In 1974, when research on teachers’ cognitions became a major focus of inquiry, teaching, like the practice of medicine, was described as a form of clinical problem solving. The concept of clinical problem solving was invoked primarily in terms of its traditional definition: the diagnosis and treatment of dysfunction. Today, after 14 years of studying teachers’ cognitions, educational researchers are beginning to provide models of classroom teaching that reflect a very different, hierarchical kind of clinical problem solving. Empirical models of each kind of clinical problem solving (diagnosis/treatment vs. hierarchical) are described and compared, as a means of evaluating the distance researchers of teachers’ cognitions have traveled. The author suggests that the transition from the initial to the newer model represents a shift in perspective that has effectively moved teaching farther from science and closer to art. The implications of this shift are discussed in terms of how pre- and inservice teachers are evaluated and trained.