usually implied by an unqualified “preferentially.”

Evidence table 2: the significant findings and the genuine nulls

The main accuracy dose-by-valence interactions were also reported as nonsignificant, but the source did not furnish a numerical p-value for those specific omnibus tests in this passage. Rather than inventing one, the table includes an explicitly quantified, closely related hit-rate interaction at p = .053. Do not convert p = .052, .053 or .081 into significant results by calling them a trend. A null comparison means the study did not demonstrate an effect under that test; it is not proof that every conceivable effect is zero.

Did the drugs make pleasant memories disproportionately stronger?

No reliable positive-versus-negative memory bias was demonstrated. The authors’ separate hypothesis was that drug-induced positive affect might favor subsequent recognition of pleasant material over unpleasant material. The main findings did not establish that predicted bias for either substance. This matters because an improvement in memory for some positive pictures does not automatically establish that the participant preferred good experiences or suppressed negative ones.

In the amphetamine group, ratings of picture positivity and arousal increased under the higher dose, yet those shifts did not provide a demonstrated mechanism establishing selective positive-memory accuracy. In the THC group, subjective drug sensations and some picture arousal ratings changed, but the memory findings likewise did not show that pleasant recollections reliably crowded out unpleasant recollections. Generalized accounts of euphoric memory distortion cannot be assumed from these tasks.

Were subjective drug effects interchangeable with memory performance?

No. A participant’s report of feeling stimulated, sedated, high, pleased or displeased was a separate dependent measure from the fraction of previously studied pictures recognized correctly. In the amphetamine group, the 20-mg condition increased reported sense of drug effect, stimulation and preference. The study reported that 24 of 25 participants would take that higher dose again, compared with 12 of 25 for 10 mg. Those counts concern subjective willingness to repeat the experimental substance, not memory-test validity, dependency diagnosis or drug concentrations.

In the THC group, participants generally felt active drug effects at both doses, but preference varied substantially. Eleven of 25 participants reported willingness to take 15 mg again, while 15 of 25 did so for 7.5 mg. Those responses do not give a calibration curve for recognition accuracy or biological THC levels. The original exploratory correlations did not provide a clear explanation of each participant’s memory effect from their mood and drug-effect ratings alone.

The ability to feel intoxicated and the capacity to perform a specific memory task are related empirical questions but not identical measurements. A report of increased alertness is neither a chemical assay nor a substitute for the actual hit-versus-false-alarm calculation. Similarly, absence of a striking subjective effect does not mean a chemical was absent, or that a different cognitive outcome could not change.

What did heart rate and blood pressure actually measure?

Researchers repeatedly measured blood pressure and pulse using routine cardiovascular instruments. Their original source tables reported mean changes from pre-capsule measurements around picture viewing. In the amphetamine experiment, the 20-mg condition yielded a mean systolic pressure change of +17.73 mmHg compared with −8.44 mmHg under placebo; the respective lower-dose change was +8.33 mmHg. In the THC experiment, heart-rate change was +8.54 beats per minute at 15 mg compared with −6.15 under placebo.

These are physiological group observations, not quantitative amphetamine or THC concentrations in blood. A person can have a heart-rate change for many reasons unrelated to psychoactive exposure, and individuals vary in cardiovascular response even when the administered compound and dose are known. The protocol did not validate using pulse or pressure to date an unknown exposure or to decide whether a particular driver was legally impaired.

Why do high baseline hit rates limit the conclusions?

The placebo hit-rate means were often in the .92–.96 range across categories. That leaves relatively little upward room for a drug to improve the number of correctly identified old pictures. False alarms were also generally low. Ceiling and floor effects can distort an apparent drug-by-category difference because one category may start nearer a measurement boundary than another, regardless of whether the underlying biological process is category-specific.

These design conditions do not invalidate the reported main effects, but they matter when translating a numerical finding into a general claim about memory. A task whose participants already identify nearly all studied pictures is not necessarily equally sensitive to every kind of cognitive change. Emotional memory in a real event may involve encoding, source attribution, retrieval, confabulation, stress and social influence, none of which is fully captured by that high-accuracy forced-choice procedure.

What do the dose comparisons establish—and what do they not?

Within each experiment, participants received placebo, a lower dose and a higher dose, allowing dose-related within-person analysis. That arrangement improves efficiency by partially controlling stable differences between participants. However, it does not create a blood-concentration response model. For example, the 10-mg dextroamphetamine memory condition and the 7.5-mg oral THC memory condition do not produce directly comparable pharmacological exposures just because their masses are written in milligrams.

