The sensitivity and specificity of the polymerase chain reaction (PCR) for the detection of HIV-1 proviral DNA was determined in five laboratories with extensive experience in PCR testing. Five panels consisting of 105 HIV-1-seronegative specimens from regularly repeating blood donors with no risk factors for HIV infection and 99 HIV-1-seropositive and culture-positive specimens from a cohort of homosexual/bisexual men were sent under code to each laboratory. Amplification procedures and testing algorithms by which specimens were judged positive, negative, or indeterminate varied between laboratories. The average sensitivity for the five laboratories was 99.0%, with two laboratories achieving 100%. The average specificity was 94.7%, varying between 90.5 and 100%. The overall false-positive rate was 1.8%, the false-negative rate was 0.8%, and the indeterminate rate was 1.9%. Of 1,005 determinations made by the five laboratories, 32 (3.2%) were misclassifications. Most of the classification errors occurred in specimens from uninfected individuals and were distributed among the laboratories in such a way as to indicate laboratory error rather than the inherent reactivity of some samples. This emphasizes the need for standardization of PCR testing and caution in interpreting positive PCR reactions in HIV-1-seronegative persons.