Lessons were designed to deal with students' alternative conceptions in three areas of mechanics: static normal forces, frictional forces, and Newton's third law for moving objects. Instructional techniques such as class discussions of the validity of an analogy between a target problem and an intuitive anchoring example, and forming a structured chain of intermediate bridging analogies were used. There were large differences in pre-posttest gains in favor of the experimental group. In formulating a model of learning processes that can explain these results, it is argued that (a) the lessons have a more complex structure than a simple model of analogy use; (b) rational methods using analogy and other plausible reasoning processes that are neither proof based nor directly empirical can play a very important role in science instruction; (c) much more effort than is usually allocated should be focused on helping students to make sense of an analogy; and (d) researchers and curriculum developers should be focusing at least as much attention on students' useful prior knowledge as they are on students' alternative conceptions.