This wonderful article on the science behind data visualization was written by Graham Odds and posted on .Net Magazine. You can read an excerpt from it below or the full article here.
Graham Odds outlines principles that will help you to design more compelling data visualisations
Over the last couple of centuries, data visualisation has developed to the point where it is in everyday use across all walks of life. Many recognise it as an effective tool for both storytelling and analysis, overcoming most language and educational barriers. But why is this? How are abstract shapes and colours often able to communicate large amounts of data more effectively than a table of numbers or paragraphs of text? An understanding of human perception will not only answer this question, but will also provide clear guidance and tools for improving the design of your own visualisations.
In order to understand how we are able to interpret data visualisations so effectively, we must start by examining the basics of how we perceive and process information, in particular visual information.
System 1 vs System 2
Daniel Kahnemann, in Thinking, Fast and Slow, introduces the terms System 1 and System 2 to differentiate between the information processing that occurs in our sub-conscious and conscious minds respectively. The former encapsulates the functions that are uncontrolled, always-on and effortless, while the latter refers to those functions that are controlled but require effort to engage.
To better understand the differences between System 1 and System 2, consider Figure 1. In the photograph on the left we immediately perceive an angry man and probably associate loud noise and aggressive movement with the depicted scene. This exceedingly sophisticated interpretation of mere pixels is almost immediate, requires no effort and comes completely naturally. Contrast that with the multiplication on the right. We instantly recognise what is being asked of us and that we are able to work it out, but most will not attempt the mental arithmetic involved because of the conscious effort required. The initial reactions in both cases are pure System 1, while the mental arithmetic is an example of System 2.
Figure 1. Perception tasks that primarily trigger System 1 and System 2 respectively
We have evolved these separate systems so that our conscious minds do not become swamped with mundane processing. Our System 2 can focus on more complex comprehension and calculation tasks, with System 1 feeding System 2 with the necessary information for such tasks. In data visualisation, we should seek to encode as much information and understanding as possible in a way that is perceived correctly by our System 1, which then frees up System 2 for more involved understanding and analysis of the data.
Having introduced a high-level, abstract view of how we process information, we can now turn attention to the problem of how the information to be processed enters our minds in the first place, via our senses. A significant amount of the human brain is dedicated to visual processing, resulting our sight having a sharpness of perception far surpassing our other senses. As you can see from Figure 2, more information enters our minds at any given time through sight than through any of our other senses, both at the sub-conscious and conscious level. In fact, roughly 70% of the body's sense receptors relate to sight.
Figure 2. Sensory bandwidths reaching sub-conscious and conscious mind, from Tor Norretranders' The User Illusion
We can also see from Figure 2 that visual information, like all the sensory information, is heavily reduced between our sub-conscious and conscious. This is not because information is simply discarded, rather it is distilled by our System 1 so that our System 2 receives less, but richer information more relevant to whatever task we are currently undertaking. Sight's combination of bandwidth and processing power is why it is more suited to comprehending data sets than our other senses.
Continue reading here