This post is part of a larger series focused on exploring the fundamental principles of data visualization. Eventually, the collection may grow into something larger and more coherent. For now, each post simply picks up and plays with one idea related to how we represent data visually. Other posts in this series can be found using the Form to Data tag.
Most charts use more than one kind of encoding at once. Whenever two or more encodings are used in the same chart, we need to consider how they work together and their relative importance on the page. The visual importance of an object is called its perceptual salience, which describes how much it stands out from the rest of the items in the chart.
For example, red is a particularly salient color, and so is likely to stand out strongly from other things on the page. Dark or very bright colors tend to have high salience, while lighter or more neutral colors tend to fade into the background.
Of course, simply increasing the salience of everything doesn’t help: you just end up with a lot of shouting, like a street full of neon lights. Instead, it is helpful to think about creating a sense of hierarchy to tell the reader what is most important and help them navigate through the data.
We already saw examples of combined encodings in the visual variables post. The images below use both shape and color encodings to draw a set of marks. In the first image, both shape and color represent identity, and they are tied to the same information channel: purple circles have one identity, and green stars have another.
The second image uses the same two visual variables, but applies them to different identity channels. Now there are green shapes and purple shapes (color encoding for identity channel 1), and circles and stars (shape encoding for identity channel 2), but we’re not sure how those two sets of identities relate to each other.
It is still possible to read the chart by focusing on just one variable at a time, looking for all the stars, or all the purple shapes. Psychologists call this selective tuning, which is basically a fancy word for being able to focus on just one thing at once in a visually cluttered world. Visualizations that apply the Gestalt principles in a straightforward way work with our automatic classification systems and do not require much effort to read. These charts feel “intuitive” and can be read “at a glance.” Selective tuning is a slower, more deliberate process that often requires conscious effort. Visualizations that require viewers to use conscious selective tuning can often feel “cluttered,” “busy,” or “hard to read.”
Our brains are astonishingly good at separating out layers of information from a busy image. In the example below, your brain is supported in this task by the fact that the words themselves have meaning, giving it a trail to follow as it works its way through the information. (Text courtesy of Wikipedia.)
Here, the layers of information were given relatively equal weights (though the large text is a bit stronger than the rest because larger size tends to increase the salience of an object). We can change the emphasis by adjusting the weight on the page of different parts of the image. The image works best when the layers of information are separated into a clear hierarchy, so that your brain can easily tell which things belong together. Adjusting only the gray values and drawing order can put the chart on top, or make the large text the focus of the image.
Color, type weight and style, and other graphical elements can help to further separate the layers, making it easier to focus on just one thing at a time. Here, the color helps your brain to group the different objects together, making it easier to separate out the different layers of information.
Using two encodings to represent the same channel is called double encoding, and it can be helpful for increasing the legibility of a chart that has a lot going on. The image below uses just a shape encoding. It is possible to pick out the stars or the circles, but if you try you will probably find yourself mistaking a star for a triangle and a circle for a hexagon. This happens because your brain uses “pointiness” and “roundness” as first-pass attempts to classify shapes into bins, and so tends to pre-screen pointier shapes into the same category, and then correct when you look more closely.
We can make selective tuning easier by adding a color encoding to reinforce the differences between shapes.
How we apply the colors also matters: above, I chose similar colors for hexagons and circles, and stars and triangles, which actually reinforces the tendency to confuse them. Using colors that are more distinct helps to separate the shapes even further, and helps you to sort them faster.
Sometimes you may need to use conflicting encodings to represent all of the important parts of a dataset. In these cases, double-encoding one (or both) of the identity channels can help to reinforce the separation that you need to create. Here, we are using shape and color to represent conflicting identity channels, as before.
Adding a position encoding helps to separate them into clearly distinct groups.
If position is not available, you could use value instead. Because the value variable is usually perceived to represent a quantity/modifier rather than an identity channel, the danger here is that people will perceive the lighter marks as less important, which might not be appropriate to the meaning of your graph.
Texture is another option to help reinforce identity.
In all of these cases, the additional encoding helps to make the layers of information as distinct as possible to support the reader’s selective tuning, making the chart faster and easier to read.