A comparison of serum ferritin determination by seven different commercial immunoassay techniques gave mean coefficients of variation of 57% for normal individuals, 41 . 4% for patients with malignant lymphoma and 43 . 1% for patients with chronic renal failure. One of the immunoradiometric assays gave consistently higher serum ferritin values in both normal and patient groups; mean values were increased (greater than 100% for normal males, greater than 50% for normal females) with respect to the other assays. Underestimation of serum ferritin by an enzyme-linked immunosorbent assay was also evident. Results were affected by storage conditions, but not by dilution of samples, in two kits. Solutions of ferritin purified from normal and malignant human and mouse tissues, and lymphoma isoferritins, were used for reference. There were overestimations of normal human liver ferritin and the most basic isoferritin. Mouse ferritin displayed minimal reactivity in all kits, and human lymphoma ferritin was often underestimated. Variability in serum ferritin determination chiefly reflects the lack of a universal standard for ferritin measurement.