A large scale community-based double blind randomized controlled trial was carried out in Chingleput district of south India to evaluate the protective effect of BCG against bacillary forms of pulmonary tuberculosis. From among 366,625 individuals registered, 281,161 persons were vaccinated with BCG or placebo by random allocation. Two strains of BCG were used, the French and Danish, with a high dose (0.1 mg/0.1 ml) and a low dose (0.01 mg/0.1 ml) in each strain. The entire population was followed up for 15 years by means of resurveys every 30 months, and selective follow up every 10 months and continuous passive case finding. There were 560 cases (189, 191 and 180 from the high dose, low dose and placebo groups respectively) arising over 15 years, among 109,873 persons who were tuberculin negative and had a normal chest X-ray at intake. This represents a small fraction of the total incidence of 2.6 per 1000 person-years most of which came from those who were initially tuberculin positive. The incidence rates in the three "vaccination" groups were similar confirming the complete lack of protective efficacy, seen at the end of 7 1/2 years. BCG offered no overall protection in adults and a low level of overall protection (27%; 95% C.I. -8 to 50%) in children. This lack of protection could not be explained by methodological flaws, or the influence of prior sensitisation by non specific sensitivity, or because most of the cases arose as a result of exogenous re-infection. The findings at 15 years show that in this population with high infection rates and high nonspecific sensitivity, BCG did not offer any protection against adult forms of bacillary pulmonary tuberculosis.