Insula cortex gates the interplay of action observation and preparation for controlled imitation

Perceiving, anticipating and responding to the actions of another person are fundamentally entwined processes such that seeing another's movement can prompt automatic imitation, as in social mimicry and contagious yawning. Yet the direct-matching of others' movements is not always appropriate, so this tendency must be controlled. This necessitates the hierarchical integration of the systems for action mirroring with domain-general control networks. Here we use functional magnetic resonance imaging (fMRI) and computational modelling to examine the top-down and context-dependent modulation of mirror representations and their influence on motor planning. Participants performed actions that either intentionally or incidentally imitated, or counter-imitated, an observed action. Analyses of these fMRI data revealed a region in the mid-occipital gyrus (MOG) where activity differed between imitation versus counter-imitation in a manner that depended on whether this was intentional or incidental. To identify broader cortical network mechanisms underlying this interaction between intention and imitativeness, we used dynamic causal modelling to pose specific hypotheses which embody assumptions about inter-areal interactions and contextual modulations. These models each incorporated four regions - medial temporal V5 (early motion perception), MOG (action-observation), supplementary motor area (action planning), and anterior insula (executive control) - but differ in their interactions and hierarchical structure. The best model of our data afforded a crucial role for the anterior insula, gating the interaction of supplementary motor area and MOG activity. This provides a novel brain network-based account of task-dependent control over the integration of motor planning and mirror systems, with mirror responses suppressed for intentional counter-imitation.