Non-Neural Sources Systematically Impact Aperiodic EEG Activity

Abstract

Aperiodic, 1/f-like EEG activity has emerged as a key index of neural population dynamics, offering insights into excitation/inhibition balance, cognitive states, and clinical conditions. However, the foundational assumption that these parameters reflect neural dynamics rather than non-neural sources remains largely untested. Here, we systematically quantify how data quality and physiological artifacts influence aperiodic parameter estimation across two independent datasets (N=99 and N=103). Findings converged across data sets and two complementary analytical approaches: experimental artifact manipulation and selective signal decomposition. We demonstrate that ocular artifacts and poor data quality systematically inflate aperiodic offsets and exponents, while muscular artifacts exert the opposite effects. These spatially widespread influences persist after state-of-the-art preprocessing and reach magnitudes comparable to reported neurophysiologically meaningful group differences. Our results reveal that differential rates of non-neural sources can drive spurious neural interpretations; consequently, we introduce and validate an approach to mitigate these biases, guiding valid inference in basic and clinical aperiodic research.