The European Centre for the Validation of Alternative Methods (ECVAM) has sponsored a large international prevalidation and validation study of three embryotoxicity tests, involving embryonic stem cells, limb bud micromass cultures, and post-implantation whole-embryo cultures. The main objective of the study was to assess the performance of these in vitro tests in discriminating between non-embryotoxic, weakly embryotoxic and strongly embryotoxic compounds. An initial part of the study was to select 20 test substances for the formal validation trial, conducted under blind conditions. A database of in vivo and in vitro developmental toxicity test results was complied on 310 chemicals that had been used in previous validation studies, or suggested for such use, or that had good quality "segment II"-type in vivo data, or for which there were human data. From this database, a shortlist of about 30 candidates was constructed. Because the ECVAM study would not include metabolic activation, chemicals known to require activation for their developmental effects were excluded as candidates, although some known stable metabolites were included. Attempts were made: to include substances of diverse mechanism; to avoid overemphasis on pharmaceuticals; to avoid biologically inert substances as non-embryotoxicants; and to make the list different from those used previously. The candidates were of three categories: Class 3, strongly embryotoxic, was defined as developmentally toxic in all species tested, inducing multiple developmental effects, and with a high A/D ratio. Class 1, non-embryotoxic, was defined as not developmentally toxic at maternally toxic exposures, but which may show some minor embryo/fetal toxicity, which cannot be separated from maternal toxicity. Class 2, weakly embryotoxic, were chemicals of intermediate activity. From this candidate list, chemicals of known receptor (androgen, oestrogen, glucocorticoid, aryl hydrocarbon) mechanisms were excluded, on the basis that simple tests for such activity are already available. In addition, chemicals not freely available were excluded, and an emphasis on human data was applied. The final list of 20 chemicals was: Class 3--6-aminonicotinamide, 5-bromo- 2'-deoxyuridine, hydroxyurea, methylmercury chloride, methotrexate, all-trans-retinoic acid; Class 2--boric acid, dimethadione, lithium chloride, methoxyacetic acid, valproic acid (VPA), 2-propyl-4-pentynoic acid (4-yn-VPA), salicylic acid sodium salt; and Class 1--acrylamide, D-(+)-camphor, dimethyl phthalate, diphenhydramine hydrochloride, 2-ethyl-4- methylpentanoic acid (isobutyl-ethyl-VPA), Penicillin G sodium salt, saccharin sodium hydrate.