Some equilibrium dialysis determinations of free thyroxine (T4) vary directly with thyroxine-binding globulin (TBG) concentration. This apparent TBG dependence has been limited to methods involving radiolabeled T4 added to the dialysis system (tracer dialysis). In this study we compared tracer dialysis with direct dialysis for determining free T4 and obtained the following results (mean +/- SD) for patients with hypothyroxinemia of nonthyroidal illness (23.8 +/- 10.7 vs 24.2 +/- 10.9 pmol/L, P greater than 0.8), patients with congenital TBG deficiency (11.4 +/- 2.2 vs 16.2 +/- 7.1 pmol/L, P greater than 0.05), normal control subjects (32.7 +/- 6.5 vs 18.5 +/- 5.8 pmol/L, P less than 0.001), and pregnant women (31.2 +/- 8.7 vs 12.1 +/- 2.6 pmol/L, P less than 0.001). Direct dialysis determinations were independent of TBG and total T4. Tracer determinations were greater than direct determinations in normals, a discrepancy that increased in pregnancy. Tracer determinations correlated significantly with total T4 and TBG concentrations (P less than 0.001). TBG and total T4 dependence in the tracer method was attributable to small overestimations of the free fraction of T4. Similar overestimations multiplied by increasing total T4 concentrations resulted in greater errors. Relative to results for normal sera, the tracer method overestimated free T4 when total T4 was increased and underestimated free T4 when total T4 was decreased.