Conference Agenda

Perceptual content is substantially colored by prior beliefs according to predictive processing. Some priors are particularly ‘stubborn’, as exemplified by bistable perception paradigms such as binocular rivalry: a phenomenon triggered when distinct stimuli overlap spatio-temporally, resulting in perceptual alternation rather than a veridical continuous mixed percept. Eastern contemplative traditions propose techniques to reduce the influence of priors and expand the boundaries of conscious perception. In this study, we examined whether volitional control over rivalry is possible via meditation-induced attention modulation. Twenty-four advanced meditators were exposed to flickering rivalry stimuli in focused attention (FA), open monitoring (OM) and no-meditation conditions, each consisting of self-report, no-report, and localizer blocks. We hypothesized FA – by upweighting attention to the currently perceived stimulus – would increase individual percept duration, whereas OM – by being attentive equally to all aspects of experience – would induce longer mixed percepts. This, in turn, should decrease perceptual switches in both meditation conditions. Switches were computed using behavioral self-reports and two complementary EEG analysis methods: frequency-tagging and pattern-classification. In line with the hypotheses, behavioral reports revealed fewer switches in both meditation conditions. However, frequency-tagging estimated switch rates showed no effect of meditation in self-report blocks; in no-report blocks, switches increased only with FA. Interestingly, initial pattern-classification results show high decoding accuracy and high correlation between estimated switches and self-reports in the no-meditation condition, demonstrating a potential for rivalry-tracking. Results from larger sample will be presented to illuminate the extent to which control over conscious perception is possible through meditation.

A growing body of theoretical and empirical work posits complexity science as a powerful approach for consciousness research, yielding remarkable results in characterizing global conscious states. However, neural complexity in meditation remains underexplored, existing literature yielding seemingly ambiguous results.

To address this gap we combine two complementary efforts: 1) a systematic literature review on neural complexity in meditation; 2) a MEG neuro-phenomenological study (n=45) exploring Lempel-Ziv complexity (LZc) and Transfer Entropy (TE) in waking rest vs. meditative states of maintaining and reducing self-boundaries.

Our review shows increased complexity in meditation vs. rest (d=0.6) and a trend of decreased baseline (trait) complexity following regular meditation practice. Accordingly, our empirical results show widespread LZc increases in meditation vs.rest (t=6.15, p<0.001), and specific orbitofrontal cortex decrease highly correlated (r=-0.64, p<0.001) with self-dissolution phenomenology. Similarly, TE increases globally in meditation vs. rest (t=5.98, p<0.001), while specific TE decreases in default-mode and fronto-parietal networks correlate with self-dissolution phenomenology (r=-0.46, p<0.01).

Our review and novel analysis both confirm meditation as a global conscious state characterized by increased neural complexity, yet highlight nuance showing subtle phenomenological aspects correlate with localized reductions in complexity. Although further theoretical work is needed, these findings align with predictive processing hypotheses, suggesting meditation may reduce high-level priors while increasing precision-weighting of sensory input. Taken together, our findings both solidify existing evidence and shed new light on the neuroscience of consciousness, highlighting meditation as a unique state where the interplay of global and local patterns of neural complexity underlie a unique phenomenology.

