300 children considered to have had adverse reactions to penicillin were examined. Informed consent was obtained from the parents. Skin tests were conducted by the scratch/prick and intradermal techniques, using benzylpenicilloyl polylysine conjugate and a mixture of minor determinants of penicillin. Specific anti-penicillin IgE antibodies were estimated by the radioallergosorbent test. There was a good correlation between the two methods. The overall frequency of positive tests was 19%. 11 children showed cutaneous reactivity only to the minor determinants mixture. Positive results were found more often in those with accelerated adverse reactions, particularly anaphylaxis, serum sickness, angio-oedema, or urticaria. The validity of penicillin-negative results was confirmed by drug challenge in 56 subjects, only 2 of whom showed a slight skin rash. Of 5 patients with positive tests, inadvertent administration of penicillin produced accelerated urticaria in all. 14 of 42 children with positive tests had lost hypersensitivity to penicillin one year later. In a separate group of 50 children with a history of adverse response to ampicillin, the overall frequency of positive tests was 12%; 38% showed evidence of recent E-B virus infection. It was concluded that penicillin allergy is often overdiagnosed. The diagnosis can be reliably confirmed by skin tests using major and minor determinants of benzylpenicillin and by the radioallergosorbent test; such hypersensitivity is not permanent.