A two-site (liquid-phase) immunoradiometric assay (IRMA) for dimeric inhibin has been developed using antibodies raised against synthetic peptide sequences corresponding to the N-terminus (1-32) of the α subunit and the C-terminal region (82-114) of the βA subunit of Mr ∼30,000 human inhibin. Highly-purified Mr 32,000 bovine inhibin (standard) gave a dilution curve parallel to those for bovine follicular fluid (bFF), human (h)FF and rat ovary extract. Whilst the assay detected both Mr 56,000 and 32,000 inhibin forms in bFF, little reaction with higher Mr forms was evident. Cross-reaction of 'free' inhibin subunit (Mr 25,000 form) and recombinant human activin A in the IRMA were minimal (< 0.1 and < 2% respectively). Although the detection limit of the IRMA (∼ 50 pg/tube) was similar to that of several reported radioimmunoassays (RIA), the IRMA was unable to detect dimeric inhibin in jugular or utero-ovarian vein plasma of heifers. Similarly, when assayed by IRMA, bFF, hFF and rat ovary extract contained 8-58 times less inhibin than was indicated by RIA. These observations are consistent with earlier evidence that the ovary secretes a substantial excess of 'free' inhibin α subunit and that this material reaches the peripheral circulation. Surprisingly, however, the inhibin contents of bFF, hFF and rat ovary extract determined by in vitro bioassay were 8-23 times greater than the corresponding IRMA values, being similar to those derived by RIA. It is suggested that this quantitative discrepancy between inhibin contents estimated by IRMA and bioassay may be due to (1) loss of bioactivity of the inhibin standard during its purification and/or storage, (2) failure of the IRMA to detect high Mr forms of bioactive inhibin and/or (3) cross reaction of follistatin and other FSH-suppressing substances in the in vitro bioassay.