The effects of genotoxic substances on ecosystems should be assessed using various test systems with multiple genetic end points. The most widely used test system has been the specific-locus test developed by W.L. Russell, using the mouse. We are developing a new, nonmammalian test system using the Japanese medaka, Oryzias latipes. We have examined 625,926 embryos that correspond to 1,586,649 loci. In the medaka test system, four genetic end points are evaluated: dominant lethals, total mutations, viable mutations, and malformations. Because the medaka is an oviparous experimental animal, we were able to determine that approximately 90% of spontaneous as well as gamma-ray-induced total mutants died during development, irrespective of spermatogenesis stages at the time of exposure. Exposure of sperm and spermatids to ethylnitrosourea (ENU) also resulted in embryonic death of approximately 90% of total mutants. In sharp contrast, approximately 90% of total mutants recovered from ENU-exposed spermatogonia became viable mutants. These results indicate that the quantitative relationship between induction of specific-locus mutations and dominant lethals remains the same among spermatogenesis stages for gamma-rays, while it is biased excessively to the induction of specific-locus mutations in ENU-exposed spermatogonia. Thus, the assessment should integrate at least two factors, agent-specific and species-specific effects.