Fig. 2. A common set of electrodes involved in perceptual switching and maintenance across two ambiguous images.

A From each image presentation, time periods were extracted wherein the same percept is maintained (>1 s from button presses, ‘maintain’), and wherein the percept switched (<0.5 s from a button press, ‘switching’). For the Granger causality analysis in Figs. 3 and 4, these periods were split into 250 ms trials to improve data stationarity (see Methods section). B Left: example time courses of high-gamma amplitude (log-transformed, baseline-corrected; see Methods section) for two electrodes that show higher gamma activity during switching periods (top) or maintenance periods (bottom). Right: each switch period was paired with its subsequent maintenance period seen in the violin plot (thick circle indicates median, thick black line is the inter-quartile range, and thin black line extends to the most extreme data points not considered outliers); significance was assessed using a two-sided paired t-test. Electrode locations are marked in C. Data from participant 1 viewing the Necker cube image. C Locations of electrodes showing ‘switch’ and ‘maintain’ behavior. Lighter shades indicate electrodes with significant ‘switch’ or ‘maintain’ behavior for both ambiguous images; darker shades indicate electrodes with significant ‘switch’ or ‘maintain’ behavior for one image. D Joint distribution of the number of electrodes designated ‘switch’, ‘maintain’, or non-significant (NS) across cube and face-vase images (large font, ‘actual’), as well as the expected numbers of electrodes if category designation is independent between the two ambiguous images (small font, ‘expected’). A χ² test was applied against the null hypothesis that the category of an electrode for one image was independent of its category for the other image, which was highly significant (p = 1.28e-80), suggesting that category designations have significant overlap between the two images. Source data are provided as a Source Data file.