In short, it takes an empirical look at all minimal guidance instruction (by whatever name you want to call it), and when it comes down to the numbers--which to me, are what matters in research--it can't stand up.
I guess what I like most here (other than the conclusion ;-) ) is how systematically this was approached. I liked the memory approach. It makes perfect sense.
I think the authors were fair as well. They did distinguish between novel information (information new to the learner) and information for learners who already have some measure of expertise within the subject-matter area. In short, the minimal guidance approach can actually lead to a loss of learning with novel information (hey, they cite the sources of the studies), can produce more errors, does not produce expertise, and has no more effect than guided instruction in experienced learners.
One interesting question for those proponents of minimally guided instruction brought up in the article. Handlesman et al. (2004--and I'm not citing this here, just look at the .pdf) ask:
". . . why do outstanding scientists who demand rigorous proof for scientific assertions in their research continue to use and, indeed defend on the bias of intuition alone, teaching with methods that are not the most effective?"
Good question. I don't get it either.
I WHOLLY RECOMMEND THIS ARTICLE.