Five natural, pharmaceutical, or xenobiotic chemicals [17beta-estradiol (E2), ethynylestradiol (EE2), diethystilbestrol (DES), methoxychlor (MXC), nonylphenol (NP)] were tested in two in vitro assays [yeast estrogen screen (YES), MCF-7 breast tumor cell proliferation (E-Screen)], and compared with previously reported results from two in vivo male sheepshead minnow vitellogenin (VTG) production studies. The purpose of this investigation was to determine how accurately the two in vitro assays predicted responses observed in live animals. EC50 values for all five chemicals were approximately one order of magnitude less sensitive in the YES assay than in the MCF-7 assay. Based on the EC50 values, DES was 1.1 (YES) to 2.5 (MCF-7) times more potent in these receptor binding assays than was E2, while EE2 was slightly less potent than E2 in the YES assay (0.7) and nearly twice as potent (1.9) as E2 in the MCF-7 assay. EE2 and DES were of approximately equal potency in the 13-day sheepshead minnow VTG production bioassay. Both MXC and NP were 10(7) times less potent than E2 in the YES assay, MXC was 10(5) times less estrogenic than E2 in the MCF-7 assay, while both were approximately 100 times less potent than E2 in the live animal bioassay. The in vitro tests were substantially less sensitive (at least 1000 times) than the sheepshead minnow VTG assay for estimating estrogenic potency of the two xenobiotic chemicals, which suggests that in vitro-based, large-scale screening programs could potentially result in many false negative evaluations.