Radiogrammetry of the second metacarpal has long served as a measure of normal and abnormal bone growth and aging, functional asymmetry, and fracture risk in osteoporosis. The method relies on algebraic interpretation (circular or elliptical models) of uni- or biplanar radiographic images. This paper tests the conformance of these models with actual measures of metacarpal geometric variation in a sample of 356 bones from an historic archaeological sample. Both the circular and elliptical models significantly over-estimate actual values for all variables (e.g., total area (3.05% and 9.42%, respectively) and cortical area (7.25% and 12.25%), bending rigidity about the mediolateral (13.88% and 20.92%) and anteroposterior axes (17.35% for the elliptical model). The greater degree of systematic bias found for the elliptical model is contrary to results of an earlier study (Lazenby, 1997), and suggests that sample composition can influence the method error associated with a particular approach. The import of radiogrammetric bias will depend on the degree of shape variation among samples compared (e.g., left vs right, male vs female, etc.), which in most cases cannot be predetermined. Consequently, reduced major axis equations were derived from regressing actual on estimated values for total area (TA), cortical area (CA) and two measures of bending rigidity (Ix, Iy) in order to permit adjustment for radiogrammetric error associated with these algebraic models. Application of these formulae to a hold-out sample showed no significant differences between actual and predicted values. Am. J. Hum. Biol. 10:747-756, 1998. © 1998 Wiley-Liss, Inc.