Psilocybin and 4-Bromo-2,5-Dimethoxyphenethylamine (2C-B) at Encoding Distort Episodic Familiarity

Psychedelics, defined here as hallucinogenic 5-HT2A agonists, are being investigated both for therapeutic potential and for their pronounced effects on consciousness. Prior human and animal work indicates that psychedelics can differentially affect subprocesses of episodic memory: they tend to impair hippocampal-dependent recollection (retrieval of specific contextual details) while in some cases sparing or even enhancing cortical-dependent familiarity (a feeling of knowing without recollection). This pattern contrasts with other drug classes (e.g., alcohol, ketamine, THC) that more uniformly impair both recollection and familiarity. Emotion interacts with drug effects on encoding in other pharmacological studies, with negative and positive memories often being differentially modulated compared with neutral memories. Doss and colleagues set out to test whether two chemically distinct psychedelics—psilocybin (a tryptamine) and 4-bromo-2,5-dimethoxyphenethylamine (2C-B, a phenethylamine)—produce similar effects on the encoding of emotional episodic memories. The study aimed to compare effects on recollection, familiarity, and metamemory (the correspondence between confidence and accuracy) for negative, neutral, and positive stimuli. The hypothesis was that, relative to placebo, both drugs at encoding would impair recollection, enhance familiarity, and possibly enhance metamemory, with preferential effects on emotional (negative and positive) memories.

This was a double-blind, placebo-controlled, within-subjects repeated-measures study with three experimental arms (placebo, psilocybin, 2C-B). Twenty-two participants were recruited, with 20 completing all memory conditions (10 males; age M = 25.00, SD = 4.91). Inclusion criteria included age 18–40, prior psychedelic experience, and BMI 18–28 kg/m2; exclusions included psychiatric history, recent psychedelic use, certain medical contraindications, pregnancy, and recent substance use. Participants were medically screened, including ECG and urine/blood tests, and gave informed consent under ethical approval. Each participant completed a preparatory session with a practice memory task. Experimental sessions were separated by at least 14 days and drug administration orders (six possible sequences) were counterbalanced across participants. On each first session of an arm participants received a capsule containing placebo, 15 mg psilocybin, or 20 mg 2C-B and performed the encoding phase roughly 185 minutes after dosing; memory was tested ≈24 hours later while participants were sober. These doses were chosen as moderate and roughly comparable in experiential duration. The emotional episodic memory task used 360 images from standard affective picture sets, each paired with a 3–5 word label. Images were categorised as negative, neutral, or positive by valence ratings and divided into six matched stimulus sets; target and lure assignments were counterbalanced across participants. The encoding phase presented 120 label–picture trials (half the trials contained the picture paired with its label) and required arousal ratings for labels and pictures, although these ratings were lost due to a computer error. The following day participants completed a cued recollection test (labels presented, yes/no whether a picture had been paired with the label, plus 4-point confidence) and a picture recognition test (pictures presented, yes/no recognition plus remember/know judgements). One participant did not complete the picture recognition test in all conditions, leaving n = 19 for those analyses. Analyses focused on drug (placebo, psilocybin, 2C-B) × emotion (negative, neutral, positive) repeated-measures ANOVAs (α = .05). For recollection measures, hit rates, false alarm rates, accuracy, and high-confidence variants were computed. For recognition, hit/false alarm rates were computed separately for "remember" and "know" responses, and an independence remember/know (IRK) correction was applied to estimate familiarity in the absence of recollection. Dual-process signal detection (DPSD) modelling of confidence-based receiver operating characteristic (ROC) curves was used to estimate recollection (y-intercept) and familiarity (curvilinearity). Because of low trial counts, a bootstrapping procedure sampled N participants with replacement 10,000 times per drug×emotion cell to generate distributions of DPSD estimates; between-condition contrasts were computed from difference distributions. Metamemory was estimated using meta-d' (HMeta-d toolbox) and expressed as metacognitive efficiency (meta-d'/d'); default MCMC settings were used (3 chains, discard first 1,000 samples, then 10,000 samples retained). Additional robustness checks included collapsing emotion conditions and re-analysing after excluding participants with chance-level cued recollection (reported in Supporting Information).