The distinction extends to drug formulation. Dronabinol capsules are not a model of whole-plant cannabis containing variable cannabinoids or terpenes. Dextroamphetamine is not chemically identical to racemic amphetamine, lisdexamfetamine, methylphenidate or methamphetamine. A route change from oral capsules to inhalation or injection modifies absorption kinetics, so an analogous nominal dose cannot be assumed to produce the same exposure timing or behavioral outcome.

The drug conditions in this experiment were administered before photographs were first studied. It was not a test of drug consumption immediately before the delayed recognition test, not a chronic treatment trial, and not a demonstration that a dose taken after an event could retrospectively make all memories more or less accurate.

Was any drug measured in venous whole blood, plasma or serum?

No quantitative post-dose concentration of parent dextroamphetamine, THC, or their metabolites in legally relevant human venous whole blood is reported. Nor did the published experiment present an individual concentration-time profile in separated plasma or serum. The article recorded capsule mass, timing, picture decisions, feelings, pressure and heart rate. Admission procedures included a qualitative urine drug screen and a breath-alcohol check to determine visit eligibility, not serial confirmatory measurement of the experimentally administered study drug.

Human venous whole blood is not equivalent to plasma or serum obtained after separating cellular components. Urine, oral fluid, hair, sweat, vitreous, tissue and exhaled aerosols are still different matrices with distinct collection and interpretive characteristics. An ethanol breath-alcohol result is likewise not an assay of blood THC or amphetamine. The site’s independently published biological matrices guide explains why the actual physical specimen and method must remain attached to any result.

The absence of chemical assay findings is a limitation, not a reason to invent a value. Nothing in this source supports an individual THC ng/mL cutoff, an amphetamine concentration in plasma, an enantiomer-specific blood ratio, or a scientifically validated interval since last dose. A dose of 15 mg swallowed does not equal a measured 15 mg/L in circulating blood, and no case-specific drug concentration can be reverse-engineered from a group memory score.

How does a visit urine screen differ from confirmed drug identification?

An eligibility urine screen addresses a predefined screening decision about suspected recent substances. It is not automatically an identification and quantitative measurement of all psychoactive analytes of interest. Interpretation depends on assay targets, method validation, calibration, cross-reactivity and the reporting threshold. The original study did not report a post-dose confirmed whole-blood amphetamine or THC analytical series simply because it performed visit drug screening.

In a legal case, the laboratory’s underlying chromatographic, mass-spectrometric or other validated analytical records would have to be reviewed separately, if they exist. An observed photograph-memory change cannot retrospectively confirm that a drug was present in a specimen. The site’s freshly verified confirmatory testing guide

Can this experiment establish whether a real eyewitness is accurate?

Not on its own. Volunteers knowingly viewed standardized photographs under controlled conditions, then recognized familiar pictures two days later. Actual witnesses may encounter dynamic events without expecting a memory test; they may have different visibility, fatigue, divided attention, intoxication histories and intervening conversations. Identifying a scene from photographs is not equivalent to reconstructing who said what, recalling a sequence of actions or distinguishing honest errors from deliberate statements.

The study did not measure eyewitness veracity, memory for a particular crime, cross-examined testimony, identification lineup accuracy, source-monitoring susceptibility or exposure to suggestive questioning. Even its emotional-versus-neutral specificity test failed to demonstrate the predicted interaction. It would therefore be unsound to assign a probability that an actual account is true or false from the group-level p-values listed here.

Research may help evaluate the plausibility of mechanisms or flag overbroad expert claims. But an evidentiary conclusion about one witness would require information about the witnessed event, independent corroboration, drug identity, route, timing, specimen findings, sleep and competing influences. That investigation is separate from what this experiment tested.

Does a memory result establish driving impairment or collision causation?

No. No driver performed an on-road test under these study conditions. There were no validated driving endpoints such as lane deviation, hazard response, divided-attention driving tasks, collision frequency or a legally relevant individual performance threshold. Positive and negative photo recognition at a two-day test cannot be mechanically translated into safe vehicle operation at the time of dosing.

Moreover, the reported statistics describe means and inferential results for separate groups of 24 memory-analyzed people. They do not provide person-specific sensitivity, specificity, predictive intervals or validated classification cutoffs. The study’s unmeasured human whole-blood concentrations make it impossible to link a particular toxicology result with any photograph-memory change by direct experiment. A case-causation claim would need evidence independent of this controlled recognition study.

What methodological limits matter most in a forensic review?

