Emergent Features in Bar Graphs

Abstract

The first purpose of the present study was to find out to what extent bar graphs may contain emergent features that can facilitate performance. In an experiment bar graphs with three different types of emergent feature were presented to each of 39 subjects who were asked to compare the sum of half of the values with that of the remaining values. The emergent features merely resulted from the relative size and position of the bars in relation to the subjects' task. A fourth type of bar graph did not contain any of these emergent features and served as a control (baseline). A second purpose of the study was to investigate the effect on performance of training graph readers in the use of these emergent features. Results of the experiment in which 50 bar graphs of the four above mentioned types were presented to the subjects showed that there were significant differences between the four graph types in both reaction time and accuracy. However, the effects for reaction time were largely in opposite directions from those for accuracy. Further analysis of this phenomenon showed that a composite measure, obtained by summing reaction time and accuracy, significantly discriminates between the four graph types. Apparently, subjects responded quickly but inaccurately on some graphs, whereas they responded slowly but accurately on other graphs. Half of the subjects received more extensive training on the use of the emergent features. This additional training did not affect performance on any of the graph types, however. Implications of these results for the theory of graphical perception and the practice of graph design are discussed.