Alcohol is a teratogenic drug and the effects appear to be grossly dose related. The severest effects are observable clinically as the fetal alcohol syndrome and are associated with heavy prenatal alcohol exposure and a history of chronic maternal abuse of alcohol. Hypotheses for subtler behavioral effects associated with lower levels of exposure can be generated from observation of the behavioral effects in FAS. Behavioral effects associated with various levels of prenatal alcohol exposure in humans include poor sucking and poor habituation in the newborn, poorer mental and motor development in infancy, and attentional and reaction time effects at four and seven years of age. Human behavioral teratology studies are necessarily complex due to the large number of covariates that affect behavior, modify the effects of teratogens, and influence interpretation. Challenging problems exist in assessing exposure, outcomes, and covariates. Common to assessment of all these classes of variables are the multiplicity of measurement and the indirect nature of the measurement. Answering specific questions about timing and dose effects demands careful statistical modeling procedures and large, complex data bases. Large data bases and indirect measurement problems suggest factor analytic extensions of current regression methodology, which we have proposed in this paper.