Fig. 3: Investigation of timescale gradient and localization in the model.
From: A hierarchy of time constants and reliable signal propagation in the marmoset cerebral cortex
a Simulated resting state activity, with white-noise input to all brain areas. Lower-level areas resonate at higher frequencies, whereas higher-level areas show more gradual fluctuations. b The autocorrelation function of each area’s activity in the resting state. c Top panel: The timescale derived from the Power Spectral Density (PSD) of simulated resting state activity, forming a gradient spanning from 50 ms to 250 ms. Bottom panel: Strong correlation of timescale between model and experiment (Pearson r = 0.94, p = 6.57 × 10−16). d Significant correlation between the simulated timescale and spine number (Pearson r = 0.95, p = 2.18 × 10−4). e Visualization of the dynamical system eigenmodes. The timescale localization for each area is portrayed by the sparseness of the eigenmodes. f The activity of selected areas following a stimulus to V1 in the model. Decay time constant increases along the signal propagation pathway. g The observed timescale gradient following a stimulus to V1, as extracted by fitting to the exponential function.
