A traditional concern with drugs administered during pregnancy has been teratogenicity or the production of gross structural malformations. Beginning in the 1970s, it became increasingly evident that the issue of drug safety and risk assessment went far beyond structural defects. During the 1980s, the newly emerged research specialty of "developmental toxicology" came to encompass a wide range of adverse toxic outcomes that include not only birth defects but also neurobehavioral and other functional effects as well. Substances of use and abuse--the opiates, cocaine, and cannabis--have come to exemplify a diverse group of compounds that produce a broad spectrum of developmental outcomes. Unlike alcohol, neither the use of heroin nor methadone during pregnancy is associated with an increased risk of birth defects but both produce a neonatal abstinence syndrome that can persist for as long as 6 months; follow-up to preschool years suggests possible risk of attention deficit and problems of fine motor coordination. Methodologic weaknesses of opiate animal models, especially with respect of appropriate dosing schedules, have hampered meaningful extrapolation of these studies to human risk assessment. Given the renewed interest in methadone maintenance as an important therapeutic intervention to reduce exposure to the human immunodeficiency virus, better designed animal studies are needed urgently to assess developmental risk, but these must incorporate techniques that better model human pharmacokinetics. Animal models of early cocaine exposure, driven by human reports of serious risk to the fetus and newborn, have found reproductive hazard, risk of neurobehavioral effects as well as altered CNS function. Whereas animal studies need to explore routes of administration other than sc and ig, particularly the volatilized form of cocaine, to date it appears that the processes of somatic growth and morphogenesis in rodents are not as sensitive to cocaine as is the functional development of the CNS. Finally, animal studies of cannabis have taught us some major methodologic and interpretive lessons for the continuing development and refinement of animal models of drugs of abuse. Of particular importance is that poorly controlled experiments that do not adequately consider the confounding influences of maternal toxicity, both prenatally and postnatally, are likely to yield a high rate of false-positive results. This is well illustrated by those studies of cannabis that antedated the current concern for pair-feeding and surrogate fostering. Nearly all of the studies that failed to include nutritional and fostering controls found neurobehavioral effects that included changes in activity as well as impairments in learning and memory.(ABSTRACT TRUNCATED AT 400 WORDS)