Time varying functional connectivity detected using task-free fMRI exhibits recurring patterns that constitute distinct dynamic connectivity states (DCS)¹. Although the range of functional connectivity states is reduced with lowered levels of consciousness^2,3, few studies have directly correlated DCS with behavior. Here we used spontaneous eyelid-closures (SEC) on sleep deprived persons as a proxy for vigilance to identify DCSs that correlate with greater and lesser levels of arousal^4,5. We then sought to test if spontaneous shifts into high or low arousal DCS would correspond to behavior during a vigilance task.Young adults underwent task-free fMRI scans as well as scans performed during an auditory vigilance task after being kept awake for ~24 hours. Video recordings of the eyes during task-free scans were assigned scores ranging from 1-closed to 9-open. Whole-brain dynamic functional connectivity between 126 predefined regions⁶ in task-free fMRI were estimated using 40s wide, sliding-window correlation, shifted in steps of 1 TR (2 sec). This was followed by K-means clustering to estimate centroids of DCSs. We then correlated the occurrence probability of each DCS with the concurrent SEC ratings to identify if any of them were associated with SEC changes. DCSs derived from task fMRI were estimated using an identical approach and were matched to their task-free fMRI counterparts. Matched DCSs were used to predict task performance.From the 9 DCSs estimated from task-free fMRI, we identified 2 extreme DCS (Figure 1). At one extreme was a DCS corresponding to a state of low arousal with high likelihood of behavioral lapses (associated with closed eyelids). At the other extreme was a DCS corresponding to a state of high arousal (associated with wide open eyelids) characterized by increased intra-network connectivity involving default mode, control, attention and salience networks, and increased anti-correlations between default mode and attention networks. The high and low arousal task-related DCSs were highly similar in configuration to their task-free counterparts. Participants who spent more time in the high arousal DCS had fewer behavioral lapses (rho = -0.548, p = 0.003), whereas more time in the low arousal DCS predicted a higher lapse rate (rho = 0.465, p = 0.022). The fluctuation between high and low arousal DCSs correlated with fluctuation in response times.Functional connectivity during periods of high arousal (proxied by way of eyelid opening) is characterized by elevated levels of intra- and inter-network connectivity and anti-correlation between networks that have been shown to be crucial in management of attentional resources, detection of salient stimuli and coordination of externally and internally oriented cognitive processes^7-9. The distinct patterns between high and low arousal DCSs resemble functional connectivity changes reported in sleep deprived subjects^10-12. We demonstrate that a significant behavioral accompaniment of fluctuating functional connectivity is fluctuation in arousal that is in turn accompanied by eyelid closure and slower responding.