Data visualization is a powerful tool to illustrate and communicate information graphically to help understand the data better and more easily. Human perception, as stated by William S. Cleveland and Robert McGill in their theory on Graphical perception[i], is the visual decoding of information encoded in graphs, and understanding it is very important for the success of data visualization as it depends on what our eyes can discern and brains can understand. An understanding of human graphical perception can greatly improve the quality and quantity of the information being displayed. Jacques Bertin, in his book published in 1967 on “The Semiology of Graphics”, explains his discovery that visual perception operates according to rules that can be followed to express information visually in ways that can represent it intuitively, clearly, accurately, and efficiently[ii].

Christopher G. Healey in his paper on Perception in Visualization[iii], states that there are a limited set of visual properties that are detected quickly and accurately by the low-level visual system. This is referred to as Preattentive processing. Visual preattentive features include line orientation, length, width, closure, size, curvature, density, number, color, flicker, direction and velocity of motion, etc. An example of a preattentive task is the detection of a red circle in a group of blue circles:

Other preattentive visual tasks include target detection, boundary detection, region tracking, counting and estimation. In data visualization, If the low-level visual system is harnessed, it can be used to effectively draw focus to areas of potential interest in a display. This can be done by assigning different visual features such as color, shape, size, texture, direction, etc. to different data attributes while keeping in mind that for certain tasks, the visual system favors one type of visual feature over another.

Cleveland and McGill[i] identified a set of elementary perceptual tasks and the order in which they are carried out, based on a combination of psychophysical theory and experimental results, when a viewer extracts information from a graph:

  1. Position along a common scale
  2. Positions along nonaligned scales
  3. Length, direction, angle
  4. Area
  5. Volume, curvature
  6. Shading, color saturation

In the ordering above, they found through their experiments that position judgements are more accurate than length judgements which are more accurate than area judgements which, in turn, are more accurate than volume judgements. They also suggested alternate graphing charts such as bar charts as replacements for pie charts, framed rectangle charts as a replacement for statistical maps with shading, grouped dot/bar charts as replacement for divided bar charts.

Bar chart instead of Pie Chart to represent quantitative data more accurately:

Framed rectangle charts instead of statistical maps with shading for easier visual comprehension:

Grouped dot chart instead of divided bar chart:

Michael Bostock and Jeffrey Heer found similar results in their paper using crowdsourcing on Amazon’s Mechanical Turk[iv] where they tried to replicate Cleveland and McGill’s theory.

The founders of the Gestalt School of Psychology studied how humans perceive pattern, form and organization in what we see and found that we organize what we see in particular ways to make sense of it. They defined the following Gestalt principles of visual behavior[ii]:

  1. Proximity - Objects that are close together are perceived as a group
  2. Similarity - Objects that share similar attributes such as shape or color are perceived as a group
  3. Enclosure - Objects that appear to have a boundary around them such as those formed by a line are perceived as a group
  4. Closure - Open structures are perceived as closed, complete, and regular whenever there is a way that they can be reasonably interpreted as such
  5. Continuity - Objects that are aligned together or appear to be a continuation of one another are perceived as a group
  6. Connection - Objects that are connected, say, by a line, are perceived as a group

In choosing a visualization, it is important to consider the above principles of human perception in providing an unambiguous and easily consumable representation of the data. Consider the following two visualizations of the same data:

Pie Chart:

Bar Graoh:

For the given data, the bar graph displays the information more effectively than the pie chart as the quantities are more accurately represented making it easier to compare the quantities, view the ranking order of quantities and provide distinct information to the viewer on the spread of the data as well as the take away from it.

Hence, human perception has an important role in visualization as it supports the cognitive associated process and an understanding of the same helps in development of effective data visualizations.

References:

[i] Cleveland, William S. and McGill, Robert. Graphical perception: Theory,experimentation, and application to the development of graphical methods. Journal of the American Statistical Association, Vol. 79, No. 387. (Sep., 1984), pp. 531-554
[ii] Few, Stephen. The Encyclopedia of Human-Computer Interaction, 2nd Ed
https://www.interaction-design.org/literature/book/the-encyclopedia-of-human-computer-interaction-2nd-ed/data-visualization-for-human-perception
[iii] Healey, Christopher G. Perception in Visualization. Department of Computer Science, North Carolina State University
https://www.csc2.ncsu.edu/faculty/healey/PP/#Table_1
[iv] Heer, Jeffery and Bostock, Michael, Crowdsourcing Graphical Perception: Using Mechanical Turk to Assess Visualization Design. Computer Science Department, Stanford University