It is a well recognized problem that sample derived transferrin-bound iron (Tf-Fe) interferes with radio-iron incorporation into heme in the in vitro assay of erythropoietin (Ep) using fetal mouse liver cells (FMLC). This paper describes a mathematical procedure to correct for the unknown quantity of "cold" Tf-Fe in a plasma sample, being assayed for Ep in the FMLC. Excellent recovery of Ep activity was obtained with this method of correction, when testing solutions containing Step III sheep Ep with and without human Tf-Fe. The distorted dose response curves, obtained, when testing specimens of rat plasma, showed, after correction, linearity and parallelism to a Step II Ep dose response curve, permitting valid estimation of the potency ratio. Reproducibility of this method was found to be excellent. Accuracy was acceptable with plasma samples containing more than 50 mU Ep/ml. The FMLC and the exhypoxic polycythemic mouse assay were compared using several batches of rat plasma. Good agreement was found between the results in both assays. As only a small amount of sample is needed (0.2 ml plasma) this method enabled us to estimate EP activity in single rats serially without disturbing their plasma volume significantly.