Fig. 4.
Bayesian model predictions of the confidence in causal estimate (CCE), which is our proposed measure for SoA. a Our Bayesian model predicts CCE to increase with a stronger causal prior. Furthermore, CCE differs for each condition even with equal prior strengths. This can be attributed to the difference in the amplitude of the jitter in the self-generated vs TMS-induced movement (muscle twitches) and audible clicks. b When plotted as functions of the trial-to-trial temporal disparity τO − τA, with the specific causal priors obtained for each condition, marked in a, CCE has a higher peak in the voluntary condition, but much lower values in the sham condition. Furthermore, CCE diminishes as the temporal disparity in sensory inputs moves further away from the prior mean |τO − τA − μAO|. This falling of the CCE is faster when the causal prior is weaker and the uncertainty in the action input signal is higher