Studies on the role of analogies in science classrooms have tended to focus on analogies that come from the teacher or curriculum, and not the analogies that students generate. Such studies are derivative of an educational system that values content knowledge over scientific creativity, and derivative of a model of teaching in which the teacher's role is to convey content knowledge. This dissertation begins with the contention that science classrooms should encourage scientific thinking and one role of the teacher is to model that behavior and identify and encourage it in her students. One element of scientific thinking is analogy. This dissertation focuses on student-generated analogies in science, and offers a model for understanding these. I provide evidence that generated analogies are assertions of categorization, and the base of an analogy is the constructed prototype of an ad hoc category. Drawing from research on categorization, I argue that generated analogies are based in schemas and cognitive models. This model allows for a clear distinction between analogy and literal similarity; prior to this research analogy has been considered to exist on a spectrum of similarity, differing from literal similarity to the degree that structural relations hold but features do not. I argue for a definition in which generated analogies are an assertion of an unexpected categorization: that is, they are asserted as contradictions to an expected schema.