In plate impact experiments, the elastic precursor is attenuated as it propagates through the shocked specimen due to plastic straining at the shock front. However, the plastic strain rates required to explain the observed precursor decay are much higher than the strain rates that are predicted for known initial dislocation densities in the unshocked specimen, regarded as homogeneous. This anomaly, which poses a significant difficulty for the acceptance of the micromechanical theory of plasticity, has motivated this investigation of the effects of dislocation generation at surfaces and subgrain boundaries on precursor decay in single crystals of high‐purity LiF. Surface damage effects have been minimized by preparing damage‐free surfaces and by using fluid layers to protect the specimen surfaces during impact. These precautions reduce the precursor decay significantly compared to the decay measured for lapped crystals impacted directly (normal stresses of 5 kbars). However, computer simulation of the experiments based on subsonic dislocation motion, a linear viscous drag model, and nonlinear elasticity (assuming a homogeneous specimen), does not predict the measured wave profiles within experimental uncertainty. When dislocation generation at the surfaces is included in the computation, the calculated precursor amplitudes are comparable with measurements only for thin (3‐mm) specimens. Reexamination of recovered specimens from plate impact experiments suggests that subgrain boundaries are important sources for dislocation generation in precursor decay experiments. When subgrain boundaries are included as additional sources for dislocation generation, the computed velocity‐time profiles at 3 and 6.6 mm are in good agreement with measured profiles. Thus, it appears that the inclusion of dislocation generation at impact surfaces and subgrain boundaries provides a means for explaining the precursor decay anomaly, at least for high‐purity LiF.