First, the primary comparisons were obtained from small groups of healthy adults with some cannabis-use history, not from a random sample of motorists, trauma patients or courtroom witnesses. Second, the drugs were evaluated separately; there was no experiment quantifying combined THC-amphetamine exposure. Third, pre-exposure administration and delayed recognition cannot identify an isolated encoding-versus-consolidation mechanism without further tests. Fourth, high baseline hits and low false alarms restrict performance range.

Fifth, the authors used a number of category-specific and exploratory tests. The failure of the principal drug-by-valence interactions to achieve significance is particularly relevant in the presence of positive within-category comparisons. Sixth, the source reports group means and standard errors for some endpoints, but not a verified person-by-person relationship between chemically quantified concentrations and outcomes. Seventh, the controlled picture task omits central features of everyday event observation and later witness interviewing.

These considerations do not erase the observed overall drug effects. They establish the proper boundary of inference: significant performance differences under particular laboratory conditions do not imply a reliable general-purpose drug test, validated witness classification or personal impairment determination.

What scientific evidence would be needed to answer a stronger question?

A study claiming emotional-memory specificity would benefit from a prespecified interaction contrast, adequate sample size to estimate that interaction, a wider task difficulty range and careful control of multiple tests. If the question concerned concurrent use, it would require a suitable factorial or controlled coadministration design. If the question concerned real witnessed events, more ecologically valid encoding and recall protocols, explicit source monitoring and controlled intervening information would be essential.

Any investigation proposing a forensic dose-to-effect relationship would also need validated analyte-specific measurements in a declared matrix, specimen collection times, appropriate internal standards and calibration, and a scientifically defensible concentration-effect model. It would still have to distinguish population relationships from predictions about a particular person’s decision-making, memory or culpability. No such additional measurements or validation are supplied by this original experiment.

Seven frequently asked forensic questions

1. Did amphetamine improve emotional picture recognition?

The dextroamphetamine experiment reported improved overall accuracy and a significant positive-picture category result, with no demonstrated main effect on neutral-picture accuracy. But the formal dose-by-valence interaction was nonsignificant, and the negative-picture accuracy result p = .052 did not establish significance. The conclusion must retain those qualifications.

2. Did oral THC impair recognition of every kind of picture?

No. Overall accuracy decreased, and positive-picture accuracy showed a significant category-specific impairment. Negative-picture accuracy p = .081 was nonsignificant; neutral accuracy was not significantly altered. The experiment does not establish that every memory or all picture categories are equally affected.

3. How many people were actually analyzed for memory?

There were 25 independently recruited people in each of two separate drug studies, but one memory-task exclusion in each group left 24 analyzable people per experiment. Each person’s three dose visits are repeated measures, not three independent people. The memory-analysis total was 48, not 150.

4. Did the researchers test amphetamine and THC together?

No. Volunteers in the dextroamphetamine experiment did not receive an experimental THC-plus-amphetamine combination, and THC-group participants did not undergo amphetamine dosing. Therefore the publication cannot establish a joint pharmacodynamic interaction or compensating effect of one drug on the other.

5. Did drug-induced positive mood make pleasant memories dominate?

The hypothesized selective positive-versus-negative memory bias was not demonstrated. Picture-valence-specific findings and drug preference responses should not be treated as proof that pleasant personal recollections overwhelm distressing recollections under either substance.

6. Were human whole-blood THC or amphetamine concentrations measured?

The article reports no quantitative post-dose parent-drug concentration series in human legal venous whole blood or separated plasma and serum. Capsule milligrams, visit urine screening, breath alcohol checks, pulse, blood pressure and memory accuracy are different measurements. There is no validated individual blood-concentration cutoff derived here.

7. Can these results decide whether a driver or witness was impaired?

No. There was no actual driving test, crash investigation, truthful-testimony classifier or individual drug-level performance calibration. The study may constrain broad expert claims about memory under controlled drug exposure, but real witness accuracy, impairment and event causation must be determined from separate case-specific evidence.

Bottom line: what may—and may not—be inferred?

The controlled experiment demonstrates oppositely directed overall delayed-recognition findings in two separate groups: oral dextroamphetamine improved accuracy, while oral dronabinol impaired it. Category-specific positive-picture comparisons were significant, but the primary evidence does not establish a significant dose-by-valence interaction or a predictable positive-versus-negative memory bias. There were genuine null comparisons. Nothing in the source establishes a numerical parent-drug blood concentration, a personal ingestion clock, a lawful driving-impairment cutoff or the truthfulness of an actual eyewitness.

Original primary study: Ballard ME, Gallo DA, de Wit H. Pre-encoding administration of amphetamine or THC preferentially modulates emotional memory in humans. Psychopharmacology. 2013;226:515–529. https://doi.org/10.1007/s00213-012-2924-5.