Meditation includes contemplative practices that are informed by ancient wisdom traditions. Interest in meditation, and particularly mindfulness, is burgeoning, and it has been shown to be effective in improving mental and physical health. Our Meditation Research Program at Massachusetts General Hospital and Harvard Medical School is leading investigations into advanced meditation, that is, skill, states, stages, and transformations that unfold with meditation mastery and practice over time. We will present several of our recent studies including using human electrophysiology (MEG-EEG) and neuroimaging (7T MRI) combined with neurophenomenological approaches to examine the neuroscience and phenomenology of advanced concentrative absorption, insight, and meditative endpoints including jhana, vipassana, and cessation. Absorption is characterized by profound ecstasy, clarity, and openness; insight includes perceived deep understanding of one’s self and mind; and cessation is complete discontinuation of consciousness. We have intensively sampled these states and events in cohorts of advanced meditators. Our results include spectral power, linear and non-linear connectivity, graph network, directed information, entropy, machine learning, microstates, and criticality insights derived from electrophysiology; and activity, static and dynamic connectivity, graph network, gradient, and eigenmode insights derived from whole-brain 7T fMRI (including brainstem and cerebellum). By relating advanced meditation to objective and intrinsic measures of the brain and rich phenomenology, these results provide powerful evidence for advanced meditators voluntarily and radically modulating their own consciousness. These studies lay foundations for studying these unique states using neuroscience toward health-related applications in both clinical and non-clinical contexts, and deep insights in the science of consciousness.

Introduction.

Hypnosis, a form of top-down regulation that can produce pronounced changes in consciousness, provides a valuable model to test theoretical frameworks such as the Global Neuronal Workspace theory (GNWT). We leveraged two hypnotic paradigms (deafness and blindness) to examine GNWT’s predictions: (1) preservation of the early (<200ms) perceptual processing predicted to correspond to an unconscious stage; (2) abolished or severely reduced late-stage of processing predicted to index conscious access (late ERP component and time-generalization measures); (3) decreased long-range functional connectivity reflecting decoupling between GNW and the corresponding sensory area.

Methods

We reanalyzed two existing odd-ball EEG datasets from healthy volunteers with varying hypnotic suggestibility. The first (auditory) included 45 subjects (23 highly suggestible) tested under baseline and hypnotic deafness; the second included 60 subjects (20 highly suggestible) tested under baseline and hypnotic blindness. We assessed stimulus-induced responses (ERPs, time-frequency, and temporal decoding) as well as ‘state markers’ (spectral power, algorithmic complexity, and functional connectivity).

Results

Results were consistent across modalities. Crucially, as predicted, late-stage (>300 ms) stimulus-induced activity was significantly reduced in hypnotic conditions (as revealed by ERPs, time-frequency, and temporal decoding analyses), while early (<300 ms) processing was unaffected. Moreover, in highly suggestible individuals, hypnosis modulated algorithmic complexity and functional connectivity, whereas low suggestible individuals showed no significant changes.

Discussion

These findings support GNWT predictions that hypnosis acts by modulating late-stage neural processing, although alternative interpretations remain possible.

The subjective experience of time flowing forward remains one of consciousness's most puzzling features. While predictive processing theories suggest this experience emerges from the brain "distrusting the present" in anticipation of predicted changes (Hohwy et al., 2016), the precise computational mechanisms underlying this phenomenon remain unclear. Here, we propose that continuous Bayesian model comparison between the hypothesis "the world changed" and the null "did-not-change" is a fundamental mechanism of perception through which temporal consciousness emerges. Our framework builds on a hierarchical event-based architecture (Zakharov et al., 2022) where the perceptual system continuously evaluates evidence for environmental change versus stability. This process implements the theoretical notion of "distrusting the present" through concrete probabilistic computation. Importantly, this mechanism serves multiple crucial functions: it enables event-based perception by segmenting continuous experience into discrete events, promotes optimal hierarchical organization of perceptual processing into naturally evolving nested timescales, and enables energy efficiency. When implemented in machine learning systems, this architecture achieves state-of-the-art performance in compression and video prediction tasks, suggesting its fundamental computational importance. Combined with previous work on surprise-based duration estimation (Fountas et al., 2022), our model accounts for how we segment continuous experience and process duration while also elucidating why consciousness exhibits its characteristic temporal flow. Furthermore, it suggests testable predictions about how alterations in this hypothesis-testing process might manifest in conditions with disturbed temporal awareness, such as depression (Kent et al., 2019). This work bridges computational and philosophical approaches to temporal consciousness, offering a unified account of time perception and its phenomenology.