Numerous studies have suggested that single-day intraperitoneal (IP) injection of inorganic arsenic results in failure of neural tube closure and other malformations in rats, hamsters, and mice. Most of these studies involved treatment of limited numbers of animals with maternally toxic doses of arsenic (generally As V ), without defining a dose-response relationship. In the present Good Laboratory Practicecompliant study, sodium arsenate (As V ) was administered IP and arsenic trioxide (As III ) was administered either IP or orally (by gavage) on gestational day 9 to groups of 25 mated Crl:CD(SD)BR rats. Only at dose levels that caused severe maternal toxicity, including lethality, did IP injection of arsenic trioxide produce neural tube and ocular defects; oral administration of higher doses of arsenic trioxide caused some maternal deaths but no treatment-related fetal malformations. In contrast, IP injection of similar amounts of sodium arsenate (based on the molar amount of arsenic) caused mild maternal toxicity but a large increase in malformations, including neural tube, eye, and jaw defects. In summary, neural tube and craniofacial defects were observed after IP injection of both As V and As III ; however, no increase in malformations was seen following oral administration of As III , even at maternally lethal doses. These results demonstrate that the frequently cited association between prenatal exposure to inorganic arsenic and malformations in laboratory animals is dependent on a route of administration that is not appropriate for human risk assessment.