Sources of bias were examined in a study of neurobehavioral effects of intrauterine exposure to methyl mercury in the Faroe Islands. The cohort of 1000 children was born during a 21-month period and did not differ from Faroese births in general as regards major obstetrical parameters. However, high mercury concentrations in the cord blood was associated with increased birth weight, presumably because other constituents of marine fish may cause a prolongation of the gestation period. Thus, children with high mercury exposures were somewhat protected against low birth weight and its associated neurobehavioral risks. Less than 25% of the women indicated occasional alcohol drinking during pregnancy, thus suggesting a limited fetal exposure to this neurobehavioral risk factor. However, maternal alcohol drinking caused a decrease in mercury concentrations in cord blood, probably because of a toxicokinetic interaction between ethanol and mercury. Any alcohol-related effect on neurobehavioral development would then be associated with lower levels of mercury exposures. The effects of these confounders would tend to bias the results of the study toward the null hypothesis.