The interactive effect of Gestalt laws of perceptual organisation and task demands on graph comprehension

I describe a series of seven experiments investigating how undergraduate students' comprehension of 2x2 `interaction' bar and line graphs widely used to present data from two-way factorial research designs is affected by both the graph format and the nature of the interaction with them.

The first four experiments investigate how different Gestalt principles of perceptual organization operate in the two graph formats and demonstrate the effects of these principles (both positive and negative) on graph comprehension. In particular, Gestalt principles are shown to hinder significantly students’ comprehension of data presented in line graphs compared to bar graphs and that the patterns of errors displayed by students are systematic. The analysis also informs the development of two modified line graphs, one of which improves data interpretation significantly to the level of the bar graphs.

The final three experiments investigate more deeply how the processes involved in different types of interaction with graphs affect users’ comprehension of the data depicted. In the first four experiments, participants attempted to understand the graphs while thinking aloud. However, a subsequent study (Experiment 5) demonstrated that writing an interpretation produced significantly higher levels of comprehension for line graphs than when thinking aloud. The final two experiments sought to identify the cause of this difference by isolating demands specific to the verbal protocol condition.

The results of this research show that (a) in certain circumstances the Gestalt principles of perceptual organization that operate in different graph formats can significantly affect the interpretation of data depicted in them but that (b) these effects can be attenuated by the nature of the interaction. The implications of this research are that identifying an appropriate method of interaction as well as ensuring appropriate display design ensures that the majority of users will be able to interpret these graphs appropriately and so recommendations can be made for graph use in educational settings.