In the Graphical Perception process, people’s visual system decoding information encoded on graphs. This is considered as a Psychophysics behavior. It is important because it helps human being find patterns of data and use analysis result to create models. Models can then help with human being’s life by increasing production, saving energy, increasing efficiency and so on so forth. No learn, no grow. Perception from visualization can always help people learn.
Cleveland and McGill proposed a whole process of graphical perception. Their approach was first, to hypothesize a set of elementary perceptual tasks that people employ to extract quantitative information from visualization. Second, they hypothesized a hierarchy of the perceptual tasks based on their efficacy. Third, they evaluated the quality of their theoretical constructs through empirical tests.
“The value of identifying basic elements and their interactions is that we thus develop a framework to organize knowledge and predict behavior.” Cleveland and McGill said this in their article to explain the first step. This is a perfect summary of the reaction when human being see a statistic graph. We try to find the secret in it, then analyze it, and eventually find the relationship between different variables. The author identified primary elementary visual task of each different common charts. Also identified other variables which contribute to the the final perception.
But the accuracy human being can extract quantitative information from different elementary perceptual tasks are different. They think length judgments are more accurate than area judgments, which is more accurate than volume judgments. This is based on a lot of real experimental.
They ran two experiments to check their hypothesis, one for bar charts, the other for pie charts. The result of data analysis approved their hypothesis.
When applying their theory to graph design, Cleveland and McGill suggested people to use elementary perceptual tasks as high in the hierarchy as possible. The authors prefer dot charts instead of bar charts and pie charts. They think human being is easier to identify pattern between data from dot charts. And they also recommended to replace bar chart with a grouped bar chart. But they prefer grouped dot chart more. I like the sample they gave in Figure 26 and 27. Its not easy to identify the vertical distance difference by human vision. So having a visualization of the distance trend is very helpful.
Heer Jeffrey and Michael Bostock recommended Crowdsourcing for gathering graphical perception. They consider this method can reduce both cost and time. But they also identified the concerns of crowdsourcing, which is lack of control. They did a analysis on Mechanical Turk and found crowdsourced graphical perception studies can be viable.
I also read two additional studies on perception and data visualization. One is Displaying proportions and percentages wrote by I. Spence and S. Lewandowsky. They found that comparisons among multiple segments take longer and have lower accuracy. Pie charts fared the worst except when multiple segments had to be compared. Tables were found to be inferior to everything except for communicating absolute values, despite what Tufte advises. The other one was wrote by Barbara Tversky and Diane Schiano in 1989 called Perceptual and conceptual factors in distortion in memory for graphs and maps. They found that when participants encountered the diagrams introduced as charts, they remembered the lines closer to the imaginary 45-degree line. Further, when the same line was presented to them as a map, this distortion did not occur. Therefore charts were remembered as being more symmetrical than they actually were.
There are thousands of studies scientists did on graphic perception and visualization. Factors impact human being’s perception on visualization. Therefore change the final analysis result.