Do Scalp EEG Measurements Allow Causal Inference?

Rapidly developing methods of causal inference are being extensively applied to analyze electroencephalographic (EEG) measurements. However, there is considerable uncertainty surrounding the interpretation of the results. Is what the methods reflect from brain activity truly causality? In this study, we first simulate a ground-truth causal process in the cortex. Next, we project the cortex activity onto the surface of the head with a forward model, thereby obtaining data closely mimicking real noninvasive EEG signals. We argue and demonstrate that if the correct interpretation of the EEG measurement from any single electrode is a linear combination of all source cortical signals, a standard causal inference may fail on fundamental levels.