We present two complementary quantitative approaches to the problem of characterizing morphometric variations between two distinct populations. The case presented focuses solely on local size variations, but the general method can easily be applied to other scalar morphometric quantities. The first method uses a statistical parametric map (SPM) to ascertain a P value, which indicates whether any statistically significant differences exist between the populations. The second method focuses on finding the best single measurement which can be used for classifying the two populations. For our case study midsagittal cross sections of the corpora callosa from a population of normal males and females are nonrigidly registered (spatially normalized) to an atlas. The resulting deformations are then used to ascertain (i) whether there are any statistically significant differences between the populations and (ii) whether these differences allow one to perform classification. We make use of the Jacobian of the deformation field and normalize it to account for overall volume changes allowing us to focus on differences which are more related to morphometry than scale. From the (SPM) approach to the problem we find evidence of statistically significant differences in the morphology between the populations. Using a linear discriminant function we find that these differences do not appear to be useful for classification. Thus, this dataset provides an example of how statistically significant effects may not be of much diagnostic value. They may be of interest to the research community, but of little value to the clinician.