In an observational study of experienced Santo Daime ayahuasca users, acute pre-encoding administration of ayahuasca enhanced hippocampal-dependent episodic recollection, hit rates and overall memory accuracy while not altering familiarity-based memory or susceptibility to false memories. The authors propose beta-carboline activity and repeated prior use as possible explanations for this selective recollection enhancement, which contrasts with prior psychedelic findings.
Background
Ayahuasca is an Amazonian brew with 5-HT 2A -dependent psychedelic effects taken by religious groups globally. Recently, psychedelics have been shown to impair the formation of recollections (hippocampal-dependent episodic memory for specific details) and potentially distort memory while remembering. However, psychedelics spare or enhance the formation of familiarity-based memory (cortical-dependent feeling of knowing that a stimulus has been processed).
Aims
Given the growing literature on the plasticity-promoting effects of psychedelics, we investigated the acute impact of ayahuasca on recollection, familiarity, and false memory in an observational study of 24 Santo Daime members with >500 lifetime ayahuasca uses on average.
Methods
Participants completed a false memory task at baseline and after they consumed a self-selected dose of ayahuasca prepared by their church (average dose contained 3.36 and 170.64 mg of N,N -dimethyltryptamine and β-carbolines, respectively).
Results
Surprisingly, pre-encoding administration of ayahuasca enhanced hit rates, memory accuracy, and recollection but had no impact on familiarity or false memory. Although practice effects cannot be discounted, these memory enhancements were large and selective, as multiple measures of false memory and metamemory did not improve across testing sessions. β-carboline activity potentially accounted for this recollection enhancement that diverges from past psychedelic research. Although ayahuasca did not impact familiarity, these estimates were generally elevated across conditions compared to past work, alluding to a consequence of frequently driving cortical plasticity.
Conclusions
When encoding and retrieval took place under acute ayahuasca effects in experienced ayahuasca users, susceptibility to memory distortions did not increase, potentially owing to enhancements in memory accuracy.
Papers cited by this study that are also in Blossom
Barbosa, P., Strassman, R. J., da Silveira, D. X. et al. · Comprehensive Psychiatry (2016)
Barrett, F. S., Carbonaro, T. M., Hurwitz, E. et al. · Psychopharmacology (2018)
Bershad, A. K., Schepers, S. T., Bremmer, M. P. et al. · Biological Psychiatry (2019)
Bouso, J. C., Fábregas, J. M., Antonijoan, R. M. et al. · Psychopharmacology (2013)
Doss and colleagues situate this study within growing interest in how psychedelics affect episodic memory, that is, the conscious re-experiencing of past events. They note ayahuasca is a brew containing DMT (from Psychotria viridis) and β-carbolines (from Banisteriopsis caapi); β-carbolines render DMT orally active by inhibiting monoamine oxidase A (MAO-A) and have their own psychoactive properties. Previous human and animal work shows psychedelics and other psychoactive drugs can differentially affect memory phases (encoding, consolidation, retrieval) and distinct memory processes, especially recollection (hippocampal-dependent retrieval of contextual detail) and familiarity (a cortical-dependent sense that a stimulus has been encountered). Those earlier findings are mixed: pre-encoding psilocybin and MDMA often impair recollection, whereas familiarity can be spared or even enhanced, and drugs administered around retrieval have been associated with increased false memories for some compounds. This observational study aimed to characterise how ayahuasca acutely influences true and false episodic memory, recollection versus familiarity, and metamemory in a sample of experienced Santo Daime members. The investigators tested participants both sober and while acutely intoxicated by a self-selected ceremonial ayahuasca batch, using a three-phase false-memory paradigm (encoding, misinformation, retrieval) designed to probe suggestion-driven distortions as well as objective measures of recollection and familiarity. They emphasised examining memory when drug effects spanned encoding, misinformation, and retrieval, a situation ecologically relevant to frequent ceremonial use but rarely modelled experimentally.
