Blood donations in the United States have been screened for antibody to human T-lymphotropic virus type I (HTLV-I) by HTLV-I enzyme immunoassay (EIA) since November 1988. Specimens repeatedly found to be reactive by EIA undergo confirmation by supplementary serologic tests. We assessed the accuracy of blood center testing of 994 HTLV-I EIA repeat-reactive specimens in five US blood centers between November 1988 and December 1991. Of 410 confirmed HTLV-I/II donations, 407 (99.3%) were infected with HTLV-I/II, as determined by polymerase chain reaction (PCR) (403 cases) and by repeat serologic testing (4 cases). The three false- positive results occurred in the first year of testing. Of 425 HTLV-indeterminate specimens, 6 (1.4%) were found to be infected by PCR (5 with HTLV-II and 1 with HTLV-I). None of 159 confirmatory test-negative donations was PCR positive. Of HTLV-I/II-seropositive specimens, 80.2% to 95.4% could be typed as HTLV-I or HTLV-II by type-specific serologic assays. These results support recommendations that HTLV-I/II- seropositive donors should be advised that they are infected with HTLV- I, HTLV-II, or HTLV-I/II (depending on results of type-specific assays). HTLV-indeterminate donors should be advised that their results only rarely indicate HTLV infection. HTLV confirmatory test-negative donors should be reassured that they are not infected with HTLV-I or HTLV-II.