(a, Left) Excursion from resting-state manifold structure in individual networks. Networks are sorted by mean early excursion on day 1 in order to make the trends in the data more clear. Note the peak in excursion after rotation onset on day 1, most prominent in frontal and parietal cognitive networks. We also observed sustained excursion in the visual network (comprising V1 and its periphery). Although we cannot speak definitively as to its cause, it may reflect some degree of top-down modulation of the visual cortex during adaptation, so as to orient spatial attention away from the location of the stimulus, or visual feedback processing (such as the feedback related to visual errors). We also contrast the high excursion in the ventral attention network (VAN) with the near absence in the dorsal attention network (DAN). In our embedding of the cognitive network, both components predominantly encoded changes in functional connectivity between the DAN and frontoparietal networks (FPN), with the VAN loading relatively weakly. Together, these results may suggest that excursion in our cognitive network reflects excursion in the VAN and FPN individually, or in the connectivity between FPN and DAN. We note that the DAN is frequently implicated in the top-down orientation of spatial attention, and so the change in functional connectivity between DAN and FPN may reflect the learning of the visuomotor rotation. (Right) Mean excursion during early/late learning. Error bars denote standard errors. (b) For each subject, we computed the mean excursion during early/late learning for each network and day. We then performed principal component analysis in order to characterize the dominant patterns of variability in excursion across all networks. The top panels show scores on the first two components for each subject, colored by behavioral group. Note that, as in our main analysis, the FF and SF groups are highly similar. The middle panel displays the loadings for each network across days and learning epoch. The bottom panel displays the change in loading from early to late learning on each day. Note that the first component appears to encode the overall degree of excursion in each network, while the second encodes changes in excursion from early to late learning (particularly on day 2).