Normative data are needed to create a reference that indicates optimal development of weight in relation to height and age, particularly in the face of the unfolding obesity epidemic. The body-mass index (BMI) has some serious limitations: it is a relatively poor predictor of current and future fatness. Currently, however, there are few available alternatives, with the possible exception of waist circumference or skinfolds. The use of cross-sectional references to construct a BMI-reference curve is problematic when there are period and cohort effects. Ideally, a reference would be based on longitudinal data in populations with little underweight, overweight, and obesity. In the meantime cross-sectional data in appropriate populations could be used to construct BMI percentiles linking BMI values at age 5 to those at age 18 (or 21) that would correspond with adult BMI values reflecting optimal health (e.g., that would correspond to adult BMI values between 21 and 23 kg/m2).