Development of problem solving performance and structural knowledge in physics problem solvers.

摘要

Prior research indicates that experts not only possess more knowledge than novices, but that their knowledge is organized differently as well. Expert knowledge is organized as schemas which allow domain problems to be categorized and solved efficiently. The goal of this dissertation was to examine how various training manipulations affect the development of schemas. Presenting multiple training problems allows the comparison and/or contrasting of problems, whereas presenting only a single training problem does not. Further, presenting training problems that are solved similarly (i.e., same-type problems) in a blocked fashion encourages comparison of same-type problems, whereas presenting training problems that are solved differently (i.e., different-type problems) in a mixed fashion encourages contrasting of different-type problems. Extant theories of schema acquisition make different predictions regarding how these training manipulations may affect schema acquisition. Analogical induction theories propose that comparing problems that are solved similarly allows schema induction. Others, however, suggest that contrasting problems that are solved differently may aid schema acquisition by highlighting the problem features which allow same-type problems to be solved similarly. Explanation based learning theories propose that intra-problem processing of a single problem is sufficient for schema acquisition. In order to test these predictions, single or multiple physics problems were presented in either blocked or mixed order to physics-naive students during training. Later, schema acquisition was assessed with both near and far transfer problems, as well as with a structural knowledge measure that is assessed independently of problem solving performance. Four major findings were obtained: (1) Single, as opposed to multiple, training problems led to better performance on near, but not far, transfer problems. (2) Presenting training problems blocked by type, as opposed to mixed, led to worse performance on all transfer problems, but only if multiple different training problems were used. (3) Findings 1 and 2 above held only when training consisted of problems to be solved rather than solved examples to be studied. (4) None of the training manipulations had an effect on structural knowledge.;These findings extend both explanation based learning and analogical induction theories. Implications for physics instruction are discussed.