During the period 1949-51 a controlled trial of BCG vaccination was initiated among a study population of 191827 children in Puerto Rico. There were 82269 children who were classified as reactors to 10 Tuberculin Units of PPD. Nonreactors comprised 31586 children who did not wish to be vaccinated 27338 controls and 50634 vaccinees. Follow-up consisted of matching officially reported cases to the study records; this was discontinued in 1969. Tuberculosis rates were highest among the children who were most sensitive to tuberculin initially and about 75% of the subsequent tuberculosis came from the reactors. The rates were 30% lower among vaccinees than among controls. About 8% of the tuberculosis problems in this population could have been prevented by vaccinating all nonreactors. It is the opinion of the authors that BCG vaccination is indicated only when there is a high risk of becoming infected with tubercle bacilli. Because there is doubt about the possibility of identifying currently potent strains of BCG they recommend comparative trials in man using freeze-dried BCG in areas with a high risk of tuberculosis infection in order to determine which strains confer the greatest protection. (authors)