We compared three serum ferritin radioimmunoassay kits and one noncommercial RIA for ferritin quantitation, with regression analysis of results for sera from 35 ostensibly healthy subjects. There was a good correlation (p less than 0.001) between these various RIAs, but the slope of the regression line varied widely, most probably because of lack of standardization of the serum ferritin assay. Determination of the ferritin content of purified samples of tissue ferritin revealed that the kits differ in specificity, differences for purified human spleen ferritin and human liver ferritin being larger than those for normal sera. Removing the iron from purified liver ferritin increased antiserum binding in two of the kits by twofold, but had no effect in two other kits. We conclude that commercial RIA methods for serum ferritin differ in specificity, and that this difference is related to the source of ferritin used in the production of the antibodies.