A tetravalent rhesus-human reassortant rotavirus (RRV-TV) vaccine (4 x 10(4) plaque-forming units/dose) was evaluated for safety, immunogenicity and efficacy in a prospective, randomized, double-blind, placebo-controlled trial involving 540 Brazilian infants. Doses of vaccine or placebo were given at ages 1, 3 and 5 months. No significant differences were noted in the occurrence of diarrhoea or vomiting in vaccine and placebo recipients following each dose. Low-grade fever occurred on days 3-5 in 2-3% of vaccinees after the first dose, but not after the second or third doses of vaccine. An IgA antibody response to rhesus rotavirus (RRV) occurred in 58% of vaccinees and 33% of placebo recipients. Neutralizing antibody responses to individual serotypes did not exceed 20% when measured by fluorescent focus reduction, but exceeded 40% when assayed by plaque reduction neutralization. There were 91 cases of rotavirus diarrhoea among the 3-dose (vaccine or placebo) recipients during two years of follow-up, 36 of them among children given the vaccine. Overall vaccine efficacy was 8% (P = 0.005) against any diarrhoea and 35% (P = 0.03) against any rotavirus diarrhoea. Protection during the first year of follow-up, when G serotype 1 rotavirus predominated, was 57% (P = 0.008), but fell to 12% in the second year. Similar results were obtained when analysis was restricted to episodes in which rotavirus was the only identified pathogen. There was a tendency for enhanced protection by vaccine against illness associated with an average of 6 or more stools per day. These results are sufficiently encouraging to warrant further studies of this vaccine in developing countries using a higher dosage in an attempt to improve its immunogenicity and efficacy.