If we were to design Information Visualization from scratch, we would start with the basics: understand the principles of perception, test how they apply to different data encodings, build up those encodings to see if the principles still apply, etc. Instead, visualization was created from the other end: by building visual displays without an idea of how or if they worked, and then finding the relevant perceptual and other basics here and there.
This approach has the problem that we end up with a very patchy understanding of the foundations of our field. More than that, there is a good amount of unproven assumptions, aesthetic judgments, etc.~mixed in with the evidence. We often don't even realize how much we rely on the latter, and can't easily identify them because they have been so deeply incorporated into the fabric of our field.
In this paper, I attempt to tease apart what we know and what we only think we know, using a few examples. The goal is to point out specific gaps in our knowledge, and to encourage researchers in the field to start questioning the underlying assumptions. Some of them are probably sound and will hold up to scrutiny. But some of them will not. We need to find out which is which and systematically build up a better foundation for our field. If we intend to develop ever more and better techniques and systems, we can't keep ignoring the base, or it will all come tumbling down sooner or later.