1 Hz is consistent with a recent study of bilateral primary auditory cortex (Nir et al., 2008). However, our findings suggest Docetaxel in vitro that in contrast to the current view on the predominant contribution from gamma

activity, low-frequency oscillations are a major contributor to large-scale network connectivity. Slow oscillations (<0.1 Hz) are commonly thought to signal general changes in network excitability (Hughes et al., 2011; Monto et al., 2008), whereas oscillations on a faster timescale (>1 Hz) may be better suited to more specific information exchange between areas. To measure interactions between network areas on a faster timescale, we calculated the coherence between the “raw” LFP signals (cf. power time series in the previous section) in each pair of network areas. The coherence measures the linear association between the LFPs as a function of oscillation frequency. For each recording session, we used multitaper methods (three tapers and ±4 Hz bandwidth) to estimate the coherence in every 500 ms time window for which there was no eye movement (excluding 0–200 ms after any preceding

PI3K inhibitor eye movement). The population mean coherence spectrum for each ROI pair showed the peak coherence at low frequencies (<20 Hz; Figure 4). Within a specified frequency band, we counted the number of sessions showing significant coherence for each pair of ROIs (jackknife variance estimates, p < 0.001). There was significant coherence in the 4–20 Hz range for 41–55 sessions (range across the six pairs of ROIs) out of the total of 58 sessions, whereas only 9–29 out of 58 sessions showed significant coherence in the 30–100 Hz range. Notably, the rank of connection strengths based on mean alpha coherence was similar to that seen in BOLD connectivity (Figure 2). For example, alpha coherence and BOLD connectivity both showed the strongest connection between the out pulvinar and

V4 and the weakest connection between the TEO and LIP. With respect to the greater effects at low versus high frequencies, these coherence results were consistent with that observed in the slow-wave power correlations. Thus, the coherence of neural activities on a fast timescale may give rise to the power correlation of band-limited neural activities at the slow fMRI timescale. Specifically, low-frequency oscillations (<20 Hz) may predominantly contribute to resting-state functional connectivity. Different frequencies of neural oscillations may be useful for different temporal and spatial scales: high frequencies like gamma for local computation, and lower frequencies like alpha for large-scale interactions. Because low-frequency oscillations have been shown to modulate high-frequency activity (Buzsáki and Wang, 2012; Canolty and Knight, 2010; Jensen and Colgin, 2007; Schroeder and Lakatos, 2009), such cross-frequency coupling may integrate functions across multiple spatiotemporal scales.