Pressure-driven shock waves in solid materials can cause extreme damage and deformation. Understanding this deformation and the associated defects that are created in the material is crucial in the study of a wide range of phenomena, including planetary formation and asteroid impact sites, the formation of interstellar dust clouds, ballistic penetrators, spacecraft shielding and ductility in high-performance ceramics. At the lattice level, the basic mechanisms of plastic deformation are twinning (whereby crystallites with a mirror-image lattice form) and slip (whereby lattice dislocations are generated and move), but determining which of these mechanisms is active during deformation is challenging. Experiments that characterized lattice defects have typically examined the microstructure of samples after deformation, and so are complicated by post-shock annealing and reverberations. In addition, measurements have been limited to relatively modest pressures (less than 100 gigapascals). In situ X-ray diffraction experiments can provide insights into the dynamic behaviour of materials, but have only recently been applied to plasticity during shock compression and have yet to provide detailed insight into competing deformation mechanisms. Here we present X-ray diffraction experiments with femtosecond resolution that capture in situ, lattice-level information on the microstructural processes that drive shock-wave-driven deformation. To demonstrate this method we shock-compress the body-centred-cubic material tantalum-an important material for high-energy-density physics owing to its high shock impedance and high X-ray opacity. Tantalum is also a material for which previous shock compression simulations and experiments have provided conflicting information about the dominant deformation mechanism. Our experiments reveal twinning and related lattice rotation occurring on the timescale of tens of picoseconds. In addition, despite the common association between twinning and strong shocks, we find a transition from twinning to dislocation-slip-dominated plasticity at high pressure (more than 150 gigapascals), a regime that recovery experiments cannot accurately access. The techniques demonstrated here will be useful for studying shock waves and other high-strain-rate phenomena, as well as a broad range of processes induced by plasticity.