Contemporary models of multifactorial inheritance are described and justified from the perspective of their intended use in genetic epidemiology and their developmental sequence. Substantial empirical data and statistical theory support the practical adequacy of the assumptions of path analysis for most multifactorial traits that show vertical inheritance. The choice of scale for quantitative traits must be considered on an individual basis. From both biological and statistical perspectives, transformations of scale may be more appropriate for analysis than the measurements are themselves. Recent criticism of contemporary models and computational procedures is based on a caricature of path analysis rather than on the method as it is actually practiced. The utility of path analysis is primarily limited by an investigator's biological insight and analytical skill, not by the method's assumptions. Model-free descriptive statistics are inherently inadequate to characterize the stable and autonomous features of the underlying mechanisms that generate observable variation in multifactorial traits. In contrast, path analysis has led to remarkably stable estimates of structural parameters for a wide variety of important biological traits. While exploratory methods can be useful for preliminary data inspection, they cannot substitute for formal tests of hypotheses based on explicit, falsifiable models.