Traditional research in neuroscience has studied the topography of specific brain functions largely by presenting stimuli or imposing tasks and measuring evoked brain activity. This paradigm has dominated neuroscience for 50 years. Recently, investigations of brain activity in the resting state, most frequently using functional magnetic resonance imaging (fMRI), have revealed spontaneous correlations within widely distributed brain regions known as resting state networks (RSNs). Variability in RSNs across individuals has found to systematically relate to numerous diseases as well as differences in cognitive performance within specific domains. However, the relationship between spontaneous fMRI activity and the underlying neurophysiology is not well understood. This thesis aims to combine invasive electrophysiology and resting state fMRI in human subjects to better understand the nature of spontaneous brain activity. First, we establish an approach to precisely coregister intra-cranial electrodes to fMRI data (Chapter 2). We then created a novel machine learning approach to define resting state networks in individual subjects (Chapter 3). This approach is validated with cortical stimulation in clinical electrocorticography (ECoG) patients (Chapter 4). Spontaneous ECoG data are then analyzed with respect to fMRI time-series and fMRI-defined RSNs in order to illustrate novel ECoG correlates of fMRI for both local field potentials and band-limited power (BLP) envelopes (Chapter 5). In Chapter 6, we show that the spectral specificity of these resting state ECoG correlates link classic brain rhythms with large-scale functional domains. Finally, in Chapter 7 we show that the frequencies and topographies of spontaneous ECoG correlations specifically recapitulate the spectral and spatial structure of task responses within individual subjects.