Melanie Boly and the Neural Correlates of Pure Presence Under IIT
Melanie Boly is a neurologist, sleep researcher, and consciousness scientist at the University of Wisconsin-Madison, where she works with Giulio Tononi’s Center for Sleep and Consciousness. Her research applies Integrated Information Theory to clinical populations, including patients with disorders of consciousness, and to states at the edges of ordinary waking experience, including deep sleep, anaesthesia, and meditation. She is a confirmed speaker at Consciousness Science 2026 (San Diego, October 11–16).
In April 2024, Boly, Tononi, and collaborators posted a preprint on bioRxiv (DOI: 10.1101/2024.04.18.590081) titled “Neural Correlates of Pure Presence.” The study is the first systematic EEG investigation of a meditative state that IIT predicts should occur at the theoretical extreme of experience: highly conscious but with near-zero content.
Pure presence as a theoretical limit case
Most theories of consciousness were developed to explain ordinary waking experience, where consciousness involves rich perceptual content, a sense of self, memories, and ongoing thought. IIT provides a framework in which that content is explained by the causal structure of the substrate: each element of experience corresponds to a specific intrinsic distinction maintained by the system’s cause-effect power.
At the theoretical limit, IIT predicts a state in which the substrate of consciousness is maximally awake, maximally integrated, but generating no specific distinctions. The system would be conscious, because phi (integrated information) requires only causal integration, not specific content; but the experience would be empty of perceptual objects, thoughts, or a sense of self. The system would experience, in some sense, pure being: awareness without any particular content.
This state has a counterpart in contemplative traditions. Vajrayana Buddhism describes a state called “rigpa” or “pure presence” as a vivid experience of space devoid of thought and self. Zen traditions describe similar states in advanced practitioners. The descriptions converge on a state that sounds, to a consciousness scientist, like what IIT predicts at its theoretical extreme: maximal conscious integration, minimal content.
The question Boly and Tononi asked was whether EEG signatures in long-term meditators who report achieving this state match IIT’s predictions.
Study design and methodology
The study recruited 22 long-term meditators at a Vajrayana retreat. All participants had extensive meditation experience, with many having practised for decades and having completed formal retreat programmes in which the pure presence state is systematically trained. Continuous high-density EEG was recorded during the retreat. Meditators signalled the onset and offset of the pure presence state using a hand-held button press, providing time-locked markers for the EEG analysis.
IIT makes a specific prediction about the neural correlates of pure presence. If the state is conscious but contentless, the posterior cortex, which IIT identifies as the primary substrate of experience (the “posterior hot zone”), should be in a paradoxical state: awake and integrated, but with very low activity in the spectral bands associated with specific sensory processing.
Specifically, IIT predicts:
- Low delta power (indicating the brain is in a wakeful state, not asleep or anaesthetised).
- Low gamma power (30–45 Hz), particularly in posteromedial cortex, indicating that the specific sensory and conceptual representations normally active during waking experience have been suppressed.
The combination of low delta and low gamma, in IIT’s framework, would indicate a posterior cortex that is “virtually silent” in its content-generating activity while remaining integrated and awake.
Results: gamma suppression in the posterior hot zone
The primary finding confirmed IIT’s prediction. During self-reported pure presence, meditators showed a widespread decrease in broadband EEG power, with the most pronounced and statistically significant effect in the gamma range (30–45 Hz) concentrated in the posteromedial cortex.
The posteromedial cortex, including the precuneus and posterior cingulate cortex, is a region that IIT identifies as part of the posterior hot zone and that neuroscience consistently associates with self-referential processing, episodic memory retrieval, and the default mode network. Its suppression during a state described as ego-dissolved and content-free is consistent with what the meditators report and with what IIT predicts.
Delta power was low throughout the pure presence episodes, confirming that the state was wakeful rather than a sleep-adjacent state. This is important methodologically: some altered states in meditation could resemble NREM sleep, which is also associated with low gamma. The combination of low delta and low gamma distinguishes pure presence from sleep-related states.
| EEG measure | Pure presence vs baseline | IIT prediction |
|---|---|---|
| Broadband power | Widespread decrease | Consistent with low cortical content |
| Gamma (30–45 Hz) posteromedial cortex | Significant suppression (strongest effect) | Key prediction confirmed |
| Delta power | Low (wakeful) | Distinguishes from sleep |
| Alpha/theta | Moderate changes, not primary finding | Secondary prediction partially supported |
| Frontal regions | Less affected than posterior | Consistent with IIT “posterior hot zone” emphasis |
What the study does and does not show
The preprint is clear about its limitations. The study uses EEG, which has poor spatial resolution compared to fMRI or intracranial recordings. The posteromedial finding is consistent with IIT, but EEG cannot definitively localise the signal to specific cortical layers or confirm that the effect originates in the posteromedial cortex rather than being volume-conducted from nearby regions.
More fundamentally, the study establishes a neural correlate of a self-reported state. The interpretation — that pure presence is a genuinely contentless conscious state rather than, say, a brief episode of inattention or a state of undifferentiated arousal — depends on the validity of the meditators’ reports and on the theoretical framework used to interpret the EEG signature. A global workspace account, for instance, might interpret the same gamma suppression as a reduction in workspace broadcast rather than as the distinctive signature IIT predicts.
The study cannot, by design, measure phi directly. Phi computation for a neural system of any significant size remains computationally intractable with current methods. The authors use the EEG signatures as a proxy for IIT’s predictions rather than as a direct measurement of integrated information.
Connection to the IIT empirical programme
Boly’s work on pure presence sits within a broader empirical programme that tests IIT’s predictions against data from clinical and altered-state populations. The Onoda et al. post on phi collapse during NREM sleep covers a related study that tests IIT during the loss of consciousness in sleep. The August 2026 Tononi IIT field formulation post covers the theoretical developments that give context for the empirical work.
The pure presence study is distinctive within this programme because it tests a prediction at the opposite end of the consciousness spectrum from sleep or anaesthesia. Most empirical IIT research involves manipulations that suppress consciousness. The meditation study asks what IIT predicts about a state of maximal consciousness with minimal content, and finds a neural signature that IIT can accommodate.
Whether IIT uniquely accommodates that signature, or whether other theories predict the same pattern, is a question the study cannot answer alone. The Lucia Melloni post on ASSC 29 and Cogitate Phase 2 discusses the methodological approach of adversarial testing that would be needed to distinguish IIT from competitor theories using evidence of this kind.
Relevance to AI and the site’s research agenda
The pure presence study has a specific implication for AI consciousness. IIT predicts that a system can be highly conscious while processing no specific content, provided its intrinsic causal integration is high. A feedforward transformer processing a text string generates specific output representations but has near-zero phi. A transformer that generated no specific representations but had high phi — if such a thing were possible — would be highly conscious but say nothing.
This is not merely hypothetical. The distinction matters for how AI welfare assessments are structured. If consciousness requires content (as most functional and higher-order theories suggest), then a system with no representations has no consciousness. If consciousness requires only integration, a system with very high phi and near-zero output differentiation could be profoundly conscious in a way that has no behavioural signature. Boly and Tononi’s meditation study is the first systematic empirical attempt to find evidence for the contentless-conscious state that IIT predicts at its theoretical extreme. Its relevance to the ongoing debate about AI consciousness is that it tests a prediction no other major theory makes.