Rumen nitrogen metabolism values were estimated by the use of a single injection of 15(NH4)2SO4 into the rumen of sheep and consecutive 15N enrichment measurements in the rumen ammonia pool, rumen non-NH3-N (NAN) pool, rumen purine pool and blood urea-N (BUN) pool for a period of 24 h. Synthesis and degradation of N compounds in the rumen and passage of N to and from the rumen were evaluated on a chemical rather than a microbial basis; microbial fractions were not separated. This model was examined in two experiments. In Expt 1 a ram (55 kg) was given a semi-synthetic diet (1067 g dry matter (DM), 22.8 g N) in which soya-bean meal provided over 90% of the N. In Expt 2, two rams (45 kg) were given in three consecutive periods a semi-synthetic basal diet containing: (1) roasted soya-bean meal (SBM, 725 g DM, 14.8 g N/d); or (2) fishmeal (FM, 728 g DM, 15.5 g N/d); or (3) raw soya-bean meal (RSBM, 724 g DM, 13.8 g N/d). In all these rations, the main protein source provided over 90% of the N. In Expt 1, 68.3% of N intake was degraded directly to NH3 in the rumen, 21.2% escaped rumen degradation and 10.5% was incorporated into stable N compounds in the rumen. Net NH3 transfer to the blood was 30.4%, NH3 flow from the rumen was 6.6% and rumen NAN output was 63% of N intake. In Expt 2, rumen NAN output was larger (7.67, 14.36 and 8.89 g N/d for diets containing SBM, FM and RSBM respectively; P < 0.05) and net NH3 loss to the blood was smaller (6.1, 0.39 and 4.17 g N/d for diets SBM, FM and RSBM respectively; P < 0.05) for diet FM as compared with the soya-bean diets. The percentage of rumen NAN that was synthesized from NH3 was larger for diet RSBM (36.4, 40.3 and 49.1 for diets SBM, FM and RSBM respectively; P < 0.05) than for the other two rations. NH3 pool sizes (g N) were 0.463, 0.385 and 0.301 for diets SBM, FM and RSBM respectively (P < 0.05), while their hourly turnover rates were 15.8, 26.1 and 5.12 for diets SBM, FM and RSBM respectively (P < 0.01), indicating no correlation between pool size and its turnover rate. The existence of a small and rapidly turning over NAN pool to which NH3 is first bound (SNAN1) was indicated by the difference in the shape of the NAN and purines 15N-enrichment curves in the first 3 h after label injection. The fact that the rate of NH3 initial binding to SNAN1 pool was related neither to ammonia pool size nor to SNAN1 pool size suggests the existence of an unknown control mechanism at this site.