Sub-Second Fluctuation between Top-Down and Bottom-Up Modes Distinguishes Diverse Human Brain States.

The human brain shifts between two directional modes on a sub-second timescale: a top-down mode where anterior regions drive posterior activity and a bottom-up mode with reverse directionality. These shifts are most distinct during full consciousness and become less pronounced as awareness fades. Simultaneous EEG-fMRI recordings show the top-down mode coincides with higher-order cognitive network activity, while the bottom-up mode aligns with sensory system activity. An inattentive ADHD cohort exhibited imbalances in these transition dynamics compared to typically developing individuals. A coupled-oscillator model of the structural brain network reproduced these patterns, suggesting they arise naturally from inter-regional neural interactions.

Study at a glance

Abstract

Information continuously flows between regions of the human brain, exhibiting distinct patterns that dynamically shift across states of consciousness, cognitive modes, and neuropsychiatric conditions. In this study, we introduce Relative Phase Analysis (RPA), a method that leverages phase-lead/lag relationships to reveal the real-time dynamics of dominant directional patterns and their rapid transitions. We demonstrate that the human brain switches on a sub-second timescale between a top-down mode-where anterior regions drive posterior activity-and a bottom-up mode, characterized by reverse directionality. These dynamics are most pronounced during full consciousness and gradually become less distinct as awareness diminishes. Furthermore, we find from simultaneous EEG-fMRI recordings that the top-down mode is expressed when higher-order cognitive networks are more active while the bottom-up mode is expressed when sensory systems are more active. Moreover, comparisons of an attention deficit hyperactivity disorder (ADHD) inattentive cohort with typically developing individuals reveal distinct imbalances in these transition dynamics, highlighting the potential of RPA as a diagnostic biomarker. Complementing our empirical findings, a coupled-oscillator model of the structural brain network recapitulates these emergent patterns, suggesting that such directional modes and transitions may arise naturally from inter-regional neural interactions. Altogether, this study provides a framework for understanding whole-brain dynamics in real-time and identifies sub-second fluctuations in top-down versus bottom-up directionality as a fundamental mechanism underlying human information processing.