Advances in neuroimaging have greatly improved our understanding of human sleep from a systems neuroscience perspective. However, cognition and awareness are reduced during sleep, hindering the applicability of standard task-based paradigms. Methods recently developed to study spontaneous brain activity fluctuations have proven useful to overcome this limitation. In this review, we focus on the concept of functional connectivity (FC, i.e. statistical covariance between brain activity signals) and its application to functional magnetic resonance imaging (fMRI) data acquired during sleep. We discuss how FC analyses of endogenous brain activity during sleep have contributed towards revealing the large-scale neural networks associated with arousal and conscious awareness. We argue that the neuroimaging of deep sleep can be used to evaluate the predictions of theories of consciousness; at the same time, we highlight some apparent limitations of deep sleep as an experimental model of unconsciousness. In resting state fMRI experiments, the onset of sleep can be regarded as the object of interest but also as an undesirable confound. We discuss a series of articles contributing towards the disambiguation of wakefulness from sleep on the basis of fMRI-derived dynamic FC, and then outline a plan for the development of more general and data-driven sleep classifiers. To complement our review of studies investigating the brain systems of arousal and consciousness during healthy sleep, we then turn to pathological and abnormal sleep patterns. We review the current literature on sleep deprivation studies and sleep disorders, adopting the critical stance that lack of independent vigilance monitoring during fMRI experiments is liable for false positives related to atypical sleep propensity in clinical and sleep-deprived populations. Finally, we discuss multimodal neuroimaging as a promising future direction to achieve a better understanding of the large-scale FC of the brain during sleep and its relationship to mechanisms at the cellular level.