Spontaneous blood-oxygen-level-dependent (BOLD) signals acquired at the resting state have recently been found to fluctuate coherently within many anatomically-connected and functionally-specific brain networks, and it may reflect an orderly organization of ongoing brain activity. Understanding this phenomenon may help us not only to understand some fundamental mechanisms of brain functions but also to find its applications in clinical field. However, the mechanisms underlying this phenomenon remains elusive, and even its neural origin is still controversial. This dissertation aimed to understand spontaneous BOLD fluctuation from its neurophysiological basis, its modulation under different brain states, and its role in brain functions. With five projects performed both on animals and humans, we have found that i) spontaneous BOLD fluctuation under deep burst-suppression anesthesia originates from underlying spontaneous neural activity, ii) spontaneous BOLD fluctuation is sensitive to changes in anesthesia depth, reflecting reorganization of ongoing brain activity at different consciousness level, iii) the resting-state visual network is spatially reorganized into activated and non-activated coherent network under continuous stimulation, and iv) the correlation strength within individuals' resting-state network can affect their evoked response to identical stimulations. These findings clearly support the functional significance of spontaneous BOLD fluctuation widely observed in animals and humans brain and provide new insights into its underlying mechanisms.