Two methods for estimating the flow of microbial protein synthesized in the rumen to the duodenum were compared: one uses microbial nucleic acids entering the duodenum, and the other uses allantoin excreted in the urine. Ten ewes were fitted with rumen and duodenum cannulae, as well as Foley catheters for collection of urine. The experiment was carried out using two series of treatments with two replications each. The ewes were randomly divided into five groups, which were assigned to one of five diets. (In the second series sheep were excluded from diets received in the first series.) The diets, differing in protein and energy content, were as follows: (1) low protein, low energy (LPLE); high protein, low energy (HPLE); (3) maintenance for protein and energy (MPME); (4) low protein, high energy (LPHE); and (5) high protein, high energy (HPHE). The rates of rumen microbial protein synthesis were 3.34, 7.00, 9.44, 4.47 and 13.44 g microbial nitrogen (N) d−1 for diets 1–5, respectively. Results indicated a high correlation between allantoin and total purine derivatives (allantoin, uric acid, xanthine and hypoxanthine) excreted in the urine and the amount of microbial nucleic acids entering the duodenum. A regression equation y = exp (0.830 + 2.089x), using allantoin N (g d−1) excreted in the urine, was proposed for estimating microbial N synthesis (g d−1) in the rumen. The ratio of allantoin N to creatinine N in the urine samples collected at 6-h intervals varied markedly. Key words: Sheep, rumen, microbial protein, allantoin, purine derivatives