Gestalt Principles for Data Visualization
Gestalt refers to the patterns that you perceive when presented with a few graphical elements. So, for example, from three lines you might see a triangle.
These rules, which are commonly presented along with interesting optical illusions that they can produce, are particularly important when it comes to creating data visualization. I’m not saying anything radical — any good book or seminar about data visualization will begin with an introduction to gestalt principles of perception. Understanding gestalt is useful for improving traditional charts like bar charts and line charts. But it is critical to understand gestalt when you are creating more complex data visualization products like network visualization or hierarchical diagrams.
A bar chart or a line chart only uses a few gestalt principles, usually based on shape and negative space. Coupled with our own familiarity with bar charts and line charts as abstractions, it’s not such a big deal. But complex data visualization methods use more and different channels for expressing patterns in the data and thus provide more opportunity to confuse readers.
A while back, I put together a four part series of explorations of gestalt principles as they apply to data visualization.
In Part 1, I deal with principles of similarity, proximity and enclosure. This is probably the most broadly applicable section because proximity is a channel always present in data visualization charts, while enclosure is correlated with annotations and similarity is the principle by which color encoding works. Hierarchical charts especially rely on enclosure to convey information, particularly treemaps and circle packing.
For part 2, I deal with common fate, parallelism and connectedness. This introduces some factors that come into play with animation and how that relates to what lines can unintentionally imply. More than anything, exploring gestalt reminds you that charts are made of graphical primitives, and those primitives (like lines or circles) mean different things in different visual context. As experts in data visualization, it becomes too easy to learn a chart and forget how arcane or abstract that chart can look to a reader unfamiliar with it. Being able to break down a chart into those constituent graphical elements can help immensely with user-centered design of those charts.
Part 3 explores factors particular to animated network diagrams. Proximity returns from Part 1 with new and exciting ways to confuse your readers. It’s common knowledge that network visualization is hard to read, but a big reason for that is neglecting to account for gestalt principles and how they might be sending signals to your readers that you didn’t intend.
Finally, in Part 4, I look at how these principles are utilized to produce valuable charting methods like the difference chart. The point here, and hopefully with all of these examples, is that this is not only of academic or theoretical value. If you understand gestalt it will improve the charts you create, the kinds of charting types you decide to use, and the innovative charting methods you eventually develop.