Since the discovery of the 'readiness potential', various studies have examined neural correlates of volitional behavior by analyzing 'self-initiated' or 'internally generated' movements, despite the obvious contradiction between experimental control and volitional in-determination. There is evidence for distinct neural sub-networks involved in different aspects of movement generation from planing of (1) specific movements to be executed or movement choice and (2) the timing of movement initiation or execution timing. Many studies have experimentally separated different aspects of volitional movement generation applying a multitude of imaging methods. In this study, we conducted an 'Activation Likelihood Estimation' meta-analysis on 78 fMRI and PET experiments to map out the spatial layout of the functional networks involved. The key feature of experiments to include was the presence of at least 2 alternative choices, which were not determined by the experimental setup and resulted in a response. We compared the assessed volitional aspect, either a 'choice' between alternative movements or the internal 'timing' of movement execution with respect to non-volitional control conditions. As PET and fMRI block-design measure sustained activity across an experimental condition, within this sub-sample we examined task-set effects of volitional movement generation. Event-related fMRI analyses were pooled to dissociate movement preparation and execution. Volition includes sub-networks for internal choice and internal timing, with choice containing a state-related component for planing and a more phasic component linked to movement preparation. Timing consisted of a state- and an event-related component. A core network was identified prominently including anterior mid-cingulate cortex and pre-supplementary motor cortex.