A serum panel comprising 19 samples of known (five positives and 14 negatives) but undisclosed HIV-1-antibody content was distributed to 30 national reference laboratories for HIV serology. In order to simulate normal circumstances of referral, participants were asked to test the panel for HIV-1-antibody status using their normal procedures. Results of testing were returned by 28 participants. There were great variations in the number and combinations of tests used. The number used ranged from one to five assays per laboratory and none of the 24 laboratories using two or more tests employed the same combination. A high average success rate of 99% was seen with the positive samples. More errors occurred with the negative samples, with an average of 87% correct negative reports. Only four of the 14 negative specimens were reported as negative by all participants.