To evaluate factors that might influence the accuracy of tuberculin tests in identifying elderly persons recently infected with Mycobacterium tuberculosis, we performed as many as 3 sequential administrations of 5 tuberculin units of purified protein derivative in 1,726 persons residing in chronic care facilities. Significant reactions (greater than or equal to 10 mm of induration) to 1 of 3 tests were found in 702 (40.7%) persons. Of these, 68% were found with Test 1, 22.5% with Test 2, and 9.5% with Test 3. Of 1,146 persons with nonsignificant reactions to Test 1, 13.8% had significant reactions on Test 2, and of 769 persons with nonsignificant reactions to Tests 1 and 2, 8.7% had significant reactions on Test 3. Males, nonwhites, and persons between 50 and 79 yr of age had a greater proportion of significant reactions for each of the first 2 tests but not for the third test. These data indicate that continued boosting of the tuberculin reaction occurs in a substantial number of persons who receive a third sequential test. Marked increases in the size of reactions caused by boosting may explain high apparent conversion rates found in facilities where the third test is delayed for one year.