Recent findings have sparked intense interest in birth weight as a predictor of type 2 diabetes mellitus. In some populations, there is an inverse association between birth size and disease; however, in the Pima Indians, a population with very high rates of gestational diabetes, high birth weight and low birth weight are associated with diabetes. This results in a U-shaped curve with higher diabetes rates in both tails of the distribution. Similarly, there are reports of an inverse association between stature and prevalence of gestational diabetes, but, in California Latina women with gestational diabetes, there is a positive association between stature and glucose concentration. Could these disparate findings also represent the two ends of a U-shaped curve? The disparities are likely due to differences in the environment, both in utero and postnatally, that different groups experience.