The thrust of the majority of chapters in this volume is to demonstrate the advantages of regression-based linear causal models for theory testing and construction in the social sciences. As has already been discussed, such approaches to making causal inferences from nonexperimental data place a number of constraints on the analyst. In particular, he must make explicit most of the assumptions underlying both his model and analysis operations. Among other things, he must close the theoretical model and make assumptions about the influences of outside variables, distinguish between measured and unmeasured variables, and specify the relations between theoretical constructs and indicators. All such constraints seem likely to improve the quality of nonexperimental research and enhance the possibilities of cumulation in the social sciences.