Neurobehavioral effects of prenatal exposure to polychlorinated biphenyls (PCBs) have been investigated most extensively in two prospective longitudinal studies--one in North Carolina, the other in Michigan. Based on the Webb-McCall methodology available when these studies were initiated, a majority of the cord serum PCB concentrations in both cohorts were below laboratory detection limits. Prenatal exposure was, therefore, assessed in North Carolina in terms of maternal body burden and, in a recent 11-year follow-up in Michigan, by averaging detectable PCB values from cord and maternal serum and maternal milk samples. The new composite prenatal exposure measure used in Michigan at 11 years was more valid in relation to maternal contaminated fish consumption and more sensitive in detecting 4-year cognitive deficits than the cord serum measure used in earlier phases of the study. During infancy, the North Carolina study found poorer gross motor function in relation to prenatal PCB exposure; the Michigan study found poorer infant visual recognition memory, an effect confirmed in a sample of Taiwanese infants exposed to PCBs and polychlorinated dibenzofurans (PCDFs) due to maternal ingestion of highly contaminated rice oil. Cognitive deficits found at 4 years in Michigan were not seen in North Carolina, possibly due to a different pattern of exposure or a different congener mix. An examination of the incidence of functionally-significant impairment (defined as > 1 SD below the sample mean) showed that the more highly exposed Michigan children were twice as likely to perform poorly than others in the sample and that there are marked individual differences in vulnerability to this exposure. Thus, a small reduction in a mean IQ score may reflect little effect on a majority of the sample, accompanied by a substantial deficit in a small number of more vulnerable children.