Materials research does not necessarily need to eliminate defects, but rather to characterize them, and to understand and control their effects. In most cases chacterization of defects means making structural or dynamic measurements of their properties. To understand these measurements in order to predict material and defect properties outside the range of the measurements is a much harder problem. Ideally a theory is required. However in the materials examples considered in this review, point defects in uranium oxide, copper clusters in steel, grain boundary aggregations, and stress concentrations, a true analytic theory is beyond our capabilities. Here computer modelling is often able to make the progress needed. This review considers the complementary nature of experimental characterization and computer simulation in our understanding of defects in materials.