This was a within-subject, fixed-order observational study in which 24 Santo Daime members completed assessments first while sober (baseline) and then in a second session 24 hours later during which they consumed a self-selected dose of ayahuasca in a ceremonial context. The sample comprised 24 participants (14 males) with extensive prior ayahuasca exposure (mean membership duration 14.2 years; mean reported ceremony attendance 563 events), mean age 55.2 years. Exclusion criteria included pregnancy, recent use of interacting drugs or other psychedelics, recent alcohol use, and MRI contraindications; urine and breath tests were used to confirm recent substance abstinence where relevant. A single batch of ayahuasca prepared by the church was analysed by high-performance liquid chromatography–mass spectrometry; the sample concentrations were reported as 0.14 mg/mL DMT, 4.50 mg/mL harmine, 0.51 mg/mL harmaline, and 2.10 mg/mL tetrahydroharmine, yielding average doses per participant of 3.36 mg DMT and 170.64 mg combined β-carbolines (average liquid dose 24 mL, range 11–40 mL). The primary cognitive probe was a three-phase false-memory task (encoding, misinformation, retrieval). During encoding, participants viewed three of six illustrated scenes (e.g., classroom, beach, ayahuasca ceremony) for 40 s each and rated scene familiarity. After a short filler, participants listened to three narratives purportedly from previous participants that described the scenes but included suggested items (four items actually present and four semantically related items that were not). After another filler, retrieval comprised 72 self-paced cued recollection trials: for each test item the participant indicated whether the object had been present (yes/no) followed by a 4-point confidence rating. Item categories per session were targets, suggested targets, suggested lures, related lures, and unrelated lures; due to a computer error some participants were missing single-item data points in each condition. Analyses focused on hit rates, several false-alarm rates (suggested, related, unrelated), and derived measures intended to control for response bias such as accuracy (hit rate minus unrelated-lure false alarm rate) and precision (hit rate minus related-lure false alarm rate). The investigators also computed suggested-precision and measures isolating the effects of suggestion and relatedness. Repeated-measures t-tests (α = 0.05) compared baseline versus ayahuasca conditions. To estimate recollection and familiarity they applied the dual-process signal detection (DPSD) model, using a bootstrapping procedure (10,000 iterations sampling 24 participants with replacement) to derive distributions and 95% confidence intervals for difference scores. Metamemory (metacognitive efficiency) was estimated via hierarchical Bayesian meta-d′ modelling (HMeta-d Toolbox) with 3 Markov Chain Monte Carlo chains, discarding the first 1,000 samples and retaining 10,000 samples per chain; multiple permutations of targets and lure sets were explored. Exploratory correlations between memory measures and plasma alkaloid concentrations and consumed dose were reported in supplemental material and interpreted cautiously given multiple tests.
Familiarity ratings collected during encoding did not differ between sober and ayahuasca sessions, indicating similar visual exploration of scenes across conditions. Contrary to the pre-registered prediction that pre-encoding ayahuasca would impair true memory, several indicators showed enhanced memory under ayahuasca: hit rates increased (CI = [0.01 to 0.12], t(23) = 2.44, p = 0.023, d = 0.50), overall memory accuracy (hit rate minus unrelated false alarms) increased (CI = [0.01 to 0.13], t(23) = 2.30, p = 0.031, d = 0.47), high-confidence hit rates increased (CI = [0.04 to 0.15], t(23) = 3.34, p = 0.003, d = 0.68), and high-confidence accuracy increased (CI = [0.05 to 0.17], t(23) = 3.90, p < 0.001, d = 0.80). There was a near-significant reduction in high-confidence unrelated-lure false alarms (CI = [-0.00 to 0.04], t(23) = 1.82, p = 0.081, d = 0.37), but these rates were low overall. DPSD modelling indicated a clear enhancement in recollection under ayahuasca (mean difference M = 0.20, SD = 0.08, CI = [0.08 to 0.34], p < 0.001). Familiarity estimates did not differ between conditions (M = 0.07, SD = 0.14, CI = [-0.21 to 0.30], p > 0.250), though the authors note that familiarity estimates at baseline and under ayahuasca were relatively elevated compared with prior datasets. Metamemory (meta-d′/d′) showed no reliable effect of ayahuasca across several operationalisations (e.g., all targets/all lures M = 0.07, SD = 0.12, CI = [-0.17 to 0.26], p > 0.250; other target-lure combinations similarly non-significant). Across the false-memory metrics (suggested, related, unrelated lures), ayahuasca did not increase false alarms; the planned analyses therefore combined scenes and treated the null findings on false-memory measures as robust within this experienced-user sample. Exploratory correlations between plasma alkaloids and memory measures were reported in supplemental material; the authors highlight a negative correlation between plasma harmine and improvements in high-confidence hit rates but advise caution in interpretation.
Doss and colleagues interpret their principal finding—an enhancement of recollection-based episodic memory under pre-encoding ayahuasca in experienced Santo Daime users—as surprising given prior reports of memory impairment from pre-encoding administration of prototypical psychedelics such as psilocybin and MDMA. They emphasise that the enhancement selectively affected hit rates, accuracy, high-confidence hits, and DPSD-derived recollection estimates, while familiarity and metamemory were unchanged. Importantly, susceptibility to false-memory formation (suggestion- and relatedness-driven false alarms) did not increase when encoding, misinformation, and retrieval all occurred under acute ayahuasca effects in these frequent users. The investigators consider several mechanisms. One possibility they outline is pharmacological: the relatively high combined β-carboline dose (170.64 mg) and low measured DMT dose (3.36 mg) might favour MAO-A inhibition or β-carboline-specific actions (for example, harmine-related effects) that enhance hippocampal-dependent encoding. Alternatively, frequent ceremonial use may produce receptor adaptations—such as downregulation of inhibitory hippocampal 5-HT2A receptors—that reduce the encoding-impairing influence otherwise seen with acute 5-HT2A activation. The authors also note the paradox that measured plasma DMT concentrations and subjective effects were comparable to studies with higher DMT doses, leaving multiple pharmacodynamic explanations open. Several limitations and uncertainties are acknowledged. The fixed-order observational design lacked a placebo control and a comparison group of infrequent or naïve users, so effects might reflect practice or other session-order confounds despite attempts to minimise practice effects (use of a practice version prior to baseline, counterbalancing of stimuli). The authors argue practice alone is unlikely to explain the magnitude and selectivity of the recollection enhancement, citing power considerations and comparisons with prior repeated-testing datasets, but they do not rule it out. Another limitation is that ayahuasca was active across encoding, consolidation, and retrieval in the second session, which complicates temporal attribution of effects; the short encoding–retrieval delay also reduces the capacity to isolate consolidation-specific effects. The sample was relatively homogenous and highly experienced, so the findings cannot be generalised to inexperienced populations. For future work the authors recommend controlled trials with placebo and infrequent-user control groups, studies that isolate the timing of drug administration (pre-encoding, post-encoding, pre-misinformation, etc.), and dose–response investigations to disentangle contributions of DMT versus β-carbolines. They note that if β-carbolines can enhance hippocampal-dependent encoding, co-administration strategies might be relevant for therapeutic contexts, but this possibility requires rigorous testing given mixed correlations in the present data and the observational design.
