Two hundred and eighty-eight subjects with a history of allergy to penicillin were studied for objective proof of their allergy. On the basis of skin tests, specific IgE antibody measurements and direct challenge tests. 64 patients (22%) were shown objectively to be allergic to one or more penicillins. The following tests were carried out: skin tests to benzyl-penicilloyl poly-L-lysine (BPO-PLL), minor determinant mixture (MDM), amoxycillin (AX) and ampicillin (AMP), in-vitro IgE antibody measurement to benzyl-penicilloyl (BPO) and AX and challenge with benzylpenicillin (BP), phenoxy-methyl-penicillin (PV) and amoxycillin. Forty-four cases were found to respond to benzyl or phenoxymethyl-penicillin, however, 20 were shown to be sensitive to amoxycillin and unresponsive to tests with other penicillins. The contribution that any individual test gave for establishing the diagnosis was 21.8% for skin testing with BPO-PLL, 9.3% with MDM and 12.5% with AX. Nine point three per cent were RAST positive to BPO and 1.5% to AX; 7.8% developed a positive response after challenge to BP, 7.8% to PV and 14% to AX. In 16% of the 64 positive cases more than one test was found to be positive. The challenge tests suggested that not all the penicillin-sensitive subjects had IgE-mediated reactions implying other immunological mechanisms. These results clearly demonstrate the importance of side chain-specific diagnostic reagents and challenge tests. Thirty-one per cent of the positive group or 6.9% of the total group would have been missed in this study using benzyl or phenoxymethyl-penicillin diagnostic reagents alone.