A routing radioimmunoassay for human C-peptide in serum is described. Antibodies against human C-peptide were raised by immunizing guinea pigs with human b-component. Nine out of 12 animals produced useful antibodies within 6 months. Insulin antibodies coupled to Sepharose were used to bind human proinsulin and insulin in the serum and after centrifugation C-peptide was determined in the supernatant. The detection limit of the assay (calculated as 2 SD from zero) was about 0.003 pmole of C-peptide (in 100 mul). The main sources of error were: (1) Normal and diabetic sera devoid of C-peptide gave a displacement of 125I-Tyr-C-peptide varying from 0 to 0.16 nM (6 different antisera). Only one antiserum (M 1181) showed no displacement, and the values of C-peptide determined with this antiserum in normal and diabetic sera were lower than the values determined with another antiserum, which gave a value of 0.07 nM in the sera free of C-peptide. It is suggested that displacement found with most antisera is due to substances in serum that are not related to C-peptide or proinsulin. (2) Serial dilutions of pancreatic extracts and sera may yield dilution curves slightly different to those of the synthetic standard. Possible explanations are discussed. These sources of error can be eliminated or reduced by the proper selection of antisera. Fasting sera from 15 normals, 8 maturity-onset diabetics and 10 insulin-requiring diabetics showed the following concentrations of C-peptide: (M 1181) 0.35 +/- 0.09, 0.74 +/- 0.51 and 0.21 +/- 0.14 (nM, mean +/- SD). One hour after 1.75 g/kg oral glucose the values increased to 2.24 +/- 0.71, 2.34 +/- 0.24 nM.