In this article, the authors first indicate the range of purposes and the variety of settings in which design experiments have been conducted and then delineate five crosscutting features that collectively differentiate design experiments from other methodologies. Design experiments have both a pragmatic bent—“engineering” particular forms of learning—and a theoretical orientation—developing domain-specific theories by systematically studying those forms of learning and the means of supporting them. The authors clarify what is involved in preparing for and carrying out a design experiment, and in conducting a retrospective analysis of the extensive, longitudinal data sets generated during an experiment. Logistical issues, issues of measure, the importance of working through the data systematically, and the need to be explicit about the criteria for making inferences are discussed.

In this study of the development of scientific reasoning, 10 5th-6th-grade children (5 boys and 5 girls) and 10 noncollege adults conducted experiments over 6 half-hour sessions to explore the causal structure of 2 physical science domains. Feedback in these systems, though relevant to discriminating among hypotheses, was noisy as a result of varying effect sizes and measurement error. After 2 hr on each task, both age groups demonstrated changes in their understanding of the content and in their strategies for generating and interpreting evidence. In general, the adults outperformed the children. Neither valid strategies nor correct beliefs alone was sufficient to guarantee success, suggesting that regarding experimentation either as domain-general induction or as domain-specific learning may oversimplify its complexity.

No abstract

This study investigates the hypothesis that when children are engaged in science experiments, the goal of which is to understand relations among causes and effects, they often use the engineering model of experimentation, characterized by the more familiar goal of manipulating variables to produce a desired outcome. Sixteen fifth-and sixth-graders worked on two experimentation problems consistent with the engineering and science models, respectively. The context in which these problems were framed was also varied, to encourage adoption of either an engineering or science model. Over six 40-min sessions, the group achieved significant increases in the percentages of inferences about variables that were both correct and valid. Improvement was greatest for those who began with the engineering problem and then went on to the science problem. The science model was associated with broader exploration, more selectiveness about evidence interpreted, and greater attention to establishing that some variables are not causal. The findings suggest that research on scientific inquiry processes should attend not only to the science content students are reasoning about, but also to their beliefs about the goals of inquiry.

No abstract