Dream phenomenology as a dynamic interaction of neural networks: a refined model in the context of hierarchical predictive coding

Objective: to develop, theoretically substantiate, and empirically test a neuro-phenomenological model of dreaming based on three components: Vividness (V), Emotionality (E), and Lucidity (L).

Method: the components are mapped onto three large-scale brain networks: the Default Mode Network with the posterior hot zone (V), the Salience Network (E), and the Executive Control Network (L). The central hypothesis is dynamic interaction and functional segregation. Because these networks are known to be antagonistically coupled, strengthening any two components goes together with weakening the third. The hypothesis was examined in a diary pilot (N = 30). A design for an extended study (N ≥ 150) with an adapted Dream Phenomenology Questionnaire (QPD) is developed. A computational formalization within hierarchical predictive coding is proposed to move beyond correlation.

Results: in the pilot, the combined intensity of any two components correlated negatively with the third.

Conclusion: the model translates subjective dream experience into hypotheses about neurophysiological processes. It can serve as a basis for personalized lucid dreaming therapy protocols for nightmares, PTSD, and alexithymia. These applications are proposals for further research.

Vividness (V) is the perceptual richness and immersion of a dream. The manuscript ties it to the default mode network together with the posterior hot zone, a region identified in high-density EEG studies as a correlate of conscious dream experience. The posterior hot zone produces the perceptual content, and the default mode network integrates it into a narrative with autobiographical context.

Emotionality (E) is the strength of the emotional charge, tied to the salience network, with the anterior cingulate and the insula as key nodes and close co-activation with the amygdala. In REM sleep these limbic regions are highly active, which fits the typical emotional saturation of dreams.

Lucidity (L) is awareness that one is dreaming, tied to the executive control network, with the dorsolateral prefrontal cortex as a key node. In typical REM sleep this network is markedly deactivated, and lucid dreaming goes together with its reactivation.

High vividness and emotionality with low lucidity (V+, E+, L−). The most common state: a vivid, emotional dream without the awareness that it is a dream. A nightmare is its extreme form. It matches the canonical picture of REM sleep, with the simulation and salience networks active and the executive control network off.

High vividness and lucidity with low emotionality (V+, L+, E−). The dreamer knows they are dreaming and explores a vivid world with emotional distance or analytic curiosity. The executive control network is reactivated while the simulation network keeps generating the world.

High emotionality and lucidity with low vividness (E+, L+, V−). A thought-like dream: a strong, often abstract emotion that the dreamer reflects on, with little visual narrative. The manuscript links this pattern to the reports of sparse, schematic dreams in alexithymia.

The manuscript formalizes the components in terms of predictive coding. Vividness corresponds to the precision of generative expectations. Emotionality corresponds to the precision weighting of prediction errors. Lucidity corresponds to the activation of a meta-model that judges the expectations themselves and concludes that the current model is a dream.

In this reading a nightmare is a state locked in a high-precision error with an inactive meta-model. Lucidity with low emotionality is a top-down lowering of the precision of the error signal. These are hypotheses that need empirical verification.

The extended study (N ≥ 150) uses the Dream Phenomenology Questionnaire (QPD). Its blocks are adapted from the Vividness of Visual Imagery Questionnaire (V), the Mannheim Dream questionnaire and mDES (E), and the Lucidity and Consciousness in Dreams scale (L). All answers use one 11-point scale from 0 to 10.

Participants keep a diary for 14 days and fill in the questionnaire within 10 minutes of waking. Correlations are analyzed with Spearman's coefficient. A formal model comparison (AIC, BIC, or Bayesian model comparison) will test the model against the activation-synthesis model and the neurocognitive model. Sleep will be monitored with portable EEG devices. The design also looks at individual differences such as personality traits, anxiety, and alexithymia.

The manuscript proposes that stable dream profiles may serve as a non-invasive proxy for the tonic balance between the three networks in wakefulness, and it outlines two protocols for further study: mnemonic induction of lucid dreams with reality testing for nightmares and PTSD, and dream incubation with intensive visualization before sleep for alexithymia.