Sometimes we run across a concept that seems like it should be simple, but for some unknown reason, we struggle to grasp it. In those times when contradictions and conflicting explanations abound, it is best to get back to fundamentals and, as my physics professors in college would say, "prove it to yourself."
In my field of data visualization, there is one topic for which nobody seems to be able to agree: how to properly scale bubbles for bubble charts.
The general consensus seems to be that we should scale them by area, since that is what gets printed on the page and that is what our eyes "see." Nathan Yau of the well-regarded FlowingData blog, as well as Edward Tufte himself-- grandfather of data visualization-- and even Wikipedia, agree.
That is why I was surprised to read Alberto Cairo's dissenting opinion this evening in his excellent book "The Functional Art." In chapter 2, Dr. Cairo argues that what the brain perceives, even in the case of a circle, is length in one direction only. That is, despite the printed page presenting us with 2 dimensions' worth of data, our brains are not good at processing them both at the same time and instead will naturally hone in on diameter (length in a single direction).
He illustrates this point further with an analogy to the relevant information our brains would process in a bear attack, relating that we don't size up an approaching animal by both width and height to discern its danger. A towering height alone is enough for our brains to conclude that we would be dead meat-- convenient, since we only have milliseconds to figure that out!
In the past, whenever I came across an argument for how to size bubbles I would get flustered because the conventional wisdom (encode to area) was at odds with what my brain intuitively knew to be true (the brain "sees" diameter rather than area). Rather than continue to rely on words as explanations, I decided to flex my Illustrator muscle and make an infographic that explains how to make bubble chart infographics (how very meta)!
Cairo goes on to recommend that we never use bubble charts when attempting to compare numbers precisely. Since our brains only perceive a single length dimension anyway, it is a more direct solution to encode to bars in a bar chart, and leave bubble charts for when we are making rough, relative comparisons. Good point, but I don't see a reason why we shouldn't encode values to diameter in order to keep even relative comparisons more accurate.
Still can't believe that it took me at least 2 years to figure out that a visual topic could be best explained... visually!