Ayahuasca is a psychedelic Amazonian brew containing N,Ndimethyltryptamine (DMT) from Psychotria viridis and βcarbolines from Banisteriopsis caapi. Although DMT degrades rapidly, the β-carbolines allow it to be orally active by inhibiting monoamine oxidase A (MAO-A). Like psychedelics such as psilocybin and lysergic acid diethylamide (LSD), activation of 5-HT 2A receptors largely contributes to ayahuasca's psychoactive effects, and DMT (Dand ayahuascaare showing promise for the treatment of depression. Nevertheless, DMT possesses a unique pharmacology including σ receptor activation that may induce hippocampal neurogenesisand therefore impact episodic memory. Moreover, whereas standard MAO-A inhibitors (e.g., moclobemide) do not produce substantial subjective effects, β-carbolines can be psychoactive on their ownand to some degree act as 5-HT 2A agonistsand GABA A negative allosteric modulators. Ayahuasca is consumed by several religious groups who take it multiple times per month, and research on these groups can contribute to understanding the impact of long-term psychedelic use on neurocognitive processes (e.g.,. Recently, there has been growing interest in the impact of psychedelics on episodic memory, the conscious re-experiencing of information from the past. Episodic memory is thought to rely on hippocampal and cortical plasticity, and the therapeutic efficacy of psychedelics has been proposed to rely on their induction of plasticity, especially in the cortex. States of heightened plasticity may allow for the revision of maladaptive memories, though they may also provide fertile ground for the formation of memories for non-experienced details of events or false memories. How ayahuasca may impact the formation (encoding), stabilization (consolidation), and remembering (retrieval) of memories is complicated by its polypharmacology. On the one hand, βcarbolines may enhance encoding, as increasing synaptic monoamines prior to the presentation of stimuli can enhance subsequent memory for these stimuli (e.g., dextroamphetamine,, and the MAO-A inhibitor moclobemide can reverse memory impairments from depression, aging, and pre-encoding administration of scopolamine. However, pre-encoding psilocybin administration impairs memory, suggesting that DMT should also impair encoding. Likewise, pre-encoding ±3,4-methylenedioxymethamphetamine (MDMA) administration impairs memoryin a 5-HT 2A -dependent fashion, even though MDMA increases synaptic monoamines, suggesting that 5-HT 2A activation via MDMA's R-enantiomer underlies its encoding impairment. Note that pre-encoding drug effects persist after stimuli are encoded, thereby potentially impacting consolidation, and many studies have administered a drug prior to encoding and tested memory without a sobering delay, thereby also impacting retrieval. To isolate drug effects to consolidation, some studies have administered drugs immediately postencoding and tested memory after a ⩾24-h delay to preclude drug effects on retrieval. Whereas post-encoding increases in monoamines via dextromethamphetamine or methylphenidate administration did not impact memory using this design, post-encoding LSD administration enhanced memory, consistent with animal work. Although no study has isolated drug effects of a prototypical psychedelic to retrieval on an objective memory test (i.e., sober encoding, delay for consolidation, retrieval under drug effects), one study found subjective vividness ratings under psilocybin during autobiographical memory retrieval were enhanced. Nevertheless, the veracity of these memories was not assessed (cf., and various psychoactive drugs, including MDMA and dextroamphetamine during retrieval increase false memory. Susceptibility to false memory formation may especially be true of psychedelics, which enhance mental imagery; but seeand primary suggestibility, factors linked to increased false memory. Another complexity regarding the effects of psychedelics on episodic memory is differential modulation of distinct memory processes, specifically, recollection and familiarity. Recollection is the hippocampal-dependent retrieval of details involving self-specific information (e.g., when and where one experienced an event or idiosyncratic associations one made during encoding), whereas familiarity is the cortical-dependent feeling of knowing that a stimulus has been experienced without retrieving corroborating evidence (e.g., recognizing a face but not remembering how one knows this individual). Pre-encoding administration of GABA A sedatives (e.g., alcohol), NMDA dissociatives (e.g., ketamine), and THC impairs recollection, as well as familiarity. Like these other drugs and consistent with the expression of 5-HT 2A receptors on inhibitory neurons in the hippocampus and the input to the hippocampus (i.e., entorhinal cortex;, pre-encoding psilocybin and MDMA administration impaired memory on a recollectionbased task (verbal-free recall;and recollection estimates from the dual process signal detection (DPSD) model. However, DPSD familiarity estimates tended to be enhanced by pre-encoding psilocybin and MDMA administration, perhaps owing to 5-HT 2A expression on excitatory cortical neurons. Moreover, MDMA administered prior to the encoding of Deese-Roediger-McDermott lists (e.g., bed, rest, awake, etc.) was found to increase false alarms to related lures that may seem familiar (e.g., tired), whereas false alarms to critical lures (e.g., sleep;, which typically involve both recollection and familiarity, were unimpacted (cf.. Interestingly, non-pharmacological studies have found that when recollection fails and familiarity is high, phenomena sometimes reported under psychedelics can occur such as déjà vu, presque vu (illusory feelings of insight,, and premonition. Together, these findings highlight the importance of delineating drug effects on memory phases, memory processes, and stimuli with varying degrees of overlapping content, as such effects can conspire to produce cognitive illusions (cf.. In the present observational study, we examined how ayahuasca impacts true and false memory in members of the Santo Daime church, who have consumed ayahuasca multiple times per month for several years. While sober and under ayahuasca, participants completed an episodic memory task containing a misinformation phase between the encoding and retrieval phases aimed at distorting memory (cf.. How psychedelics could impact false memory when drug effects are present across these three phases is an important question, as a significant event can happen during acute effects that is soon followed by elaboration and querying one's memory. Such a situation may be particularly relevant to frequent ayahuasca users who interact with each other under acute effects during ceremonies. In addition to measuring recollection, familiarity, and false memory, we measured metamemory, the capacity to understand one's own memory, as the etymology of "psychedelic," being "mind revealing," would suggest that psychedelics increase insight into one's own cognitive processes. Although psychedelics have not been found to impact metacognition, increasing synaptic monoamines via dextroamphetamine and dextromethamphetamine during encoding or retrieval has been shown to enhance metamemory.
