Our aim was to determine whether density of immunolabeling can be used to estimate the amount of an antigen in a tissue. The biological model was the pancreatic insulin-containing B cell. The insulin content of the pancreas of Wistar rats was decreased by five injections of glibenclamide (0.5, 1, or 2 mg/kg) every 12 hours. After resection of the whole pancreas specimens were taken for insulin extraction and measurement by radioimmunoassay and for immunocytochemistry. The sections were treated either by a polyclonal anti-insulin serum at 1/500 or 1/3000 and peroxidase-antiperoxidase complex or by a monoclonal anti-insulin serum at 1/500 and indirect immunoperoxidase. Peroxidase was revealed by diaminobenzidine. The density of immunostained B cells was determined with an automatic image analyzer (Ibas 2000, Kontron, FRG). Compared with controls, pancreatic insulin concentration was decreased by about 40, 60, and 85% in rats treated by the three doses of glibenclamide. A strong correlation was found between the insulin concentration and the optical density of islets under certain conditions: with the monoclonal anti-insulin serum (r = 0.90) and with the polyclonal anti-insulin serum at a high dilution (r = 0.95) but not at a low dilution (r = 0.13). With the latter, the optical density was high even in islets with reduced insulin content. In conclusion, a low dilution of antiserum should be used to detect cells with a small amount of antigen, whereas a higher dilution makes it possible to estimate the antigen concentration in the tissue. Thus, under appropriate conditions, a linear relationship exists between the optical density of the immunostained material and the concentration of immunoassayable antigen. This technique may thus prove useful in evaluating the functional state of cells, in particular secretory cells, under normal or pathological conditions.