A total of 568 lateral spine radiographs of 191 generally healthy white perimenopausal women were used to determine the limits of normal for shape and dimensions of the vertebral bodies most often involved in osteoporotic fracture. Anterior and posterior vertebral heights for T7-L4 were used to compute wedge shape, relative posterior height, and relative serial change. Wedge shape was defined as anterior height minus posterior height divided by posterior height. Relative posterior height was defined as the posterior height of a vertebra minus the posterior height of the vertebra superior divided by the posterior height of the vertebra superior. The degree of normal wedging depended on position: wedge values progressed down the spine, from a mean of -0.106 at T7 to 0.048 at L4. The minimum normal wedge value ranged from -0.209 at T7 to -0.083 at L4. Radiographs on a cohort of 28 osteoporotic women were evaluated both by ordinary clinical reading of the radiographs and by using the standards developed from the normal subjects. Initial agreement between the two modes of assessment was 85.4%, and in resolving the remaining disagreements the clinician agreed that he had initially misread the films in all but 3.2%. On the basis of this limited experience, we conclude that the use of such objective standards, in a computer-operated algorithm, is more accurate than routine radiographic assessment. Serial x-rays spanning a 15-20 year period in these women permitted determination of both reproducibility and longitudinal change in vertebral dimensions in the perimenopause. The mean relative serial change was not significantly different from zero for all vertebrae.(ABSTRACT TRUNCATED AT 250 WORDS)