Simulation is of continually increasing importance in training. Cost, energy, and safety are among the driving factors. The goal of simulator development is to provide increased training effectiveness. This is frequently considered to be almost synonymous with increased simulation realism. Actually, the relationship between simulation realism and training effectiveness is far from simple. Realism itself is not a simple scalar - there are types of realism, and functions of realism must be considered in multidimension space. This applies to cost as well as to training effectiveness. Questions in this area may not be difficult to answer, once they are asked. There is a continual danger that important decisions will be made, and made incorrectly, due to failure to ask the pertinent questions. This paper considers results from simulation experiments and training activities over the years. In particular, surprising and non-intuitive results are examined. The major goal is to illuminate areas of pertinent questions, although in some cases tentative answers are presented.