Four fluorescent treponemal antibody absorbed (FTA-ABS) test kits were evaluated with 86 treponemal and 84 non-treponemal sera selected with enzyme immunoassay (EIA). Agreement between all four kits was 63% for treponemal, and 50% for non-treponemal, sera. Discrepancies with treponemal sera were associated with low levels of antibody characterised by T. pallidum haemagglutination assay (TPHA) titres < or = 160 and a negative Venereal Diseases Research Laboratory (VDRL) test. Discrepancies with non-treponemal sera were significantly associated with false reactivity on screening with EIA. Possible reasons for the differences obtained between kits, the importance of screening policies for pre-selecting sera, and the significance of equivocal (borderline) reactions on positive and negative predictive values are discussed.