Specifications for the Design of Problem-Solving Assessments in Science

In Specifications for the Design of Problem-Solving Assessments in Science, CRESST researcher Brenda Sugrue draws on the CRESST performance assessment model to develop a new set of test specifications for science. Sugrue recommends that designers follow a straightforward approach for developing alternative science assessments. “Carry out an analysis of the subject matter content to be assessed,” says Sugrue, “identifying key concepts, principles, and procedures that are embodied in the content.” She adds that much of this analysis already exists in state frameworks or in the national science standards. Either multiple choice, open-ended, or hands-on science tasks can then be created or adapted to measure individual constructs, such as concepts and principles, and the links between concepts and principles. In addition to measuring content-related constructs, Sugrue’s model advocates measuring metacognitive constructs and motivational constructs in the context of the content. This permits more specific identification of the sources of students’ poor performance. Students may perform poorly because of deficiencies in content knowledge, and/or deficiencies in constructs such as planning and monitoring, and/or maladaptive perceptions of self and task. The more specific the diagnosis of the source of poor performance, the more specific can be instructional interventions to improve performance. Sugrue’s model includes specifications for task design, task development, and task scoring, all linked to specific components of problem-solving ability. An upcoming CRESST report will discuss the results of a study designed to evaluate the effectiveness of the model for attributing variance in performance to particular components of problem solving and particular formats for measuring them.