Learning to Distinguish Between Representations of Data:

Students often fail to learn crucial distinctions between different representations of data. For instance, many students learning about scatterplots consistently create representations which have the surface features of scatterplots but with informational content more appropriate for discrete bar graphs. Schwartz and Bransford (1998) have found that combining feature-based conceptual instruction with contrasting cases is an effective way to help students make conceptual distinctions. We adapt their approach to the domain of data representation and incorporate it into a cognitive tutoring curriculum. We show that this new curriculum improves learning more than a curriculum where the contrasts are not present.