We have devised a practical, sensitive, and reliable assay for measurement of free T3 concentration in serum. The assay employs a convenient and disposable plastic equilibrium dialysis cell and a buffer that resembles the in vivo biochemical environment (Nelson JC, Tomei RT 1988 Clin Chem 34:1737). A 200-microliters aliquot of serum was dialyzed against 2.4 mL buffer at 37 degrees C for 18 +/- 2 h and T3 was quantified by RIA of about 1.0-mL aliquot of the dialysate buffer. The detection threshold of the RIA approximated 2 pg/ml permitting accurate measurement of > 200 pg/dL of free T3 directly. Serum specimens that contained less free T3 were spiked with 200 ng/dL of non-radioactive T3 prior to dialysis. Free T3 in the dialysate of these samples was divided by total T3 in serum (after spiking) to determine percent free T3. Free T3 was calculated by multiplying percent free T3 and serum total T3 (before spiking). Free T3 concentration (pg/dL) did not differ appreciably in a serum pool when tested both with and without spiking with exogenous T3. The between assay coefficient of variation of three specimens tested over an 8-month period averaged 20%. Serum free T3 concentration (pg/dL) was [mean +/- SD (n), range, p] [293 +/- 12 (39), 154-440] in normal subjects. It was significantly increased [742 +/- 87 (13), 525-1700, p < 0.001] in hyperthyroidism and significantly decreased in nonthyroidal illness [NTI, 138 +/- 26 (9), 53-320, p < 0.001], cord blood serum [124 +/- 7.5 (11), 93-353, p < 0.001], and third trimester of pregnancy [214 +/- 26 (8), 93-253, p < 0.02]. Serum free T3 concentration varied widely in hypothyroidism 274 +/- 92 (10), 10-923, NS]. We have described a practical method and initial results of direct measurements of free T3 concentration in health and disease.