Temporal dynamics of spectral power

We compared the peak times of the spectral power time series for LOC, PT, PAR, and MTL in order to examine the temporal relation of high frequency activity through the visual hierarchy. We identified the peak time for a particular region by calculating the time during which the average spectral power time series across participants reached its maximum value. We then computed the difference in the peak times between two regions to identify the temporal relation of spectral power between them. To assess whether this difference was significant, we generated a chance distribution to which the true difference in peak times could be compared. We computed this chance distribution by randomly switching the spectral power time series for one region with the spectral power time series of the other region in each participant. Hence, in each permutation, some participants would retain their original power time series traces in their original regions, and some participants would have the labels for the regions randomly switched. We then averaged these shuffled spectral power time across participants for each region and then computed the difference in peak times between the two regions in each permutation. We repeated this procedure 1000 times to generate a shuffled distribution of differences in peak times. We assigned p values that characterize the difference in peak times between any two brain regions by comparing the true difference in peak times to the shuffled distribution of differences.

To estimate how quickly high frequency activity increased in each brain region and how this compared across LOC, PT, PAR, and MTL, we computed the instantaneous slope of the increases we observed in the time series of high frequency spectral power. We computed the difference in spectral power between adjacent time bins (200 ms overlapping bins incremented by 100 ms) and then averaged these estimates of instantaneous slopes across all time points within the first 500 ms after image presentation. Within each brain region in each participant, we computed the average instantaneous slope across all visually responsive electrodes. We compared the distribution of average values across participants between two brain regions using an unpaired t test (p < 0.05) in order to assess whether the rise in high frequency activity was different between the regions across participants.

To determine whether the differences in high frequency 80–120 Hz spectral power that we observed between conditions arose at different times in different brain regions, we performed two analyses. In both cases, we explicitly generated a time course of the average difference between conditions for each electrode that showed any significant difference between conditions in each brain region in each participant. In the first analysis, we used the rise in the average time series across all significant electrodes of the differences in spectral power to estimate the first time point when this difference deviates from zero and to estimate the time point when the increase in high frequency power reached 50% of its peak. We used this approach to generate a more temporally precise estimate of when this signal first increased above baseline since in our main analysis we generated the time series using overlapping 200 ms bins incremented every 100 ms. To estimate this initial time of deviation, we identified the time point of the peak difference between conditions and the time point of the local minimum that immediately preceded the peak difference. We then fit a line using all points in between these two time points and identified the time point when that line intersected with zero. We designated this as the time point at which the difference between conditions first deviates from baseline. We compared the distribution of these first time points across participants between brain regions using an unpaired t test. We similarly identified when the rise of spectral power reached 50% of the peak and compared the distribution of 50% time points across participants between each brain region. We also compared the estimated time points at which we first observed a rise in the difference in high frequency power between conditions to the time points at which we observed overall increases in high frequency power in the MTL across conditions. In a similar manner, we used the average time series of spectral power across significant MTL electrodes to estimate the first time point when overall 80–120 Hz power deviated from baseline in the MTL. In the second analysis, we identified the time points that exhibited the first significant difference in spectral power between conditions in each electrode within a region. We then averaged these first time points across all electrodes within each region in each participant. We compared the distribution of these time points of first differences across participants between brain regions (unpaired t test, p < 0.05).