Sixteen mixed rations (TMR) for dairy cows were tested with a semi-continuous rumen-simulation. An 8-day period of collecting samples of the feed residues, the liquid efffluent, and microbes was followed by a 7-day period of adaptation. It was the aim of the experiments to investigate which effects the high variation in TMR composition exerts on the fermentation of nutrients, on the amount and amino acid composition of the microbial crude protein (CPTB), and on the efficiency of the microbial protein synthesis. Based on these data multiple regressions were calculated in order to identify factors of TMR influencing the mentioned parameter. To determine the amount of microbial protein, the buffer solution contained 15N enriched ammonium chloride. Two diets were used to investigate the dynamics of 15N-enrichment of the microbes during continuous 15N application via the buffer. It could be shown that, using this in vitro system, the maximum 15N-enrichment in TB was not achieved before 7 days of a continuous infusion of 15N. Furthermore, a separation of the microbial population into reference microbes (TB), liquid associated microbes (LAB), and solid associated microbes (SAB) was carried out. The fermented proportion of organic matter (OM), crude protein (CP), and crude fibre (CF) varied between 35 and 48%, 25 and 63%, and 9 and 30% after a 48 h incubation. The in vitro fermentation of OM could carefully be predicted by the parameters metabolizable energy, CF and CP (r2=0,79; sy.x=4,9%). The efficiency of the microbial protein synthesis was in a wide range between141 and 286 g XPTB per kg fermented OM using different TMR. A potential factor determined to affect the efficiency of microbial protein synthesis was the content of CP. The use of the parameters CF and "inclusion of grass silage in dry matter" of a mixed ration can improve the accuracy of prediction. It turned out that there were clear differences in the amino acid pattern of the TB's and the SAB's protein. Within one microbial fraction there were significant differences in the composition of the microbial protein between TMR's. Multiple regressions did not clearly indicate any parameter influencing the amino acid pattern.