Twenty-four Santo Daime members (14 males; see Tablefor demographics) were enrolled, an N at least as large as similar pharmacological investigations of episodic memory. Exclusion criteria comprised MRI contraindications, pregnancy (confirmed with urine tests), use of drugs in the past 24 h that interact with MAO-A inhibitors, use of a psychedelic (other than ayahuasca; note that 20 of the 24 participants had participated in a ceremony in the 3 weeks prior to testing) in the past 3 months, use of other psychoactive drugs in the past 7 days (confirmed with urine tests), and use of alcohol in the past 24 h (confirmed with breath tests). These participants were experienced ayahuasca users with a mean (standard deviation (SD)) membership duration of 14.2 (8.3) years, attendance of 563 (650) ceremonies (range: 50-2700), age of 55.2 (10.2) years, and weight of 75.5 (12.6) kg. All participants provided informed consent, and the study was conducted in accordance with the Declaration ofamended in Fortaleza (Brazil, October 2013) and with the Medical Research Involving Human Subjects Act (WMO) that was approved by the Academic Hospital and Maastricht University's ethics committee.
This observational study used a within-subject, fixed-order design and is described in detail elsewhere. Participants completed measures while sober (baseline) and 24 h later under the acute effects of a self-selected dose of ayahuasca (M = 24 mL, SD = 8.16, range = 11-40 mL) consumed in a ceremonial context with other Santo Daime members who were also tested in this study (six members per ceremony). A single batch of ayahuasca was prepared by the Santo Daime church, and the research team was not involved in producing or administering the ayahuasca. Alkaloid concentrations in a sample were determined after dissolution in 25 mL of water using high-performance liquid chromatography coupled to mass spectrometry, which was calibrated with pure reference substances of DMT (Cerilliant, Round Rock, TX, USA), harmine, harmaline (Aldrich Chemistry, St. Louis, MO, USA), and tetrahydroharmine (LGC Standards, Wesel, Germany). This sample contained 0.14, 4.50, 0.51, and 2.10 mg/mL of DMT, harmine, harmaline, and tetrahydroharmine, respectively. Thus, the average doses of DMT and combined β-carbolines were 3.36 and 170.64 mg, respectively.
This task is based on a previous study with several changes. During the baseline session, participants completed a practice version of the task before the full version, and during the ayahuasca session, they began the full version of the memory task ≈160 min post-ayahuasca ingestion. Participants completed other measures before and after the memory task that are reported elsewhere. The memory task consisted of three phases: encoding, misinformation, and retrieval. During each encoding phase, participants viewed three of six illustrated black and white scenes comparable in complexity (classroom, beach, funeral, surveillance room, ayahuasca ceremony, and ayahuasca brewing). Scenes were presented in randomized order each for 40 s and were not repeated between sessions. Participants were instructed to attend to details because their memory would be tested. After each scene, the name of the picture was displayed, and participants had to rate how familiar they were with the scene on a 0-100 scale. Scenes were counterbalanced across participants such that they occurred an equal number of times in baseline and ayahuasca sessions (stimuli and counterbalancing procedure can be found at). After the encoding phase, participants completed a 1-min filler task (Raven's Progressive Matrices)followed by the misinformation phase (cf.. Three narratives were presented each describing a scene from the encoding phase for ≈30 s. Participants were told that these narratives were from former participants describing the scenes from memory and that some information in these narratives may be inaccurate. Each narrative suggested four objects that were actually presented in the scene and four objects that were not presented in the scene but were semantically central to the scene (e.g., a beach towel suggested for the beach scene). After the misinformation phase, participants completed another 1-min filler task (a different version of Raven's Progressive Matrices) followed by the retrieval phase. The retrieval phase consisted of a self-paced cued recollection task containing 72 trials in randomized order. On each trial, participants were asked if they saw a particular object in one of the scenes (yes/no) followed by a confidence rating (4-point scale). Participants were told to only respond based on what they saw in the scenes and not based on the narratives of the misinformation phase. For each scene, there were eight objects presented in the scene and not suggested in the misinformation phase (targets), four objects presented in the scene and suggested in the misinformation phase (suggested targets), four objects not presented in the scene but suggested in the misinformation phase (suggested lures), four objects not presented in the scene or suggested in the misinformation phase but related to objects presented in the scene (related lures), and four objects not presented in the scene, suggested in the misinformation phase, or related to objects in the scene (unrelated lures). Thus, for each participant and drug condition, there were 24 targets, 12 suggested targets, 12 suggested lures, 12 related lures, and 12 unrelated lures. Due to a computer error, 12 participants in each drug condition were missing data for 1 target, and 6 participants in each drug condition were missing data for 1 unrelated lure.
