Robust Scaling in Human Brain Dynamics Despite Correlated Inputs and Limited Sampling Distortions

Whether brain dynamics operate near a critical regime remains a central question in neuroscience, with potential implications for information processing and computational flexibility. However, conventional approaches are susceptible to artifacts introduced by temporal correlations, spatial dependencies, and subsampling, which can create the illusion of scaling in noncritical systems. Here we introduce an analytical and numerical framework centered on the covariance matrix and its spectrum, combined with a phenomenological renormalization group (PRG) approach, and extended to incorporate colored inputs, temporal and spatial correlations, and robust inference and control strategies for empirical data. Applying this framework to pooled resting-state fMRI, we find that collective brain activity is slightly subcritical yet close to criticality. The extracted exponents are robust and align with predictions from recurrent firing-rate models in the long-time correlation limit. Beyond these results, our Letter provides methodological tools for more reliable tests of criticality in neuroscience and complex systems.