Figure 3.

Lag thread motifs, or ‘one-way streets’, correspond to resting state network topographies. (a) Schematic of two lag threads propagating through four voxels. The overall pattern of propagation differs in the two threads, but the sequences through voxels 1 and 2 are identical (red box). Thus, voxels 1 and 2 represent a lag thread motif or ‘one-way street’. By contrast, the propagation directions between voxels 3 and 4 (blue box) are reversed across the two example threads, depicting a ‘two-way street’. (b) The lag threads in (a) realized as time series. (c) Temporal ordering of the voxels in the illustrated lag thread schematic. (d) Correlation of temporal orderings, across lag threads, for voxels 1 and 2. Note that the correlation is positive in the case of a ‘one-way street’ or motif. (e) Correlation of temporal orderings, across lag threads, for voxels 3 and 4. Note that the correlation is negative in the case of a ‘two-way street’. (f) Voxel-wise correlation matrix computed over latency values in the four lag threads derived from real data, as shown in figure 1d. Blocks of high correlation are found on the diagonal, or intra-RSN, blocks, implying that BOLD signal propagation is largely unidirectional within networks. By contrast, negative correlations (and lower positive correlations) are found in the off-diagonal, or inter-RSN, blocks, implying that cross-RSN propagation is generally bidirectional.