This study evaluates five different criteria of true conversion in relation to the annual tuberculosis infection rate or risk of infection. It also examines prospective results of a general population, 16 to 69 yr of age, who had less than 6 mm of induration at first PPD test. This population received a second PPD test between 1976 and 1986. In about 95% of cases, the interval was 1 to 5 yr. In 1,355 bacillus Calmette-Guerin (BCG)-vaccinated subjects, none of the criteria is acceptable for true conversion. Among 1,834 nonvaccinated BCG, only the criterion of a change from less than 10 to 18 mm or more with an increase of 12 mm or more, or the criterion of change from less than 10 to 18 mm or more with an increase of 18 mm or more, is compatible with true conversion and with the correspondent annual risk of infection. But we must accept that with smaller increases, recently infected persons may escape diagnosis of infection. A third test was performed in 81 subjects with two negative test results. In all cases except one, the annual conversion rate observed was much higher than the annual risk of infection. These data reveal shortcomings of the current criterion for conversion. They show that the tuberculin conversion at retesting is not epidemiologically related to new infection and that none of the criteria assures the discovery of true converters. Other factors must be taken into account.