02) (Figure S4). Thus, in contrast to long-range synchronization, which predicted perception before the stimulus became ambiguous, changes in power rather seemed to reflect a consequence Ibrutinib of the establishment of the different percepts. In summary, our results demonstrate highly structured large-scale cortical networks of oscillatory synchronization: up to seven anatomically confined cortical areas synchronized their activities across several centimeters and multiple processing stages along the sensorimotor pathways. Synchronization within these networks was temporally well localized to the cognitive event of interest and was linked to specific frequency

ranges that differed across multiple octaves between networks (beta and gamma). Although much progress has been made studying neural population activity in individual cortical areas, it remains difficult to characterize large-scale neural interactions across the entire brain. This is largely due to methodological problems. On the one hand, it is difficult to simultaneously record from multiple brain regions in invasive experiments. On the other hand, although EEG and MEG sample neural activity from a large part of the brain, estimating cortical interaction on the basis of these extracranial signals remains difficult. A further important obstacle is the lack of tools to efficiently analyze

cortico-cortical interactions in a high-dimensional space over with the ensuing substantial multiple-comparison problem. Our cluster-permutation–based approach may provide a valuable new tool to address

see more these problems and to identify large-scale networks of interacting sources. In particular, it goes beyond imaging neural activity across a singular cortical space and provides a framework to characterize interactions in a full pairwise cortico-cortical space. In principle, the approach is not limited to the study of synchrony, as demonstrated here, but may be applied to any bivariate parameter defined across the brain. Furthermore, the approach can be applied to a broad spectrum of experimental designs, including simple condition differences as well as complex parametric models. Moreover, no a priori assumptions need to be made about the structure of cortical networks. The method is robust to oversampling of the pairwise interaction space. This allows for directly imaging the extent of networks in space, time, and frequency. This approach well complements recent applications of graph-theoretical measures that provide powerful tools to quantify the global structural properties of large-scale connectivity (Bressler and Menon, 2010, Hagmann et al., 2008 and Palva et al., 2010). Our results provide strong evidence for the functional relevance of synchronization within the identified large-scale cortical networks.