The consensus is that class size is not a particularly important factor when the goal of instruction is the acquisition of subject matter knowledge and academic skills. This appears to hold true across various class types (e.g. lecture, discussion). However, smaller class sizes are somewhat more effective than larger ones when the goals of instruction are motivational, altitudinal, or higher-level cognitive processes. (Pascarella & Terenzini, 1991, p87).
Putting this in perspective with the training world - it depends. For example, a trainer in the UK used to train heavy construction equipment (e.g. bulldozers, earth-moving scrapers, scooploaders, backhoes). he used a 4:1 learner to trainer ratio when instructing the practical exercises. When you are in a muddy field and you got 10 inches of red clay sticking to your boots and you are trying to train, coach, provide real life experiences, and at the same time be safety conscience due to the number of things that can happen with a beginner on a mega-ton piece of equipment, then you start to get the picture for such a low ratio.
Also, the 4:1 ratio is used in other aspects of training, but I believe it has more to do with being symmetrical (the trainer in the middle with a student at each corner), than it has to do with any real learning advantage.
During some diversity training I facilitated, the class sizes were kept small, under 10 learners, as the designers wanted to allow room for a lot of interaction to take place. So while one instance might call for small class sizes due to safety and the complexity of the training, another might be kept small to allow for certain activities to take place.
I believe the best approach is to analyze your training requirements, check on classroom facilities, and then choose your class size accordingly.