Twenty participants contributed to the primary analyses (19 for picture recognition). The cued recollection measures provided modest and mixed evidence for impaired recollection: there were trends and some significant contrasts for reductions in high-confidence hit rates, high-confidence accuracy, and modelled DPSD recollection estimates, but many omnibus effects did not reach conventional significance (several reported main effects/interactions had F < 2.00, p > .100). A drug × emotion interaction for remember hit accuracy trended toward significance (F(4,72) = 2.13, p = .086, ηp2 = .11) and reached significance in supplementary analyses; psilocybin and 2C-B showed their largest recollection impairments on neutral stimuli in some analyses. By contrast, consistent drug effects emerged in measures related to false alarms and familiarity. Overall false alarm rates were increased after psilocybin and 2C-B at encoding (F(2,36) = 3.33, p = .047, ηp2 = .16). This increase was driven by IRK-corrected familiarity false alarms on the picture recognition test (F(2,36) = 5.99, p = .006, ηp2 = .25), whereas remember-based false alarm rates did not show the same drug effect. DPSD-derived familiarity estimates were reduced under psilocybin and 2C-B, a finding the authors note could result from misattributed heightened familiarity rather than a true loss of familiarity strength. Valence did not significantly interact with drug in omnibus tests (drug × emotion F(4,72) = .97, p > .250), yet elevated familiarity-based false alarms were apparent selectively for negative and positive pictures in follow-up contrasts. These familiarity-related effects remained robust in supplementary analyses reported in the Supporting Information. Metamemory results were heterogeneous: psilocybin and 2C-B at encoding tended to reduce metamemory for negative and neutral memories while tending to enhance metamemory for positive memories, though these effects were weaker and less reliable in additional analyses. Bootstrapped DPSD and HMeta-d procedures were used to obtain distributions and contrast p-values; the authors also report that excluding participants with chance performance did not qualitatively change the main findings (details in SI).

Doss and colleagues interpret their findings as evidence that both psilocybin and 2C-B distort episodic familiarity when administered at encoding. While only modest and inconsistent evidence was observed for impaired recollection—contrary to some prior reports—the study found clearer drug-related increases in familiarity-based false alarms and reductions in DPSD-derived familiarity estimates. The authors argue that these results are best explained by a misattribution mechanism: psychedelics may heighten familiarity for stimuli (including labels), leading participants to infer that associated pictures had been seen and thereby producing more low- and moderate-confidence false alarms on recognition tests. The investigators place these results in the context of earlier pharmacological memory research, noting differences in prior findings that may stem from methodological factors. One important procedural difference is timing: encoding here occurred ≈185 minutes after dosing, later than in some prior studies that tested near peak effects (≈120 minutes), which could alter whether familiarity is enhanced or impaired. The use of labels as lures (so lures were not entirely novel) may have amplified misattribution effects, and the 24-hour test delay leaves open contributions from consolidation or lingering effects at retrieval. The authors also note that prior reports of enhanced familiarity following psychedelics (or MDMA) sometimes tested at different delays or with different lure constructions, complicating direct comparisons. Key limitations acknowledged by the authors include the inability of the design to isolate encoding-specific effects from post-encoding consolidation or retrieval influences, the modest sample size limiting power to detect subtle differences between drugs, and task-specific factors such as lost arousal ratings and non-novel lures. They recommend larger and more fine-grained comparisons across psychedelic compounds and manipulations that can dissociate encoding, consolidation, and retrieval. Finally, the authors suggest that a familiarity-based misattribution may be a common neurocognitive mechanism across chemically distinct psychedelics and could relate to phenomenological features such as the "noetic" quality, déjà vu and presque vu, though they frame such links as hypotheses for further research rather than established conclusions.