Fig. 3: Temporal dynamics of neural activity in the PCG.

a, Coronal T1 MRI slice through the PCG and central sulcus in the left hemisphere, with overlaid shared representation map, showing RMA area sites. b, Coronal T1 MRI slice with overlaid somatotopic delineation map. c, Timecourse of normalized broadband activity (65–115 Hz) for RMA, foot and hand channels (white arrowheads in a and b) for a full experimental run (30 trials). Background shading indicates EMG-defined movement periods of the hand (red), foot (purple) and tongue (green). d, Magnified inset from c, with EMG included. e, Brain activity averaged to onset of tongue, foot and hand EMG. This subject had no somatotopic tongue site to display. f, Brain activity averaged to onset of tongue, foot and hand EMG across all subjects. As each channel’s broadband is normalized to itself, the magnitudes of these averaged traces are not meaningful, but the shapes are. All traces were shifted to start at a z-scored power of 0 to allow for more natural comparison. g, The latency between brain and movement was calculated as the peak of a sliding dot product between sEEG broadband and rectified EMG. h, Histograms of latency between somatotopic hand/foot/tongue sites and their paired EMG traces, as well as RMA site to each EMG show that all brain sites lead EMG (P values determined by two-sided one-sample t-test versus 0). Note that there was no significant difference in latency when comparing somatotopic and RMA sites (P = 0.24, unpaired t-test of 369 somatotopic latencies versus 144 RMA latencies). Panels a–e show data recorded in subject 2.