Research on the effectiveness of graphical displays for information acquisition and retention lacks a system for classifying graph information and generating test items to assess learning. The purpose of this study was to validate a system based on two types of information and three types of informational units. Results of an analysis of variance indicated differences in learning predictable from the classification system; however, a multitrait-multimethod matrix analysis failed to provide evidence of trait validity for the system's informational constructs. In light of these results, a graph information processing strategy was proposed in which subjects utilize data point information. (Author)