Human placentas were perfused in vitro through fetal vessels of a single cotyledon with a solution of tritiated androstenedione (A) or tritiated testosterone in oxygenated Krebs-Ringer bicarbonate buffer containing glucose. The same buffer, without labeled steroids, was used to perfuse the maternal side of the placenta. Fractions of unrecycled "fetal" and "maternal" perfusates were collected to estimate the steady-state rates of output into each perfusate of labeled estrone (E1) and estradiol (E2), formed by aromatization of the perfused androgens. An unequal distribution of the two estrogens was observed, i.e., a significantly larger proportion of E2 than of E1 was released to the maternal perfusate. Addition of unlabeled A or ethinyl estradiol (10(-5)M) to the fetal perfusion buffer resulted in a redistribution of the E2 output toward the fetal perfusate. These in vitro results are similar to the findings of Walsh and McCarthy, who detected by tracer experiments an unequality in the in vivo fetomaternal distribution of the output of placentally produced E1 and E2 in rhesus monkeys. These results also support the possibility of the existence of saturable specific E2 carriers in the syncytiotrophoblast, postulated by the same authors to explain their findings. However, an extrasyncytial conversion of E2 to E1 could also account for the disproportion in the relative output of the estrogens toward mother and fetus. The perfusion system used in these studies may be useful for the elucidation of factors determining the relative rates of placental secretion of steroids toward mother and fetus.