Testing with antigens that elicit delayed-type cutaneous hypersensitivity reactions is commonly used to evaluate immune competence in persons infected with the human immunodeficiency virus; however, the reliability of such testing has not been determined. We performed serial testing with tuberculin, mumps, and Candida antigens in 491 HIV-infected persons and found that 30% of persons who initially had no reaction (0 mm) to any of the three antigens, and, thus, were considered to be anergic, had reaction to the mumps or Candida antigen when they were retested 12 months later. We also examined the results of mumps antigen tests in 50 subjects who had a negative tuberculin tests after an initial positive test. The mumps antigen test was positive in 39% of the subjects when the tuberculin test was falsely negative. We conclude that tests commonly used to define anergy cannot reliably identify the anergic state. Moreover, using the mumps antigen to aid in the interpretation of the tuberculin test will often lead to erroneous conclusions. These data indicate that the results of anergy testing should not be used to make individual patient decisions concerning preventive therapy for tuberculosis.