Brain state-contingent stimulation to probe task-dependent functional networks.

Transcranial magnetic stimulation (TMS) is a fundamental tool for causal inference, but identical stimulation produces different outcomes depending on the instantaneous brain state, undermining interpretation and reproducibility. EEG-guided TMS studies to date have primarily addressed this by timing stimulation to local oscillatory features over circumscribed regions during the resting state. However, because these approaches do not control concurrent large-scale state dynamics, each stimulus is delivered within unconstrained and highly variable network regimes. We provide a theoretical and methodological framework for real-time control of large-scale functional networks, enabling stimulation to be delivered when a predefined network pattern, modulated by task demands, is expressed. By framing this neural variability as an informative resource, our approach aims to strengthen conditional causal inference through the real-time estimation of task-relevant connectivity. We define a workflow, from offline operationalisation and validation to real-time state-contingent stimulation, to promote shared standards and reproducibility in the study of human behaviour and to inform its clinical translation, from TMS-EEG biomarkers of network dysfunction to individualised, network-targeted neuromodulation.