The percentage of tuberculin reactors, by age, is sometimes used in the epidemiology of tuberculosis as one measure of the tuberculosis problem in a population. Whereas such data are useful for determining the load of infection in communities, it is shown in this paper that only in very few instances do they lend themselves to a quantitative determination of the risk of infection. Such data are, therefore, of limited epidemiological value, and the author suggests other types of data which could give more useful information.The results of tuberculin-testing obtained in sixteen mass BCG campaigns, from 1948 to 1951, have been used to illustrate the problems under discussion.