Ruth Clark claims that "games don't teach," an obviously false statement. She has some legitimate points about matching the game design to the learning outcomes, but her claim that no research supports using games for anything other than "drill and practice" type activities is clearly incorrect. She makes this claim without addressing any work by Squire, Aldrich, etc., so it appears she didn't do a literature review prior to writing.
She cites one study with two games that were less effective at helping learners remember, and she believes that discounts the dozens of other studies on the topic. First, maybe those games were poorly designed. Second, if you're just measuring "transfer and retention" rather than application, I wouldn't be surprised that a game didn't do as well. Games are often better at moving from recall to application--but of course, she didn't measure application.
The author calls this a new learning theory combining behaviorism & cognitivism. I see a new instructional design model that combines elements from a number of different sources, but I'm not sure I see a new learning theory. The model seems very complex; how long would you have to work with this before you internalized all the separate parts of the model?
Student results were better using this model. However, the control group was tested before doing a roleplaying game and the experimental groups did the game prior to testing. This could just show that roleplaying helps students understand characters in the Aeneid. Free registration required.