Brain networks and dreaming
⸻
V. Toward Network-Based Models: Contemporary Alternatives to the Lateralization Framework
While the study of hemispheric specialization has provided valuable insights into how the brain organizes complex functions, recent advances in cognitive neuroscience suggest that the architecture underlying REM sleep and dreaming is better conceptualized in terms of dynamic, large-scale brain networks rather than static lateralized modules. Network-based models, grounded in empirical findings from functional neuroimaging and computational neuroscience, offer a more nuanced and systemically integrated perspective on the generation and modulation of dream experiences.
A. Dreaming as Emergent from Functional Brain Networks
One of the most compelling frameworks for understanding REM dreaming arises from the study of intrinsic brain networks—collections of anatomically distributed but functionally coordinated regions that co-activate during rest, task performance, and sleep. Among these, the Default Mode Network (DMN), Salience Network (SN), and Frontoparietal Control Network (FPCN) have garnered particular interest in dream research.
The DMN, comprising medial prefrontal cortex, posterior cingulate cortex, precuneus, and medial temporal lobes (including the hippocampus), is centrally involved in internally directed cognition such as self-referential processing, autobiographical memory, and mental simulation. Numerous studies indicate that DMN activity is preserved—or even enhanced—during REM sleep. This suggests that dreaming may represent a continuation or transformation of the brain’s waking capacity for imagination and narrative construction, albeit in a decoupled and internally generated mode. Importantly, DMN activity does not show a strict hemispheric bias, reinforcing the view that dreaming is an emergent property of distributed, bilateral network interactions.
The Salience Network, which includes the anterior insula and dorsal anterior cingulate cortex, plays a key role in tagging emotionally significant stimuli and facilitating transitions between the DMN and task-positive networks. Its activation during REM may help explain the intense emotionality often associated with dream content, especially in dreams emerging from traumatic or stressful experiences. While some asymmetries have been noted in the anterior insula, these tend to vary by individual and task context rather than supporting a consistent lateralization model for dream affect.
In contrast, the Frontoparietal Control Network, responsible for executive functions and metacognitive regulation, tends to be less active during REM sleep, consistent with the reduced cognitive control and bizarre narrative logic typical of non-lucid dreams. However, partial activation of this network—particularly in prefrontal regions—has been observed in lucid dreaming, suggesting that its engagement may correlate with heightened reflective awareness and voluntary control within the dream state. Again, while there may be lateralized contributions in specific cases, the phenomenon appears to depend more on the degree of network integration than on hemisphere-specific activation.
Overall, these findings suggest that dreaming arises not from isolated hemispheric modules but from interactions among distributed networks, whose activity patterns vary with the phenomenological qualities of the dream, such as self-reference, emotional salience, and volitional control.
B. Predictive Coding and Dream Generation
Another influential model reinterprets dreaming through the lens of predictive coding and active inference. In this framework, the brain is understood as a hierarchical prediction machine, constantly generating and updating models of the world to minimize the error between predicted and actual sensory inputs. During waking life, this process is tethered to external input; in REM sleep, however, the brain operates in a decoupled mode, generating internally driven simulations without sensory feedback—a process that may underlie the dream experience.
Dreams, in this view, are simulations in which the brain tests its generative models in the absence of external constraint. They reflect an attempt to optimize internal models by resolving prediction errors through internally constructed narratives. Emotional salience during REM may indicate the “training” of affective predictions, possibly serving adaptive functions related to emotion regulation and memory consolidation.
Crucially, this framework de-emphasizes lateralization in favor of hierarchical interactions among sensory, emotional, and executive regions. The architecture of predictive coding is implemented across both hemispheres, and while certain layers or functions may exhibit asymmetries (e.g., language or affective biases), the key dynamic is precision-weighted inference across levels, not hemisphere-specific dominance.
C. Integration with Contemporary Theories of Consciousness
These network-based perspectives also align with leading theories of consciousness. Global Workspace Theory (GWT), for instance, proposes that conscious experiences arise when information is globally broadcast across a network of specialized processors. REM dreaming, under this model, involves a shift in global workspace content from external sensory input to internal simulation, with altered gating mechanisms. While hemispheric contributions may vary depending on the type of content (verbal, visual, affective), no systematic lateralization is necessary to generate conscious dream states.
Similarly, Integrated Information Theory (IIT) posits that consciousness corresponds to the capacity of a system to integrate information. REM sleep, characterized by high cortical activation and rich internal dynamics, may meet the criteria for high “phi” (Φ), or information integration. This would explain the vivid and immersive nature of dreams. Again, lateralization plays a secondary role here, subordinate to the system’s overall capacity for causal integration across cortical areas.
⸻
In sum, network-based and predictive models of dreaming represent a significant paradigm shift away from hemisphere-centric thinking. While lateralization continues to offer valuable insights into component processes such as emotion, language, and spatial cognition, the phenomenon of dreaming—especially as it occurs during REM sleep—is best understood as a distributed, emergent property of coordinated neural networks. These networks are dynamic, context-sensitive, and shaped by individual differences, rendering simplistic left/right brain models not only insufficient but misleading. The complexity of dreaming demands models that can accommodate its rich phenomenology, neural diversity, and evolutionary depth.
⸻