Pre-registered hypotheses and data can be found at. We predicted that ayahuasca would impair (true) memory for stimuli that were presented (due to acute effects on encoding) and increase false alarms across suggested, related, and unrelated lures (due to acute effects during retrieval). We expected there to be a particularly robust increase in false alarms to suggested lures given that psychedelics can increase suggestibility and acute drug effects were present during the misinformation phase. Finally, we predicted that enhanced visual imagery would mediate these false alarms and that false alarms would be overall larger for ayahuasca-related stimuli given the familiarity of these contexts. However, false alarms were not increased (for any of the three types of lures), visual imagery was not enhanced, and there were no differences in memory performance between ayahuasca and non-ayahuasca scenes. Thus, analyses were combined across scenes to minimize comparisons and increase power (cf.. Memory performance for each drug condition was first analyzed for hit rates (p["yes"|target]), suggested hit rates (p["yes"|suggested target]), suggested lure false alarm rates (p["yes"|suggested lure]), related lure false alarm rates (p["yes"|related lure]), and unrelated lure false alarm rates (p["yes"|unrelated lure]). We also explored the data using measures that controlled for response bias. A standard measure of memory accuracy comparable to memory paradigms without misinformation can be computed from the difference between hit rates and unrelated lure false alarm rates. A more fine-grained memory measure ("precision") comparable to tasks containing related lures with a high incidence of false alarm rates (e.g.,) can be computed from the difference between hit rates and related lure false alarm rates. Likewise, precision can be computed for items with suggested information ("suggested precision") from the difference between suggested hit rates and suggested false alarm rates. To determine the effect of suggestion on false memory that controls for an item's relatedness to stimuli from encoding ("suggestion"), related lure false alarm rates were subtracted from suggested lure false alarm rates. To determine the effect of relatedness on false memory that controls for a general response bias ("relatedness"), unrelated lure false alarm rates were subtracted from related lure false alarm rates. High-confidence versions of all these measures can be computed by using only "yes" responses given the highest level of confidence. To test the effects of ayahuasca on these measures and average familiarity ratings of the scenes during the encoding phase, repeated measures t-tests were conducted (α = 0.050). Exploratory correlations between all these measures and plasma concentrations of DMT, harmine, harmaline, tetrahydroharmine, and total β-carbolines, as well as amount of consumed ayahuasca (previously reported inare reported in the Supplemental Material. Given the large number of tests, these correlations should be interpreted with caution. Note that many of the memory performance metrics are related and could be influenced by the same latent processes. For example, an accurate recollection for the scenes could increase hit rates and decrease all types of false alarms, and high familiarity for scenes or misinformation could increase false alarms to related and suggested lures, respectively. Moreover, a good understanding of one's memory or the task (especially after already performing the task once) could influence confidence ratings. Thus, confidence data were submitted to a DPSD analysisto estimate recollection and familiarity and a meta-d′ analysis to estimate metamemory. The receiver operator characteristic (ROC) Toolbox for Matlab was used for DPSD modeling and is discussed in detail elsewhere. Briefly, hit rates are plotted against false alarm rates in a cumulative fashion starting with the highest level of confidence to the lowest level of confidence with the final point being (1,1). A ROC curve is then fit to these points using maximum likelihood estimation, but unlike a standard ROC curve, the y-intercept is allowed to vary. Recollection estimates come from this y-intercept, whereas familiarity estimates come from the curvilinearity of the function. Because the trial count per participant was low for model-fitting, we ran a previously used bootstrapping procedure. For each drug condition, 24 participants (this study's N) were sampled with replacement 10,000 times. On each iteration, confidence counts were collapsed across the sampled participants and fit with the DPSD model to generate distributions of recollection and familiarity estimates. Distributions from the baseline and ayahuasca conditions were subtracted to compute 95% confidence intervals (CIs) of the difference distributions and p-values (the proportion of the distribution greater than 0). Because recollection estimates are zero-inflated when the incidence of high-confidence