Searching for Extreme Portions in Distributions: A Comparison of Pie and Bar Charts

Abstract

Aggregated data visualizations are often used by collaborative teams to gain a common understanding of a complex situations and issues. Pie and bar charts are both widely used for visualizing distributions. The study of pie versus bar charts has a long history and the results are seemingly inconclusive. Many report authors prefer pie charts while visualization theory often argues for bar graphs. Most of the studies that conclude in favor of pie charts have focused on how well they facilitate the identification of parts to the whole. This study set out to collect empirical evidence on which chart type that most rapidly and less erroneously facilitate the identification of extreme parts such as the minimum, or the maximum, when the distributions are similar, yet not identical. The results show that minimum values are identified in shorter time with bar charts compared to pie charts. Moreover, the extreme values are identified with fewer errors with bar charts compared to pie charts. One implication of this study is that bar charts are recommended in visualization situations where important decisions depend on rapidly identifying extreme values.