Three studies were conducted to evaluate a method for estimating grass and legume fractions in extrusa samples from esophageally-fistulated animals, based on different concentrations of neutral detergent fiber (NDF), Ca and lignin in grasses and legumes. In experiment 1, NDF, Ca and lignin concentrations were measured in mixtures of six combinations of red clover-orchardgrass and alfalfa-tall fescue (100:0; 80:20; 60:40; 40:60; 20:80; 0:100). The R² values were .90, .80 and .22 for Ca, NDF and lignin, respectively. In experiment 2, fresh tall fescue and red clover were collected during 8 d, mixed in the same proportions as in experiment 1, and each fraction was fed to esophageally-fistulated steers. Extrusa samples were collected without loss of saliva. Regression equations developed were y= 1.08 - .008x (R²= .91) for Ca and y= 50.9 + .22x (R²= .74) for NDF, where x represents % grass. In experiment 3, two esophageally-fistulated steers were allowed to graze in 20 4 X 4 m² plots of red clover and tall fescue after fasting overnight. The grass/legume proportions of the extrusas and the grab samples were extrapolated from slopes. The botanical composition of grab samples differed (P<.001) from that of extrusas for both components and for both animals. Extrapolation from NDF values tended to overestimate the proportion of grass in extrusa samples and underestimate it in grab samples. Calcium was not affected by salivation and mastication and was more reliable than NDF. Lignin was not an accurate predictor. / M.S.