Digital meditation at scale: a randomized citizen science trial comparing four meditation practices

This preregistered randomized controlled trial evaluated whether app-delivered meditation improved psychological, cognitive, and physiological outcomes relative to an active control, whether four meditation techniques differed in their effects, and whether individual differences moderated response. A total of 1,211 adults were randomized to four weeks of app-guided loving-kindness, mantra, body scan, or nondual meditation, or to a classical-music active control. Self-report (well-being, mood, self-transcendence, nondual awareness, interoception, compassion), cognitive (Stroop, N-back), and physiological (electroencephalography (EEG), heart rate variability (HRV), heartbeat-evoked potentials (HEP)) outcomes were assessed before and after the intervention and analyzed with linear mixed models and false-discovery-rate correction. No preregistered meditation-versus-control outcome survived FDR correction, and Bayes factors additionally favored the null for the temporal EEG measures. None of the four preregistered technique-specific hypotheses was supported. Two self-report outcomes showed arm differences surviving correction on non-predicted contrasts (self-compassion isolation and sleep quality, both q = 0.050), and mantra participants showed faster working-memory reaction time than body scan participants. The four techniques did not differ from one another. In exploratory moderation analyses, trait neuroticism moderated the meditation effect on working-memory reaction time, with higher-neuroticism participants showing greater benefit in cognitive speed than controls. Four weeks of app-based meditation was not superior to an active music control across psychological, cognitive, and physiological measures, and no preregistered technique-specific hypothesis was supported. The neuroticism finding is hypothesis-generating and warrants replication. Endpoint data were available for 43% of randomized participants and missingness was non-random, which limits inference. This study was preregistered on the Open Science Framework (https://osf.io/xdjfq/overview) on October 24, 2024. International Standard Randomised Controlled Trial Number, ISRCTN10867561, retrospectively registered on August​ 12, 2026.