A common approach for introducing students to a new science concept is to present them with multiple cases of the phenomenon and ask them to explore. The expectation is that students will naturally take advantage of the multiple cases to support their learning and seek an underlying principle for the phenomenon. However, the success of such tasks depends not only on the structure of the cases, but also the task that students receive for working with the examples. Two studies used contrasting cases in the context of teaching middle-school students about projectile motion. Using a simulation and the same set of cases for all students, students completed a traditional "compare and contrast" approach, or an instructional method called "inventing," where students try to produce a single general explanation. The results show that inventing led to superior learning. Examination of student worksheets revealed that the "compare and contrast" instruction led students to focus mostly on the level of discrete, surface features of the phenomenon. Rather than trying to account for the variation across cases, students simply noticed each instance of it. In contrast, the inventing task led students to consider how the variations across the cases were related. As a result, "invent" students were more likely to search for and find the unifying functional relation. Driving towards an overall explanation is a fundamental tenet of science, and therefore, it is worthwhile to teach students to do the same.