19 20 Although it is well established that fronto-parietal regions are active during action observation, 21whether they play a causal role in the ability to "mindread" others' actions remains controversial. 22 In experiments reported here, we combined offline continuous theta-burst stimulation (cTBS) 23with computational modeling to reveal single-trial computations in the inferior parietal lobule 24 (IPL) and inferior frontal gyrus (IFG). Participants received cTBS over the left IPL and IFG, in 25 separate sessions, before completing an intention discrimination task or a kinematic 26 discrimination task unrelated to intention. We found that transient disruption of activity of the 27 IPL, but not the IFG, specifically impaired the observer's ability to judge intention from 28 movement kinematics. Kinematic discrimination unrelated to intention, in contrast, was largely 29unaffected. Computational analyses revealed that IPL cTBS did not impair the ability to 'see' 30 changes in movement kinematics, nor did it alter the weight given to informative versus non-31 informative kinematic features. Rather, it selectively impaired the ability to link variations in 32 informative features to the correct intention. These results provide the first causal evidence that 33 IPL maps kinematics to intentions. 34
Introduction 35 36When watching others in action, we readily infer their intentions from subtle changes in the way 37 they move. Theoretical work 1-5 and related experimental findings (e.g., [6][7][8][9] ) suggest that this 38 ability to 'mindread' the action of others is mediated by the fronto-parietal action observation 39 network. Despite two decades of research, however, the specific neural computations involved in 40 the ability to read the intention of an observed action remain unclear and causally untested. 41 42A major difficulty in studying action mindreading is the ever-changing nature of movement 43 kinematics 10,11 . Movement is "repetition without repetition" 12 . Averaging across repeats of 44 nominally identical, but actually different motor acts, as done in standard trial-averaged analyses, 45can obscure how intention information is encoded in trial-to-trial variations in movement 46 kinematics 13 . More importantly, the brain does not operate according to an average response over 47 averaged kinematics. Real-world action mindreading requires real-time readout of intention-48 information encoded within a specific motor act. Thus, studying action mindreading with single-49 trial resolution is critical for understanding how intention readout maps to the multiplicity and 50 variability of kinematic patterns. 51 52Here, we developed a novel analysis framework to capture intention mapping at the single-trial 53 level. This framework was inspired by recent advances in our understanding of how sensory 54 information encoded in a neural population is read out to inform single-trial behavioral 55 choice 14,15 . In this study, we extended this approach to investigate neural computations 56 performed in the le...