We consider the problem of automatic visualizing information contained in a database and propose a framework that addresses the issues involved. Our approach does not depend on the particular data model used, the size of the instance set, or the nature of the data being represented. The aim of the framework is to produce a correct, complete, and possibly highly effective visual representation. Such properties of the visual representation are made possible by the consistency of a set of logic rules contained in a knowledge base. These rules express facts and relationships about the visual domain, the data domain, the mapping between both domains, and the perceptual domain.