Two radioimmunoassay (RIA) procedures were used to measure human myelin basic protein (HBP) in cerebrospinal fluid (CSF): (1) an inhibition RIA, with the use of TNP-conjugated anti-BP IgG, 125I-labelled HBP, and anti-TNP-coated polystyrene beads, and (2) a non-competitive two-site RIA, with the use of Sepharose-coupled anti-BP antibodies and 125I-labeled anti-BP IgG. The two-site RIA detects less HBP in CSF than the inhibition RIA, partly due to the presence of HBP fragments in CSF that are detected by the inhibition assay, but less by the two-site RIA. The correlation was improved when in the two-site RIA Sepharose-coupled anti-BP antibodies were changed. Because certain substances (such as autoantibodies to HBP) may give false-positive results in the competitive RIA but not in the two-site RIA, we conclude that a combination of the (more sensitive) inhibition RIA with the (more specific) two-site assay provides a more reliable HBP assay than either assay alone.