Mapping Content and Dynamics in the Stream of Consciousness: a Latent Brain State Analysis
Yuhua Yu, Quentin Raffaelli, Delaney Freveletti, Jack Joseph Brossart, Ramsey R Wilcox, Matthew D. Grilli, Jessica R Andrews-Hanna
Study at a glance
AI-extracted from the abstract| Characteristics | Observational neuroimaging study |
|---|---|
| Population | Individuals speaking their unprompted thoughts aloud during fMRI scanning |
| Measures | latent brain state modelling of fMRI data, large-language-model derived metric of semantic perplexity |
| Key findings | The authors report that unprompted thoughts are supported by a multistage brain architecture: a core default mode network state tracks early internally oriented thought, an executive control and dorsal attention network state tracks externally oriented thought, and a medial temporal lobe and dorsal attention network state engages later during verbalization and is associated with novel content generation. They propose that default mode subsystems interact recurrently with attention and control networks to scaffold the fluid progression of spontaneous thought. |
Abstract
Human thoughts often arise unprompted, forming a stream of consciousness. Despite their ubiquity, it remains unknown how the brain supports unprompted thoughts as an integrated experience encompassing diverse content embedded in rich temporal dynamics. Using latent brain state modelling of fMRI data collected while individuals spoke aloud their unprompted thoughts, we identified recurring states linked to the moment-to-moment thought content and dynamics. A brain state with activation in core default mode network (DMN) regions tracked early emergence of internally oriented thought, while a state involving executive control and dorsal attention networks tracked externally oriented thought. Another state involving the medial temporal lobe and dorsal attention networks engaged later during internal thought verbalization, suggesting roles in elaboration. The latter state was also associated with generating novel content, indexed by a large-language-model derived metric of semantic perplexity. Our findings reveal a multistage architecture of unprompted thoughts, where DMN subsystems interact recurrently with attention and control networks to scaffold their fluid progression.