Perception Essay
Data Visualization
Mohnish K Singh - ANLY 512-50
08-17-2017
This ‘perception essay’ intends to explore a) What do we know about the role of perception in visualization? And b) Why is it important?
Graphical perception is defined as the visual decoding of the information encoded on graphs. Just like we learn grammar before fully understanding a language, it is important to understand the basic rules of human perception before beginning to visualize the data. Choosing effective visual encodings and effective visualization design requires knowledge of visual perception. Researchers William Cleveland and Robert McGill’s in their seminal 1984 paper on Graphical perception: Theory, experimentation, and application to the development of graphical methods - identified set of elementary but critical perceptual tasks (or building blocks or elementary graphical encodings) that people carry out when out when they extract quantitative data from graphs. These tasks are critical factors in determining the performance of a graph. Then they ordered these tasks on the basis of how well people performed them during the experiments set up by Cleveland and McGill. These ordering were based on a combination of psychophysical theory and experimental results.
1. Position along a common scale 2. Position along nonaligned scales 3. Length, Direction, Angle 4. Area 5. Volume,Curvature 6. Shading, color saturation
The power law of theoretical psychophysics (Stevens 1975) says that p1/p2 = (a1/a2) ^ z, where p = perceived magnitude, a = actual magnitude. Per Baird (1970), for visual perception this power law of theoretical psychophysics is a good description of reality. When z = 1, the perceived scale is the same as the actual physical scale. Baird (1970) through his experiments found a pattern that z is reasonably closer to 1 for length judgments, smaller than 1 for area judgments and even smaller for volume judgments.
Prior to 1984 little in terms of theory existed concerning graphical perception. Cox (1978) argued “There is a major need for a theory of graphical methods” and Kruskal (1975) stated “.in choosing, constructing, and comparing graphical methods we have little to go on but intuition, rule of thumb, and a kind of master-to-apprentice passing along of information.there is neither theory nor systematic body of experiment as a guide”. Cleveland and McGill through their paper put forward certain definitive guidelines about how to design better graphs to allow better accuracy when people decode the information encoded on graphs and to redesign old graphical forms and to design new ones. In simple words, the visualization methods are only successful if the decoding is effective. Per J.D Mackinlay (1986, 2007) - assessing the importance of visual encodings on graphical perception enables designers to optimize their visualizations and is vital to the design of automatic presentation software.
Based on their 1984 study, Cleveland and McGill offered alternative graphical forms as replacements to old but popular graphical forms. For example, dot charts and dot charts with grouping replaced bar charts and divided bar charts. An important aspect of their work was to develop basic principles of graphical perception and to develop a framework to organize knowledge and predict behavior. They advocated the construction of a graphical form that uses elementary perceptual tasks as high in the hierarchy as possible. The hypothesis Cleveland and McGill (1984) used was that by selecting as high as possible, they will elicit judgments that are as accurate as possible and as a result the graph will maximize a viewer’s ability to detect patterns and extract the quantitative information. Through their experiments, they were able to demonstrate that there is an increased ability to perceive patterns as a result of the increased accuracy of perceptions. Cleveland and McGill’s (1984) perceptual theory also serves well as a guide for designing graphical methods for statistical analyses.
Jeffery Heer and Michael Bostock (2010) used the experiments results of Cleveland and MacGill to further validate those in the current world of automated design and crowdsourcing techniques and to test their relevance in a different environment and settings. Heer and Bostock (2010) hypothesized and validated that graphical perception is affected by other design parameters and data characteristics such as contrast effects, aspect ratio, and scale among others. Heer and Bostock’s (2010) chart height and gridlines experiment suggests optimized parameters for displaying charts on the web; gridlines should be spaced at least 8 pixels apart and increasing chart heights beyond 80 pixels provides little accuracy benefit on a 0-100 scale.
Appendix
References