AbstractThe first objective of this experiment was to investigate the effect of using different bacterial isolates on the estimation of microbial protein production in semi-continuous fermenters (RUSITEC) given four diets, and to test whether a ‘mixed’ bacterial pellet was representative of the whole bacterial population in the fermenters. A second objective was to compare two different microbial markers (nitrogen-fifteen (15N) and purine bases (PB)). Diets consisted of neutral-detergent fibre from grass hay (10 g/day) and sugar-beet pulp (2 g/day) and 280 mg/day of one of four N forms (isolated soya-bean protein, soya-bean peptides, amino acids blended to profile soya-bean protein and NH4Cl). Two 14-day incubation runs were carried out and in each run each of the four different diets were given to two vessels. On days 12 and 13, total digesta (effluent plus nylon bags residues) was collected for analyses of non-ammonia N,15N enrichment and PB concentration, and for isolation of total mixed bacterial pellets (TB). On the last day of each incubation run, the system was stopped for isolation of liquid- (LAB) and solid-associated (SAB) bacteria. Microbial N flow was estimated from the15N enrichment and PB concentration in both total digesta and in the three different bacterial pellets (TB, LAB, and SAB). For all diets, LAB presented a greater (P < 005)15N enrichment and PB: N ratio than SAB, with TB having an intermediate value. For both markers, the use of LAB produced the lowest (P < 005) estimates of microbial N flow and the use of SAB produced the greatest (P < 005) estimates. The use of TB produced intermediate values with all diets, suggesting that TB consisted of SAB and LAB. For all bacterial pellets, PB produced greater (P 005) values of microbial N flow than15N. However, there was a positive relationship (r = 0·883; P 0001; no. = 15) between the values of microbial N flow determined with the two markers when TB were used as reference.