Women with vertebral deformities caused by osteoporosis have more back pain and disability and are at higher risk for subsequent vertebral deformities than women without deformities. Despite the importance of vertebral deformities, there has been a great deal of controversy about how to identify or define them. In order to compare methods for defining vertebral deformities, we studied spinal radiographs from women in the Study of Osteoporotic Fractures (SOF), a cohort study of 9704 non-black women over age 65 recruited from population-based listings in four clinical centers. Using radiographs obtained at the baseline exam, we compared five methods for defining vertebral deformities: one based on a semiquantitative reading by a radiologist and four using vertebral morphometry. The semiquantitative method was compared with the other methods in a random sample of 503 films, while the morphometric methods were compared with each other in a larger sample of 9575 films. We tested a system of "triage" in which only those films with evidence of deformity were assessed by morphometry. We compared the relationship between deformity, defined by each method, and a variety of clinical criteria including bone mineral density at the lumbar spine, height loss since age 25, back pain, and incidence of subsequent deformity. Semiquantitative reading and three of the four morphometry-based methods provided similar relationships to clinical criteria. The fourth morphometry method (based on ratios of each vertebral height to the corresponding height at T4) produced significantly weaker relationships for several of the clinical validation criteria. Triage of radiographs rarely resulted in missed deformities and did not reduce the performance of any of the methods. We conclude that use of any of the similar methods, with or without triage, provides a valid approach to defining vertebral deformities.