The latest controlled trial of BCG vaccination in southern India showed that two vaccines failed to confer protection against pulmonary tuberculosis. This result cast serious doubt on the effectiveness of BCG vaccination of the newborn, which is widely applied in developing countries. Therefore, WHO initiated a global research study to evaluate current programmes in developing countries. Part of this study was carried out in Lomé, Togo, in which child contacts of newly detected patients were followed up with clinical and radiological examinations. All observations were recorded according to a scoring system. Concomitant observations were made to verify the comparability of the vaccinated and unvaccinated children. Of the child contacts of 352 index cases, 1421 completed the examinations. The distribution of the final score made it possible to distinguish 175 children likely to suffer from tuberculosis: 113 among the 546 unvaccinated and 62 among the 875 vaccinated children. Significant incomparability was observed in respect of intensity of exposure: the vaccination coverage was relatively low, and the risk of disease relatively high, if a parent was the index case or the child shared the bedroom of the index case (which very often coincided). The other variables studied, including age and sex, turned out to be practically irrelevant as regards comparability. The estimate of the protective effect against all types of tuberculosis combined is 61.5%, which is slightly lower than suggested by the raw data (66%). The protective effect, however, appeared to increase considerably with severity of disease. In children of 5 years and older it was lower than in the younger children. Tuberculin testing failed to reveal any sensitivity induced by BCG in the vaccinated children. The distribution of the tuberculin reactions correlated poorly with the other diagnostic findings. Small reactions were only slightly more frequent in healthy than in sick children; only the very large reactions were associated with a higher risk of disease. This confirms that the tuberculin test is of very limited diagnostic value in young children.