MDMA Impairs Both the Encoding and Retrieval of Emotional Recollections

Recent human and clinical research suggests that ±3,4-methylenedioxymethamphetamine (MDMA) modulates emotional processing and has renewed interest as an adjunct in post‑traumatic stress disorder (PTSD) treatment. Existing evidence indicates that MDMA attenuates threat detection and amygdala responses and that it can impair memory for emotionally neutral material, but prior studies often tested memory while participants remained intoxicated and therefore could not cleanly separate encoding from retrieval effects. At the same time, classic stimulants can selectively enhance emotional memory, and MDMA’s mixed pharmacology (serotonergic actions plus stimulant-like noradrenergic effects) leaves its net impact on emotional episodic memory uncertain. Doss and colleagues therefore designed a double‑blind, placebo‑controlled experiment to disentangle MDMA’s effects on encoding versus retrieval of emotional memories and to distinguish recollection (retrieval of event‑specific details, hippocampal-dependent) from familiarity (a sense of knowing without specific details, cortical). The primary prediction was that MDMA would attenuate recollection‑based emotional memory when administered either at encoding or at retrieval. The study tested healthy participants with negative, neutral and positive pictures and used dual‑process analyses to estimate recollection and familiarity separately.

Sixty healthy volunteers (three groups of 20, 50% male, ages 18–34) were recruited and screened with physical exam, ECG and a semistructured psychiatric interview. Exclusion criteria included current Axis I disorder, substance dependence, very heavy smoking, history of psychosis or mania, low education or insufficient English fluency, BMI outside 19–30 kg/m2, uncontrolled hypertension or abnormal ECG, routine medication use (other than contraceptives), pregnancy or lactation. Eligible participants reported between 4 and 40 prior uses of MDMA without adverse events. Women not on hormonal contraceptives were tested during the follicular phase. Participants were told they might receive a stimulant, sedative, cannabinoid or placebo to minimise expectancy effects, and they followed pre‑session abstinence rules for alcohol, marijuana and other drugs; compliance was checked by breath and urine tests. The active drug was MDMA prepared at 1.0 mg/kg in opaque capsules, with dextrose placebo capsules. Participants were randomised, double‑blind, to one of three groups: MDMA during encoding and placebo at retrieval (Encoding), MDMA during retrieval and placebo at encoding (Retrieval), or placebo at both sessions (Placebo). All took two morning sessions separated by 48 hours. Ninety minutes after capsule ingestion in each session participants completed the task battery. Stimuli were 180 images from the International Affective Picture System (IAPS) with 2–3 word labels, balanced into two sets for counterbalancing; images spanned negative, neutral and positive valence. During the encoding session participants saw all 180 labels, with half followed by the corresponding picture, and provided liking, valence and arousal ratings. The retrieval session (48 h later) included two surprise tests, administered 90 minutes after capsule ingestion: a cued recollection test (labels presented; yes/no seen judgment plus 5‑point confidence) and a picture recognition test (old/new decisions for pictures plus remember/know judgments for recognised items). Both tests were self‑paced. Primary dependent measures included hits, false alarms and accuracy by valence for cued recollection and recognition, high‑confidence responses, and estimates of recollection and familiarity. Two methods estimated recollection and familiarity: a dual‑process signal detection (DPSD) analysis using receiver operating characteristic (ROC) curves from confidence data, and an independence remember/know (IRK) procedure for recognition data. Because the trial counts per condition were relatively small (≈30 studied and 30 nonstudied items per valence), confidence data were collapsed across participants to generate aggregate ROC curves and parameter reliability was assessed with non‑parametric bootstrapping (10 000 iterations). Statistical analyses compared Encoding vs Placebo and Retrieval vs Placebo separately using 2 (group) × 3 (valence) ANOVAs for hits, false alarms, accuracy and recollection/familiarity estimates, applying Greenhouse–Geisser corrections when appropriate and conducting pairwise t‑tests for contrasts.

