Figure 2. Brain rhythms define distinct interaction networks.

(A-D) Data from Monkey 1 (see Fig. S2 for Monkey 2). Note that color scales are logarithmic.

(A) The four matrices in this column show coherence (lower triangular matrix) and power correlation (upper triangular matrix) for the frequency bands listed to their left. Each matrix entry corresponds to the respective FC value of one pair of recording sites, calculated across all available Post-cue data epochs (see Methods), and averaged over the frequency bins in the respective frequency bands (see Results, Methods). Matrix entries with non-significant FC are masked in gray (non-parametric randomization test by shuffling data epochs, corrected for multiple comparisons across site pairs). The axes list the cortical areas, from which the sites have been recorded, with the areas ordered according to their hierarchical level (Chaudhuri et al., 2015). Area boundaries are indicated by gray lines on the matrices. Each area, and its corresponding recording sites, is given a color code. The sites will maintain these area-specific colors, when they are reordered in the modularity analysis shown in (B, D).

(B) Same FC values as in (A), but re-ordered according to modules obtained from a consensus modularity analysis (see Methods). Modules are separated by gray lines. The modularity analysis was performed separately for coherence and power correlation (i.e. separately on those triangular matrices), and consensus was obtained over the four frequency bands. The color codes on the margin indicate per site the respective cortical area as introduced in (A); note that those color codes are separate for the upper and the lower triangular matrix.

(C) Similar to (A), but for GC. Note that GC is a directed metric, requiring the full matrix. Each matrix entry corresponds to the GC from a site in the cortical area listed on the y-axis to a site in the area listed on the x-axis.

(D) Similar to (B), but for GC. GC modularity analysis was performed on the full matrix, and consensus community structure was obtained over the four frequency bands.

(E-G) Data averaged over both monkeys.

(E) Distinctiveness (1-R²; see Results) between patterns of FC of a given type (as listed per row), for all combinations of frequency bands (listed per column). The patterns of FC are the triangular matrices shown in (A) for coherence and power correlation, and the full matrices shown in (C) for GC. Values in parentheses are the Distinctiveness after partialization for distance on the cortical surface.

(F) Distinctiveness (1-R²; see Results) between patterns of FC of a given type (as listed per row in (E)), and the pattern of the product of power at the respective sites (specifically$lo{g}_{10}(\sqrt{powe{r}_{site1}\times powe{r}_{site2}})$), and in the frequency bands listed per column. Values in parentheses are the Distinctiveness after partialization for distance on the cortical surface.

(G) Same as (F), but only for GC and replacing the product of power by the power at the sending site, i.e. the site from which the GC originates.

See also Figure S2.