While there has been general agreement among modern philosophers of science that a purely a priori method is inappropriate to the task of establishing a theory of science, there has, unfortunately, been little comparable agreement regarding the method that is appropriate. I try to lay the foundations for such agreement. I first set out reasons for a purely empirical method for establishing a theory of science, and defend such a method against charges raised by Giere. I then develop some very (...) basic criteria for the evaluation of alternative empirical methods for establishing a theory of science, and use these criteria to evaluate the two such methods that have dominated philosophic discussion in recent years--those of Lakatos and Laudan. I end by defending a revised version of Lakatos' proposal. (shrink)