Here's a couple of quotes from one thread:
Random Review wrote:emer1ca wrote:Unless your goal is to speak the language ASAP, you're better off sticking to the passive learning plateau.
I do think a long phase of mostly passive work is useful. I see so many people here rushing to speak ASAP and ending up speaking terrible Chinese in a horrible accent.
I'm never quite sure that shows anything -- correlation is not causation, and I've met more learners with terrible language and terrible accents than I've ever met with good ones, and I've never seen any evidence of one particular school of thought resulting in better long-term acquisition, although people claim that for pretty much every one of them. I also believe I've got good results in most of my languages and I've always focused on production early on.
My accent in Italian isn't great (Italians often actually think I'm Spanish) and Italian is the only language I've learned without any real conscious thought about pronunciation (except consonant gemination, which even now I have to think about). I'd say I've spent a lot less time on Italian than any other language, but the lack of conscious pronunciation work is the biggest difference from what I usually do.
I think the "speak first!" "no listen first!" "no read first!" argument is the wrong way of looking at it.
I've said it before, and I'll probably say it another thousand times before I die: language is redundant, which means you rarely have to be accurate to be understood, and you can understand a piece of language without noticing all the details.
For example, if a foreign guy was to tell me he went somewhere "in the bus", I would understand him. Conversely, if I told him I'd gone somewhere "on the bus", he wouldn't need to notice that I'd used the preposition "on" instead of "in" and he could simply understand it as "on the bus" and never notice I'd said anything different. So neither "speaking" or "listening" leads to correct grammar.
It's also the same with pronunciation, because there are very few minimal pairs that can't be disambiguated from context.
There's a lot of talk in the second language literature about "noticing" -- the students that are most successful in immersive settings seem to be the ones that are consciously aware of the language they're hearing and repeating. The students that are most successful in classes based on parroting seem to be those who are thinking about the components of the phrases they're repeating. There's also the notion that "focus-on-form", activities which require direct attention to wordforms and grammatical patterns while also dealing with meaning, consistently appears to be a feature of successful learners and effective classrooms.
It may be my own confirmation bias, but as I recall it, I've seen various papers giving figures to support the notion that conscious thought is required throughout the early stages of the process to learn a language well, and all I've seen arguing the other way is philosophising and assertions without evidence.
I believe early speaking is successful if you are speaking something that's a good approximation of "correct" from day one, and early listening is successful if you're able to pay attention to the details well and build up a notion of "correct".
However, I think the problem is that a great many teachers and course writers have a philosophy of "survival first, accuracy later", and whatever your mode of learning, if you're using shortcuts instead of learning the underlying mechanics of the language, then you simply aren't learning the language. You'll learn yourself into a dead end, boxed in by all the mistakes you've collected on the way.
Self-teachers often fall into this same trap of "survival first", often because it seems logical to serve your immediate needs, but also because as someone who doesn't know the language, it's hard for you to lead yourself to accuracy. (Which is where I feel that my admittedly limited study of linguistics helps me -- I'm much better able to understand a grammar book or a pronunciation guide than I was before the course, so I can get a fairly good idea of what is correct and accurate, and if I feel an explanation in a book is not telling me the whole story, I know the sort of things I need to look up the fuller answer.) This is one of the reasons I think a good teacher or tutor is worth the expense -- someone who knows what accuracy is can teach you it.
So my view, partly arising from my opinions but supported by a lot of literature, is that noticing and focus-on-form are the most important factors in success as a learner whatever the mode and medium of learning.
My personal extension to that is that I believe accuracy can only really be reliably practiced in spoken production, because many tasks in listening, writing and reading can all be accomplished seemingly successfully without necessarily needing to be completely accurate and notice everything.