Decoding self-reported meditative depth from EEG recordings is feasible. Expert Vipassana meditators (34 people) reported their depth on a 1–5 scale during two sessions, using either traditional probing or a novel spontaneous emergence method. Machine learning models fused spatial, spectral, and connectivity information from theta, alpha, and gamma bands to predict depth across unseen sessions. The spontaneous emergence method produced more frequent reports and correlated better with post-session outcomes than probing. No single EEG channel or default mode network region captured the complex neural dynamics; multivariate patterns were necessary. The findings suggest potential improvements for neurofeedback in meditation.
Meditation depth can be decoded from brain activity measured by EEG in expert Vipassana meditators. A novel 'spontaneous emergence' method, where meditators report their depth on a 1-5 scale only when they feel a shift, outperformed traditional periodic probing and correlated more strongly with post-session outcomes. A new machine learning approach that fuses spatial, spectral, and connectivity information achieved the best accuracy in predicting self-reported depth across separate sessions. Conventional EEG channel-level methods and default mode network regions were insufficient to capture the complex neural dynamics. The findings demonstrate the feasibility of decoding personally defined meditative depth and suggest that 'spontaneous emergence' is a less obtrusive, ecologically valid sampling method.