To evaluate the sensitivity and specificity of HTLV-I/II assays, serum from 1100 pregnant Haitian women was tested with seven commercially available HTLV I/II assays. Serum that was found to be reactive in any assay was analysed by western blot and all indeterminate samples were further characterised by radioimmunoprecipitation assays (RIPA). 59 (5.4%) samples were HTLV I/II antibody positive by western blot and/or RIPA. The sensitivity of these seven assays ranged from 93.2% to 100%. with the 'Recombinant HTLV-I' (Cambridge Bioscience) and 'Serodia HTLV-I' (Fujirebio) assays having the highest sensitivity (100%). The specificity of these assays ranged from 98.4% to 100%, with the Abbott assay having the highest specificity (99.5%, 100%) according to two different methods of evaluation. Whether the antigens used in any assay were whole disrupted virus or recombinant gene products made no difference. The low positive predictive values of some of these assays (71.8-91.7%), even in a high prevalence population, and the need for RIPA to test indeterminate sera, indicate that for routine screening of blood donors there is still room for improvement both in screening and confirmatory assays for HTLV-I/II.