The objective of this study was to compare the estrogenic potency of environmental estrogens at two testing tiers: at the initial level of in vitro screening assays, and at the level of definitive fish reproduction tests in vivo. The in vitro tests included a recombinant yeast estrogen receptor (ER) assay, a competitive radioreceptor assay using the hepatic ER of carp (Cyprinus carpio), and assays on vitellogenin induction in cultured hepatocytes of rainbow trout (Oncorhynchus mykiss) and carp. In vivo, full life cycle tests with zebrafish (Danio rerio) were performed, using fertilization success as estrogen-sensitive reproductive endpoint. The test compounds included the natural estrogen 17beta-estradiol (E2) (only applied in the in vitro assays); the synthetic estrogen ethynylestradiol (EE2); and two xenoestrogens, 4-tert-octylphenol (OP) and bisphenol A (BPA). Among the in vitro assays, differences were observed in the relative ranking of the test substances, and in the absolute sensitivity (EC50 values), although the interassay differences of EC50 values were within one order of magnitude. The in vivo activity of the test compounds was not accurately predicted by the in vitro assays, with respect to neither sensitivity nor ranking. The in vitro assays tended to overestimate the relative potency of the xenoestrogens; i.e. the ratio between the activity of the reference compound, EE2, and that of the test compound. The best prediction of the in vivo fish test results was obtained from the recombinant yeast assay.