In the course of eight successive Brucella abortus experiments, 394 cows were followed by clinical, serological and bacteriological examinations. Cultures were made from vaginal swabs and colostrum at delivery and from 10 pools of lymph nodes or organs sampled at slaughter, about 6 weeks after calving. From these cows, 346 had been experimentally challenged with 1.5 X 10(7) B abortus strain 544 via the conjunctiva, 3 congenitally infected and 45 bought from known naturally infected herds. Males were more numerous (64%) in aborted foetuses and less numerous (34%) in premature live calves than females, the sex ratio being normal in calves born at term. Recovery of Brucella infection occurred between calving and slaughter in 13/230 cows (5.7%). It appeared that localization, spread and level of infection in different lymph nodes and organs, duration and outcome of pregnancy, and agglutination titers were strongly correlated. Comparisons between experimentally and naturally infected cows evidenced the fact that both types of infection were very similar. Confidence for field application of experimental results was thus reinforced.