Reject analysis is an accepted standard of practice for quality assurance in conventional radiology. The need for reject analysis has been challenged by the introduction of computed radiography (CR) because of low reported reject rates and because criteria for improperly exposed images were lacking. Most CR systems include quality control (QC) workstations that are capable of modifying the appearance of images before release, and also of deleting poor images before they are analyzed. Texas Children's Hospital has been using computed radiography since October 1995, and now conducts essentially filmless imaging operations using a large-scale picture archival and communications system (PACS) with fourteen CR units. The QC workstation is a key element of our CR operation; however, the extensive software tools of the workstation are limited in terms of avoiding repeated examinations. Neither the QC workstation nor the PACS itself is designed to support reject analysis, so our task was to design a system that accommodates identification, isolation, and archiving of repeated examinations, making use of our electronic imaging systems. We had already developed transcription codes for our radiologist's examination critique, so we adopted these as codes for rejected images. The technologist at the QC workstation appends the critique code to patient demographic information, modifies other fields to indicate that the image is rejected, and archives as usual. Modified routing tables prevent the release of rejected images but ensure they are available for review. Our frequency and reasons for repeated examinations are comparable to other reports of reject analysis in the literature. The most frequent cause of a repeated examination is mis-positioning. The process of developing the method for capturing repeat, collecting the data, and analyzing it is only one-half of the battle; to achieve an improvement in services, it is necessary to feed back the results to management and staff and to implement training as indicated. It is our intention to share our results with PACS and CR vendors in the hope that they will incorporate some mechanisms for reject analysis into the design of their systems.