A cluster sampling survey was performed in 1989 in Libreville, Gabon, to determine HTLV-I and HTLV-II prevalence and to compare the efficacy of polymerase chain reaction (PCR) and serology in detecting HTLV-I and HTLV-II infections. A total of 322 sera from adults were tested by ELISA and by Western blot (WB). The WB patterns were interpreted according to WHO criteria and those of the manufacturer. PCR analysis using primer pairs in the gag and pol region, with a specific probe for HTLV-I and HTLV-II, was performed on the lymphocytes of the 322 adults. In addition, 134/322 samples were re-tested with tax primers, in a second laboratory. Using WHO criteria, 8/322 (2.5%) samples were positive on WB and 25 were indeterminate; with the criteria of the kit, 26/322 (8.1%) were positive and 7 were indeterminate by WB. By PCR, 13 (4%) samples were positive, including 12 for HTLV-I (3.7%) and one for HTLV-II (0.3%). All 8 seropositive samples (by the WHO criteria) were positive by PCR, as were 4 out of 25 indeterminate samples. Only one out of 289 seronegative samples was positive by PCR. In contrast, only 12/26 positive samples by the kit criteria were confirmed by PCR. These results confirm the relatively high HTLV-I/II seroprevalence observed in Gabon. HTLV-I infection is preponderant, but HTLV-II is also present. The WHO criteria for WB give a better fit with PCR results than the kit criteria for WB. In the absence of a specific confirmatory test and based on the uncommon "seronegative" HTLV-I/II infection, the indication for PCR appears limited to the positive WB samples (to differentiate HTLV-I and II infection) and to the indeterminate WB samples.