According to one definition¹, engineering is making things based on scientific principles – as opposed to the intuitive making that defines a craft. Information visualization (InfoVis) is practiced like a craft today, based mostly on practical examples, but not on theoretical basics. Here is a sketch of not only InfoVis as an engineering field, but InfoVis as a science.
We could model the science of InfoVis after physics, a well-established science. If we assume that we already have the engineering part, then there’s also theoretical physics, experimental physics, and computational physics. That last one is of course less interesting for a field that can’t exist without computers, so we’ll just treat it as a part of the other two. Other than that, this seems to be a good model for visualization. Work certainly already exists in theoretical and experimental visualization (even though most user studies for testing techniques would be better considered engineering), but a better understanding of the different types of work would help steer that research, and also make it easier to put work into context.

The following is a rough sketch of the work in those different areas, and what the connections between them would be:

• Theoretical InfoVis. Propose theories for a better understanding of visualization, based on observed phenomena (like experiments and user studies) and other theoretical work from visualization, psychology, aesthetics, semiotics, etc. These models not only need to explain what has been observed, but also allow the prediction of properties that new techniques should or will have to have to be successful.
• Experimental InfoVis. Devise experimental methods for testing theories from theoretical InfoVis as well as understanding observed phenomena in InfoVis engineering. Perform experiments and propose theories and further experiments based on their results.
• InfoVis Engineering. Application development and user testing. Design new techniques and applications for specific user tasks, and test their success in user studies. Devise approaches to basing application development on the results of theoretical and experimental InfoVis. Feed back results into the other two areas.

Given the size of the field, there would (and should) certainly not be a clear distinction between these areas of research in terms of the people involved. A clearer distinction for the purposes of paper submissions and discussions would be helpful, however.
It should also be noted that all of these fields have a meta-role as well: they need not only develop their theories, applications, etc., but also the methods for doing that work. and for translating their results for the other fields.
One specific difficulty in visualization (as opposed to physics and other hard sciences) is that it is much more difficult to move research results between the different fields. There is no well-defined underlying universal basis (like mathematics) to allow translation of results: the common factor are human perception and cognition, fields that are still not very well understood, and won’t be for the foreseeable future.

Some of the above certainly also applies to scientific visualization and visual analytics, and theoretical work would hopefully lead to a much better understanding of the considerable overlap between these fields.
InfoVis will not be able to survive as a field without becoming a proper science. The methods for this science still need to be developed, and we need to do more than just copying other fields (like we do to an extent right now with perception, computer graphics, etc.). But the field needs to move beyong being a craft.

¹Mitcham C., 1994. Thinking through Technology: The Path between Engineering and Philosophy. Chicago,University of Chicago Press

( by Robert Kosara @ eagereyes.org)