In the frontal regions, no increase in IPC is apparent (Figure 5). Therefore, an increase in IPC is one characteristic of LFP signals in the temporal lobe that contributes to classification performance and is clearly different from the behavior of frontal regions. The statistical significance of the IPC measurement can be tested by asking the following question: At what point in time during the response are the phases statistically different from a uniform distribution? If the distribution is approximately
uniform, the “mean” phase will be the result of noise and will have no meaning. Palbociclib In the temporal lobe, a Rayleigh test of uniformity shows that the phases during both correct and incorrect trials are nonuniform just after the stimulus is presented and remain nonuniform for about 1 s (Figures 6A and 6B, black lines). Both mean p values are at or below 0.05 during the time interval t = 119–944 ms. Phases in the frontal lobe electrodes are, on average, uniform over the entire interval and thus do not reach statistical significance
( Figures 6A and 6B, blue lines). Next, given that there is a distribution of phases around each mean, we can ask whether the phase distributions for correct and incorrect responses have different median values. In the temporal lobe, the correct and incorrect trials have statistically different medians (circular Kruskal-Wallis GDC-0199 cost test, p < 0.05) during the interval 483–762 ms after the onset of the second image (Figure 6C, black line). Again, the electrodes in the frontal lobe never reach a level of statistical significance (Figure 6C, blue line). The results of these statistical tests yield some insight into the dynamics of the phase difference between
correct and incorrect trials. In the temporal lobe, the mean phase difference across electrodes varies smoothly over time (Figure 6D, dashed black line). The phase difference is zero 90 ms after the image appears, which roughly corresponds to the beginning of the time interval when the phase distributions are statistically nonuniform (Figure 6D, dark gray line). Therefore, there is an alignment of the correct and incorrect only phases early in the presentation of the second image. Over time, the phase difference increases, and its peak value at ∼π corresponds to the time interval where the median phase values are statistically different (Figure 6D, green line). We hypothesize that this similarity in correct and incorrect trials just after the presentation of the stimulus serves as a common starting point for the unique neural responses to the stimulus itself, analogous to the reset of an integrator. We can verify that the zero mean phase difference is not an artifact of averaging by looking at the fraction of electrodes with a large mean phase difference (Figure 6E).