Although generalized equations to predict body composition exist, the validity of these equations when applied to a homogeneous group of athletic women is uncertain. Sixty-five women runners (age = 28.3 ± 6.4 yrs; ht = 166.8 ± 6.2 cm; wt = 58.9 ± 6.6 kg; body density (Db ) = 1.047 ± 0.009 g · cc-1 ; percentage body fat (%BF) = 22.1 ± 3.7%; V̇O2 max = 56.0 ± 5.3 ml/kg · min-1 ; training volume = 45.8 ± 16.0 km · wk-1 ) were used to cross-validate the Jackson, Pollock, and Ward (1980) generalized quadratic skinfold equations (JPW) and the Tran and Weltman (1989) generalized girth equation (TW). Additionally, the average of TW and the JPW quadratic skinfold equations were calculated and compared to hydrostatic weighing (HW). None of the JPW equations nor the TW equation accurately predicted Db (P <0.05). All JPW equations underpredicted %BF with values ranging from 16.3-19.7%. Correlations ranged from r = 0.73 to r = 0.79. Standard errors (SE = [∑(Y - Y')2 /N]1/2 ) for predicting %BF ranged from 6.6% to 3.9%. The TW equation underestimated body density (1.041 ± 0.008) with a correlation of r = 0.54 and a SE of 4.2%. When the results of the JPW sum of 4 skinfolds (or the sum of 7 skinfolds) was averaged with TW, no significant mean differences in Db or %BF were observed and the standard errors were 2.7%. Additionally, a population specific equation was derived using skinfolds and girths (n = 40) and was shown to accurately predict body density in a cross-validation sample of women runners (n = 25). This equation yielded a correlation of r = 0.72 and a SE = 2.7%. It was concluded that the combination of skinfold and girth techniques is superior to either method alone when predicting body composition in women runners. © 1993 Wiley-Liss, Inc.