Data collected in a national external quality assessment program for free thyroxine (fT4) and free triiodothyronine (fT3) were analyzed to evaluate the performance of 10 method/kits with 26 control samples distributed to approximately 170 laboratories. The control materials were normal serum pools, pooled sera supplemented with thyroid hormones, a pregnancy serum pool, serum pooled from patients with familial dysalbuminemic hyperthyroxinemia (FDH), and a normal serum pool progressively diluted. The between-laboratory variability (CV) was approximately constant in normal and supplemented pools for fT4 (15.3%) and fT3 (24.0%) but markedly increased in diluted, pregnancy, and FDH pools (21.9-35.2% for fT4 and 28.6-66.5% for fT3) because of increases in systematic between-kit differences in control samples with altered binding-protein capacity. Moreover, free hormone concentrations measured in progressively diluted sera averaged lower than in undiluted samples. This decrease of concentration was less for back-titration or labeled-antibody techniques and greater for labeled-analog methods; only the method involving adsorption to cross-linked dextran (Sephadex) was unaffected by dilution. Evaluation of the reproducibility of the method/kits showed between-assay, between-laboratory precision ranging from 7.8% to 17.0% for fT4 and from 9.8% to 20.3% for fT3.