For a long time, I did my level best to avoid speaking English in my Spanish classes. I would use pictures, gestures, comparison charts, whatever, to avoid telling students what a given word or phrase meant. The idea was for me to create "Aha!" moments*, situations in which it was easy and natural for students to engage, and for engaged students to accurately guess the meanings of words and phrases. Somewhere around here I have an article that suggests that this is what language learning is--the transition from confusion to creating meaning. Inherent in that idea is confusion: you have to start out not knowing what's going on, and use the tools available to you to create first a broad sense of a communication, then increasingly more accurate detail as you get a better base of the language and more skill at creating meaning. I saw speaking in English the same way I see shouting at students as a classroom management technique: it does more harm than good, it's a failure of good practice that nevertheless happens because we are after all only human.
When I went to the TPRS training (see an earlier post), I had a minor crisis of faith. It was fairly public, and it wasn't pretty. Everything I thought was essentially undermined--constant translation, followed by repetitions ad nauseam, was sort of the root of how to do it right. There was no confusion, just a continuum of processing speed. You start out as a slow processor: you have to look up at the translation, listen as the teacher enunciates every word very clearly, watch carefully as she points out every word she says, and in general try to pretend you're paying attention to the meaning. If the story hook is good enough, you will. As you go through, you have to look at the translations less and less, and your confusion between similar-sounding but very important words starts to clear up.
There's another article in support of what TPRS does, and what I'm trying to do. (The link is to a German website, but the article is in English.) It feels a little heavy on the interpretation and a little light on supporting "clinical" trials, but it has enough to be a credible source. It argues in favor of using a learner's first language to support learning a second language. I find it faintly troubling that the author appeals to 2000 years of language teaching practice as a reason to ignore recent developments--what did the Romans know about language acquisition? He does, however, take pains to point out that the learner needs to use the new language: "We do not learn any language by using another one."
At any rate, another point of research in a growing body.
*Like "teachable moment," I hate this phrase. I'm beginning to think that I just don't like the word "moment."