Graph comprehension: Modelling and related processes

The usefulness of graphs for visualisation has offen been reported when readers have to learn from texts. Educators have emphasised how important it is that students leam how to interpret graphs and diagrams as well as to produce them from texts. Graphs and diagrams help us thinking and solving problems.In the proposed project we will investigate the comprehension of hierarchical graphs. Everyday examples of hierarchical graphs inelude, among others, the structure of computer file systems, family trees, preference trees etc.The objeetives of the project are to identify psychological processes underlying comprehension, and to test a model of graph comprehension. We will do so by using state-of-the-art eye-tracking techniques.The results of the projeet will provide design criteria for how to draw hierarchical graphs to optimise comprehension by humans.

These days, methods of graphical visualisation are commonly applied in various fields like teaching, management and the design of computer user interfaces. Because there are many different ways to visualise the same information in a graph, psychological investigation of how different graphs are comprehended and of what constitutes a comprehensible graph is central to their continued utility. Suppose a user who is confronted with a graphical file browser has to verify the question "Is folder a contained in folder b?". How do readers understand the respective graphic representation? What processing might underlie graph comprehension and how can they best be described in a coherent and simple model? One subtask of graph comprehension might be to identify and locate the labelled nodes in the graph that represent the objects (folders) mentioned in the question: the target nodes. The other subtask might involve relating the nodes to each other using information about their location and inter- connecting lines. To test these and further assumptions we used the full potential of modern eye-tracking techniques in sophisticated experimental designs. We tested the model of graph comprehension proposed by Körner (2004). It postulates three stages - two initial search stages and a third stage during which the graphical reasoning is carried out. Information about interconnecting lines is ignored during the first two stages and only relevant in the third one. The results of our first set of experiments (Körner, submitted) provided clear support for this three stage model. In the second set of experiments (Körner, Höfler & Tröbinger, 2006; Körner, Höfler & Tröbinger, in preparation; Körner, 2006a, 2006b, 2006c) we conducted gaze-contingent display changes which changed the relationship between the two target objects. The three-stage model stood even up to this hard empirical test. Based on more specific results of our experiments, we modified the three-stage model. The resulting "soft" three-stage model suggests that some reasoning may be carried out during the search stages. The main part of the process of reasoning and decision making, however, occurs in a dedicated third stage. Due to the importance of visual search in graph comprehension we further investigated the role of short-term memory in repeated visual search in a related set of experiments. With simple search tasks (Is object a present or absent?) we could segregate effects of distance to previous fixations from actual memory effects (Höfler, 2006; Körner & Gilchrist, 2006; Höfler & Körner, in preparation).