Previous studies have shown conflicting results with regard to how spatiotemporal action sequences such as well-timed finger movements are represented in the CNS, specifically whether the sequential timing (temporal feature) is represented independently of the sequential order of movements (spatial feature). My recent studies employed precisely timed sequences of finger presses in externally (SRT) and internally driven paradigms to probe the conditions under which temporal feature transfer can be expressed. The findings support the notion that the CNS forms a modular representation of newly acquired action sequences which allows the subjects to transfer the spatial or temporal features to a novel sequence independently, with the temporal knowledge only expressed when integrated with the spatial representation non-linearly. Notably, subjects with more pronounced independent spatial and temporal transfer also show higher sequence-specific performance after training, suggesting a beneficial component of this modularity. I will present neuroimaging data (fMRI) that supports the idea of a hierarchical motor skill encoding with abstract temporal and spatial features of sequences encoded independently in overlapping patches of the bilateral premotor cortices and integrated spatio-temporal representations only present at the output level of the cortical motor hierarchy, in the contralateral primary motor cortex. Finally, I will discuss our current approach for examining how such representations are activated dynamically during action preparation and production (MEG).