Pie Chart Research
This post is an aside to my main post on pie charts.
In my quest to understand this controversy I ended up reading a number of research papers on the topic. Here in chronological order are a few notes from relevant ones:
1984: I covered Cleveland and McGill’s seminal work on graphical perception1 in my main article on pie charts. In the paper, they investigated a number of representations for accuracy in various tasks. Overall they found pie charts and stacked bar charts to be roughly equivalent in performance, with neither being particularly accurate. Instead they recommended the use of a bar chart or dot chart to encode the same information.
1991: Spence and Lewandowsky2 found bar charts and pies to be mostly equivalent. Bar charts were better at estimating direct magnitude and pies were better for comparisons of combinations of proportions (is segment A bigger than segments B + C?).
1998: Hollands and Spence3 looked at performance judging proportions of a whole and found pie charts the most accurate and equally speedy as stacked bars. Bar charts were least accurate and their speed decreased with the number of bars. They suggest that the poor bar chart performance is due to the users needing to mentally sum the bar lengths to determine the overall length, so perhaps adding an explicit 100% marker to the chart might avoid this issue.
2006: There’s a lot of interesting research looking at the use of visualizations to convey medical health risk information. Ancker et al4 compared stacked bars, pair of numbers, icon arrangements and pie charts. They found pie charts and random icons to be the slowest and least accurate when judging the larger of two quantities. Pie charts were superior when mental summation of slices was required.
2008: Schonlau et al5 compared bar charts, pie charts and tables, asking participants to identify largest/smallest categories and estimate the perecentage of the largest. They found that tables were the most accurate and bars pies roughly equally less accurate. They also found that adding a gratuitous 3rd dimension negatively affected performance for pie charts but not for bar charts.
2010: Heer and Bostock6 retested Cleveland and McGill’s findings asking participants to identify the smaller of two values and estimate the percentage the smaller was of the larger. Again they found that bar charts (position) performed better than both pies (angle) and stacked bars (length), which all performed better than sized circles or rectangles (area).
2012: Schonlau revisited7 the issue with Peters in 2012. Graphs were found better for estimating differences and tables better for equality and sums. Stacked bar charts are to be avoided (“Comprehension… was worse for all the but the easiest tasks”). Pie charts got the backhanded compliment of “pie charts never assisted comprehension, they impaired comprehension on fewer than half of the tasks.” They were as good as tables for judging which of two categories was larger and worse for estimating percentage or difference of percentage. Of note, their paper includes a concise summary of some previous research (some mentioned here).
2016: Recently Skau and Kosara looked at variations of pie charts to see how design choices affect their performance. They found both pie and doughnut charts to perform fine and listed a few design guidelines (e.g. avoid exploded pie charts if segment summation is required; keep the radius consistent between segments).
- William S. Cleveland and Robert McGill, Graphical Perception: Theory, Experimentation, and Application to the Development of Graphical Methods, Journal of the American Statistical Association, Vol. 79, No. 387 (Sep., 1984), pp. 531-554 ↩
- Spence, Ian, and Stephan Lewandowsky. “Displaying proportions and percentages.” Applied Cognitive Psychology 5.1 (1991): 61-77. ↩
- Hollands, J. G., and Ian Spence. “Judging proportion with graphs: The summation model.” Applied Cognitive Psychology 12.2 (1998): 173-190. ↩
- Design features of graphs in health risk communication: a systematic review. Ancker JS, Senathirajah Y, Kukafka R, Starren JB, J Am Med Inform Assoc. 2006 Nov-Dec; 13(6):608-18. ↩
- Schonlau, Matthias, and Ellen Peters. “1RAND Corporation, 4570 Fifth Avenue, Suite 600, Pittsburgh, PA 15213; email.” (2008). ↩
- Crowdsourcing Graphical Perception: Using Mechanical Turk to Assess Visualization Design, Jeffrey Heer and Michael Bostock, CHI 2010, April 10–15, 2010, Atlanta, Georgia, USA. ↩
- Schonlau M, Peters E. (2012).Comprehension of Graphs and Tables Depends on the Task: Empirical Evidence from two web-based studies. Statistics, Politics and Policy. 3(2); Article 5. ↩