The impact of 2 doses of a Saccharomyces cerevisiae were evaluated, 5 × 10(10) cfu/kg of feed (L1) and 5 × 10(11) cfu/kg of feed (L2) against a control (CON) with no added yeast, using an in vitro model [colon simulation technique (Cositec)] to mimic digestion in the pig colon. The L2 (but not L1) dose significantly improved DM digestibility compared to CON (61 v 58%) and increased NH(3) concentrations (+15%). Volatile fatty acid concentrations increased with L2 compared to CON--isobutyrate (+13.5%), propionate (+8.5%), isovalerate (+17.8%), and valerate (+25%)--but only valerate was increased with L1 (+14.2%). The analysis of microbiota from the liquid associated bacteria (LAB) and solid associated bacteria (SAB) revealed an interaction between the fraction and treatment (P <0.05). Indeed, L2 had a significant impact on SAB and LAB (P <0.01) whereas L1 only tended to change the structure of the population in the SAB (P <0.1). Overall, this study showed that a live yeast probiotic could improve digestion in a colonic simulation model but only at the higher dose used and this effect was associated with a shift in the bacterial population therein.