Ayahuasca-inspired DMT/harmine formulation alters creative thinking dynamics during artistic creation

Creativity is increasingly recognised as a crucial capacity for adaptability, innovation and problem solving, and impaired cognitive flexibility is implicated in several mental health conditions. Psychedelic drugs are hypothesised to promote cognitive flexibility and creative thinking, but empirical findings are mixed. Prior studies have reported enhancements, impairments or null effects on divergent and convergent thinking depending on drug type, dose, setting, timing of measurement and study design. The authors note a further gap: most laboratory studies assess creativity with isolated tasks rather than tracking how creative thinking unfolds over time during real-world creative work. To address this gap, Suay and colleagues examined acute effects of an ayahuasca-inspired pharmaceutical formulation combining N,N-dimethyltryptamine and harmine (DMT/HAR), harmine alone (HAR), and placebo, using a double-blind, placebo-controlled, within-subject design. They assessed micro-level creative cognition with standard tasks (convergent thinking via the picture concept task, and divergent thinking via the alternative uses task) and macro-level creative process dynamics during a 21-minute painting task using the creative process report diary (CRD) with experience sampling. The study also tested whether subjective altered-state experiences (measured with the 5D-ASC subscales) predicted changes in task-based creativity, aiming to integrate cognitive, phenomenological and process-level perspectives on psychedelic effects on creativity.

The study used a double-blind, placebo-controlled, within-subject design with three experimental sessions (DMT/HAR, HAR, placebo) separated by at least two weeks. Thirty-six healthy, right-handed male volunteers aged 20–40 were recruited according to strict medical and psychiatric inclusion criteria; after dropouts and a technical failure, 30 participants were included in the analyses. On each study day participants underwent urine and breath alcohol screening, received a 100 mg harmine or placebo orodispersible tablet sublingually, and then an intranasal spray (DMT or placebo) delivered intermittently in ten 10 mg doses at 15-minute intervals (one exception with a 30-minute interval during a task). Creativity testing began 110 minutes after the final nasal spray; the setting was a comfortable, soundproof laboratory room with standardised background music and medical supervision throughout. Micro-level creativity was assessed with the picture concept task (PCT) for convergent thinking and the alternative uses task (AUT) for divergent thinking. The PCT involved 12 timed pictorial stimuli with one correct association per stimulus; PCT parallel versions were used. The AUT presented three object prompts (pen, towel, shoe) administered in separate weeks; four AUT dimensions (fluency, flexibility, originality, elaboration) were scored by two independent raters and averaged. Macro-level creativity was measured with a 21-minute free painting task on a Wacom tablet using Adobe Fresco; every 3 minutes participants completed the CRD experience sampling indicating which of six creative stages they had engaged in (Preparation, Incubation, Ideation, Illumination, Production, Validation). Transition counts between stages and stage-selection frequencies were treated as dependent variables. Subjective effects were assessed with the Five-Dimensional Altered States of Consciousness (5D-ASC) scale approximately 360 minutes after first dose, focusing on subscales hypothesised to relate to creativity (insightfulness, visionary restructuralization, changed meaning of percepts, impaired cognition, vigilance reduction and overall ASC). Statistical analyses used JASP, Python and R. Repeated-measures ANOVA (rmANOVA) assumptions were checked (Shapiro–Wilk, Greenhouse–Geisser where needed), p-values were adjusted for multiple comparisons (Bonferroni; post-hoc tests with FDR correction for some analyses) and significance was set at p = 0.05. Mixed-effects models and intraclass correlation coefficients (ICC) quantified individual variability. For transition analyses a two-step data-driven filter was applied: retain transitions with Cohen’s d > 0.5 and one-way ANOVA p < 0.1, then perform rmANOVA and paired tests (final transition sample for these analyses was n = 28 after excluding two participants who selected all options). For CT and DT predictive models, difference scores ([DMT/HAR − PLA] etc.) were used in multiple linear regressions with backward elimination.