false alarms is high (the first point of the ROC curve is pulled away from the y-intercept), the DPSD model cannot accommodate false memory effects unless one assumes impaired recollection. Contrary to this assumption, false memory manipulations that produce such reductions in recollection estimates are not accompanied by reductions in hit ratesor performance on recollection-based tasks (e.g., verbal free recall,. Thus, the DPSD analysis only contained targets and unrelated lures. The HMeta-d Toolbox for Matlab was used for hierarchical Bayesian meta-d′ modeling and is discussed in detail elsewhere. Briefly, the probability of a correct response (i.e., p["old"|target + "new"|lure]) is plotted against the probability of an incorrect response (i.e., p["new"|target + "old"|lure]) in a cumulative fashion starting with the highest level of confidence to the lowest level of confidence with the final point always being (1,1). A ROC curve is then fit to these data. However, such "type 2" ROC curves scale with overall memory (i.e., type 1) performance. That is, the metacognitive measure is confounded with how good one's memory is when one should be able to have a good memory and not know this about themselves or a bad memory and know this about themselves. Because each type 1 d′ (i.e., z[hit rate] -z[false alarm rate]) has a "best case" type 2 ROC curve if one were to have perfect correspondence between their type 1 and type 2 performance, a type 1 d′ can be interpolated from the type 2 ROC curve produced by the actual data that assumes such perfect correspondence between type 1 and type 2 performance. This interpolated type 1 d′, termed meta-d′, is in the same units of the d′ produced from the actual data, thereby allowing type 1 performance to be taken into account. The ratio between meta-d′ and d′ is termed metacognitive efficiency and is referred to here as "metamemory." For each drug condition, we used default settings of the HMeta-d toolbox, specifically 3 Markov Chain Monte Carlo chains, discarding the first 1000 samples of each chain to allow for model convergence followed by 10,000 samples. Like the recollection and familiarity estimates, distributions from the baseline and ayahuasca conditions can be subtracted to compute a 95% CI and p-value. Due to the multiple types of targets and lures, we explored several methods for estimating metamemory including all targets (non-suggested and suggested) and all lures (suggested, related, and unrelated), targets and unrelated lures, targets and related lures, and suggested targets and suggested lures.
Tableand Figuredisplay the results of the analysis of memory performance. Familiarity ratings for scenes during encoding did not significantly differ between drug conditions (CI = [-7.11 to 9.91], t(23) = 0.34, p > 0.250), suggesting that participants visually explored scenes similarly sober and on ayahuasca. As shown in Figure(a) and (b), pre-encoding ayahuasca administration (further referred to as pre-encoding ayahuasca) surprisingly enhanced memory, as evidenced by increased hit rates (CI = [0.01-0.12], t(23) = 2.44, p = 0.023, d = 0.50), memory accuracy (CI = [0.01-0.13], t(23) = 2.30, p = 0.031, d = 0.47), and especially, high-confidence hit rates (CI = [0.04-0.15], t(23) = 3.34, p = 0.003, d = 0.68) and high-confidence accuracy (CI = [0.05-0.17], t(23) = 3.90, p < 0.001, d = 0.80). Further evidence for a memory enhancement from pre-encoding ayahuasca came from a near-significant reduction in high-confidence unrelated lure false alarm rates (CI = [-0.00 to 0.04], t(23) = 1.82, p = 0.081, d = 0.37; Figure), though these were near floor. Pre-encoding ayahuasca did not impact other measures, including measures of memory distortion (Tableand Figure).). As would be expected from an increase in high-confidence hit rates without an increase in high-confidence false alarm rates, recollection estimates were enhanced by ayahuasca (M = 0.20, SD = 0.08, CI = [0.08-0.34], p < 0.001; Figure). Although pre-encoding ayahuasca did not impact familiarity estimates (M = 0.07, SD = 0.14, CI = [-0.21 to 0.30], p > 0.250; Figure)), we note that familiarity estimates at baseline and under ayahuasca were rather elevated compared to past work, a point we return to in the Discussion.
Figuredisplays the distributions of metamemory estimates from the meta-d′ analyses (see also Table). Pre-encoding ayahuasca did not impact metamemory using all targets and all lures (M = 0.07, SD = 0.12, CI = [-0.17 to 0.26], p > 0.250; Figure), targets and unrelated lures (M = 0.15, SD = 0.14, CI = [-0.13 to 0.38], p = 0.148; Figure), targets and related lures (M = 0.08, SD = 0.19, CI = [-0.28 to 0.39], p > 0.250; Figure)), and suggested targets and suggested lures (M = 0.12, SD = 0.20, CI = [-0.29 to 0.45], p > 0.250; Figure). Thus, regardless of how metamemory was estimated, there was no evidence that pre-encoding ayahuasca impacted metamemory, consistent with the effects of psilocybin and MDMA at encoding or retrieval.
