Fast and slow architectures of cortical dynamics in human brain

The human brain navigates a vast temporal landscape, bridging rapid millisecond dynamics of perception and action to the slow patterns of internal physiological regulation. This multi-scale capacity is traditionally attributed to a unitary, hierarchical gradient of intrinsic timescales. However, this unidimensional model fails to explain how local circuits can simultaneously sustain rapid behavioral responses and slow, body-coupled dynamics. Here, we resolve this fundamental tension by demonstrating that the human brain is organized into two parallel and functionally dissociable temporal architectures. Using a novel method, SPLIT (spectral piecewise-linear inference of timescales), applied to large-scale stereo-electroencephalography (8,619 contacts from 185 individuals), we dissociate fast ([~]10-100 Hz) and slow ([~]1-10 Hz) temporal components. We show that the fast-component timescales exclusively organize into the canonical sensory-to-transmodal hierarchy and capture fine-grained anterior-posterior gradients within the hippocampus and insula. In contrast, the slow-component timescales lack hierarchical structure and are selectively coupled to cardiac-related physiological signals. This dual architecture persists across wakefulness, resting state, sleep, and anesthesia, suggesting an intrinsic organizational principle. Our findings overturn the unitary view of brain timescales, establishing a bipartite framework where parallel temporal components simultaneously orchestrate perception, cognition, and action while maintaining neurovisceral regulation.