Microdosing psychedelics has no impact on cognitive function in naturalistic settings

The study used an observational within-subjects longitudinal design over 28 days. Recruitment occurred in two waves (February 20–March 5 and May 12–21, 2020) via a university and social media; 20 participants started and 18 completed the protocol. The final sample comprised 4 male and 14 female participants aged 19–57 (median 22) from Europe, North America and Africa. All participants had prior microdosing experience and provided written consent; ethical approval was obtained from the authors' university review board. Microdosing behaviour and psychological variables were recorded daily using the EthicaData smartphone app. Participants reported whether they had taken a microdose that day (MD 1), the day before (MD 2) or neither day (MD 0), and specified substance and quantity (LSD reported in micrograms; psilocybin mushrooms/truffles in grams). Sleep quality was rated each morning on a 0–3 scale. An index of positive and negative emotion was computed from repeated emotion items. Participants completed an instructional practice session before data collection to reduce practice effects. Cognitive function was assessed remotely using the CNS Vital Signs (CNSVS) battery administered on participants' personal computers. Tests included Symbol-Digit Coding (processing speed), a 4-Part Continuous Performance Test (sustained attention and 2n-back working memory), Stroop (inhibitory control), Shifting Attention Test (cognitive flexibility), and CNSVS verbal and visual memory tests (with delayed recall). Participants were instructed to distribute assessments across MD 1, MD 2 and MD 0 days and to keep testing conditions consistent; up to four assessments per week were allowed (maximum 16 total). Typical battery duration was ~25 minutes. Analyses were performed in RStudio using the lme4 package. Multilevel linear models with random intercepts for participants were used to examine whether Microdose Day (three levels: MD 1, MD 2, MD 0) predicted each cognitive outcome. Model building compared null, linear and unconditional intercept models using ANOVA; microdose day was only added if the unconditional model fit best. Sleep quality and positive/negative emotion were entered stepwise to control for motivational or sleep effects. Intraclass correlation coefficients (ICCs) were reported to indicate the proportion of variance between individuals.

Adherence and dosing: Eighteen completers produced a total of 100 microdosing occasions (each participant microdosed between 1 and 9 times, median 5). Across 40 mushroom occasions the dose range was 0.05–0.3 g (mean 0.163 g, median 0.15 g). For 39 LSD occasions the range was 5–30 µg (mean 13.6 µg, median 15 µg). Two truffle occasions involved 1 g each. The initial 20 participants completed 148 cognitive tests with at least one valid domain score; after matching test days to microdosing reports and excluding invalid scores, 100–104 usable test occasions remained. Per participant the number of usable batteries ranged from 2 to 15 (median 6, mean ~7.4). Timing of remote testing varied widely across and within participants. Processing speed: Analysis used 104 observations nested in 17 individuals. The unconditional multilevel model fit better than simpler models and the ICC indicated about 34% of variance was between individuals. Adding microdose day as a fixed effect did not improve model fit, indicating no detectable effect of microdosing day on processing speed. Sustained attention: The dataset comprised 100 observations nested in 17 individuals. The unconditional model again outperformed alternatives and ICC was about 30%. Microdose day did not improve fit, and this null finding remained when controlling for sleep quality (n=87) and for positive and negative emotion (n=63). Inhibitory control: With 103 observations nested in 17 individuals, the unconditional model was superior and ICC was approximately 50%. Inclusion of microdose day did not improve model fit; controlling for sleep and emotions (n up to 90 and 66 respectively) did not change this result. Cognitive flexibility: Analyses of accuracy and reaction time used 103 observations nested in 17 individuals. Unconditional models fitted better and ICCs were 52% (accuracy) and 54% (reaction time). Microdose day did not account for additional variance in either accuracy or speed, and this null result held when sleep and emotion covariates were added. Working memory: The 2n-back composite used 100 observations nested in 17 individuals. The unconditional model explained the data (ICC ~29%) and adding microdose day did not improve fit. Controlling for sleep quality and emotions (n=87 and n=63) did not alter the null finding. Visual memory: Analyses included 104 observations nested in 17 individuals. ICCs were about 32% for accuracy and 30% for reaction time. Microdose day did not improve model fit for either outcome, and the result persisted when controlling for sleep and affect (n up to 91 and 67). Verbal memory: Using 104 observations nested in 17 individuals, ICCs were roughly 50% for accuracy and 21% for reaction time. Adding microdose day did not improve explanatory power; controlling for sleep and emotions (n up to 91 and 67) again did not change the null result. Overall, across all seven neurocognitive domains tested, the researchers found no evidence that microdosing days (same day or day-after) produced measurable improvements or impairments in cognitive performance in this naturalistic sample.

Dinkelacker and colleagues interpret their findings as evidence that, in a naturalistic setting among people with prior microdosing experience, microdosing does not produce acute improvements or impairments in a broad range of cognitive domains. This outcome contrasts with some prior work that reported dose-dependent improvements in a subset of individuals or small effects on processing speed and sustained attention, but it aligns with other studies that observed no change at low doses. The authors note that medium and higher doses show dose-dependent cognitive impairment in laboratory studies, and their results suggest that typical microdoses in daily practice do not reach thresholds that alter tested cognitive functions. The researchers discuss several possible explanations for the discrepancy between widespread subjective reports of cognitive benefit and the null objective results. One possibility is that beneficial effects accrue cumulatively after prolonged use, which would require longer follow-up than the 28-day window used here; they cite evidence from long-term ayahuasca users showing better executive function over years. Another explanation is that standard neurocognitive tests may not capture the kinds of psychological changes users subjectively experience; subjective improvements might instead be driven by expectation, set and setting, or other non-neurocognitive mechanisms. The authors highlight prior work showing expectancy effects predict well-being changes with microdosing. Strengths emphasised by the authors include the ecological validity afforded by remote, self-organised testing, the within-subjects longitudinal design and the multilevel analytic approach which captures intra- and inter-individual variability. Limitations they acknowledge are important: lack of experimental control over testing conditions (for example, hardware and distance to monitor), wide variability in testing times, modest sample size and variable adherence, absence of placebo control and no direct measurement or control of expectation effects. The investigators also note that remote testing requires incentives and larger samples to achieve adequate power for longitudinal analyses. In closing, the authors suggest that naturalistic remote testing combined with experience sampling offers a promising route to study pharmacological and psychological effects in ecologically valid contexts, but they caution that larger, controlled studies that address expectation and dosing variability are needed to resolve remaining uncertainties about microdosing's effects.