The rise of data visualization has largely affected the way we “digest information”. Graphical display provide “powerful tools” for “sense-making” (data exploration) and communication (data presentation) (Schriger & Cooper, 2001). For example, compared to numerical tables, scatterplot has much higher efficiency in conveying information – numerical tables are extremely time-consuming especially when with a large sample size (Legge, Gu & Luebker, 1989). As the saying goes, “a picture is worth a thousand words”, but this can only be done by a well-designed picture. Thus, how can the information be accurately communicated via graphics and how to decide the type of graphics to best do this? Understanding graphical perception can give us some significant implications for the “display of data” (Cleveland & McGill, 1985).
Actually, graphical perception is an information visually decoding process. According to Cleveland and McGill, human graphical perception is a complicated process accomplished on the basis of “a set of elementary perceptual tasks”. For instance, when we look at a bar chart, the assessment of the height of each bar is the judgement of “position along a common scale”; when we look at pie chart, the information is extracted by judging the angle size, etc (1989). These elementary graphical encodings are empirically ranked as 1) position; 2) length, direction, and angle; 3) area; 4) volume and curvature; 5) shading and color saturation and the higher ranking in the hierarchy, the better perception of information conveyed (Cleveland and McGill, 1989). For example, the bar chart (position judgement) can help people more quickly and accurately understand the variance in values than pie chart (angle judgement). This finding has been testified by crowdsourced study that position encodings are significantly outperformed length and angle, which is better than area (Heer & Bostock, 2010). Moreover, this experiment done by Heer and Bostock in turn proved the viability of crowdsourcing, a cheaper and scalable means of graphical perception experiment in the future (2010).
There are many other researches done on graphical perception using Cleveland and McGill theory. Crowdsourcing graphical perception is one of them. Crowdsourcing research uses Mechanical Turk platform created by Amazon (Known as MTurk), a micros task market. Requesters can post tasks and pool of users can work on complete these tasks. Users get rewarded for each task. Main goal was to replicate previous perception experiments on crowdsourcing platform. Mechanical Turk research shows that graphical perception experiments are viable using crowdsourcing. Also it revealed new aspects of results on rectangular area judgement experiment.
Perception of circles is very important for cartographic environment. Patricia Gilmartin has well presented this topic in ‘Influence of map Context on Circle Perception’ paper. There are many factors that can affect characteristics of maps with use circles. There are different effect on visualization, because of size difference between target circle and circles around it. For e.g., target circle looks smaller when adjacent circles are larger and also target circle looks larger when placed with smaller adjacent circles. In above figure, both target/center circles are of same size but appear to be of different size.
Knowing graphical perception is critical to “effective visualization design”. It can provide guidelines for data display that interpret information like patterns and behavior more clearly. For example, as shown in Figure 1, identical reference grids can make the comparison of curves much easier to be decoded since it gives the base lengths as well as avoiding cluttering the graph.
In conclusion, data visualization is actually a method designed to convey both quantitative and qualitative information more clearly, more simple and easier for perception. These information are encoded by a set of complex elementary graphical encodings. As graphical perception theory identified these encodings and ranked them based on their accuracy and effectiveness of conveying information, guidelines can be provided for selecting graph type for better data display.
References Cleveland W.S., McGill R. (1989). “Graphical Perception: Theory, Experimentation, and Application to the Development of Graphical Methods”. Journal of the American Statistical Association. 79(387): 521-534. Cleveland, W. S. (1994). The Elements of Graphing Data. Murray Hill, NJ: AT&T Bell. Heer J., Bostock M. (2010). “Crowdsourcing Graphical Perception: Using Mechanical Turk to Assess Visualization Design”. ACM Human Factors in Computing Systems (CHI). 203–212 Legge G.E., Gu Y.C., Luebker A. (1989). “Efficiency of Graphical Perception”. Percept Psychophys. 46(4): 365-74. DOI: 10.1016/j.injury.2008.01.050 Schriger D.L., Cooper R.J. (2001). “Achieving Graphical Excellence: Suggestions and Methods for Creating High-Quality Visual Displays of Experimental Data”. Annals of Emergency Medicine. 37(1):75-87. DOI: 10.1067/mem.2001.111570