A new, automated turbidimetric-kinetic (ATK) assay was used to quantitate bacterial endotoxin in rumen fluid or in serum of Holstein steers. The ATK method used Limulus amebocyte lysate (LAL) reagent with added beta-glucan (LAL-ES) which improved specific sensitivity to endotoxin. Design of the feeding trial permitted comparison of endotoxin levels found during consumption of a basal ration with those higher levels detected at various times following the introduction of increasing percentages of rolled barley to that basal ration. Both serum and ruminal endotoxin levels were significantly higher in steers on the higher concentrate rations. Peak endotoxin levels were detected 20 days following the change to the highest concentrate ration which contained 60% barley. Endotoxin levels from both sources subsequently decreased. Ruminal endotoxin stabilized at about 10 times the level, and serum endotoxin stabilized at 2 to 4 times the level, of that previously found during feeding of the basal ration. Test protocol included sample dilution and heating in order to avoid the effects of endotoxin inhibitors. Recovery rates for added endotoxin to either serum or rumen fluid supernates ranged from 120 to 136%. Coefficient of variation for endotoxin concentration in serum was lower than 10%, and in rumen fluid only slightly higher. There was significant correlation between ruminal concentration as measured by the ATK method and an alternative chromogenic substrate assay procedure. Changes in endotoxin level in experimental steers were those predictable from experience with naturally occurring incidents of grain engorgement. The ATK assay appears to be an accurate and rather simple technique which will prove useful for experimental and clinical studies in the future.