By Xiaoyu Wang

After reading the article: Theory, Experimentation and Application to the Development of Graphical Methods written by Cleveland and McGill in 1984, I have a deeper understanding on Graphical Perception, how science of statistical graphics are connected with human visual perception and how it is important to today’s data visualization. The article has become the milestone and be the foundation theory providing a guideline for the link between data visuals and the human visual system. Cleveland and Mcgill guided us observing the data from scientific perspectives. In the research paper, Cleveland and Mcgill showed us the basic foundations of how humans perceive the elements of graphs. The figure [1] below illustrates 10 elementary perceptual tasks that people use to extract quantitative information from graphs. As we can see from the figure, humans have a better judgment on dot position than length, direction, angle, area, curvature and volume. This “elementary perceptual tasks” clearly showed how we visually-mentally process graph elements. This is how the authors order the 10 elementary tasks from the most to the least accurate. 1. Position along a common scale 2. Positions along nonaligned scales 3. Length, direction, angle 4. Area 5. Volume, curvature 6. Shading, color

knitr::include_graphics("/Users/Jade/Desktop/Harrisburg University/Second Semester/Data Visualization ANLY 512 2017LateFall/Graphical Perception/Elements.png", dpi = NA)

Figure [1]. Perception of graphical elements (Cleveland & Mcgill, 1984, P532)

This is a very important finding because the it can be used in designing of graphics when presenting data visualization. Based on this foundation, it is not hard to argue dot graphics are better than bar chart; bar charts are better than pie charts; pie charts are better than curve and shades graphics and etc.

Drew Skau and Robert Kosara, in their paper discuss the perception mechanics of pie charts, and how pie charts or donut charts are less efficient than bar plots. Dots are being perceived before length, length are perceived before area. Therefore, dot charts are better at showcasing information than the area or angles in pie charts.

knitr::include_graphics("/Users/Jade/Desktop/Harrisburg University/Second Semester/Data Visualization ANLY 512 2017LateFall/Graphical Perception/pie.png", dpi = NA)

knitr::include_graphics("/Users/Jade/Desktop/Harrisburg University/Second Semester/Data Visualization ANLY 512 2017LateFall/Graphical Perception/dot.png", dpi = NA)

Figure [2]: Dot chart versus Pie chart (Cleveland & McGill, 1984, P547)

As we discussed how length is more effective than area and volume in graphical perception. Therefore, there are a lot of research papers discussing line charts. Line graph is invented by William Playfair in 1786 to help people understand time series data. In the article of Sizing the horizon: the effects of chart size and layering on the graphical perception of time series visualization, J. Heer, N. Kong, and M. Agrawala compared line charts with horizon graphs. They introduce four visualization techniques for multiple time series: simple line graph, small multiples graph, horizon graph, and braided graph. Line graphics are widely used in data presenting and line graphics are more effective for human to perceive for time series data.

Nowadays, the theory developed by Cleveland and McGill is still very important to designers, data scientists and data story tellers. Designers need to know what degree human can accept and perceive their graphs when it comes to effective visualization design and data presenting. In the research paper Crowdsourcing graphical perception: Using mechanical turk to assess visualization design, the authors Jeffrey Heer and Michael Bostock conducted further research and expanded on Cleveland and McGill’s graphical perception theory through experiments on human judgment on dots, length, area and volume in graphs. With its low cost and scalability, crowdsourcing presents an attractive option for evaluating the large design space of visualization; however, it first requires validation. In this paper, they assess the viability of Amazon’s Mechanical Turk as a platform for graphical perception experiments. The authors also conduct new experiments on rectangular area perception (as in tree maps or cartograms) and on chart size and gridline spacing. The authors found out people make higher errors when they are judging graphics in bigger spacing. Effective visualization should present closer information for people to compare but not too close from preventing data and information blur together.

Conclusion:

Data visualization is a process of compiling and analyzing data and telling a visual story with processed data. Only if story tellers know how human brain perceives graphs, the story tellers will tell a good story using graphs. If the story tellers do not understand how people process visuals in brain, their information cannot be digested by audiences. The effective communication using visualization should be easy to perceive by human brain, easy to understand and easy to tell a story.

Citation

  1. Figure [1] [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 1984 experiments by McGill and Cleveland rank how accurately people assess graphic depictions of data
  2. Skau D., Kosara R.: Arcs, Angles, or Areas: Individual Data Encodings in Pie and Donut Charts. Computer Graphics Forum (Proceedings EuroVis) 35, 3 (2016)
  3. Crowdsourcing Graphical Perception: Using Mechanical Turk to Assess Visualization Design Jeffrey Heer, Michael Bostock ACM Human Factors in Computing Systems (CHI), 203-212, 2010
  4. J. Heer, N. Kong, and M. Agrawala. Sizing the horizon: the effects of chart size and layering on the graphical perception of time series visualization. In Proceedings of the ACM CHI 2009 Conference on Human Factors in Computing Systems, pages 1303-1312, 2009