There exists considerable research on the effects of prenatal maternal stress on offspring. Animal studies, using random assignment to experimental and control groups, demonstrate the noxious effects of prenatal maternal stress on physical, behavioural and cognitive development. The generalizability of these results to humans is problematic given that cognitive attributions moderate reactions to stressors. In humans, researchers have relied upon maternal anxiety or exposure to life events as proxies for the stressors used with animals. Yet, the associations between maternal anxiety or potentially non-independent life events and problems in infants are confounded by genetic transmission of temperament from mother to child. We summarize the literature on prenatal maternal stress and infant cognitive development, leading to the conclusion that the human literature lacks the ability to separate the effects of the objective exposure to a stressor and the mother's subjective reaction. We then describe our prospective Project Ice Storm in which we are following 150 children who were exposed in utero to a natural disaster. We demonstrate significant effects of the objective severity of exposure on cognitive and language development at age two years with important moderating effects of the timing during pregnancy. The implications of our findings are discussed.