A critical analysis of the literature base regarding the reproductive and developmental toxicity of arsenic compounds, with emphasis on inorganic arsenicals, was conducted. The analysis was stimulated by the great number of papers that have purported to have shown an association between exposure of pregnant laboratory animals to arsenic compounds and the occurrence of offspring with cranial neural tube defects, particularly exencephaly. For the most part, the literature reports of arsenic developmental toxicity in experimental animals are inadequate for human risk assessment purposes. Despite the shortcomings of the experimental database, several conclusions are readily apparent when the animal studies are viewed collectively. First, cranial neural tube defects are induced in rodents only when arsenic exposure has occurred early in gestation (on Days 7 [hamster, mouse], 8 [mouse], or 9 [rat]). Second, arsenic exposures that cause cranial neural tube defects are single doses that are so high as to be lethal (or nearly so) to the pregnant animal. Third, the effective routes of exposure are by injection directly into the venous system or the peritoneal cavity; even massive oral exposures do not cause increases in the incidence of total gross malformations. Fourth, repetition of similar study designs employing exaggerated parenteral doses is the source of the large number of papers reporting neural tube defects associated with prenatal arsenic exposure. Fifth, in five repeated dose studies carried out following EPA Guidelines for assessing developmental toxicity, arsenic was not teratogenic in rats (AsIII, 101 micromol/kg/d, oral gavage; 101 micromol/m3, inhalation), mice (AsV, 338 micromol/kg/d, oral gavage; est. 402 micromol/kg/d, diet), or rabbits (AsV, 21 micromol/kg/d, oral gavage). Data regarding arsenic exposure and adverse outcomes of pregnancy in humans are limited to several ecologic epidemiology studies of drinking water, airborne dusts, and smelter environs. These studies failed to (1) obtain accurate measurements of maternal exposure during the critical period of organogenesis and (2) control for recognized confounders. The lone study that examined maternal arsenic exposure during pregnancy and the presence of neural tube defects in progeny failed to confirm a relationship between the two. It is concluded that under environmentally relevant exposure scenarios (e.g., 100 ppm in soil), inorganic arsenic is unlikely to pose a risk to pregnant women and their offspring.