There is a need to identify vertebral fractures from radiographs taken at a single point in time, but considerable controversy surrounds the methods to be used. We extended a data set to comprise baseline radiographs of the thoracic and lumbar spine on an age-stratified random sample of 762 Rochester, Minnesota, women and used revised methods to define vertebral deformities morphometrically. Changes in the method of measuring vertebral heights, changes in the source of normal values for vertebral measurements and changes in the criteria for assessing vertebral deformity had little impact on estimated prevalence and incidence in this population. The prevalence of any vertebral deformity was estimated at 25.3 per 100 Rochester women aged 50 years and over (95% CI, 22.3-28.2), while the incidence of a new deformity in this group was estimated at 17.8 per 1000 person-years (95% CI, 16.0-19.7). Projected nationally, these data suggest that over 500,000 white women in the United States develop vertebral deformities for the first time each year and that over 7 million white women aged 50 years and over might be affected at any given time. These estimates are limited by the absence of a reliable 'gold standard' with which to determine false positive and false negative rates associated with this or any other morphometric method. Information on the health consequences of vertebral deformities of various sorts would be most helpful in choosing between alternative approaches to defining them.