Accurate measurement of central fat patterning is difficult to obtain by conventional anthropometry. Direct measurement of intra-abdominal fat area by magnetic resonance imaging, while accurate, is impractical for large-scale observational studies. This report examines the sex-specific associations of conventional anthropometric indices with intra-abdominal fat and subcutaneous fat areas measured by magnetic resonance imaging. A total of 157 volunteers (97 men and 60 women) aged 48-68 years of predominately white ethnicity had intra-abdominal fat and subcutaneous fat areas measured as part of the Atherosclerosis Risk in Communities (ARIC) Study. Weight, body mass index, waist circumference, waist : hip ratio, and subscapular skinfold thickness were measured or calculated by a standardized protocol. On average, women had a lower intra-abdominal fat area than men (109.5 cm2 vs. 152.9 cm2) but a higher mean subcutaneous fat area (287.8 cm2 vs. 214.6 cm2). After adjustment for age, intra-abdominal fat area was quadratically associated with body mass index, waist circumference, weight, and subscapular skinfold thickness in men; in women, these associations were best modeled by a positive linear equation. Waist : hip ratio was linearly related to intra-abdominal fat area in both sexes. In general, anthropometric measures predicted lower percentages of the total variance in intra-abdominal fat area for men than for women. For subcutaneous fat area, all anthropometric indices were linearly associated and predicted more of the variance in subcutaneous fat area than in intra-abdominal fat area. These results indicate that among men, greater intra-abdominal fat deposition rates occur at relatively low body weights and fat is more uniformly deposited at higher weights. Women appear to deposit intra-abdominal fat at a constant rate as they gain weight, even after menopause. The authors conclude that when waist circumference or body mass index is used as a surrogate for intra-abdominal fat area in men, a quadratic term should be included in the analysis as a predictor variable. Subcutaneous fat area can be estimated well by linear measures commonly employed in epidemiologic studies.