Data from developmental toxicology present several problems from the standpoint of statistical testing. The concurrence of the litter effect (the correlation of responses of littermates), the large number of endpoints to be evaluated, and the fact that most developmental toxicology endpoints are quantal raises the need for unique statistical solutions for hypothesis testing. The paper reviews methods that have been developed to accomodate these problems, particularly those having to do with the litter effect, compares them with respect to power, and speculates on the structure of models that may incorporate more biological information about the nature of the litter effect.