Across analyses, emotional valence strongly influenced raw memory performance: negative pictures produced higher hits and accuracy than neutral or positive pictures in both cued recollection and recognition tests. Physiological and mood measures were collected but the extracted text does not clearly report their outcomes here. Encoding versus Placebo: On the cued recollection test raw hits and accuracy showed the expected valence effects but no main effects or interactions involving group. However, DPSD‑derived recollection estimates from bootstrapped aggregate ROC curves were reduced in the Encoding group for negative items (difference 95% CI: 0.004 to 0.297; p = 0.022) and for positive items (95% CI: −0.008 to 0.207; p = 0.034), with the positive effect noted as less reliable. Neutral recollection and familiarity estimates did not differ between groups (all p's > 0.10); there was a trend toward greater positive familiarity in the Encoding group (p = 0.078). Recognition (IRK) analyses comparing Encoding and Placebo again showed no significant group effects on hits, false alarms or overall accuracy, although valence marginally affected false alarms. Recollection estimates from the IRK procedure showed a main effect of valence and a marginal main effect of group (group F(1,38) = 5.289, p = 0.054). Exploratory contrasts indicated that recollection estimates for both negative (t(38) = 2.154, p = 0.038, d = 0.681) and positive (t(38) = 2.164, p = 0.037, d = 0.684) pictures were lower in the Encoding group than in Placebo, with no reliable difference for neutral items. There were no significant effects on familiarity estimates. Retrieval versus Placebo: Cued recollection hits and accuracy again showed robust valence effects but no significant main effects of group. The Retrieval group’s DPSD recollection distributions lay intermediate between Encoding and Placebo; contrasts between Retrieval and Placebo were not significant for negative (95% CI: −0.075 to 0.198; p = 0.193) or positive recollection estimates (95% CI: −0.077 to 0.157; p = 0.235). A trend was observed for reduced neutral familiarity estimates in the Retrieval group (95% CI: −0.057 to 0.421; p = 0.066), but this did not reach conventional significance. Recognition analyses for Retrieval versus Placebo found no main effects on hits, but a main effect of valence on false alarms due to more false alarms for positive pictures. There was a non‑significant trend for the Retrieval group to produce more false alarms to positive stimuli (t(38) = 1.816, p = 0.077, d = 0.075). Exploratory within‑group contrasts in the Retrieval group showed reduced recollection estimates for negative (t(19) = 2.146, p = 0.045, d = 0.480) and positive (t(19) = 2.393, p = 0.027, d = 0.535) items compared with neutral items, an effect not observed in Placebo. No reliable group differences emerged for familiarity estimates. Across tests, MDMA did not consistently change raw proportions of hits or overall accuracy on the cued recollection test, but dual‑process measures indicated selective weakening of recollection for emotional items when MDMA was present at encoding, with subtler and less consistent effects when MDMA was present only at retrieval.

Doss and colleagues conclude that MDMA reduces the recollection of details associated with emotional events, with the strongest evidence when MDMA was administered at encoding and weaker, more variable effects when administered at retrieval. The effect was specific to recollection‑based memory estimates derived from both DPSD ROC analyses and the remember/know IRK procedure; familiarity measures and gross accuracy measures were largely unchanged. The authors interpret this pattern as MDMA weakening the vividness or precision of retrieved details—quality—rather than eliminating the ability to recognise that an event occurred—quantity. The discussion situates these results relative to prior work: whereas earlier studies found MDMA‑related amnesia for neutral material, the present study isolated encoding and retrieval phases and found effects largest for emotionally valenced material. The results contrast with the emotional memory‑enhancing effects of some stimulants (for example dextroamphetamine), and the authors suggest MDMA’s serotonergic actions or 5‑HT2A receptor activity may oppose noradrenergic influences that otherwise boost emotional memory. MDMA at retrieval also tended to increase false alarms for positive items; the authors propose that enhanced visual imagery or fluency for positive content under MDMA may drive such false recollections. Clinical implications are discussed cautiously. On one hand, attenuating recollection of traumatic details could help reduce the intensity of traumatic memories and generalisation of fear, supporting MDMA’s therapeutic potential in PTSD. On the other hand, MDMA’s concurrent reduction of positive recollection might be undesirable in PTSD patients with comorbid depression, a condition already associated with diminished positive memory. The authors note practical concerns: a strong amnestic effect that prevents successful encoding or retrieval could hinder psychotherapy that relies on memory reactivation and updating. They recommend further research to characterise precisely how MDMA modifies memory representations, whether higher doses produce different effects, and how MDMA‑induced distortions compare with those produced by other pharmacological agents.