Fig. 6: Experiment 3 design and results.

a Experimental setup of the non-invasive BMI paradigm, whereby healthy participants used motor imagery to control a virtual hand and provide agency ratings. b Timeline of a cued BMI trial, constituting the core of the experiment. These trials were two-thirds of the total and used sham BMI without participants being aware of it, to minimise variability related to decoder performance. One-third of the trials (not shown) were self-paced and used actual BMI. They were only used to keep participants convinced that they were in control of the virtual hand. In total, 5 blocks of 60 trials were collected. c Uncorrected log p-values for the phase opposition product in the − 0.5/0 s, 8–13 Hz range, computed over 114 ROIs after eLORETA source reconstruction via one-sided comparison with 10000 permutations. Red arrows point at the location of the cluster of two ROIs surviving whole-brain cluster correction, in the left contralateral SMA and M1. d Time-frequency plot of the phase opposition product (negative log p-value from one-sided comparison with 10000 permutations) for the most significant ROI, corresponding to the posterior part of the left-contralateral SMA. The black rectangle denotes the selected time and frequency range, the dashed line the onset of sensory feedback. e Empirically measured Phase Opposition Product (8–13 Hz, − 0.5/0 s) for the left SMA compared to 10000 permutations with shuffled agency ratings. f Correlation between the individual alpha peak and the frequency of strongest phase opposition in the left SMA. The shaded area indicates the 95% confidence interval for the regression line. Data of the tetraplegic participant from Experiments 1-2 (not included in the regression) is shown for comparison (red dot).