Neural responses to sensorimotor conflict in an embodied agency task.

Visuomotor conflicts that reduce the sense of agency—the feeling of controlling one's actions—are linked to increased alpha- and theta-band brain activity. In an EEG study using virtual reality, participants moved a finger while viewing a virtual hand that either matched their movement, differed anatomically (same movement, different finger), or spatially (same finger, shifted angle). Conflicts robustly decreased self-attribution. Altered trials could be decoded from unaltered ones with up to 68% accuracy starting around 200 milliseconds after movement onset. Cross-decoding revealed shared neural patterns across conflict types emerging around 500 milliseconds, suggesting a temporal progression from early, condition-specific sensorimotor responses to a later domain-general component, consistent with a two-step model of agency.

Study at a glance

Abstract

The Sense of Agency (SoA), the subjective experience that 'I am in control of my actions', has been proposed to involve both early implicit sensorimotor processes (feeling of agency) and later explicit higher-level processes (judgment of agency). Even though SoA is fundamental to our interactions with the external world and to our construct of the self, its underlying neural mechanism remains elusive. In this pre-registered EEG study, we used time-frequency analysis and Multivariate Pattern Analysis (MVPA) to investigate the neural characteristics of sensorimotor conflicts within an agency paradigm. Using an established embodied virtual reality paradigm, we modulated visual feedback to examine neural responses to conflicts between predicted and perceived sensory feedback. Participants moved their finger while viewing a virtual hand that either mimicked their movement, differed anatomically (identical movement, different finger), or spatially (identical finger, angular shift), and then rated their SoA while brain activity was recorded. In accordance with our pre-registered hypothesis, visuomotor conflicts, which were associated with robust decreases in self-attribution, were linked to increased alpha-band power, and exploratory analyses further revealed increased theta-band power. We show that trials containing a sensorimotor alteration could be reliably decoded from unaltered trials with up to 68% accuracy starting around 200 msec after movement onset. Cross-decoding further revealed shared neural patterns across anatomical and spatial manipulations, emerging around 500 msec post-movement. Together, our results indicate a temporal progression from early, condition-specific sensorimotor responses to a later domain-general component, consistent with the two-step model of agency.