ABSTRACT
Congress both considers a wide variety of different substantive issues and represents the diverse constituencies of 435 congressional districts and fifty states. If the authors succeed in accounting for individual roll call decisions with a parsimonious model, it follows that considerable constraint operates across issues. The basic implication of the constraint hypothesis is that all issues tend to be mapped onto a fixed ordering or placement of legislators. This fixed ordering can be thought of as the underlying, "basic" or "predictive," dimension. The authors adopted a simple spatial model with probabilistic voting. Assuming this model is a correct model of actual behavior, they developed a method for recovering the positions of legislator and roll call outcomes solely from observed individual roll call decisions. That is, the method is blind to any external information, such as political parties, about the legislators and the roll calls.
