Task-evoked flexibility of intrinsic neural timescales is hierarchically organized, behaviorally relevant, and refined across two developmental stages

Cortical regions integrate temporal information over different timescales at rest, ranging from fast in sensorimotor regions to slow in the association cortex. Here, we study how far regions adapt these timescales to meet task demands, as well as how this \emph{timescale flexibility} is organized across the cortical hierarchy, and how it develops in humans. We quantified rest-to-task changes in intrinsic neural timescales, referred to as \emph{timescale deflections}, using fMRI data from four independent datasets totaling 1,712 participants aged 8--64 years. We found that timescales reduced under cognitive load in all four datasets, most strongly in association cortex, which partially flattened the resting-state timescale hierarchy. Individual differences in timescale deflections were organized along the sensorimotor-association (S-A) axis and tracked cognitive performance, both in adults and in youth. Next, in youth, deflections followed a two-stage developmental course: a generalized reduction throughout childhood, followed by a selective expansion in association cortex from mid-adolescence. Furthermore, the timing of these stages aligned with cortical gradients of inhibitory gene expression and intracortical myelin content, both features implicated in the opening and closing of critical periods of neural plasticity. Overall, our results demonstrate that task-evoked timescale flexibility is hierarchically organized, behaviorally and neurobiologically relevant, and refined across adolescence.