As Ehrenberg stated in 1975, the purpose of a graph is not to convey the numbers with decimal accuracy. If that were the goal, tables would be better. The power of graph is its ability to enable one to take in quantitative information, organize it, and see patterns and structure not readily revealed by other means of studying data. And this is where graphical perception becomes critical to help read and analyze the graphs appropriately.

Graphical perception is the ability of the viewer to understand the data encoded in graphs. Understanding how the viewer will decode the information helps designers to optimize their visualizations. A formal definition could be defined as interaction between statistical graphics and our visual system to create human visual perception. Graphs are developed using quantitative data and qualitative data and this data is decoded visually when a person perceives the respective graph. Human perception plays a key role for effective decoding of the graphs. For example, one of the key findings from Cleveland and McGill’s paper was that some elementary perceptual tasks like length are better and more accurate than area or volume in graph analysis. These elementary tasks are showed in Figure 1.

Figure 1: The Elementary Perceptual Tasks

Figure 1: The Elementary Perceptual Tasks

Also, Zubiaga and Namee described two studies to understand graphical perception based on Heer and Bostocks research using crowdsourcing platform. Their initial research experiment showed, majority of their users achieved 97 percent accuracy with histograms, in comparison to other charts. In subsequent experiments, their results showed, histograms can be more reliant than density trees for data visualization. To elaborate this further, let us take another example. For instance, nonaligned scaled graphs are better for inferences compared to length, which is turn better than area which is better than volume visualizations. The comparison of nonaligned scales and length is showed in Figure 2.

Figure 2: Bar Charts and Framed Rectangles

Figure 2: Bar Charts and Framed Rectangles

In this figure, the top chart shows two bars with equal widths and unequal heights. If bar height was an important aspect, the elementary perceptual task of judging the length would be hard in determining which bar is longer. In the bottom chart, the same bars are shown which are surrounded by frames of equal size and construction. Each symbol is a graph with scale and portrays a number. The elementary perceptual task now is to judge position along nonaligned scales. Now, it is easy to see that the bar on right represents a larger quantity than the left.

Through these experiments and analysis, Cleveland and McGill proved that dot charts are superior to bar charts and framed rectangle charts are better than statistical map with shading but we see very little use of dot charts or framed rectangle charts in industry and media. This seems to indicate that either society’s understanding of the role of perception in visualization is still very primitive and/or the use of visualization techniques is very sticky. Understanding the role of perception is a constant task.

In conclusion, it is important to understand the role of perception in visualization to make better visualizations as well as keep evolving our understanding of the role of perception. Data visualization is also very important to understand data analysis results. Data analysis can be interpreted differently by different types of audiences, based on their expertise. That is why perception study becomes critical. All the different researches and experiments show importance of graphical perception in data visualization. This helps us build graphical representations that can enable similar data interpretation for all type of audiences.

References

  1. Heer, J., & Bostock, M. (2010, April). Crowdsourcing graphical perception: using mechanical turk to assess visualization design. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (pp. 203-212). ACM.

  2. Cleveland, W. S., & McGill, R. (1984). Graphical perception: Theory, experimentation, and application to the development of graphical methods. Journal of the American statistical association, 79(387), 531-554.

  3. Healey, C. G. (2007). Perception in visualization. Retrieved February, 10, 2008.

  4. Healy C. (1999). Perception in Visualization. Retrieved August 7, 2017.

  5. Zubiaga, A., & MacNamee, B. (2016). Graphical perception of value distributions: an evaluation of non-expert viewers’ data literacy. Journal of Community Informatics, 12(3).