An assay for human islet cell antibodies (ICAs) in serum yielding numerical values and amenable to statistical evaluation has been developed utilizing fluorescence spectrophotomicroscopy (FSPM). The results of the blinded trials by FSPM were compared with those by standard indirect immunofluorescence (IFL). No false-positive or false-negative readings were obtained in 258 observations when the results by FSPM were compared with those by IFL. The intra- and interassay variabilities encountered were not enough to misclassify a specimen. The presence of anti-thyroid, anti-adrenal, or anti-nuclear antibodies did not produce false-positive readings. Twenty-eight additional specimens from diabetic children were also analyzed via three blood group type O pancreases. There was complete agreement concerning the presence or absence of ICA between IFL and FSPM analyses. Analysis of the three pancreases yielded different numbers of results positive for ICA (14/28 vs. 22/28 vs. 15/28, P = .008) in both assays. Thus, selection of pancreatic substrate may influence the outcome of assays for ICA. A matrix of fluorescent microspheres has been devised that allows calibration of the FSPM system. Now, reproducible and comparable readings for ICA values can be obtained from the various reporting laboratories. Should an international reference serum for ICA become available, the remaining problem in the ICA assay, that of substrate (pancreas) variability, should be resolved.