Visualization is good. Visualization is valuable. Let's have more visualization.
That's great, but what does visualization really mean? In the past I have always thought of visualization as primarily a mental process: you receive some knowledge (from any of various sources) and, when you understand it thoroughly, you can "create a picture of it" in your mind. Nowadays computer graphicists are trying to place this picture more directly in the mind by creating the pictures with a computer. (This, of course, has been done for some time using more conventional illustration media). The term "visualization" has come to be a proper noun referring to the actual picture or computer image itself, as in the phrase "I created a visualization of the process on the screen". Even though the visualization is on a piece of paper or a computer screen, the ultimate destination is the mind.
Neurophysiologists have found that the two hemispheres of the brain seem to process knowledge in different ways. The left side does the logical symbolic things and the right side the intuitive visual things. While this may be a gross oversimplification it at least serves as a useful metaphor for a dichotomy that arises often enough that it must mean something. The tradeoff in communication seems to be between symbolic text and intuitive visuals. I have spent much of my life creating computer graphics for scientific visualizations so I obviously think it is basically a good idea. Even so, I would like to point out some problems with visualization, situations in which text works better or where visualization is mishandled.
There are some things that pictures are simply not good at. Among these are:
Since we live in a world of three dimensions we find it difficult to visualize geometrical figures in higher dimensions. In fact, pictures really only have two dimensions; images of three dimensional shapes are only projections onto two dimensions. Four dimensions can be shown using time, either in animation or by displaying rows of images much like a strip of movie film. One can hint at five dimensions with a 2D grid of images, but this begins to be clumsy. Attempts to go much higher in dimensionality are a real problem. The problem is significant because whole new phenomena arise at higher numbers of dimensions. As a low dimensional example consider simple mathematical functions. One dimensional functions can have local maxima and local minima. Bivariate (two dimensional) functions can have these, but they can also have saddle points, a concept that is inexpressible in one dimension. Even more complex topological combinatorics can arise in ten dimensional string theory that is the forefront of theoretical physics. Mathematicians dealing with such higher dimensional constructions work almost entirely symbolically, with equations and coordinate vectors. Even in this arena, one can benefit from making diagrams in lower dimensions, but the real work is done symbolically, with equations.
Another problem with explicit visualization is its very explicitness. If you make a diagram of a geometrical theorem about right triangles that diagram must of necessity be of one particular right triangle. You lose the sense of generality when you do this. Generality is a concept that text is good at: the simple word "triangle" implicitly means a general triangle without any special restriction. Here, even though text represents the generality better most people would still like a diagram and will accept the fact that the diagram is meant to convey a more general situation than it explicitly shows.
In other cases, there may be good ways to visualize something but it is not clear that they should be visualized. The danger comes from the power of images to confuse the viewer or to distort data.
A scientist who is just beginning to work out a theory might create pictures simply to test out some ideas. The problem comes in when the quality of the graphics is better than the quality of the theory. Since it is so easy to construct flashy computer graphics, even half baked ideas are endowed with a believability beyond that which even the originator intends. Perhaps researchers should be encouraged to make low quality graphics on purpose for theories that are not yet well supported.
Another problem is with theories that deal with only a part of a system, e.g. a theory of global temperature variation on an asteroid whose precise shape is not known. To make pictures you need explicit data about the shapes of things. If that's not the part of the system the theory covers, the visualizer is forced to make up shapes for the sake of getting a picture.
A related situation occurs in the use of rendering techniques that can imply relationships that don't exist. For example, stipple patterns simulating grey levels should be kept random to avoid imposing regular patterns where they don't really exist.
Visualizations often translate physical quantities into the graphic elements of shape or color. If the observer hasn't been properly educated in this translation confusion can result. One prime example of this is the use of color banding to enhance contrast of monochromatic data. Low values will, for example, be translated into blue, medium values into green, higher values into yellow and highest into red. While this may make some features stand out it is still an inherently dangerous visual metaphor. I have actually heard a worst case scenario of this usage in connection with some colorized photographs of Halley's comet taken by a Russian space probe. A radio announcer describing the event said something like: "Halley's comet is very colorful. It has a bright red core, surrounded by a green shell which is surrounded by a larger blue shell."
Some situations demand a schematic visualization treatment rather than a realistic one. This typically happens when showing situations which simultaneously have a wide range of sizes or time scales as, for example, planetary orbits. Here, radii of planets need to be exaggerated so that they are visible on a solar system sized picture. Another example might be the representation of electron flow through a metal, where quantum mechanical effects are replaced with classical approximations. The danger comes from the crispness and mathematical precision of computer images. A viewer may think that since some parts of the image are precise, then all parts must be mathematically correct. The visualizer should use some technique, such as purposely sketchy or irregular line textures, to alert the viewer of the schematic nature of the image.
When making images, the best design advice is to simply be conscious of your design decisions. These decisions include choices of colors, line weights and textures, sizes and relative positioning of shapes. Just the action of asking yourself what each choice is saying, even subconsciously, gets your thought processes going in the right direction.
I have hit upon several tricks to help me decide if a design works. One trick is to decide, before making any images, what is intellectually the most important part of a diagram. Then make a trial image, stand about twelve feet from the screen, look away from the screen and then look back quickly. What is the first thing you see? If that is the object you previously determined was the most important thing, you have succeeded. If not, you need to change something. Good design is not just useless decoration, it can serve to direct the viewers attention to the intended message.
When visiting a museum, sketching the exhibits encourages you to pay attention to what you are experiencing. When listening to a lecture, taking notes encourages you to pay attention while the physical motion of writing reinforces learning. In a similar way, making visualizations encourages us to pay attention to details of our problem we might otherwise never think about. I have had several experiences working with scientists where the visualization process itself led to questions about the theory that they hadn't considered.
Left brain symbolic thinking may be best for some types of problems while right brain visual thinking is best for others. But to really understand something well you must be able to think about it with BOTH halves of your brain. There are many paths to enlightenment and visualization is only one of them.