Micro-level convergent thinking (PCT): Mean PCT scores were lower under DMT/HAR (M = 8.41 ± 1.59) and HAR (M = 8.82 ± 1.05) than placebo (M = 9.23 ± 1.34). The rmANOVA produced a marginally non-significant main effect of condition (p = 0.064), while a mixed-effects model indicated substantial individual variability (ICC = 40%). An interaction showed that participants with higher placebo PCT performance experienced larger declines under DMT/HAR (β = -0.79, p < 0.001). Targeted post-hoc contrast comparing DMT/HAR to placebo found a significant reduction in CT (t(42) = -2.594, p = 0.013, Cohen’s d = -0.607). Contrasts for HAR versus placebo and DMT/HAR versus HAR were non-significant. Micro-level divergent thinking (AUT): The rmANOVA found no significant overall effect of condition across AUT dimensions (p = 0.105) and no interaction between drug and AUT dimension. Mixed-effects models showed high inter-individual variability (ICC range 15.7%–48.5%), with fluency and elaboration particularly variable. Contrast analyses revealed trend-level reductions under DMT/HAR for fluency (Cohen’s d = -0.47, p = 0.07) and elaboration (Cohen’s d = -0.43, p = 0.09); these did not meet conventional significance. HAR produced small, non-significant increases in elaboration and originality. Macro-level painting and CRD: Across the six stage categories there was no main effect of drug condition and no condition × stage interaction; stage selection differed overall, with Production selected more often than several other stages (p < 0.001). Exploratory stage-specific analysis identified an effect of condition on Incubation selection (F(2,54) = 4.072, p = 0.022) and trend-level reductions in Incubation under DMT/HAR (t(27) = -1.987, p = 0.057, p corr = 0.085) and HAR (t(27) = -2.405, p = 0.023, p corr = 0.067) relative to placebo, though these did not survive strict multiple-comparison correction. Transition-matrix visualisations showed distinct patterns between conditions: placebo had balanced persistence and switching, HAR appeared more restricted, and DMT/HAR more diverse. Using the pre-specified filter (d > 0.5 and p < 0.1) several transitions were prioritised for analysis. Confirmatory tests showed a significant reduction in Incubation → Incubation self-transitions under both DMT/HAR and HAR compared to placebo (p corr = 0.026). Trend-level reductions were observed for Incubation ↔ Preparation transitions (p corr values around 0.055–0.065). Notably, the Incubation → Illumination transition showed a trend-level reduction unique to DMT/HAR relative to both HAR and placebo (p = 0.069), suggesting a disruption in the reflective-to-insight progression under the combined formulation. Subjective experiences and prediction of performance: 5D-ASC scores differed by condition (all p < 0.001), while HAR was reported as subjectively similar to placebo in the initial rmANOVA. Multiple regression analyses using difference scores found no significant predictors of CT changes. For DT, fluency was predicted by changes in meaning of percept (β = -0.827, p = 0.007) and vigilance reduction (β = 0.453, p = 0.014), with the model explaining 57.3% of variance (R2 = 0.573). Insightfulness was marginally predictive of fluency. Elaboration was predicted by changes in meaning of percept (β = -0.726, p = 0.038) and insightfulness (β = 0.619, p = 0.047), explaining 59.3% of variance (R2 = 0.593). No reliable predictors emerged for originality or flexibility. Models showed acceptable multicollinearity diagnostics (VIF < 10).

Suay and colleagues interpret their results as evidence that DMT/HAR does not uniformly enhance creativity; rather, effects differ by cognitive mode and process stage and vary across individuals. They highlight a dissociation between subjective reports of altered, often creative-feeling experiences and objective task performance: DMT/HAR significantly reduced convergent thinking ability (particularly in participants with higher baseline CT), while divergent thinking measures were largely preserved though some elaboration and fluency metrics trended lower. The authors link CT impairment to possible disruptions in top-down control networks (frontoparietal and salience networks) and 5-HT2A-mediated mechanisms, proposing that weakened goal-directed cognition may underlie the observed declines. At the same time, they suggest this transient ‘‘loosening’’ might be therapeutically useful in contexts where rigid negative thinking predominates. On macro-level dynamics, the investigators report reduced engagement with incubation-related processes under both DMT/HAR and HAR, evidenced by fewer incubation selections and a significant reduction in incubation self-transitions. They argue that the combined formulation (DMT/HAR) uniquely disrupted the Incubation → Illumination transition at trend level, implying that insight under the drug may arise via alternative, more fluid associative pathways rather than through the conventional incubation-to-insight sequence. Harmine alone produced some similar incubation effects, raising the possibility that MAO-A inhibition or harmine-specific cognitive effects contributed to changes in reflective processing. The authors acknowledge several key limitations: the all-male sample limits generalisability; substantial inter-individual variability suggests heterogeneous responses and the need to study baseline cognitive traits and neural factors; some macro-level findings required exploratory filtering and did not survive strict corrections, indicating possible power limitations or sensitivity issues in the task design; and expectancy or unblinding may have influenced results, suggesting the value of active placebos in future studies. They recommend longitudinal and comparative work (different psychedelics, dose–response) and refined psychometric and real-time assessment tools to better capture the multifaceted relationship between subjective experience and objective creative outcomes. Practically, the authors propose that psychedelic-assisted approaches may be better suited to unstructured ideation phases, with sober periods used for structured refinement and evaluation of ideas.

The study concludes that an ayahuasca-inspired DMT/harmine formulation acutely impairs structured, goal-directed convergent thinking while leaving core divergent idea generation largely intact, and that both DMT/HAR and harmine modulate incubation-related dynamics during an artistic creation task. DMT/HAR appears to alter the canonical incubation-to-illumination pathway, favouring more fluid and free-associative transitions to insight. Subjective altered-state experiences selectively predicted facets of divergent thinking (fluency and elaboration) but did not predict convergent problem-solving. The authors emphasise that psychedelics should not be assumed to universally enhance creativity; instead, their utility may depend on the stage of the creative process and individual differences. They call for future studies addressing long-term effects, dose dependency, broader samples and comparative compounds to refine understanding of how psychedelics influence creative cognition and its dynamics.