From 1949 to 1951, a controlled trial of BCG vaccinations was conducted in Puerto Rico. The 191,827 children, 1-18 years of age, initially enrolled in the study were skin-tested with tuberculin to determine their eligibility for vaccination. A total of 82,269 children were classified as reactors and not vaccinated. Of the 109, 558 nonreactors, 31,586 refused vaccination, 50,634 were vaccinated with BCG, and 27,338 were left unvaccinated as controls. We ascertained the incidence of cancer over an average follow-up period of 23.3 years in the latter two groups using the Puerto Rico Central Cancer Registry. By the end of December 1973, a total of 77 cancers had been diagnosed among the controls and 150 among the vaccinees. The overall incidence of cancer among the two groups was similar. Although a number of differences existed between the vaccinee group and the controls in regard to the incidence of cancer at various "sites", none of these differences was statistically significant. However, when cases of lymphosarcoma and Hodgkin's disease were combined for analysis, a statistically significant excess of cases occurred among the vaccinees. We concluded that BCG vaccination had no protective effect on the subsequent development of cancer in this population. The slight excess of cases of lymphosarcoma and Hodgkin's disease among the vaccinees raised the possibility that BCG may have had an adverse effect.