We found that pre-encoding ayahuasca in experienced ayahuasca users of the Santo Daime church surprisingly enhanced memory, specifically by increasing hit rates, accuracy, highconfidence hit rates, high-confidence accuracy, and DPSD recollection estimates, but had no impact on DPSD familiarity estimates, which were rather elevated at baseline in comparison to previous work. Also contrary to expectations, pre-encoding ayahuasca did not increase susceptibility to false memory formation via suggestion or relatedness. Finally, despite the acute effects of ayahuasca being present during retrieval, there was no evidence that ayahuasca increased false memories to unrelated lures, as found with several psychoactive drugs during retrieval. Despite not correcting for multiple comparisons across these three types of false alarms, there was no evidence that ayahuasca distorted memory in experienced ayahuasca users. Why ayahuasca might enhance the formation of recollectionbased memories in experienced users is perplexing considering that pre-encoding psilocybin and MDMA typically impair memory, and experienced psychedelic users typically exhibit less cognitive impairment than non-psychedelic users, not enhancement, during acute psychedelic effects. Moreover, hippocampal and entorhinal 5-HT 2A receptors are mostly expressed on inhibitory neurons. Although a single dose of psilocybin downregulates 5-HT 2A receptors in both the cortex and hippocampus, one possibility is that frequent ayahuasca use causes greater downregulation of inhibitory hippocampal 5-HT 2A receptors that would otherwise attenuate recollection. Nevertheless, it remains unclear why recollection was enhanced rather than not impaired by pre-encoding ayahuasca. One clue may come from the doses of DMT and β-carbolines. Whereas 170.64 mg of βcarbolines was similar to prior work, 3.36 mg of DMT was 10-20 times lower. Despite this low dose of DMT, plasma concentrations of DMT and subjective effects (seewere similar to those from studies administering ayahuasca containing higher DMT doses. Thus, one possibility is that DMT does in fact enhance the encoding of recollection-based memory in experienced users who purportedly exhibit "reverse tolerance," which would allow even low doses of DMT to impact cognition. Alternatively, MAO-A inhibition via β-carbolines could have driven the recollection enhancement (cf., and some work has found that harmine specifically can enhance memory in animals (Dos Santos and Hallak, 2017). Perhaps even the low dose of DMT contributed to this effect, though microdoses do not typically enhance cognition; but seefor studies with selective enhancements in which caffeine was not permitted to be consumed). If βcarbolines can truly enhance memory encoding, their coadministration in psychedelic therapy may be warranted to counteract the memory impairments otherwise produced by psychedelics. Nevertheless, given the negative correlations between plasma harmine and improvements in high confidence hit rates (the strongest relationship across all correlations, see Supplemental Material but interpret with caution), there may be diminishing returns or even memory impairments with higher doses of β-carbolines. Future work should explore the possibility that β-carbolines could attenuate the psychedelic-induced impairments of forming hippocampal-dependent memories. Another observation was that baseline familiarity estimates in this study were rather high, especially considering the use of a task that taxes recollection. Prior work has found experienced ayahuasca users to have better executive functioningand verbal free recallcompared to non-using populations, though one study found worse verbal free recall in ayahuasca-using versus non-using adolescents. Although it is difficult to assess the significance of the elevated familiarity in the present study, familiarity estimates were ≈25% greater than a reanalysis of several datasets containing cued recollection tasks. Across 40+ placebo and drug manipulations in this reanalysis, familiarity enhancements produced by pre-encoding psilocybin and MDMA were of the few cases that rivaled the elevated familiarity estimates of the present study. Note that enhanced familiarity can increase false memory formation, though the enhancement of recollection may have precluded such effects. A possible explanation for elevated familiarity may be cortical plasticity. Familiarity is thought to be an aggregate signal of perceptual and semantic processing, and a greater number of connections in the cortical processing hierarchy could allow for more cortical spread during encoding, thereby enhancing subsequent familiarity. This study had several limitations such as the observational study design that lacked a placebo condition and control group of infrequent users. Furthermore, because the ayahuasca session always took place after the baseline session, the recollection enhancement under ayahuasca could be due to practice effects. However, while some degree of practice effects was likely present, we believe that practice effects are unlikely to completely account for the recollection enhancement for a few reasons. First, we attempted to minimize practice effects by implementing a practice version of the task prior to the baseline assessment and tested memory for different stimuli that were counterbalanced between the two sessions. Second, the recollection enhancements from pre-encoding ayahuasca were large enough such that a power analysis of this effect on high-confidence accuracy (α = 0.05, power = 0.80, d = 0.80) suggested that it could have been detectable with only 15 participants. The enhancement of recollection was also larger than any drug-mediated enhancement of recollection in a reanalysis of 10 datasets. This reanalysis also contained a dataset in which memory was tested on the same task 16 times with the effects of zolpidem occurring on tests 6-8 and 10-12, yet