Evaluating Models of Visual Comprehension

This paper evaluates the predictions of two existing models of graph comprehension: BOZ (Casner, 1990) and UCIE (Lohse, 1993). Each model was implemented in Java and was used to make predictions about the relative efficiency of different graphical presentations of numerical data for use in different tasks. These predictions were then compared with the results of human subjects performing the same tasks using the same presentations. The results of the human study do not correspond to the predictions of either model. In particular, while both models predict that the tabular presentations would have the worst results, in practice the tables actually proved to be the best presentation type. A possible explanation for this result is that the models capture optimal, expert performance, while the subjects used less efficient techniques.