Crystal structure analysis using X-ray diffraction is, in many cases, the most advanced method available for obtaining high-resolution structural information about biological macromolecules. The ways in which X-ray diffraction data are collected and refined have a strong impact on the final quality of the structural models and the type and magnitude of their associated errors. It is becoming increasingly necessary for both structural and non-structural biologists to judge the reliability and accuracy of these models, which are being used in many aspects of research, including structure-based drug design, and to address detailed functional biological questions. In this article, we discuss how errors in these models arise and how they can be evaluated, and we argue for even more stringent validation checks and documentation of structures before deposition with the Protein Data Bank.