In the paper by Cleveland and McGill, graphical perception is described as the visual decoding encoded on graphs. To explain the role of graphical perception, Cleveland and McGill has identified a set of elementary perceptual tasks that are carried out when people extract quantitative information. Through experiment with elementary perceptual tasks, Cleveland and McGill wants to show that a graphical form that involves elementary perceptual tasks that lead to more accurate judgments than another graphical form will result in better organization and increase the chances of a correct perception of patterns and behavior. The theory could be used to analyze and redesign several much used graph forms. For example, use dot charts and bar charts as replacements for divided bar charts and pie charts, and use framed-rectangle charts as replacements for statistical maps with shading lead to far more accurate judgments.
Following Cleveland and McGill’s study, many researchers have applied experimental methods to test differences across chart types. The studies measure how an individual encoding variable affects the accuracy and response time of estimating values of the underlying data. Foe example, in 1987, the study by Simkin and Hastie predicted that graph types and judgmental types would interact to determine the speed and accuracy of quantitative information extraction. The prediction was confirmed by the finding that a comparison judgment was most accurate when judgment required assessing position along a common scale. In 1991, the study by Spence and Lewandowsky found that pie chart compares favorably when tasks other than direct magnitude estimation re required.
Later in 2010, Heer and Bostock did research and experiment on crowdsourcing graphical perception. With its low cost and scalability, crowdsourcing presents an attractive option for evaluating the large design space of visualizations. Through the experiment, Heer and Bostock assessed the feasibility of using Amazon’s Mechanical Turk to evaluate visualizations and use the methods to gain new insights into visualization design. Replication results of prior laboratory studies using crowdsourcing techniques suggest that crowdsourcing is viable for testing graphical perception. Also, the study shows that qualification tasks and verifiable questions help ensure high quality responses. Such study results will generalize to a variety of crowdsourcing studies.
Reference: 1. Cleveland, McGill - 1984 - Graphical Perception Theory, Experimentation, and Application to the Development of Graphical Methods 2. D. Simkin and R. Hastie. An information-processing analysis of graph perception. J. Am. Stat. Assoc., 82:454–465, Jun 1987. 3. I. Spence and S. Lewandowsky. Displaying proportions and percentages. Applied Cog. Psych., 5:61–77, 1991. 4. Heer, Bostock - 2010 - Crowdsourcing Graphical Perception Using Mechanical Turk to Assess Visualization Design