The formation of fragments due to avalanche-like growth of damage under impulsive forces is a process central to numerous studies ranging from shaped charge jet break up and rock blasting to bolide impacts, and, more recently, earthquake rupture. In the latter case, pulverized rocks found millimeters to tens of meters from the principal slip zones of large faults have been associated with fast, even supershear, rupture propagation. It has been postulated that earthquake source characteristics directly affect the degree of fragmentation, and the study of fragment size distribution may shed light on the energy budget of individual earthquakes as well as long-term effects on fault zone properties. The actual fragmentation process, and the partitioning of dissipated energy at fast loading rates, however, are still enigmatic. We use modified Split Hopkinson Pressure Bar experiments, in which we can control stressing rate, amplitude, and duration, as a laboratory analog for the complex natural prototype source processes. In our experiments, we characterize the velocity distribution of ejected fragments from Westerly Granite specimens resulting in a range of fragmentation states, from weakly fragmented to pulverized. Analysis of the velocity distributions (and the related kinetic energy) reveals spatial domains that are free of ejected fragments; these so called ‘zero kinetic energy modes’ are related to the fragmentation state: increasing fragmentation corresponds to a reduction of zero mode domains. The evolution of these zero modes with strain rate reveals that the transition from low strain rate fracturing to high strain rate pulverization is a smooth, continuous transition, rather than a sharp boundary. Furthermore, our results yield important insights into the process of fragmentation in earthquake process zones, including how dissipated energy is partitioned during fragmentation, and indicate that delocalization of energy is systematically coupled with source parameters.