the recollection enhancements between the 1st and 16th test were much smaller than those reported here (cf.. Furthermore, zolpidem's amnestic effects were large enough to overcome any practice effects, whereas the large memory enhancements in the present study would suggest that even if ayahuasca impaired the encoding of hippocampal-dependent memory, which is typically a reliable effect of prototypical psychedelics and MDMA in humans and animals, such an impairment could not have been very robust. If there were general practice effects, then it begs the question why only recollection was impacted but not other measures on this task. For example, participants did not become better at avoiding false memories nor did any measure of metamemory improve (even without corrections for multiple comparisons), which might be expected if participants had learned how to perform the task better. Moreover, performance on other cognitive tasks was not impacted across sessions. The memory enhancements from pre-encoding ayahuasca were also larger than any of the improvements in executive functioning on a range of cognitive tasks in a study with a similar design (i.e., baseline followed by acute ayahuasca in experienced users,. Nevertheless, some evidence suggests that episodic memory can be more susceptible to practice effects than executive functioning. Given the inflated optimism in psychedelic research, controlled trials are needed to validate this potential encoding enhancement. Another limitation was that ayahuasca was active across encoding, consolidation, and retrieval. Considering that there was little delay between encoding and retrieval, drug effects on consolidation were likely minimized (but seefor how GABAA sedatives can retroactively enhance memory with relatively short encoding-retrieval delays). Furthermore, no drug to our knowledge enhances retrieval (i.e., when drug effects are isolated to retrieval), and when psychoactive drug effects have been isolated to retrieval, including the effects of MDMA, the only effect has been an increase in false memories. Finally, because ayahuasca enhanced recollection, it could have also enhanced memory for the misinformation. Better memory for misinformation could increase susceptibility to false memories, though such an effect may have been mitigated by the enhancement of accurate recollection. One study also suggests that frequent ayahuasca users may be less susceptible to mnemonic interference. Regardless, the present study would suggest that ayahuasca does not increase susceptibility to false memory when an event takes place and is remembered all under acute effects in frequent ayahuasca users (but note that the findings here cannot generalize to inexperienced populations, and our sample was relatively homogenous). Considering that ayahuasca is being explored for its therapeutic potential, it will be important to examine whether inexperienced users are also not more susceptible to memory distortion under the acute effects of ayahuasca. Future work with delays between memory phases (for drug clearance) and precisely timed drug administrations (i.e., preencoding, immediately post-encoding, pre-misinformation, immediately post-misinformation, and just prior to a delayed memory test) will delineate the temporal dependency of drug effects on encoding, consolidation, retrieval, and memory distortion.
Create a free account to open full-text PDFs.
Bouso, J. C., González, D., Fondevila, S. et al. · PLOS ONE (2012)
Rodríguez-Fornells, A., Ribeiro, S., Sanches, R. F. et al. · European Neuropsychopharmacology (2015)
Calder, A. E., Hasler, G. · Neuropsychopharmacology (2022)
Callaway, J. C., Mckenna, D. J., Grob, C. S. et al. · Journal of Ethnopharmacology (1999)
Carhart-Harris, R. L., Kaelen, M., Whalley, M. G. et al. · Psychopharmacology (2014)
Carhart-Harris, R. L., Leech, R., Williams, T. M. et al. · British Journal of Psychiatry (2018)
De Araujo, D. B., Ribeiro, S., Cecchi, G. A. et al. · Human Brain Mapping (2011)
Doering-Silveira, E., Lopez, E., Grob, C. S. et al. · Journal of Psychoactive Drugs (2005)
Dos Santos, R. G., Hallak, J. E. · Journal of Psychoactive Drugs (2016)
Doss, M. K., Mallaroni, P., Mason, N. L. et al. · Biological Psychiatry (2024)
Doss, M. K., Samaha, J., Barrett, F. S. et al. · Biorxiv (2022)
Doss, M. K., Weafer, J. J., Gallo, D. A. et al. · Neuropsychopharmacology (2017)
D’Souza, D. C., Syed, S. A., Safi-Aghdam, H. et al. · Neuropsychopharmacology (2022)
Family, N., Maillet, E. L., Williams, L. T. J. et al. · Psychopharmacology (2019)
Hutten, N. R. P. W., Mason, N. L., Dolder, P. C. et al. · European Neuropsychopharmacology (2020)
Inserra, A., De Gregorio, D., Gobbi, G. · Pharmacological Reviews (2021)
Kloft, L., Otgaar, H., Blokland, A. et al. · European Neuropsychopharmacology (2022)
Sumiyoshi, T., Kraehenmann, R., Pokorny, D. et al. · Frontiers in Pharmacology (2017)
Madrid-Gambin, F., Gomez-Gomez, A., Busquets-Garcia, A. et al. · Biomedicine & Pharmacotherapy (2022)
Mallaroni, P., Mason, N. L., Kloft, L. et al. · NeuroImage (2024)
Olson, J. A. · Neuroscience Insights (2018)
Palhano-Fontes, F., Barreto, D., Onias, H. et al. · Psychological Medicine (2018)
Ramaekers, J. G., Mallaroni, P., Kloft, L. et al. · Journal of Cognitive Neuroscience (2023)
Riba, J., Valle, M., Urbano, G. et al. · Journal of Pharmacology and Experimental Therapeutics (2003)
Sanches, R. F., Osório, F. L., Dos Santos, R. G. et al. · Journal of Clinical Psychopharmacology (2016)
Valle, M., Maqueda, A. E., Rabella, M. et al. · European Neuropsychopharmacology (2016)
Wießner, I., Olivieri, R., Falchi, M. et al. · European Neuropsychopharmacology (2022)
Yanakieva, S., Polychroni, N., Family, N. et al. · Psychopharmacology (2018)
Papers in Blossom that reference this study
Griesfeller, A. F., Kooman, L., Kloft-Heller, L. et al. · Psychopharmacology (2026)
Evens, R., Uyar, A., Gosslau, E. et al. · Psychological Medicine (2026)