The Effect of Configural Graphs on Concurrent and Retrospective Performance
Following up on a previous study showing the performance on integrated tasks for non-configural graphs to be superior to that for configural graphs if the memory for the graph is tested (retrospective or memory-based conditions), this paper further contrasts retrospective and concurrent (display-based) task performance. This was done by experimentally investigating the effect of various configural and non-configural static graphs on integrated task performance (requiring the consideration of lower-level graph information as well as higher-level graph information), using both retrospective and concurrent conditions. Subjects were asked to answer a question about each graph, which was phrased in terms of the domain of the data and which could not be easily anticipated. Graphs also differed in the amount of fit between graph structure and data structure (data-graph compatibility). The results confirmed the expectation that the reversal effect (inferior performance for configural graphs) is only found under memory-based conditions. Both display-based and memory-based performance were better for the configural graphs with high data-graph compatibility, although only significantly so for display-based search time. The two separable types of graphs could only be compared with respect to the amount of time needed to memorize the graphs: longer times were found for the graph type with low data-graph compatibility. However, the latter effect may also be due to a difference in data structure complexity, as this factor was confounded with data-graph compatibility in the two separable graph types. Although more research is needed to disambiguate some of the present results and to make other and better comparisons, the results of this study still show the importance of structural and semantic factors in determining the effectiveness of configurality in statistical graphs.