The ability of Bifidobacterium bifidum from a commercial bifidus milk to antagonize Salmonella enteritidis subsp. typhimurium in vivo, and to reduce the pathological consequences for the host, was determined using conventional and gnotobiotic mice. Conventional animals received daily, by gavage, 0·1 ml bifidus milk containing about 109 cfu B. bifidum and germ-free animals received a single 0·1 ml dose. The conventional and gnotobiotic groups were challenged orally with 102 cfu of the pathogenic bacteria 5 and/or 10 d after the beginning of treatment. Control groups were treated with milk. Bifidus milk protected both animal models against the challenge with the pathogenic bacteria, as demonstrated by survival and histopathological data. However, to obtain the protective effect in gnotobiotic animals, the treatment had to be initiated 10 d before the challenge. In experimental and control gnotobiotic mice, Salm. enteritidis subsp. typhimurium became similarly established at levels ranging from 108 to 109 viable cells g−1 of faeces and remained at these high levels until the animals died or were sacrificed. It was concluded that the protection against Salm. enteritidis subsp. typhimurium observed in conventional and gnotobiotic mice treated with bifidus milk was not due to the reduction of the intestinal populations of the pathogenic bacteria.