Resting-state fMRI (rsfMRI) is now a widely used method to assess the functional connectivity (FC) of the brain. However, the mechanisms underlying rsfMRI are still poorly understood. In this talk I will address several related aspects of the rsfMRI signal. The first is the global signal, which represents the whole brain average signal and has been widely used as a regressor for removing the effects of global variations in resting-state activity. I will discuss the controversy surrounding global signal regression and describe new approaches for minimizing global signal effects. A related topic concerns the origins of global activity in the brain. There is growing evidence that a considerable portion of this global activity arises from fluctuations in vigilance and arousal. I will discuss the recent findings in this area and discuss the implications for the analysis and interpretation of rsfMRI studies. Finally, I will describe recent empirical and theoretical work demonstrating the limitations of regression based methods that are widely used to minimize the effects of nuisance components in rsfMRI studies.