This study reports on what naive, novice, and expert designers do and learn when investigating simple mechanical devices and then planning their redesign. Participating in the study were 32 high school and adult subjects who did two investigate-and-redesign (I&R) tasks. Same gender pairs of subjects with similar design experiences explored, analyzed, and evaluated different brands of a device, designed experiments to compare them, and then proposed their redesign. Each two-hour session was videotaped, and portions were analyzed using methods adapted from protocol analysis techniques. Results suggest that when naive designers do I&R tasks, their learning is highly contextualized and device-specific. Naive designers made few connections from their work to key science ideas, and instead used mechanical advantage preconceptions that they did not spontaneously redress during the I&R sequence. Experts made connections to concepts and cases, inferred key design decisions, and sought “critical design problems” for the devices studied. All groups used strategies involving analysis more than those involving synthesis or evaluation. Notably, during conceptual design, opportunities for using science, present especially when subjects analyze design ideas, went underutilized by nonexpert designers. Scaffolded questions are needed to focus the learning of science embedded in design-oriented activities. All findings reported are tentative, given the limited number of cases included in this study. © 2001 John Wiley & Sons, Inc. J Res Sci Teach 38: 791–820, 2001