Temporal Dynamics of Brain Activity Predicting Sense of Agency over Muscle Movements

Abstract

Sense of agency (SoA), the conscious sensation that "I" am causing my own bodily movements, is a fundamental feature of human experience and a core facet of self-awareness. While prior work has measured participants' SoA over action consequences (e.g., a tone resulting from a button press), the neural mechanisms underlying SoA over action itself remains poorly understood since control of one's own muscles is normally unambiguous. In this work, we preempt subjects' own movements during a reaction time task using electrical muscle stimulation (EMS). On each trial, subjects report whether their movements were self-caused or caused by EMS. Then, by adaptively adjusting the timing of EMS delivery using Bayesian Optimization, we caused subjects to erroneously attribute roughly half of such involuntary (EMS-induced) movements to their own volition. Comparing trials with illusory agency to those without a sense of agency, we identify multivariate patterns of EEG activity that predict trial-by-trial SoA over EMS-caused movements across subjects. We show these EEG-derived predictors transition from transient to sustained dynamics as predicted by dual-process theories of action monitoring. In an exploratory analysis, we additionally decode SoA from nonlinear fractal measures of trial-by-trial brain activity. Results suggest that while the complexity (i.e., fractal dimension) of activity throughout the cortex is robustly higher when subjects experience agency over a behavior, but electrodes over motor regions appear to be selectively less sensitive to local perturbation (i.e., suppressed long-range correlations) when this sense of agency is absent. This finding is consistent with an account of SoA as a mechanism that gates the motor system from irrelevant sensation.

Metacognition bridges experiences and beliefs in Sense of Agency

“…), and recent controversies have cast doubt on the notion that a common cognitive or neural substrate can account for agency judgements across all these scales (Galang et al, 2021;Wang et al, 2020). Indeed, the neural predictors of agency judgements appear to meaningfully differ even between different types of sensorimotor judgements, such as those concerning muscle movements (Veillette et al, 2023) and downstream outcomes (Timm et al, 2016). If asked, however, most people might say that the "I" to which they attribute actions and consequences does not differ across these domains.…”