Insulin autoantibodies (IAA), a marker for insulin-dependent diabetes mellitus (IDDM), have been reported in other diseases such as thyroid disease and after treatment with sulfhydryl containing medications. Reported prevalences of IAA in non-diabetics vary widely, probably due in part to methodological differences between laboratories. In addition, certain sera may have a high non-specific binding to insulin. We compared a radioimmunoassay (RIA) for IAA which included non-specific binding with an RIA that incorporated a competitive displacement with cold insulin to remove non-specific binding. Using the RIA which measured specific plus non-specific binding, IAA positivity was found in 22/92 (23.9%) of sera from thyroid disease patients, 16/124 (12.9%) of random masked sera from a hospital laboratory, 27/335 (8.1%) of first degree relatives of IDDM patients, 63/178 (35.4%) of subjects with newly diagnosed IDDM, and 0/92 (0%) of normal controls. Insulin antibodies (IA) were found in 80/99 (80.8%) of insulin-treated diabetic subjects. In contrast, using the displacement assay which allowed measurement of specific binding, the frequency of IAA positivity was lower for subjects with thyroid disease (7/92 (7.6%)), random hospital sera (12/124 (9.8%)), and for first degree relatives of IDDM patients (8/335 (2.4%)), while higher for subjects with newly diagnosed IDDM (71/178 (39.9%)). Subjects with insulin-treated diabetes (78/99 (78.8%)) and normal subjects (1/92 (1.1%)) showed little change. Strikingly, three of the eight (37.5%) relatives of IDDM patients that were positive in the RIA measuring specific binding were detected only because cold displacement was utilized. We conclude: (1) subjects with thyroid disease and first degree relatives of IDDM patients frequently have high non-specific binding for IAA in an RIA not employing a cold displacement step, (2) in some newly diagnosed IDDM patients and first degree relatives of IDDM patients, IAA may be missed by an assay not optimized to measure specific binding, and (3) displacement with cold insulin increases both the specificity and sensitivity of RIAs measuring insulin autoantibodies.